Corticotropin-releasing factor and noradrenergic signalling exert reciprocal control over startle reactivity

Abstract

Corticotropin-releasing factor (CRF) and norepinephrine (NE) levels are altered in post-traumatic stress disorder and may be related to symptoms of hyperarousal, including exaggerated startle, in these patients. In animals, activation of both systems modulates anxiety behaviours including startle plasticity; however, it is unknown if they exert their actions orthogonally or dependently. We tested the hypothesis that NE receptor activation is required for CRF effects on startle and that CRF₁ receptor activation is required for NE effects on startle. The study examined the effects of: (1) α₂ agonist clonidine (0.18 mg/kg i.p.), α₁ antagonist prazosin (0.8 mg/kg), and β_1/2 antagonist propranolol (0.8, 8.0 mg/kg) pretreatment on ovine-CRF (oCRF)- (0.6 nmol) induced increases in startle reactivity and disruption of prepulse inhibition (PPI); (2) α₂ antagonist atipamezole (1–30 mg/kg) and α₁ agonist cirazoline (0.025–1.0 mg/kg) treatment on startle; (3) CRF₁ antagonist (antalarmin, 14 mg/kg) pretreatment on atipamezole- (10.0 mg/kg) induced increases in startle. oCRF robustly increased startle and reduced PPI. Pretreatment with clonidine or prazosin, but not propranolol, blocked oCRF-induced increases in startle but had no effect on oCRF-induced disruptions in PPI. Atipamezole treatment increased startle, which was partially attenuated by CRF₁ antagonist pretreatment. Cirazoline treatment did not increase startle. These findings suggest that CRF modulation of startle, but not PPI, requires activation of α₁ adrenergic receptors, while CRF₁ activation also contributes to NE modulation of startle. These data support a bi-directional model of CRF–NE modulation of stress responses and suggest that both systems must be activated to induce stress effects on startle reactivity.

Corticotropin-releasing hormone (CRH), CRF, norepinephrine, PTSD startle

Introduction

The acoustic startle reflex is a defensive behaviour characterized by reflexive contraction of the skeletal musculature in response to sudden intense acoustic stimuli, and is increased under conditions of threat. Prepulse inhibition (PPI) of the startle reflex refers to suppression of the startle reflex when a startling acoustic stimulus is preceded by a weaker prestimulus. Startle reactivity and inhibition of startle are altered in certain anxiety disorders, including post-traumatic stress disorder (PTSD) (Grillon et al.1996, 1998; Grillon, 2008; Pole et al.2009). In PTSD, neurotransmission of corticotropin-releasing factor (CRF), a neuropeptide released in response to stress, may also be altered (Baker et al.1999; Sautter et al.2003). CRF levels in cerebrospinal fluid have been found to be higher in PTSD patients compared to control subjects (Baker et al.1999; Bremner et al.1997). Given that CRF modulates the startle response (Grillon, 2008; Risbrough & Stein, 2006), one hypothesis is that increases in CRF neurotransmission may contribute to exaggerated startle behaviours [e.g. increased startle magnitude (reactivity) and reduced startle inhibition] observed in PTSD. Identifying the mechanism through which CRF affects startle could therefore be beneficial in providing a target for developing novel compounds to treat anxiety disorders like PTSD.

CRF acts in the brain at two known G-protein-coupled receptors, CRF₁ and CRF₂ (Hauger et al.2006). CRF₁ activation is required for CRF- or stress-induced increases in startle reactivity and decreases in PPI (Risbrough et al.2003, 2004, 2009). However, the central mechanism by which CRF exerts its effect on startle plasticity is unclear. One potential mechanism mediating CRF system regulation of startle and PPI may involve the noradrenergic system. The noradrenergic system, originating primarily in the locus coeruleus (LC), consists of α₂ (presynaptic/post-synaptic), α₁ (post-synaptic), and β (post-synaptic) receptors. The hypothesis that the noradrenergic system may mediate the effects of CRF on startle emerges from observations that CRF terminals synapse with dendrites of noradrenergic neurons in the LC and recent studies have labelled CRF₁ receptors in tyrosine-hydroxylase neurons of the LC (Reyes et al.2006). Additionally, CRF administration [intracerebroventricularly (icv) or directly into LC] increases norepinephrine (NE) release in the bed nucleus stria terminalis (BNST) (Curtis et al.1997; Dunn et al.2004; Van Bockstaele et al.1996), a critical site for CRF-induced increases in startle (Lee & Davis, 1997). The BNST is also densely innervated with noradrenergic fibres from the LC (Forray et al.2000; Koob, 1999) and requires noradrenergic signalling to control fear behaviours (Fendt et al.2005). Additionally, CRF receptor antagonist treatment or null mutation of the CRF gene blocks stress-induced changes in extracellular NE (Isogawa et al.2000; Jeong et al.2000), suggesting that CRF modulates NE release. Increases in NE signalling have been shown to increase startle both in PTSD patients and in some rodent models (Carasso et al.1998; Davis, 1980; Lahdesmaki et al.2004; Morgan et al.1995; Risbrough & Geyer, 2005). These findings support the hypothesis that CRF's effects on defensive startle behaviour may require activation of the noradrenergic system.

Given that CRF increases startle, NE increases startle under certain conditions, and CRF increases NE release, we predicted that blocking NE signalling would reverse the effects of CRF on startle behaviours. To test this hypothesis, we examined the effects of α₂ agonist (clonidine) pretreatment on CRF-induced changes in startle. Clonidine decreases NE release by activating inhibitory presynaptic α₂ autoreceptors. Second, we examined the contribution of the excitatory α₁ or β post-synaptic receptors to CRF effects on startle via pretreatment with an α₁ antagonist (prazosin), or a β_1/2 adrenergic antagonist (propranolol). To determine if NE receptor activation effects on startle result from CRF receptor activation, we tested the hypothesis that NE activation is sufficient to mimic CRF effects on startle and PPI via treatment with a selective α₂ antagonist that increases NE release (atipamezole) and an α₁ agonist (cirazoline). Finally, we used a CRF₁ antagonist (antalarmin) to determine whether NE signalling-induced changes in startle also require concomitant CRF₁ receptor activation, suggesting reciprocal permissive roles of these systems in this behaviour.

Methods

Subjects

Male C57BL/6J mice (aged 2–3 months; Jackson Laboratories, USA) were housed in a temperature- and humidity-controlled room (12-h reversed light/dark cycle, lights off 08:00 hours) (n=7–12/group per experiment). Food and water were available ad libitum. Mice acclimated to the vivarium for at least 1 wk prior to cannulation surgery. All testing occurred between 10:00 and 18:00 hours and was in accordance with the guidelines of the National Institutes of Health's ‘Guide for the Care and Use of Laboratory Animals’, as approved by UCSD.

Surgery and histological verification

Surgery and histological verification were as described previously (Vinkers et al.2007). A 23-gauge 7-mm unilateral guide cannula was placed above the lateral ventricle (flat skull; anteroposterior, −0.1 mm; mediolateral, ±1.1 mm; dorsoventral, −1.5 mm below dura). Mice recovered for >5 d before testing. A 30-gauge 8-mm injector was used for drug infusions. After study completion, mice were anaesthetized and cannulae placement was checked via dye infusion (1.5 µl). Nine mice were removed from analysis due to incorrect placement.

Apparatus

Startle and PPI testing were performed in ventilated, dark, sound-attenuating commercial startle chambers (39×38×58 cm, SR-LAB system, San Diego Instruments, USA) as described previously (Risbrough et al.2003). A loudspeaker provided continuous broadband 65-dB background noise and all acoustic stimuli. Beginning at stimulus onset, 65 consecutive 1-ms readings were recorded to obtain the average magnitude of the mouse's startle response.

Expt 1: time-course and dose response of ovine-CRF (oCRF)

Drugs

In order to identify a behaviourally active dose and time-point of the CRF₁-preferring agonist (oCRF) for use in expt 2, mice (n=9–10/group) were treated with 0, 0.06, 0.2, or 0.6 nmol oCRF (0.28–2.8 µg/5 µl icv) (Bachem, USA). Artificial cerebrospinal fluid (aCSF) was the vehicle. Doses were based on previous studies using the non-selective CRF receptor agonist rat/human (r/h) CRF which dose-dependently increased startle and disrupted PPI (Risbrough et al.2003, 2004).

Behavioural testing

Mice were placed in the startle apparatus immediately following drug infusion (‘time 0’). Startle testing was repeated hourly for 6 h post-infusion, with mice returned to the home cage for 40 min after each test (session duration 20 min). The startle session began with a 2-min acclimation period, followed by presentation of four 120-dB pulse trials (presented at start of the session to achieve a stable level of startle reactivity for the remainder of the session). These trials were immediately followed by a block of trials in which the prepulse intensity varied (10 startle pulse-alone trials, each at 120 dB, 40-ms duration, and 30 prepulse+pulse trials). The prepulse+pulse trials were evenly divided among the following trial types (10 each): a 69, 73, or 77-dB prepulse (20 ms) preceding a 120-dB pulse stimulus (40 ms). The time between the prepulse and pulse onset was 100 ms. This block was immediately followed by another block of trials in which the inter-stimulus interval (ISI) between the prepulse and pulse was varied. During the ISI block, seven startle pulse trials, each at 120 dB, were presented. Additionally, 73-dB prepulses preceded a 120-dB stimulus by 20, 120, 360 ms (four trials of each ISI duration). The average inter-trial interval for all trial types throughout the session was 15 s (range 6–21 s). All trial types within the prepulse intensity and ISI trials were presented in a pseudo-random order. The average startle magnitude over the record window (65 ms) was used for all data analysis.

Data analysis

To assess CRF effects over time, startle magnitude in response to the 120-dB pulses was averaged over each session. A two-way analysis of variance (ANOVA) was used with oCRF treatment (0, 0.06, 0.2, 0.6 nmol) as the between-subject and time post-infusion (0–6 h) as the within-subject factors. %PPI was calculated using the formula:

Initially a three-way ANOVA, with oCRF treatment as the between-subject factor and prepulse intensity (69, 73, 77 dB) or ISI (20, 120, 360 ms) and post-infusion time as within-subject factors was conducted. Prepulse intensity did not interact with the other factors and thus was removed from the analysis. Tukey's post-hoc analyses followed significant main or interaction effects.

Expt 2: the role of α₂, α₁, and β_1/2 noradrenergic receptors in CRF-induced alterations in startle plasticity

Drugs

Expt 2a: α₂ agonist (clonidine) effects in oCRF-treated mice

Mice (n=9–11/group) received 0.6 nmol oCRF (2.8 µg/5 µl icv) or aCSF vehicle 2 h before testing. They also received the α₂ agonist clonidine (0.09 mg/kg, 5 ml/kg i.p., Tocris, USA) or saline 30 min before and 60 min after oCRF infusion (0.18 mg/kg total dose). Dose and timing of oCRF was based on expt 1 (0.6 nmol has large effects on startle and PPI 2 h post-infusion; Fig. 1). Two injections of clonidine were given to ensure a lasting effect of clonidine over the 2-h pretest and testing interval (Risbrough et al.2003, 2004).

Fig. 1

(a) Effect of CRF on acoustic startle responding to 120-dB pulses in C57BL/6J mice. Mice received icv injections of 0, 0.06, 0.2, or 0.6 nmol oCRF diluted in aCSF vehicle. Acoustic startle responding was measured hourly from 0 to 6 h post-administration. * p<0.05 for 0.2 and 0.6 nmol, compared to vehicle group; †p<0.05 for 0.6 nmol compared to 0.06 nmol group; ‡p<0.05 for 0.6 nmol compared to vehicle, post-hoc test). (b) Effect of oCRF on %PPI (averaged across 69, 73, and 77-dB prepulse trials) at 0–6 h post administration. oCRF significantly decreased %PPI (* p<0.05, main effect of oCRF).

Expt 2b: α₁ antagonist (prazosin) effects in oCRF-treated mice

Mice (n=8–11/group) received oCRF (0.6 nmol icv) or aCSF 2 h before testing. They received the α₁ antagonist prazosin (0.4 mg/kg, 5 ml/kg i.p., Sigma-Aldrich, USA) or vehicle (3% Cremophor in water) 30 min before and 1 h after oCRF infusion for a total dose of 0.8 mg/kg. The prazosin dose was based on pilot studies and previous reports in mice (Okuyama et al.1999).

Expt 2c: β_1/2 antagonist (propranolol) effects in oCRF-treated mice

Mice (n=7–11/group) received oCRF (0.6 nmol icv) or aCSF 2 h before testing. The β_1/2 antagonist propranolol (4 mg/kg, 5 ml/kg i.p., Sigma-Aldrich) or saline was given 30 min before and 1 h after oCRF (8 mg/kg total; Goldschmidt et al.1984; Schank et al.2008). To ensure results were not due to non-selectivity of propranolol for β_1/2 receptors, we also used a 10-fold lower dose of propranolol (0.4 mg/kg per injection, 0.8 mg/kg total) in a separate cohort (n=9–11/group) (Izquierdo & McGaugh, 1985; Saha et al.1991; Zarrindast et al.2004). To confirm the antagonist activity of propranolol at β_1/2 receptors during the time-frame of our startle test, we examined the effect of this same dose and pre-injection time on clenbuterol-induced disruptions in locomotor activity in a separate group of mice (n=7–11/group). Timing of propranolol administration before locomotor testing was identical to that in the startle tests, with propranolol (4 mg/kg) administered 2½ h and 1 h before testing. Clenbuterol was administered 30 min before testing. Data were analysed over the first 30 min of the locomotor test to match the duration of the startle session (see below).

Behavioural testing

A 65-dB background was presented continuously during the session which lasted 30 min. The session began with a 5-min acclimation period followed by presentation of five 120-dB pulse trials (average inter-trial interval, 15 s) (block 1). These pulses were followed immediately by startle threshold response trials (block 2), variable prepulse intensity trials (block 3), and variable ISI trials (block 4). Five 120-dB pulse trials were presented at the end of the session (block 5) to measure startle habituation to 120-dB pulses across the session (i.e. blocks 1–5). During the threshold response block, 20 startle pulse trials varied in intensity (four each at 80, 90, 100, 110, and 120 dB, 40 ms duration) were presented. Next, the variable prepulse intensity block consisted of 12 startle pulse-alone trials (each at 120 dB) and 30 prepulse+pulse trials. The prepulse+pulse trials were evenly divided among the following trial types (10 each): 68, 71, or 77 dB prepulse (20 ms) preceding a 120-dB pulse stimulus (40 ms). The time between prepulse and pulse onset was 100 ms. During the variable ISI block, eight startle pulse trials (120 dB) were presented. In addition, 71-dB prepulses preceded a 120-dB stimulus by 25, 50, 100, 200, or 500 ms (four trials each). All trial types were presented in a pseudo-random order.

Data analysis

A two- or three-way ANOVA was used with pretreatment and/or treatment as between-subject factors and pulse intensity (80, 90, 100, 110, 120 dB), prepulse intensity (68, 71, 77 dB), ISI (25, 50, 100, 200, 500 ms) or startle block (blocks 1–5) as within-subject factors. Tukey's post-hoc analyses followed significant main or interaction effects.

Expt 3: the role of CRF₁ receptors in noradrenergic modulation of startle

Drugs

Expt 3a: α₂ antagonist (atipamezole) effects on startle

Mice (n=11/dose) were treated with the α₂ antagonist atipamezole (1, 3, 10, 30 mg/kg, 10 ml/kg s.c.; Pfizer Health, USA) or saline 30 min before testing (doses and timing based on Risbrough & Geyer, 2005).

Expt 3b: α₁ agonist (cirazoline) effects on startle

Mice (n=11/dose) were treated with the α₁ agonist cirazoline (0.25, 0.5, 1.0 mg/kg, 5 ml/kg i.p.) (Sigma) or saline 5 min before testing (Alsene et al.2006; Barr et al.2006; Carasso et al.1998; Menon et al.1990; Shilling et al.2004; Varty et al.1999). Based on expt 3b, mice were treated with lower doses of cirazoline (0.025, 0.083, 0.25 mg/kg, 5 ml/kg i.p.) or saline 5 min before testing (n=10–11/dose).

Expt 3c: CRF₁ receptor antagonist (antalarmin) effects on NE-induced increases in startle

Mice (n=10–12/dose) were treated with the CRF₁ antagonist antalarmin (14.0 mg/kg, 10 ml/kg i.p.; Sigma-Aldrich) or vehicle (19% DMSO in saline) 1 h before testing (Bale & Vale, 2003; Djouma et al.2006). Atipamezole (10 mg/kg s.c.) or saline was administered 30 min before testing.

Behavioural testing

The startle session used in expts 3a–3c was identical to that used in expts 2a–2c except it did not contain an ISI block.

Data analysis

Data analysis was identical to expt 2.

Results

Expt 1: time-course and dose response of oCRF

The effect of oCRF on average startle magnitude depended upon post-infusion time (oCRF×time: F_18,198=2.81, p<0.0002; Fig. 1a). Subsequent analysis at each time-point indicated that oCRF significantly increased startle magnitude at 0, 1, 2, and 3 h post-infusion. Post-hoc analysis indicated that at 0, 1 and 2 h post-infusion, 0.2 and 0.6 nmol oCRF significantly increased startle (p<0.05, vs. vehicle). At 1 h, 0.6 nmol also elevated startle compared to 0.06 nmol (p<0.05). At 3 h, 0.6 nmol continued to increase startle (p<0.05, vs. vehicle).

oCRF treatment significantly decreased %PPI (oCRF: F_3,33=4.15, p<0.05) (Fig. 1b) over the majority of the testing period (oCRF×time: F_18,198=1.40, p=0.13). Inspection of the data suggests that the two highest doses (0.2 and 0.6 nmol) disrupted PPI to the greatest extent at 2 h post-infusion (Fig. 1b). oCRF (0.6 nmol) similarly decreased %PPI at 120 ms ISI (oCRF×ISI: F_6,66=5.96, p=0.0001; p<0.05, 0.6 nmol vs. vehicle, post-hoc test). At the shorter ISI trial (20 ms), oCRF increased %PPI (p<0.05, 0.2 nmol vs. vehicle; p<0.1, 0.6 nmol vs. vehicle, post-hoc data; data not shown). These ISI-dependent effects of CRF on PPI are consistent with previous observation in rats (Conti et al.2009) and suggest that CRF effects on PPI may be via alterations in the speed of prepulse processing (i.e. shifting the time window of prepulse effects on startle reactivity; Graham, 1975).

Expt 2: the role of α₂, α₁, and β noradrenergic receptors in CRF-induced alterations in startle plasticity

Expt 2a: α₂ agonist (clonidine) effects in oCRF-treated mice

During the startle threshold test, oCRF increased startle magnitude to high-intensity pulses and this effect was reduced by pretreatment with clonidine, the α₂ agonist (intensity×clonidine×oCRF: F_4,140=4.32, p=0.005; Table 1). A follow-up two-way ANOVA at each intensity confirmed a significant clonidine×oCRF interaction at 120-dB pulse (F_1,35=4.53, p<0.05). Post-hoc tests indicated that startle magnitude in the oCRF/Veh group was significantly higher relative to all other groups (p<0.05).

Table 1

Effect of α₂ agonist (clonidine), α₁ antagonist (prazosin), and β antagonist (propranolol) on CRF-induced changes in startle magnitude and %PPI (mean±s.e.m.)

In the clonidine experiment, there was a significant main effect of oCRF on PPI at all prepulse intensities (see text for details).

*

p<0.05 vs. all other groups at 120 dB.

†

p<0.05 vs. all other groups at 80 dB.

‡

p<0.05 vs. Veh/Veh group only at 90 dB and 120 dB.

a

Significant main effect of oCRF on PPI at 68-dB and 71-dB prepulse trials in the propranolol experiment (see text for details).

oCRF-induced increases in startle magnitude were predominantly present in the initial blocks of the session, which was significantly reduced by clonidine pretreatment (Fig. 2a; block×oCRF: F_4,140=6.07, p<0.05; block×clonidine×oCRF: F_4,140=3.05, p<0.05). Follow-up two-way ANOVA conducted at each block confirmed a significant clonidine×oCRF interaction at blocks 1 and 2 (F_1,35>4.39, p<0.05). Post-hoc analyses indicated that startle magnitude in the oCRF/Veh group was significantly higher compared to all other groups during blocks 1 and 2 (p<0.05, oCRF/Veh vs. Veh/Veh, Veh/clonidine, oCRF/clonidine) (Fig. 2a, right).

Fig. 2

(a) Effect of clonidine on CRF-induced increases in startle in block 1. Clonidine reversed the startle increase in oCRF-treated mice (* p<0.05, oCRF/Veh vs. all other groups, post-hoc test). Right panel: Effect of clonidine on CRF-induced increases in startle across the test session. oCRF-induced increases in startle during blocks 1 and 2 was significantly blocked by clonidine treatment (* p<0.05, oCRF/Veh vs. all other groups, post-hoc test). (b) Effect of prazosin on CRF-induced increases in startle in block 1. Prazosin reversed the startle increase in oCRF-treated mice (* p<0.05, oCRF/Veh vs. all other groups, post-hoc test). Right panel: Effect of prazosin on CRF-induced increases in startle across the test session. oCRF-induced increases in startle were significantly blocked by prazosin in blocks 1 and 2 (* p<0.05, oCRF/Veh vs. all other groups, post-hoc test. Startle magnitude in the oCRF/Veh group remained significantly higher compared to Veh/Veh and Veh/prazosin during block 5 (‡p<0.05, oCRF/Veh vs. Veh/Veh, Veh/prazosin, post-hoc test). (c) Effect of propranolol on CRF-induced increases in startle in block 1 (* p<0.05, main effect of oCRF). Right panel: Effect of propranolol on CRF-induced increases in startle across the test session. Although oCRF increased startle magnitude, this effect was not reversed by propranolol during any block (* p<0.05, main effects of oCRF, see text for details).

As expected, %PPI increased with prepulse intensity (F_2,70=97.61, p=0.0001) and decreased with oCRF administration (F_1,35=17.71, p=0.001). The effect of oCRF on PPI appeared dependent upon intensity and clonidine (intensity×clonidine×oCRF: F_2,70=3.63, p<0.05). However, two-way ANOVAs at each intensity indicated that oCRF effects on PPI were not significantly altered by clonidine pretreatment (Table 1). A similar pattern of PPI results was found from the ISI test (data not shown).

Expt 2b: α₁ antagonist (prazosin) effects in oCRF-treated mice

oCRF-induced increases in startle magnitude were significantly reduced by the α₁ antagonist prazosin at specific startle pulse intensities (Table 1; intensity×prazosin×oCRF: F_4,140=3.29, p=0.01). Two-way ANOVAs conducted at each intensity revealed a significant interaction at the 80-dB and 120-dB pulses (prazosin×oCRF: F_1,38>4.10, p<0.05). At these intensities, the startle magnitude in oCRF/Veh was significantly higher relative to all other groups (p<0.05, oCRF/Veh vs. Veh/Veh, Veh/prazosin, oCRF/prazosin).

oCRF-induced increases in startle were again greatest in the initial blocks (Fig. 2b; block×oCRF: F_4,140=3.44, p=0.01). There was also an overall significant interaction between oCRF and prazosin on startle reactivity (oCRF×prazosin: F_1,35=4.33, p<0.05) (Fig. 2b, right). Post-hoc analyses indicated that startle magnitude in oCRF/Veh was significantly greater than all other groups during blocks 1 and 2 (p<0.05). During block 5, startle magnitude remained significantly greater in the oCRF/Veh group compared to Veh/Veh and Veh/prazosin (p<0.05) and there was a trend towards a significant difference compared to oCRF/prazosin (p<0.1). oCRF-induced disruption of PPI was not affected by prazosin (Table 1).

Expt 2c: β_1/2 antagonist (propranolol) effects in oCRF-treated mice

The effect of oCRF on startle magnitude was greatest at the 90-dB and 120-dB pulses (intensity×oCRF: F_4,128=6.98, p=0.0000, p<0.05 post-hoc test vs. Veh/Veh) (Table 1). As we have seen throughout, oCRF increases in startle magnitude were highest during the initial blocks (block×oCRF: F_4,128=6.65, p<0.0001, Fig. 2c). A significant main effect of oCRF was observed during block 1 (oCRF: F_1,32=15.57, p<0.0004) (Fig. 2c). Although the main effects of oCRF were maintained during blocks 2–5 (F_1,32>8.13, p<0.008), oCRF-induced startle increases were not reversed by propranolol (β_1/2 antagonist) in any block (Fig. 2c, right). oCRF-induced disruptions in PPI (intensity×oCRF: F_2,64=8.28, p<0.005; at 68 dB and 71 dB, main effects of oCRF were significant: F_1,32>4.79, p<0.05), were unaffected by propranolol (Table 1). In a similar experiment using a lower dose of propranolol (0.8 mg/kg total) we again found that oCRF-induced increases in startle reactivity and decreases in PPI were not reversed by propranolol (data not shown).

Conversely, this same dose and timing of propranolol treatment was found to block the effects of clenbuterol, a β adrenergic agonist, on locomotor activity. Clenbuterol decreased distance travelled over a 30-min period and this effect was fully blocked by propranolol pretreatment (propranolol×clenbuterol: F_1,33=4.24, p=0.04). [Mean distance travelled (cm±s.e.m.) for Veh/Veh, Veh/clenbuterol, propranolol/Veh, and propranolol/clenbuterol groups: 4613.9±172.0, 3555.6±313.4, 4427.8±267.9, 4380.8±225.2, respectively].

Expt 3: the role of CRF₁ receptors in noradrenergic modulation of startle

Expt 3a: α₂ antagonist (atipamezole) effects on startle

Similar to oCRF, atipamezole (an α₂ antagonist that increases NE release via blockade of autoreceptors) dose-dependently increased startle magnitude predominantly at high startle intensity trials (Table 2; atipamezole×intensity: F_16,188=2.58, p=0.001) (30 mg/kg atipamezole at 110 dB and 120 dB; 10 mg/kg at 120 dB, p<0.05, vs. vehicle, Tukey's test). Unlike CRF, atipamezole effects were relatively robust across all blocks (atipamezole: F_4,47=8.75, p=0.0000; atipamezole×block: F_12,141=1.99, p=0.03) (Fig. 3a, right). Post-hoc analysis on each block indicated that compared to vehicle, 30 mg/kg increased startle magnitude during all blocks and 10 mg/kg increased startle during blocks 2, 3 and 4 (p<0.05). Atipamezole did not affect %PPI (Table 2).

Table 2

Effect of α₂ antagonist (atipamezole) and α₁ agonist (cirazoline) on startle magnitude and %PPI (mean±s.e.m.)

*

p<0.05 vs. 0 mg/kg at 120 dB.

†

p<0.05 vs. 0 mg/kg at 110 dB.

a

Significant main effect of cirazoline (see text for details).

Fig. 3

(a) Dose–response effect of atipamezole on startle magnitude in block 1. Atipamezole (30 mg/kg) significantly increased startle magnitude (* p<0.05, vs. vehicle, post-hoc test). Right panel: Effect of atipamezole on startle across the test session. The highest dose of atipamezole (30 mg/kg) increased startle across the session (* p<0.05, 30 mg/kg relative to vehicle in blocks 1, 2, 3, and 4, post-hoc test), while 10 mg/kg increased startle during blocks 2, 3, and 4. (‡p<0.05, 10 mg/kg relative to vehicle in blocks 2, 3, and 4, post-hoc test). (b) Dose–response (0.25–1.0 mg/kg) of cirazoline. The highest dose of cirazoline (1.0 mg/kg) decreased startle magnitude during block 1 (* p<0.05, 1.0 mg/kg vs. vehicle, post-hoc test). Right panel: Dose–response of cirazoline on startle magnitude across the startle session. The highest doses of cirazoline decreased startle magnitude (* p<0.05, 1.0 mg/kg relative to vehicle in block 1; ‡p<0.05, 0.5 mg/kg relative to vehicle in block 2; post-hoc test). (c) When a lower dose range (0.025–0.25 mg/kg) was used, cirazoline had no effect on startle magnitude during block 1 or across the entire session (right panel).

Expt 3b: α₁ agonist (cirazoline) effects on startle (0–1.0 mg/kg)

The α₁ agonist cirazoline decreased startle magnitude regardless of pulse intensity (cirazoline: F_3,37=10.01, p=0.0001; Table 2). Post-hoc analysis conducted on the average startle response (collapsed across intensity) indicated that each dose of cirazoline decreased startle (p<0.01, vs. vehicle). Cirazoline-induced decreases in startle were most prominent in the initial blocks (cirazoline×block: F_9,111=2.69, p=0.007) (Fig. 3b). Compared to vehicle, 1.0 and 0.5 mg/kg decreased startle during blocks 1 and 2, respectively (p<0.05) (Fig. 3b, right). Cirazoline did not affect %PPI (Table 2). No effects on startle or PPI were found with lower cirazoline doses (0.025, 0.083, 0.25 mg/kg) (Fig. 3c, Table 2).

Expt 3c: CRF₁ receptor antagonist (antalarmin) effects on NE-induced increases in startle

To determine whether startle increases following NE release are modulated by CRF activation, we explored the effects of antalarmin (CRF₁ antagonist) on startle reactivity in atipamezole-treated mice. Atipamezole increased startle magnitude at the high intensity pulses and this effect appeared dependent upon antalarmin pretreatment (intensity×atipamezole: F_4,152=4.42, p=0.002; intensity×atipamezole×antalarmin: F_4,152=2.56, p=0.04) (Table 3). A two-way ANOVA conducted at each intensity confirmed a significant antalarmin×atipamezole interaction at the 120-dB pulse (F_1,38=4.34, p=0.04) (Table 3). Tukey's tests at 120 dB indicated that startle magnitude in the Veh/atipamezole group was significantly higher relative to Veh/Veh (p<0.05). The effects of atipamezole appeared to be mildly attenuated by antalarmin as the antalarmin/atipamezole group was not significantly different from Veh/Veh; however, it was also not significantly different from Veh/atipamezole. To compare with previous experiments, we also conducted an analysis on block 1 (Fig. 4), finding a significant antalarmin×atipamezole interaction (F_1,38=7.45, p=0.01) with post-hoc tests indicating a significant startle magnitude increase in the Vehicle/atipamezole group compared to Veh/Veh (p<0.05, Tukey's test). A less conservative planned simple comparison revealed a significant reduction of atipamezole effects by antalarmin pretreatment (t test between Veh/atipamezole and antalarmin/atipamezole: p=0.05). Although the effects of atipamezole on %PPI depended on prepulse intensity (atipamezole×intensity: F_2,76=3.72, p=0.03), subsequent ANOVAs conducted at each intensity separately revealed no significant main effect of atipamezole (Table 3). Antalarmin also did not affect PPI, and it did not interact with atipamezole (Table 3).

Fig. 4

Effect of antalarmin on atipamezole-induced increases in startle in block 1. * p<0.05 Veh/atipamezole vs. Veh/Veh; Tukey's post-hoc Test. ‡p<0.05 Veh/atipamezole vs. antalarmin/atipamezole, planned simple t test. Right panel: Effect of antalarmin (Ant) on atipamezole (Ati)-induced increases in startle across the session. Overall there was a significant Ant×Atip interaction which did not interact with block.

Table 3

Effect of CRF₁ antagonist (antalarmin) on atipamezole (Ati)-induced changes in startle magnitude and %PPI (mean±s.e.m.)

*

p<0.05 vs. Veh/Veh at 120 dB.

Discussion

We investigated the mechanisms by which CRF and NE signalling interact to modulate defensive startle behaviours. Acute oCRF administration robustly increased startle reactivity and reduced PPI. The α₂ agonist (clonidine) and the α₁ antagonist (prazosin), but not the β_1/2 adrenergic antagonist (propranolol), significantly reduced the effects of CRF on startle, suggesting that α₂ and α₁ receptors mediate CRF-induced increases in startle reactivity. None of these treatments affected CRF-induced disruptions in PPI. We also found a reciprocal contribution of the CRF₁ receptor to NE signalling effects on startle, with α₂ antagonist (atipamezole)-induced increases in startle attenuated by CRF₁ antagonist (antalarmin) pretreatment. These findings support the hypothesis that NE transmission is required for CRF effects on startle reactivity. These data lend support to a model of dual permissive roles for NE and CRF₁ receptor activation for stress-induced increases in startle. These findings also suggest that the neurotransmission mechanism mediating CRF effects on startle reactivity is dissociable from CRF-mediated changes in PPI.

oCRF robustly increased startle in all experiments, consistent with earlier rodent findings wherein r/h-CRF increased startle magnitude (Liang et al.1992; Risbrough et al.2003, 2004; Swerdlow et al.1986, 1989). In rats, CRF effects on startle have been localized to a hippocampal–septal–BNST circuit expressing CRF₁ and CRF₂ receptors (Davis et al.1997; Van Pett et al.2000). In mice, both receptors contribute to CRF effects on startle and PPI (Risbrough et al.2003, 2004, 2009). Early evidence suggested that NE signalling may be induced by activation of the CRF system. Lee et al. (1994) reported that CRF facilitates NE release in the hippocampal dentate gyrus. Additionally, CRF₁ pharmacological blockade and CRF gene deletion block stress-induced increases in NE release (Griebel et al.2002; Jeong et al.2000). Here we identified NE signalling via α₁ receptors as a mechanism required for CRF effects on startle behaviour, since clonidine and prazosin prevented CRF-induced increases in startle. Both drugs were effective at the 120-dB stimuli, while prazosin also reversed CRF-induced increases in startle at 80 dB (Table 1). These results hint that α₁ and α₂ receptors could mediate startle at different points in the startle circuit and/or alter neuronal excitability in different ways (reduced threshold vs. increased firing rate), although further testing is required to test this speculation. Neither clonidine nor prazosin treatment alone significantly affected startle in the absence of CRF (Table 1). Thus, it is unlikely that the blockade of CRF-induced increases in startle following clonidine and prazosin pretreatment reflected a summation of opposing effects.

The β_1/2 adrenergic antagonist propranolol had no effect on CRF-induced increases in startle, despite the fact that this pretreatment was clearly behaviourally active in blocking the effects of clenbuterol. Thus, the null effect of propranolol on CRF modulation of startle was not due to incorrect dosing or timing of propranolol treatment. However, it should be noted that the CRF-induced increase in startle in the propranolol experiment, although robust and highly significant, was smaller compared to its effect in the clonidine and prazosin experiments. The possibility that this smaller behavioural window obscured detection of a propranolol effect on CRF-induced increases in startle cannot be excluded. Nonetheless, our propranolol data corroborate previous startle findings that blockade of CRF₁, but not β_1/2 receptors, reduces stress-induced increases in startle reactivity (Adamec et al.2007, 2009). Taken together, the present data suggest that α₂ and α₁ adrenergic receptors modulate the CRF startle effect, while the β_1/2 receptors may not.

After determining that NE signalling, probably via α₁ receptors, is required for the effects of CRF on startle reactivity, we next examined the efficacy of α₂ antagonist (atipamezole) and α₁ agonist (cirazoline) treatments to increase startle reactivity. These studies were based on the hypothesis that if NE release (via blockade of α₂ autoreceptor) or α₁ activation mediate CRF effects on startle, then modulation of these receptors may be sufficient to mimic CRF-induced alterations in startle reactivity. We found that atipamezole (which releases NE), but not cirazoline (which selectively activated α₁), increased startle. The inability of cirazoline to increase startle corroborates a recent report that α₁ agonist treatments (cirazoline, methoxamine, or phenylephrine) do not affect baseline startle magnitude in rats, despite activity in other measures (PPI) (Alsene et al.2006). Thus, NE release is sufficient to mimic CRF-induced increases in startle, but direct α₁ activation is not.

These findings raise the question of how could increased NE release potentiate startle if direct stimulation of α₁ receptors has no effect? In humans yohimbine, an α₂ antagonist, induces increased CRF release (as measured by increased CRF in cerebrospinal fluid; Vythilingam et al.2000). If the NE-induced release of CRF in turn activates the CRF₁ receptor, then this would explain why atipamezole-induced increases in startle were attenuated by the CRF₁ antagonist antalarmin (expt 3). Thus, atipamezole may induce CRF release in the BNST either via its induction of NE release and subsequent activation of excitatory NE receptors (α₁ and/or β) at CRF-containing neurons, or via direct antagonist activity of inhibitory α₂ receptors at CRF-containing neurons (Lorton & Davis, 1987; Phelix et al.1994; Shields et al.2009). Our data suggest that atipamezole-induced increases in startle is likely to be via direct blockade of inhibitory α₂ or indirect activation of excitatory β_1/2 receptors at CRF-containing neurons, as activation of α₁ with cirazoline did not increase startle reactivity. These data are also consistent with previous reports suggesting that activation of CRF receptors mediates noradrenergic-related increases in anxiety behaviours. For example, α₂ antagonists (idazoxan) and α₁ agonists (phenylephrine) result in stress-like decreases in exploratory behaviour and a non-selective CRF antagonist (α-helical CRF) blocks phenylephrine-induced decreases in exploratory behaviour (Berridge & Dunn, 1989). Taken together, the present findings that α₂ agonist and α₁ antagonist treatment disrupt CRF-induced increases in startle and that CRF₁ antagonist treatment attenuates α₂ antagonist-induced increases in startle lend support to a model of dual permissive roles for CRF and NE receptor activation for stress-induced increases in startle.

To integrate these findings, we have proposed a potential circuit model of the CRF/NE signalling effects on startle reactivity (Fig. 5). Previous reports indicate that the BNST is the site of action for CRF₁ effects on startle (Walker et al.2009). However, unlike our findings following systemic prazosin pretreatment (expt 2b), we observed no effect of α₁ antagonist (4, 12, 40 µg terazosin) infusions in the BNST on CRF-induced startle increases (Gresack et al. unpublished observations) suggesting that α₁ effects on CRF modulation of startle are located elsewhere. It is possible that these effects are via activation of α₁ at the pontine reticular formation (Stevens et al.1994) which mediates the acoustic startle reflex via activation of motor neurons in the spinal cord (Koch, 1999). Intrathecal injections of α₁ agonists in rats robustly increases startle reactivity, suggesting that brainstem activation of α₁ receptors may be sufficient to increase startle (Astrachan et al.1983). α₁ receptors in the reticular formation are predominantly excitatory by facilitating initial neuronal responsiveness and increasing repetitive firing (Stevens et al.1994). Hence, CRF receptor activation may modulate startle by activating two separate pathways that converge at the nucleus reticularis pontis caudalis (PRC) of the startle circuit (Lee et al.1996): the BNST–PRC (via glutamate release at the PRC that is independent of α₁ activation) (Dong & Swanson, 2004; Holstege et al.1985), and the LC–PRC (via NE release and subsequent α₁ receptor activation at the PRC; Jones & Yang, 1985; Strazielle et al.1999) (Fig. 5). Because NE release in turn increases levels of CRF (Vythilingam et al.2000), inducing NE release via an α₂ antagonist would activate both of these pathways, while direct stimulation of α₁ receptors would not (Fig. 5). If both pathways are required to achieve a maximal startle response, this model could explain why we did not observe effects of systemic α₁ treatment on startle alone. Further site-specific injections of NE and CRF signalling blockers at the PRC are required to test this hypothesis.

Fig. 5

Schematic proposing a circuit model of CRF/NE signalling effects on startle reactivity. In this model, CRF₁ receptor activation modulates startle by activating two pathways that converge at the nucleus reticularis pontis caudalis (PRC) of the startle circuit. One pathway is the LC–PRC pathway which requires NE release and α₁ receptor activation at the PRC. The second pathway, the BNST–PRC pathway, involves CRF₁ receptor activation of glutamatergic neurons projecting to the PRC (see text for additional details). CRF, corticotropin releasing factor; NE, norepinephrine; LC, locus coeruleus; BNST, bed nucleus stria terminalis; Glu, glutamate.

In the present study, CRF-evoked disruptions in sensorimotor gating were insensitive to manipulation of the noradrenergic system, since clonidine, prazosin, and propranolol did not affect CRF-induced decreases in PPI. Thus, a dichotomy exists between CRF signalling mechanisms for startle and PPI. Downstream signalling mediators of CRF effects on PPI remain unclear. Neither D₁ nor D₂ signalling is required for CRF-induced disruption in mice (Vinkers et al.2007, but see Dirks et al.2003). Similarly serotonin signalling does not appear to be required for CRF-induced decreases in PPI in rats (Sutherland et al.2008). CRF does modulate excitatory glutamatergic synapse transmission at forebrain regions that modulate PPI (Liu et al.2004; Swerdlow et al.2001), and thus may be a system for future investigation of mechanisms underlying CRF-induced alterations in PPI.

Finally, none of the pharmacological manipulations of the noradrenergic system affected PPI. These findings diverge from previous reports in rats indicating that α₁ activation (via cirazoline) decreases PPI (Alsene et al.2006; Carasso et al.1998; Swerdlow et al.2006), and this effect is blocked by α₂ activation (via clonidine; Swerdlow et al.2006). There are few studies of NE effects on PPI in mice. Indeed, to our knowledge, the present study is the first to describe cirazoline effects in mice, which appear to be different than found in rats, perhaps due to species differences in NE receptor function or distribution (Lorton & Davis, 1987; Schultz & Daly, 1973). In mice, α_2cvs. α_2a receptor subtypes appear to exhibit opposing roles on PPI with α_2c and α_2a null mutation mice exhibiting decreased and increased PPI, respectively (Sallinen et al.1998; Lahdesmaki et al.2004). Non-selective activation/blockade of α₂ subtypes by clonidine/atipamezole could thus result in overall null effect on PPI in the present study.

The finding that NE release (α₂) and α₁ receptor activation is required for CRF effects on startle hyperreactivity and that CRF₁ activation is required for NE-induced increases in startle may have clinical implications for PTSD. The CRF and noradrenergic systems, both of which influence arousal, have been implicated in the pathophysiology of PTSD (Baker et al.1999; Bremner et al.1997; Geracioti et al.2008). Blockade of α₂ receptors selectively increases acoustic startle in combat veterans with PTSD but not combat controls (Morgan et al.1995). These clinical findings, in conjunction with the present preclinical findings, suggest that PTSD subjects may have increased vulnerability to NE effects on startle, perhaps due to increased CRF signalling (and vice versa). Prazosin has shown some promise in treating sleep disturbances in PTSD (Strawn & Geracioti, 2008) and α₂ agonists may be another novel way to treat PTSD symptoms (Krystal & Neumeister, 2009). A current clinical trial is also examining the effects of CRF₁ antagonists on PTSD symptomology (clinicaltrials.gov). Our findings support the notion that targeting NE and CRF signalling mechanisms may prove beneficial in specifically treating heightened arousal symptoms associated with PTSD.

Acknowledgements

This research was supported by MH074697. We thank Dr Mark Geyer for helpful discussions and comments on earlier versions of this manuscript and Ms. Chelsea Wallace for technical assistance.

Statement of Interest

None.