Noradrenaline is a stress-associated metaplastic signal at GABA synapses

Abstract

Exposure to a stressor sensitizes behavioral and hormonal responses to future stressors. Stress-associated release of noradrenaline enhances the capacity of central synapses to exhibit plasticity (metaplasticity). We report noradrenaline-dependent metaplasticity at GABA synapses in the paraventricular nucleus of the hypothalamus in rat and mouse that control the hypothalamic-pituitary-adrenal axis. Previously, work indicates these GABA synapses become excitatory following acute stress. The current study finds that in vivo stress exposure is also required for these synapses to undergo activity-dependent long-term potentiation (LTP_GABA). The activation of β-adrenergic receptors (β-ARs) during stress functionally upregulates metabotropic (mGluR1) glutamate receptors allowing for mGluR1-dependent LTP_GABA during afferent bursts. LTP_GABA is expressed postsynaptically and manifests as the emergence of new functional synapses. Our findings provide the first demonstration that noradrenaline release during an in vivo challenge alters information storage capacity at GABA synapses. These changes may contribute to neuroendocrine sensitization to stress.

Any threat to survival triggers rapid defense mechanisms known as the stress response^1,2. In vertebrates, the hallmark of this response is the activation of the hypothalamic-pituitary-adrenal (HPA) axis culminating in an increase in circulating glucocorticoids^1,2. Although a number of brain areas perceive and process diverse modalities of stress (e.g. psychological, physical and homeostatic stress), the neuroendocrine response requires that all stress-related signals eventually converge onto parvocellular neuroendocrine cells (PNCs) in the paraventricular nucleus of the hypothalamus (PVN)^2,3. The stereotyped recruitment of these ‘stress effector’ cells is vital for managing impending challenges, but considerable evidence indicates that flexibility in this system is also important for tuning neuroendocrine output appropriately based on previous stress experience^1,2,4. This may be achieved, in part, through plasticity in higher brain regions^1,4, but emerging evidence showing that synapses in the PVN also undergo adaptations following stress^3,5–11, suggests a key role for this structure.

Glutamate synapses onto PNCs undergo stress-dependent priming that manifests as short-term potentiation following tetanization⁹. Acute restraint stress also causes a depolarizing shift in the GABA_A receptor (GABA_AR) reversal potential (E_GABA) that is sufficient to convert GABA synapses to excitatory both during and immediately after stress^12,13. Given that GABA is the dominant synaptic input to PNCs (> 50% of all synapses)^10,11, stress-dependent changes in efficacy and plasticity of GABA synapses, together with an excitatory conversion, would powerfully enhance the excitatory inputs to the PVN and consequently sensitize the neuroendocrine output to subsequent stressors.

In addition to GABA and glutamate, PNCs also receive noradrenaline inputs from A1/A2 cell groups in the caudal medulla^14,15. noradrenaline rapidly excites PNCs and contributes to the activation of the system at stress onset^15,16. In other systems, noradrenaline acting via β-adrenergic receptors (β-ARs) links behaviors/experience to subsequent synaptic plasticity^1,17,18. We hypothesized that noradrenaline ensures specific information is extracted by PNCs during stress. Here we report that noradrenaline, through recruitment of β-ARs during stress, primes PNCs to be more sensitive to heterosynaptic glutamate signaling. This creates conditions that are permissive for the induction of activity-dependent potentiation at GABA synapses in the PVN.

Results

A single stress experience unmasks LTP_GABA

Exposure to a single stressor alters neuroendocrine responses to future challenges⁴. This suggests the stress axis is capable of learning and provides an experimental model to study how stress affects subsequent information processing and storage. We asked whether a single stress alters the ability of GABA synapses onto PNCs to undergo activity-dependent plasticity. Using whole-cell voltage-clamp recordings we examined evoked inhibitory postsynaptic currents (eIPSCs). Synaptic efficacy was assessed in response to high-frequency stimulation (HFS, 100 Hz, 1 s repeated 4 times with 20 s interval, V_m = −80 mV) delivered to slices obtained either from rats subjected to 30 min immobilization stress (IMO) immediately prior to slice preparation, or from age-matched, naïve rats. In naïve rats, this protocol had no lasting effect on eIPSC amplitude (98 ± 9% of baseline, p = 0.4; Fig. 1a,c). By contrast, we noted a robust potentiation of eIPSC amplitude in slices from stressed rats (140 ± 6% of baseline, p = 0.0001; Fig. 1b,d). This potentiation persisted for the duration of the stable recording (> 25 min after HFS). We also tested different durations of HFS on synaptic strength (Fig. 1e). In slices from naive rats, shortening (0.5 s) or lengthening (4 s) the induction protocol had no effect on eIPSC amplitude (0.5 s, 110 ± 6.5% of baseline), p = 0.1; 4 s, 112 ± 9.8% of baseline, p = 0.2). Following stress, the potentiation in response to 0.5-s HFS was not statistically significant (139% ± 20% of baseline, p = 0.05) but the, 4-s protocol elicited LTP_GABA (155.4 ± 19% of baseline,, p = 0.03). A cumulative probability distribution of normalized post-HFS (0.5–4 s) eIPSC shows that > 20% potentiation was observed in 28% of cells in naïve rats, and in 74% of cells following stress (Fig. 1g). In the majority of PNCs, a single stress exposure unmasks the ability of GABA synapses to potentiate in response to bursts of afferent activity.

Figure 1. Acute stress unmasks activity-dependent LTP_GABA.

(a, b) Top, schematics for experiments, Bottom, raw (light) and averaged (solid) traces of eIPSCs recorded from PNCs before (black) and after (blue/red) HFS in slices from (a) naïve and (b) IMO stressed rats. Scale bars represent 50 pA and 20 ms. Amplitude values were assessed at the time points indicated by shadowed bands below. (c, d) Summary time-course of eIPSC amplitude recorded in slices from naïve (blue, (n = 8 (7 rats)) and IMO stressed (red, (n = 14 (14 rats)) rats. Arrows indicate time of HFS. Data are mean ± s.e.m. (e) Summary of the changes of eIPSC amplitude, in slices from naïve and IMO stressed rats following different durations of HFS (0.5, 1, 4 s; n=7, 10, 6 (5 rats each)). (f) Summary of eIPSC amplitude change in slices from IMO stressed rats following lower frequency stimulation (10 Hz for 10 s, n=5 (5 rats)), and in slices from rats exposed to predator odor (n=5 (5 rats) or forced swim (n=7 (5 rats) following HFS. (g) Cumulative probability distribution of normalized eIPSC amplitude from naïve (blue) and IMO-stressed (red) rats following HFS shown in (e). Bin size 10 %.

LTP_GABA was also observed when lower frequency protocols (10 Hz, 10 s) that better recapitulate physiological synaptic inputs were used following stress (132 ± 10%, p = 0.03, Fig. 1f). The unmasking of LTP_GABA is not limited to IMO stress, as predator odor stress (30-min exposure to a fox feces compound) also led to HFS-induced LTP_GABA (133± 9%, p = 0.02). Forced swim stress (20-min inescapable swim), however, did not induce statistically significant LTP_GABA (121 ± 12%, p = 0.08, Fig. 1f). These results argue that the priming of GABA synapses is not limited to a specific induction protocol (100 Hz) or a specific stressor (IMO), but also it may not be generalizable to all types of stressors. Below we use IMO stress and 100-Hz, 1-s protocol to study the mechanisms underlying LTP_GABA.

β-AR signaling during stress unmasks LTP_GABA

In other brain regions, noradrenaline, through activation of β-AR, modulates activity-dependent synaptic plasticity at glutamate synapses^19–21. Since initiation of a neuroendocrine stress response requires local release of noradrenaline^15,16 we asked whether β-AR activation during stress was necessary for unmasking subsequent LTP_GABA. To test this, we injected rats with the β-AR antagonist propranolol (PRO, 10 mg kg⁻¹, intraperitoneal) 30 min before IMO and repeated experiments as above. Propranolol pre-treatment prevented the stress-dependent unmasking of LTP_GABA (93 ± 9% of baseline, p = 0.4; Fig. 2a), demonstrating that β-AR activation during IMO stress is necessary in order to enable GABA synapses to express subsequent activity-dependent LTP. We next asked if prior history of β-AR activation, specifically in the PVN in the absence of stress, was sufficient to mimic the effects of IMO. We treated naïve hypothalamic slices with β-AR agonist isoproterenol (ISO, 1 μM) for 10 min. First, we confirmed that ISO application alone caused no lasting change in eIPSCs (103 ± 4% of baseline after 20 min washout, p = 0.5). We did note a transient potentiation peaking approximately 15 min after application (118 ± 7% of baseline, n = 5 (4 rats), p = 0.06). In subsequent experiments ISO was washed out > 30 min between slice incubation and recording. Following ISO treatment, HFS induced robust LTP_GABA (149 ± 15% of baseline, p = 0.01; Fig. 2a), suggesting that transient pharmacological activation of β-AB is sufficient to unmask subsequent LTP_GABA.

Figure 2. Stress-induced β-AR signaling unmasks activity-dependent LTP_GABA.

Left, schematics for the experiments. Middle, averaged traces before (grey) and after (colored) HFS. Right, summary time-course of normalized eIPSCs. (a) Recordings in slices from IMO rats treated with β-AR antagonist PRO (purple, n = 8 (6 rats)) and in slices from naïve rats treated ex vivo with β-AR agonist ISO (orange, n = 8 (6 rats)) (b) Recordings in slices from naïve TH-Cre/ChR2-eYFP mice without treatment (gray, n = 8 (7 mice)), ex vivo flushed with blue light in the absence (blue, n = 9 (8 mice)) and presence (orange, n = 6 (4 mice)) of PRO. Scale bars represent 50 pA and 20 ms. Data are mean ± s.e.m.

The primary noradrenaline projection to the PVN originates in the caudal medulla noradrenergic populations (A1/A2)^14–16. To examine the contribution of noradrenergic inputs to GABA synapse priming, we used an optogenetic approach in which we stereotaxically injected Cre-recombinase-dependent adeno-associated viral vector (AAV) carrying channelrhodopsin-2 (ChR2) and eYFP or -mCherry into the caudal medulla of mice that expressed Cre under the control of tyrosine hydroxylase (TH) promoter (TH-Cre mice)²². Patch clamp recording from acute brain stem slices confirmed functional ChR2 expression in the caudal medulla neurons. eYFP-positive cells fired action potentials in response to 5 ms pulse of blue light at 5–50 Hz with 100% fidelity at 20 Hz and below (Supplementary Fig. 1). In the PVN, we observed efficient ChR2-eYFP expression in fibers (Supplementary Fig. 1). We stimulated ChR2-expressing noradrenergic fibers ex vivo in PVN slices prepared from naïve mice and asked whether the release of endogenous noradrenaline was sufficient to unmask subsequent LTP_GABA. We delivered 5 ms blue light pulses at 10 Hz for 20 min (to mimic stress), and then rested the slices for > 30 min before recording. Following light stimulation, HFS induced robust LTP_GABA (160 ± 18% of baseline, p = 0.004; Fig. 2c). HFS failed to induce LTP_GABA in ChR2-expressing slices that did not receive light stimulation (107 ± 6% of baseline, p = 0.5; Fig. 2c). Incubation of the slice with PRO (50 μM) 10 min before and during light stimulation attenuated the priming (118 ± 5% of baseline, p = 0.03 vs no PRO; Fig. 2b), indicating the necessity of β-AR activation. These results collectively support the hypothesis that local release of noradrenaline in the PVN primes GABA synapses for subsequent activity-dependent plasticity. In addition to noradrenaline, other neuromodulators are also released in the PVN during the onset of stress. Specifically, CRH release during IMO stress primes glutamate synapses onto PNCs to express short-term potentiation (STP)⁹. Systemic administration of CRH receptor type 1 antagonist CP-154,526 (30 mg kg⁻¹, intraperitoneal) 30 min before IMO, (a treatment that abolished IMO-dependent priming of glutamate STP⁹), failed to prevent HFS-induced LTP_GABA (130 ± 9 % of baseline, n = 5 (3 rats), p = 0.03), indicating that CRHR1 activation is not required for LTP_GABA. These results identify noradrenaline, through actions at β-ARs, as a stress-associative metaplastic signal²³ that is necessary and sufficient for the subsequent expression of activity-dependent long-term plasticity at GABA synapses.

β-AR-induced priming requires PKA activation

β-ARs activation induces metaplasticity at glutamate synapses through downstream activation of protein kinase A (PKA)^19–21, and PKA has been implicated in GABA synapse plasticity²⁴. We examined the role of PKA in the priming of LTP_GABA by first incubating slices from naïve rats with ISO in the presence of selective PKA inhibitor KT5720 (1 μM, applied 10 min before and during ISO). We observed no ISO-dependent priming of LTP_GABA (102 ± 12% of baseline, p = 0.4; Fig. 3). Conversely, we found that direct activation of PKA with 8-bromo-cAMP (200 μM for 10 min, and washout > 30 min) was sufficient to unmask HFS-induced LTP_GABA in slices from naïve rats (145 ± 14% of baseline, p = 0.02; Fig. 3). These data indicate that PKA activation is the key intracellular signaling mechanism downstream of β-AR activation.

Figure 3. β-AR-induced priming requires PKA activation.

Recordings in slices from naïve rats treated ex vivo with PKA activator 8-Br-cAMP (green, n = 7 (4 rats)) and with ISO in the presence of PKA inhibitor KT5720 (purple, n = 7 (3 rats)). Scale bars represent 50 pA and 20 ms. Data are mean ± s.e.m.

Unmasking of LTP_GABA requires stress- and β-AR-induced priming of mGluR1

We next investigated the mechanism by which IMO-induced β-AR signaling unmasks LTP_GABA. Our data suggest that β-AR → PKA activation acts as a priming signal rather than an actual LTP_GABA trigger as pharmacological activation of β-AR alone was insufficient to LTP_GABA. Furthermore, LTP_GABA induction required only a brief β-AR activation (by stress, pharmacological or optogenetic manipulation) prior to but not concomitant with HFS. In other brain areas, activity-dependent LTP_GABA requires heterosynaptic signaling through group 1 metabotropic glutamate receptors (mGluR1 and 5)^24,25, which are also expressed by PNCs²⁶. Moreover, β-AR signaling and downstream PKA prevent agonist-induced internalization/desensitization of mGluR1²⁷, and PKA activation facilitates mGluR1-dependent LTP_GABA24. Thus we hypothesized that mGluR1 signaling is required for LTP_GABA, and that IMO-induced β-AR signaling enhances group 1 mGluRs responsiveness to glutamate released during the HFS protocol. To test this idea, we first asked whether activation of group 1 mGluRs is required for HFS-induced LTP_GABA in PNCs by delivering HFS in the presence of subtype non-selective group 1/2 mGluR antagonist (S)-α-methyl-4-carboxyphenylglycine (MCPG, 200 μM) in slices from IMO stressed rats. MCPG abolished LTP_GABA (92 ± 10% of baseline, p = 0.4; Fig. 4a,b). Two structurally unrelated mGluR1 subtype-specific antagonists 7-(hydroxyimino)-cyclopropa[b]chromen-1a-carboxylate ethyl ester (CPCCOEt, 100 μM) and JNJ16259685 (750 nM) also blocked (CPCCOEt, 93 ± 13% of baseline, p = 0.6; JNJ16259685, 105 ± 11, p = 0.6, Fig. 4b) while mGluR5 specific antagonist 3-((2-methyl-1,3-thiazol-4-yl)ethynyl)pyridine hydrochloride (MTEP, 10 μM) failed to inhibit LTP_GABA (166 ± 19% of baseline, p = 0.03; Fig. 4b), indicating that heterosynaptic activation, specifically of mGluR1, is necessary for LTP_GABA. The GABA_AR antagonist bicuculline (100 μM; focally applied near the postsynaptic cell) prior to and during HFS failed to block LTP_GABA (132 ± 11%, p = 0.04), suggesting that GABA_ARs do not participate in the induction of LTP_GABA.

Figure 4. LTP_GABA requires priming of mGluR1 by β-AR activation.

(a) Left, averaged traces before (grey) and after (green) HFS recorded in slices from IMO stressed rats, with bath application of MCPG (mGluR1/5 antagonist). Right, summary time-course. Arrow indicates HFS. (b) Summary of HFS-induced eIPSC amplitude changes in control (without antagonist, adapted from experiment shown in Fig. 1d) adapted from experiment shown in Fig. 1d, and with MCPG, n = 6 (4 rats), CPCCOEt (mGluR1 antagonist, n = 7 (5 rats)), JNJ16259685 (mGluR1 antagonist, n = 7 (5 rats)), MTEP (mGluR5 antagonist, n = 5 (4 rats)) and bicuculline (GABA_AR antagonist, n = 5 (3rats)). * p < 0.03, *** p = 0.0001. (c) Left, schematics for the experiments. Middle, averaged traces of eIPSC before (grey) and after DHPG (mGluR1/5 agonist, blue/red) recorded in slices from naïve (blue, n = 8 (5 rats)) and IMO (red, n = 7 (5 rats)) rats. Right, summary time-course. (d) Left, averaged traces of eIPSC from time periods 1–3 indicated in the right graph. Right, time course of eIPSC amplitude recorded in a slice from IMO stressed rat. DHPG (orange bar) application was followed by HFS (arrow). Broken lines indicate the average amplitude before DHPG (black) and HFS (orange). (e) Summary of the effects of HFS on eIPSC amplitude after DHPG-induced potentiation, n = 6 (5 rats). (f) Left, schematics for the experiments. Middle, averaged traces of eIPSC before (grey) and after DHPG (orange/purple) recorded in slices from IMO rats injected with PRO (orange, n = 6 (6 rats)) and in slices from naïve rats treated ex vivo with ISO (purple, n = 9 (7 rats)). Right, summary time-course. Arrow and orange horizontal bars represent the time of HFS and DHPG application, respectively. Data are mean ± s.e.m. Scale bars represent 50 pA and 20 ms.

If IMO/β-AR enhancement of mGluR1 signaling is necessary for LTP_GABA induction following stress, then pharmacological activation of mGluR1, specifically following stress, should be sufficient to induce LTP_GABA independent of HFS. Bath application of mGluR1/5 agonist (S)-3,5-dihydroxyphenylglycine (DHPG, 5 μM, 5 min) to slices from stressed rats elicited long-lasting potentiation of eIPSCs in slices from stressed (141 ± 8% of baseline, p = 0.003; Fig. 4e), but not naïve rats (94 ± 11% of baseline, p = 0.6; Fig. 4e). Next, we asked whether DHPG-mediated LTP_GABA and HFS-induced LTP_GABA share a common underlying mechanism. We pretreated slices obtained from stressed animals with DHPG and washed out the drug > 30 min to allow LTP_GABA to develop. Subsequent delivery of HFS failed to induce LTP_GABA (102 ± 9% of post-DHPG level, p = 0.8; Fig. 4d,e). Finally, we examined whether β-AR signaling causes the functional upregulation of mGluR1 by IMO. Blockade of β-AR with PRO during IMO stress challenge (same treatment as above) abolished DHPG-induced LTP_GABA (98 ± 9% of baseline, p = 0.9; Fig. 4f). Conversely, ex vivo stimulation of β-AR with ISO in PVN slices from naïve rats unmasked DHPG-induced LTP_GABA (137 ± 11% of baseline, p = 0.02; Fig. 4f). Importantly, β-AR activation needs to occur prior to mGluR1 activation as reversing the temporal sequence of ISO → DHPG treatment (e.g. DHPG → ISO) did not induce LTP_GABA (108 ± 7%, n = 7 (4 rats), p = 0.3, Supplementary Fig. 2). These data support our hypothesis that β-AR activation functionally upregulates mGluR1 to enable GABA synapses to express activity-dependent LTP.

Number of functional GABA synapses increases during LTP_GABA

We next examined the mechanisms underlying the expression of LTP_GABA. To this end, we analyzed the paired-pulse ratio (PPR) and coefficient of variation (CV) of the eIPSCs shown in Fig. 1c-d (see Methods for details of analysis). Alterations of PPR indicate changes to probability of neurotransmitter release from the presynaptic terminals (p_r), while alterations to 1/CV² suggest changes to either p_r, the number of functional synapses (N), or both²⁸. Baseline PPR or 1/CV² prior to HFS were similar between naïve and IMO rats (PPR: 0.8 ± 0.04 and 0.7 ± 0.03 for naïve and IMO, p = 0.2; 1/CV²: 22 ± 6 and 34 ± 13, p = 0.5), suggesting that stress alone did not affect these synaptic properties. We found no change in PPR during LTP_GABA following stress (97 ± 3% of baseline, p = 0.4; Fig. 5a), suggesting that LTP_GABA occurs without p_r increase. By contrast, we observed an increase in 1/CV² (234 ± 28% of baseline, p = 0.0004; Fig. 5a), suggesting an increase in N during HFS-induced LTP_GABA. In naïve rats, HFS failed to alter PPR (104 ± 7% of baseline, p = 0.6) or 1/CV² (122 ± 29% of baseline, p = 0.5). To more closely examine the relationship between PPR, CV and LTP_GABA, we plotted the changes in PPR or 1/CV² against eIPSC amplitude changes for individual cells. This revealed a linear, positive correlation between 1/CV² and eIPSC changes (r = 0.7, p = 0.0009; Fig. 5b), but not between eIPSC and PPR (r = 0.4, p = 0.06; Fig. 5c). These results indicate that LTP_GABA is primarily attributable to an increase in N.

Figure 5. LTPGABA is accompanied by changes relevant to an increase of synapse number.

(a) Summary time-course of normalized PPR (green) and 1/CV² (orange) of eIPSCs recorded in slices from IMO stressed rats. Arrow indicates the time of HFS. (b, c) Plots of normalized post-HFS 1/CV² (b) and PPR (c) against amplitude, recorded in slices from naïve (blue, n=16) and IMO stressed (red, n=21) rats. Smaller symbols represent individual data and larger ones mean ± s.e.m. Lines represent least square linear best fit. (d–f) Sample traces (d) and cumulative probability plots of sIPSC frequency (e) and amplitude (f) before (grey) and after (red) HFS recorded in a slice from IMO rats. Scale bars represent 20 pA and 500 ms. (g) Summary time-course of normalized sIPSC frequency (purple) and amplitude (blue) recorded in IMO slices. (h, i) Plots of normalized post-HFS sIPSC frequency (h) and sIPSC amplitude (i) against eIPSC amplitude recorded in slices from naïve (blue) and IMO stressed (red) rats.

If recruitment of new synapses underlies LTP_GABA, such changes may be reflected as an increase in the frequency of spontaneous IPSCs (sIPSCs). Consistent with this, we found an increase in sIPSC frequency during HFS-induced LTP_GABA following stress (158 ± 14% of baseline, p = 0.002; Fig. 5d–g). Mean sIPSC amplitude did not change (110 ± 7% of baseline, p = 0.07; Fig. 5d–g), although this does not exclude the possibility that HFS expands the size of pre-existing receptor clusters only in a subset of synapses contributing to the eIPSC, as sIPSC amplitude represents the average of entire population of GABAergic inputs; consequently, subpopulation-specific changes to synapses could be underestimated. The rise in sIPSC frequency can be primarily attributed to N but not p_r increase, as linear regression analysis found that sIPSC frequency change strongly correlated with the change in eIPSC 1/CV² (r = 0.7, p = 0.0007; Supplementary Fig. 2) but not PPR (r = 0.3, p = 0.3). Consistent with our observations for eIPSCs, HFS failed to affect either the frequency or amplitude of sIPSCs in the naïve rats (frequency, 87 ± 7% of baseline, p = 0.07; amplitude, 100 ± 4 %; p = 0.9), suggesting that HFS-induced sIPSC frequency facilitation is stress-dependent. Linear regression analysis revealed that the increase in sIPSC frequency positively correlated with the increase in amplitude of eIPSC (r = 0.6, p = 0.01; Fig. 5h), while sIPSC amplitude did not (r = 0.4, p = 0.2; Fig. 5i). This indicates that, similar to the conclusion derived from 1/CV² and PPR analysis, LTP_GABA can be primarily explained by an increase in N.

LTP_GABA requires postsynaptic mechanisms

We next examined biochemical mechanisms responsible for LTP_GABA. In the hippocampus and cerebellum, glutamatergic signaling that leads to heterosynaptic LTP_GABA requires elevations in postsynaptic Ca²⁺ concentration, activation of Ca²⁺/calmodulin-dependent kinase II (CaMKII) and surface trafficking of GABA_ARs^29–31. In support of a requirement for postsynaptic Ca²⁺ signaling during LTP_GABA in our experiments, postsynaptic loading with Ca²⁺ chelator 1,2-bis(o-aminophenoxy)ethane-N,N,N′,N′-tetraacetic acid (BAPTA, 10 mM) completely blocked HFS-induced LTP_GABA (95 ± 7% of baseline, p = 0.4; Fig. 6a). Elevated intracellular Ca²⁺ triggers calmodulin binding to CaMKII, autophosphorylation of a specific threonine residue (Thr286) and generation of autonomous (Ca²⁺/calmodulin-independent) CaMKII activity. Phosphorylation of Thr286 is required for GABA_AR surface trafficking in hippocampal neurons³² and is enhanced in parvocellular subdivision of the PVN following stress³³. Inclusion of the specific CaMKII inhibitor autocamtide-2-related inhibitory peptide (AIP, 1 μM) in the patch pipette abolished HFS-induced LTP_GABA (97 ± 5% of baseline, p = 0.7; Fig. 6b). Next we evaluated the involvement of SNARE-dependent exocytosis in the postsynaptic cell, a process required for the membrane GABA_AR insertion. Inclusion of botulinum toxin C (BoNT/C, 5 μg/ml) in the patch pipette, which blocks SNARE-dependent exocytosis, prevented LTP_GABA (116 ± 8% of baseline, p = 0.2; Fig. 6c). Compared to BAPTA and AIP, however, BoNT/C was less effective in preventing the early phase of LTP (~ 10 min after HFS), suggesting that there are additional mechanisms contributing to the onset of potentiation independent of BoNT/C-sensitive exocytosis. In addition to blocking eIPSC potentiation, the infusion of BAPTA, AIP, or BoNT/C into the postsynaptic cells completely blocked the increase in sIPSC frequency (BAPTA, 98 ± 12% of baseline, p = 0.9; AIP, 102 ± 19%, p = 0.8; BoNT/C, 71 ± 9%, p = 0.02; Fig. 6d), suggesting that the increase in N requires similar postsynaptic mechanisms. Interestingly, BoNT/C, but not BAPTA or AIP, decreased sIPSC frequency to below pre-HFS levels, without altering the amplitude (102 ± 6% of baseline, p = 0.8; Fig. 6e), implying that the blockade of exocytosis unmasked an opposing elimination of functional synapses presumably downstream of Ca²⁺ and CaMKII. BAPTA and AIP did not change sIPSC amplitude (BAPTA, 96 ± 6% of baseline, p = 0.6; AIP, 107 ± 5% of baseline, p = 0.3; Fig. 6e).

Figure 6. LTP_GABA requires postsynaptic mechanisms.

(a–c) Left, average sample traces of eIPSC before (grey) and after (colored) HFS recorded in slices from IMO stressed rats where BAPTA (green, n = 9 (7 rats)), AIP (purple, n = 6 (3 rats)), or BoNT/C (blue, n = 6 (6 rats)) was included in the patch pipette. Without inhibitor group (pink, adopted from experiment shown in Fig. 1d) is shown for comparison. Right, summary time-course of normalized eIPSC amplitude. Arrows indicate the time of HFS. Scale bars represent 50 pA and 20 ms. (d–e) Summary of the effects of HFS on sIPSC frequency (d) and amplitude (e) where BAPTA, AIP or BoNT/C were included in the pitch pipette. Data represent mean ± s.e.m. * p = 0.02, *** p = 0.002.

Increase in functional synapses requires postsynaptic mechanisms

Our data strongly implicate both increase in N and postsynaptic vesicle exocytosis in the potentiation of eIPSCs (i.e. LTP_GABA). To directly examine the alterations in the quantal events composing eIPSCs and their causality with postsynaptic exocytosis, we examined the frequency and amplitude of Sr²⁺-induced asynchronous eIPSCs before and after LTP_GABA and their sensitivity to postsynaptic infusion of BoNT/C. Substitution of Ca²⁺ by equimolar amount of Sr²⁺ in aCSF alters fast-onset, synchronous synaptic responses into slow-onset asynchronous quantal events. This approach has been used to determine whether synaptic plasticity involves an increase in the number (frequency) and/or size (amplitude) of the quantal events^34,35. We first validated that Sr²⁺ substitution indeed produced asynchronous synaptic events identical to miniature events in GABA synapses onto PNCs (Supplementary Fig. 4). To compare the changes in asynchronous events associated with LTP_GABA, we next replaced Ca²⁺-ACSF with Sr²⁺-ACSF before and after the induction of LTP_GABA in slices from IMO stressed rats. Figure 6a–c shows the expression of LTP_GABA similar to that observed in Fig. 1d and asynchronous synaptic events before and after LTP_GABA induction. In line with our previous results, we found significant increase in the frequency of asynchronous eIPSCs (152 ± 18% of baseline, p = 0.03, Fig. 7b) with no change in amplitude (106 ± 10%, p = 0.6, Fig. 7c). We next examined the role of exocytosis by repeating the same experiment but with BoNT/C included in the patch pipette. As observed in the previous experiment (Fig. 6c), BoNT/C prevented LTP_GABA (Fig. 7d). In line with the blockade of LTP_GABA, BoNT/C effectively blocked increase of asynchronous IPSCs frequency (131 ± 17%, p = 0.1, Fig. 7e). The amplitude of asynchronous IPSC was not affected (112 ± 8%, p = 0.2, Fig. 7f). These results provide further support for our hypothesis that N increase contributes to LTP_GABA through BoNT/C-sensitive postsynaptic mechanisms.

Figure 7. Increase of active synapse number during LTP_GABA requires postsynaptic mechanism.

(a, d) Top, sample traces of asynchronous eIPSC in Sr²⁺ aCSF before (black) and after (red/blue) LTP_GABA from the same neuron recorded in a slice from IMO-stressed rat. Scale bars represent 50 pA and 20 ms. Arrowheads indicate the time of afferent stimulation. Bottom, summary time course of eIPSC, n = 6 (4 rats). Ca²⁺ was replaced with equimolar Sr²⁺ during the time window indicated by shadows. Cumulative histograms of interevent interval (b, e) and amplitude (c, f) of the Sr²⁺-induced asynchronous eIPSCs before (grey) and after LTP_GABA (red/blue). Experiments were conducted in the absence (a–c) and presence (d–f) of BoNT/C in the patch pipette, n = 6 (4 rats). (g) Left, raw (thin) and averaged (thick) traces of whole-cell current observed in response to focal pressure application of GABA_AR agonist muscimol in slices from IMO-stressed rat. Traces are from the time windows indicated by shadows before (black) and after (red/blue) DHPG treatment. Right, summary time course of normalized muscimol-induced whole-cell current in the presence (blue, n = 6 (4 rats)), or absence n = 7 (5 rats), of BoNT/C in the patch pipette or not (red). Scale bars represent 100 pA and 1 s. Data are mean ± s.e.m.

Although BoNT/C is commonly used to block insertion of GABA_ARs in postsynaptic membrane³⁰, it can also block the release of vesicularly packaged retrograde signaling molecule(s) which in turn acts on the presynaptic terminals⁹. To determine specific contribution of postsynaptic GABA_AR insertion to LTP_GABA, we directly activated postsynaptic GABA_ARs in slices from stressed rats with GABA_AR agonist muscimol and assessed the effects of DHPG (5 μM) application on the whole-cell current. If mGluR1 activation following stress elicits the insertion of GABA_ARs, we predict that more receptors would be available for activation after exposure to DHPG and the whole-cell current induced by direct GABA_AR activation would become larger. As predicted, following DHPG treatment, we observed an increase in the amplitude of whole-cell current in response to muscimol (145 ± 18 %, p = 0.008, Fig. 7g). Inclusion of BoNT/C in the patch pipette abolished this potentiation (99 ± 8 %, p = 0.5). These results argue that SNARE-dependent postsynaptic GABA_ARs insertion contributes to LTP_GABA and that the same mechanism is required for the increase of quantal number during LTP_GABA.

Discussion

Our data demonstrate that, immediately after stress, GABA synapses can undergo activity-dependent LTP. This requires (1) noradrenaline as an obligatory instructive signal during stress (2) activation of mGluR1 by heterosynaptically released glutamate (3) postsynaptic insertion of GABA_ARs. These results further support emerging evidence that the hypothalamic PVN is a critical site for stress memory formation^3,5–11 and identify noradrenaline as a key associative signal that allows this stress effector circuit to undergo experience-dependent plasticity.

In addition to directly altering synaptic efficacy, neuromodulators including neurotransmitters^19–21, neuropeptides⁹, steroids^8,36 and lipid mediators³⁷ are important for more flexible tuning of neural circuits. Specifically, our data provide evidence for stress-induced noradrenaline regulating metaplasticity of GABA synapses. That is, a history of β-AR activation in PNCs during stress provides a higher-order (meta) control over the ability of the synapses to express plasticity²³. Our findings provide, to the best of our knowledge, the first demonstration of β-AR-dependent metaplasticity at GABA synapses, and add to the previous description made at glutamate synapses in the hippocampus and amygdala where β-AR signaling prior to or concomitant with afferent stimulation facilitates the induction of LTP^19–21. Mechanistically, we demonstrate that two types of mGluR1 antagonists completely blocked LTP_GABA, and that priming by β-AR was necessary and sufficient for both mGluR1 agonist-induced and afferent stimulation (which triggers endogenous glutamate release)-induced LTP_GABA. Moreover, β-AR activation needs to precede mGluR1 stimulation. In effect, we propose that β-AR priming involves functional upregulation of mGluR1, whose activation in turn induces LTP_GABA. Although the exact mechanisms mediating mGluR1 functional upregulation in PNCs remain unclear, several lines of evidence suggest a key role for PKA. We provide evidence that PKA activation is necessary and sufficient for the priming of LTP_GABA. PKA activation in turn facilitates LTP_GABA via multiple mechanisms including (1) inhibition of agonist-induced mGluR1 internalization/desensitization²⁷ (2) facilitation of CaMKII activation by inhibiting protein phosphatases that negatively regulate CaMKII²⁴ (3) upregulation of translational process²⁰. Interestingly, glucocorticoids, which function as stress feedback signals to the brain, also unmask mGluR1-³⁶ and mGluR5-dependent (our coordinated submission by Wamsteeker et al.) plasticity following in vivo stress. In these cases, however, the priming of mGluRs was evident 90 min (0 min for the current study) after the termination of stress, suggesting that multiple mechanisms exist for mGluR1/5 modulation by stress that vary in time and context.

Collectively, our data suggest that, following acute stress, mGluR1 activation and subsequent SNARE-dependent exocytosis in the postsynaptic cells lead to LTP_GABA that results from an increase in the number of functional GABA synapses. SNARE-dependent exocytosis serves two mutually nonexclusive mechanisms (1) receptor insertion (2) vesicular release of retrograde messengers that in turn may facilitate presynaptic terminals. Our results argue for the substantial contribution of the former mechanism, although do not exclude the recruitment of the former, as mGluR1-induced exocytosis leads to persistent augmentation of the whole-cell postsynaptic currents in response to focal application of muscimol that reflects the amount of total (synaptic and extrasynaptic) surface GABA_ARs. Interestingly, although postsynaptic infusion of BoNT/C completely abolished the augmentation of muscimol-induced total GABA_AR current, it did not effectively block the early phase of LTP_GABA. Given that GABA_ARs in postsynaptic membranes are dynamic, with continuous turnover between synaptic and extrasynaptic loci^38,39, our results point to an intriguing possibility that LTP_GABA may initially rely on recruitment and clustering of GABA_ARs from the extrasynaptic pool and that exocytotic GABA_AR insertion may be necessary for its stable expression. Although the exact mechanisms await further characterization, our observations collectively suggest that activity-dependent remodeling of GABA_ARs in postsynaptic membranes plays a deterministic role in controlling the number of functional synapses, and thereby the strength of synaptic inhibition on PNCs. This new form of GABA synapse plasticity is reminiscent of ‘silent synapses’ described at glutamate synapses where NMDAR only (silent) synapses before LTP acquire functional AMPARs following afferent stimulation⁴⁰. Indeed, recent imaging studies showing that nascent GABAergic synapses are formed with presynaptic changes preceding postsynaptic appearance of GABA_AR clusters in hippocampal cultures^41,42, support the idea that GABA presynaptic terminals lacking postsynaptic receptor clusters do exist. Intriguingly, a recent study demonstrating that chronic stress causes an anatomical reorganization and increase of GABA synapse contacts onto PNCs¹¹, may hint at the functional end point for the plasticity we described here.

Various types of stress, including IMO, trigger noradrenaline release which in turn increases the firing of PNCs and drives release of CRH^15,16. Although previous studies identified the expression and activities of both α1-, α2- and β-ARs in PNCs^16,43–45, extensive evidence shows that the noradrenaline stimulation of neuroendocrine response is primarily mediated by α1-ARs^15,16,44,46. In line with this, we previously reported that α1-ARs activation causes, through down-regulation of K⁺ Cl⁻ co-transporter KCC2, a depolarizing shift in E_GABA and conditional GABA-mediated excitation of PNCs¹³. These findings clarify a synaptic mechanism by which α1-ARs activation initiates and sustain neuroendocrine response. The present study, on the other hand, reveals a previously unknown consequence of noradrenaline action in the PVN where a history of β-AR activation primes GABA synapses for subsequent LTP. Because GABA becomes excitatory immediately after stress onset, we speculate that this priming mechanism of LTP_GABA may contribute to stress-induced sensitization of the HPA axis. This priming for GABA potentiation, however, is likely to have a “refractory period” when an increased levels of glucocorticoids following stress signal back to the PVN. Our accompanying paper demonstrates that in response to glucocorticoid feedback signaling, the same GABA synapses can be depressed through retrograde opioid signals that decrease the presynaptic release of the transmitter following pro-longed afferent activation. These two opposing mechanisms likely work in a coordinated manner, during temporally distinct windows, to prevent excessive activation of stress axis.

More generally our data are consistent with extensive behavioral evidence that noradrenergic activities, in particular through β-ARs, have been implicated in learning and memory formation associated with emotional arousal and stress challenges¹⁸. Importantly, noradrenaline release in the PVN is not universal but varies as a function of stressor modality^15,47. Consistent with this idea, we observed stress-specificity with respect to priming of LTP_GABA. Thus, this type of plasticity would be particularly useful for sculpting neuroendocrine output based on the modality/intensity of prior stress experience. It is intriguing that β-AR has been proposed as a potential target for stress-related disorders such as posttraumatic stress disorders⁴⁸. Understanding the neurobiological mechanisms of long-term changes in stress-relevant neurocircuitry will provide potential treatment targets for mental health disorders.

Methods

Animal Handling and Stress Procedure

All experiments were approved by the University of Calgary Animal Care and Use Committee in accordance with Canadian Council on Animal Care guidelines. Male Sprague-Dawley rats (postnatal day 22–35, Charles Rivers) were group-housed on a 12 h:12 h light:dark cycle (lights on 06:00) and had free access to food and water. Animals were randomly assigned to naïve or stress groups. For IMO stress procedure, rats were confined and cervically immobilized in a plexiglass restrainer for 30 min. In some experiments, rats were injected with (±)-propranolol (PRO; 10 mg kg⁻¹, intraperitoneal, Sigma-Aldrich) or CP-154,526 (30 mg kg⁻¹, intraperitoneal, Tocris Bioscience) dissolved in saline 30 min before IMO. For predator odor stress, rats were exposed to 2,5-diydro-2,4,5-trimethyulthiazoline (TMT, Contech), a compound isolated from fox feces. A single rat was placed in a testing cage and exposed to TMT (150 μl, absorbed in Kimwipe in a scintillation vial) placed in one corner of the cage for 30 min. For forced swim stress, a single rat was placed in a plastic cylinder (40 cm diameter) filled with water (> 30 cm depth, 30°C) for 20 min. All stress procedures were performed between 10:00 and 12:00. Immediately after stress, rats were anesthetized and brain slices were prepared as described below.

Slice Preparation and Electrophysiology

Animals were anesthetised with isoflurane and decapitated. The brain was quickly removed; it was submerged and coronally sectioned on a vibratome (Leica, Nussloch, Germany) to 300 μm in slicing solution (0°C, 95% O₂/5% CO₂ saturated) containing (in mM): 87 NaCl, 2.5 KCl, 0.5 CaCl₂, 7 MgCl₂, 25 NaHCO₃, 25 D-glucose, 1.25 NaH₂PO₄, 75 sucrose. After placement into aCSF (30°C, 95% O₂/5% CO₂ saturated) containing (in mM):126 NaCl, 2.5 KCl, 26 NaHCO₃, 2.5 CaCl₂, 1.5 MgCl₂, 1.25 NaH₂PO₄, 10 glucose, hypothalamic slices recovered for at least 1 h. Once transferred to a recording chamber superfused with aCSF (1 ml/min; 30–32°C; 95% O₂/5% CO₂), slices were visualized using an upright microscope (BX51WI, Olympus) fitted with infrared differential interference contrast optics. Pulled borosilicate glass pipettes (3–6 M3) were filled with a solution containing (in mM) 108 K-gluconate, 2 MgCl₂, 8 Na-gluconate, 8 KCl, 1 K₂-EGTA, 4 K₂-ATP, 0.3 Na₃-GTP, and 10 HEPES. In some experiments, BAPTA (10 mM, Sigma-Aldrich), AIP (1 μM, Tocris Bioscience) or the light chain BoNT/C (the proteinase subunit which lacks membrane permeability, 5 μg/ml, List Biological Laboratories) were included in the internal solution. For Sr²⁺ experiment, recording chamber was superfused with regular aCSF or Sr²⁺-ACSF (in which Ca²⁺ was replaced with equimolar concentration of Sr²⁺) at a flow rate of 2 ml/min to achieve fast buffer exchange.

Whole-cell patch-clamp recordings were performed from PNCs identified by location, morphology, and current clamp fingerprint⁴⁹. PNCs were voltage-clamped at −80 mV with constant perfusion of 6,7-dinitroquinoxaline-2,3-dione (DNQX; 10 μM, Tocris Bioscience); other drugs were bath applied by perfusion pump. Pairs of GABA_AR-mediated IPSCs were evoked 50 ms apart at 0.2 Hz intervals using a monopolar aCSF-filled glass electrode placed in surrounding neuropil. For HFS, afferents were stimulated at 100 Hz for 1 s (or for 0.5 or 4 s where indicated), repeated four times with an interval of 20 s while postsynaptic cells were held at −80 mV in voltage clamp configuration. In some experiment, lower frequency stimulation (10 Hz for 10 s) was also used. In the experiments where HFS was delivered after DHPG application, half of the recordings (three out of six cells) were obtained in slices pre-treated with DHPG (followed by > 30 min wash) without recording to confirm that the lack of HFS-induced LTP_GABA is not due to intracellular dialysis by long (> 45 min) whole cell recording prior to HFS. Another three recordings were obtained from single cells treated with DHPG and then HFS as shown in (Fig. 2c). To measure the changes in postsynaptic GABA_AR surface expression, we monitored inward current induced by focal application of GABA_AR agonist. Muscimol (100 μM dissolved in aCSF) was pressure applied from a glass pipette whose tip was placed 20–30 μm away from the soma of the postsynaptic cell. Three applications (10 ms, 4 s apart) was delivered every 1 min by using a Picospritzer III (Science Products). The responses to the three successive puffs were averaged for each epoch. Access resistance was continuously monitored; recordings in which values exceeded 20 M3 or 15% change were excluded from analysis. DL-isoproterenol was obtained from Sigma-Aldrich. 8-Br-cAMP, KT5720, (s)-MCPG, CPCCOEt, JNJ16259685, MTEP, bicuculline, (s)-DHPG and muscimol were purchased from Tocris Bioscience.

Optogenetics

Tyrosine hydroxylase (TH)-Cre transgenic mice²² were obtained from the Jackson Laboratory (B6.Cg-Tg(Th-cre)1Tmd/J, Stock Number:008601), and were maintained on a C57BL/6J background. Cre transgenes were genotyped with standard PCR protocols using primers 5′-GCG GTC TGG CAG TAA AAA CTA TC-3′ and 5′-GTG AAA CAG CAT TGC TGT CAC TT-3′ to produce a 100 bp transgenic fragment. Mice were single housed in a cage on a 12 h:12 h light:dark cycle (lights on 06:00) and had free access to food and water. 7 week old mice were anaesthetized using isofluorane (5% at induction, maintained with 1–2% during surgery) with a nose cone, and were given buprenorphine (0.1 mg kg⁻¹) for post-surgery analgesia. The heads of the mice were positioned in a stereotaxic apparatus with bregma and lambda in the horizontal plane. Through a burr hole in the skull, glass capillaries were lowered into the brain at stereotaxic coordinates corresponding to the nucleus of the solitary tract (NTS). Recombinant AAV carrying ChR2-eYFP (Addgene plasmid 20298, pAAV-EF1a-double floxed-hChR2(H134R)-EYFP-WPRE-HGHpA)⁵⁰ and ChR2-mCherry (Addgene plasmid 20297, pAAV-EF1a-double floxed-hChR2(H134R)-mCherry-WPRE-HGHpA)⁵⁰ was pressure injected with Nanoject II apparatus (Drummond Scientific Company) in a total volume of 350 nl (3.4 × 10¹³ GC ml⁻¹). Mice were then allowed to recover > 14 days before ex vivo optogenetic/electrophysiological experiments. The age of mice at the time of electrophysiological experiments was 9–11 weeks old. Both male and females were used. Coronal brainstem and PVN slices (250 μm) were prepared using a vibratome as described above, and transduced neurons (in the caudal medulla) and fibres (in the PVN) were identified by the expression of fluorescent proteins. Photostimulation experiments were conducted in PVN slices from both male and female mice. In a preliminary experiment we did not observe any gender specificity in the expression of LTP_GABA. The fiber optic cable (105 μm core diameter) was placed 1–2 mm above the slice using a manipulator, and a blue-light laser (473 nm, OptoGeni 473, IkeCool corporation) was controlled via Digidata 1440A (Molecular Devices). Light intensity on tissue was 2.5 mW as measured by Photodiode Power Sensor (Thorlabs).

Analysis and Statistics

Signals were amplified (Multiclamp 700B, Molecular Devices), low pass filtered at 1 kHz, digitized at 10 kHz (Digidata 1440, Molecular Devices), and recorded (pClamp 10.1, Molecular Devices) for offline analysis. Evoked IPSCs (eIPSC) were calculated by subtraction of peak synaptic current from pre-stimulation baseline current. Spontaneous IPSC (sIPSC) events, with eIPSCs and stimulus artifacts removed, were detected using variable thresholds and confirmed by eye (MiniAnalysis, Synaptosoft). Data were normalized and expressed as percentage of the baseline values (0–5 min before treatment) within each cell. Post-treatment values were evaluated 15–20 min after treatment unless otherwise specified. Paired-pulse ratio (2nd evoke/1st evoke), sIPSC event frequency/amplitude, and coefficient of variation (CV, standard deviation/mean) were analyzed with 5 minute intervals. Asynchronous IPSCs in Sr²⁺-aCSF were evoked by a single pulse delivered at 0.2 Hz intervals. Amplitudes and frequencies of the asynchronous eIPSCs were measured during a 150 ms period, starting 5 ms after stimulus to exclude the initial synchronous synaptic events. Results were presented as mean ± s.e.m. To ensure physiological significance, recordings for each group were obtained from minimum of three animals. 4 slices were prepared from one animal and the slices were randomly assigned to drug/no-drug treatments. Experimenters were not blinded to the treatment. Sample sizes were determined post-hoc based on those used in previous studies^8,9,13. Because the effects of HFS or drugs were compared to the baseline within cells, one sample parametric t-test (two tailed) was used for all experiment, except for one experiment (TH-cre light stimulation, Fig. 2b) when sample distribution was not normal (used Wilcoxon ranked test). Gaussian distribution of the data was examined by a D’Agostino & Pearson omnibus normality test (GraphPad Prism 4). In one experiment (Fig. 2b, TH-cre light vs. TH-cre light + PRO), the values between groups were compared using non-parametric Mann Whitney test. Correlation between two variables was analyzed using linear regression analysis. p < 0.05 was considered to be statistically significant.