Acute stress markedly influences cognitive functions (de Quervain et al., 1998; Kuhlmann et al., 2005; Wong et al., 2007) through cellular mechanisms that alter excitatory synaptic plasticity in relevant brain regions (Kim et al., 1996; Joels et al., 2006). Specifically, in the CA1 region of the hippocampus, acute stress facilitates the induction of N-methyl-D-aspartate receptor (NMDAR)-dependent long-term depression (LTD), which is tentatively linked to acute stress-induced impairment of spatial memory retrieval (Xu et al., 1997; Wong et al., 2007). Furthermore, artificial facilitation of LTD induction in the hippocampal CA1 of adult rats using glutamate transporter inhibitors reproduces the effects of acute stress on hippocampus-dependent spatial memory retrieval (Wong et al., 2007). This finding suggests that LTD plays a critical role in the behavioral responses to acute stress. However, the distinct mechanisms underlying stress facilitation of LTD are not well understood.

Previous studies have revealed that LTD is involved in spatial learning and memory (Brigman et al., 2010; Ge et al., 2010) and is a possible mechanism for memory “deletion” or “forgetting” (Collingridge et al., 2010). In experiments, the expression of hippocampal LTD is generally dependent on the stimulus pattern (Buschler et al., 2012) and on the developmental stage of the animals (Wagner and Alger, 1995). Classical low-frequency stimulation (LFS, usually 1–5 Hz, 900 pulses) consistently induces NMDAR-dependent LTD at synapses of Schaffer collateral/commissural (SC) fibers on CA1 neurons in young animals (18–22 day-old) (Wagner and Alger, 1995). In contrast, a similar LFS protocol elicits only short-term depression (Buschler et al., 2012) but not LTD in adult (5–10 week-old) animals both in vivo (Xu et al., 1998; Wong et al., 2007; Buschler et al., 2012) and in vitro (Kim et al., 1996; Otani and Connor, 1998). Thus, the facilitation of LTD induction in adult animals under certain conditions, such as acute stress exposure, suggests that such conditions might alter the induction threshold of hippocampal LTD.

Hippocampal GABAergic interneurons play a central role in modulating the synaptic plasticity of excitatory synapses (Wigstrom and Gustafsson, 1983; Kerr and Abraham, 1995; Nusser and Mody, 2002). In particular, postsynaptic GABA_A receptors regulate the threshold of LTD induction by shunting (as phasic inhibition) and hyperpolarizing (as tonic inhibition) pyramidal neurons (Otani and Connor, 1998; Nusser and Mody, 2002). In adult animals, the strong effects of GABAergic inhibition on NMDAR activation during the LFS episode effectively restrict the induction of NMDAR-dependent LTD, whereas the removal of GABAergic inhibition by blockade of GABA_A receptors facilitates LTD induction (Wagner and Alger, 1995; Otani and Connor, 1998). Therefore, the reduction of the LTD threshold by acute stress suggests that an attenuation of GABAergic inhibition may be involved during stress. Indeed, previous studies have revealed that bath application of corticosteroid to hippocampal slices depressed GABAergic transmission in the CA1 region (Maggio and Segal, 2009).

The endogenous opioid system (EOS) has long been implicated in the stress response. The enhanced release of endogenous opioid peptides during stress can either attenuate or aggravate stress responses depending on the specific opioid receptor that is activated (Valentino and Van Bockstaele, 2008). Moreover, the EOS participates in stress-induced memory impairments, as supported by several lines of evidence: blockade of opioid receptors by naloxone can reverse the effect of acute restraint stress to impair the retrieval of long-term memory in the inhibitory avoidance task (Rashidy-Pour et al., 2004); the corticosterone-induced impairment of memory retrieval in the Morris water maze (MWM) task was blocked by intra-hippocampal infusions of naltrexone (Sajadi et al., 2007); and the application of naloxone can also inhibit the impairment of recognition memory retrieval caused by forced swimming stress in a novel object recognition task (Liu et al., 2016). Importantly, our unpublished data show that μ-opioid receptors (μRs) in the hippocampus are endogenously activated by increased enkephalin (one of the major endogenous opioid peptide neurotransmitters) during acute elevated platform (EP) stress. These receptors then trigger the EP stress-induced spatial reference memory impairment in the MWM task (Cao et al., 2015).

Hippocampal μRs are predominantly located at the axons, terminals, dendrites, and somata of GABAergic inhibitory interneurons (Drake and Milner, 2002). As a vital modulator of GABAergic signaling, exogenously activated μRs can disinhibit pyramidal neurons by inhibiting the firing rate and neurotransmitter release of GABAergic interneurons (Drake and Milner, 2002), therefore altering the excitability of CA1 pyramidal neurons and synaptic plasticity in the hippocampus (Dacher and Nugent, 2011). Considering the critical role of the EOS on both memory processing and stress responses (Valentino and Van Bockstaele, 2008; Shields et al., 2017), we hypothesize that endogenous activation of μRs by acute stress leads to modifications of hippocampal synaptic transmission and plasticity, which may contribute to stress-induced memory impairment. In the present study, we recorded evoked field excitatory postsynaptic potentials (fEPSPs) and population spikes (PSs) in pyramidal neurons of anesthetized adult mice, in combination with the MWM task. These studies allowed us to examine the effect of μRs on changes in synaptic transmission and plasticity in the hippocampal CA1 area following a single exposure to EP stress, as well as the contribution of these receptors to stress-induced memory impairment.

Twenty-four mice were trained in the MWM task. Subsequently, the trained mice were randomly divided into the following groups: unstressed with saline injection, unstressed with morphine injection, and stressed with saline injection (n = 8 for each group). Memory retrieval was evaluated 24 h after the last training session. Compared to the unstressed group that received the saline treatment, the unstressed group that received morphine (15 mg/kg, i.p.), a μR agonist, 30 min before the probe test spent a significantly reduced time in the target quadrant in the MWM task. Administration of EP stress before the probe test produced a reduction in the target quadrant time very similar to that in the morphine treated mice (Figure 2A). To address the underlying neuronal mechanisms of this morphine- or stress-elicited memory deficit, the effect of μR activation exogenously by systemic morphine injection or endogenously by EP stress on excitatory synaptic transmission and plasticity of the hippocampus were examined by recording the SC stimulation-evoked fEPSPs and PSs, respectively, of CA1 pyramidal neurons in anesthetized mice.

Facilitation of NMDAR-dependent LFS-LTD in the hippocampal CA1 region of the adult rodent brain is closely linked to stress-induced impairment of memory retrieval (Kim et al., 1996; Xu et al., 1997; Wong et al., 2007). This alteration of glutamatergic synaptic plasticity is mediated by the activation of glucocorticoid receptors, GluN2B-containing NMDAR, and protein synthesis (Kim et al., 1996; Xu et al., 1998; Joels et al., 2006). Here, we further demonstrate that acute stress alters GABAergic synaptic transmission via activation of hippocampal μR_GABA, as supported by the following evidence: (1) EP stress attenuated both GABAergic feedforward and feedback inhibition of hippocampal CA1 pyramidal neurons in a μR_GABA-dependent manner. (2) EP stress facilitated μR_GABA-dependent LFS-LTD induction. Such stress-induced attenuation of hippocampal GABAergic signaling alters glutamatergic synaptic plasticity and contributes to the learning and memory impairments associated with stress (Figure 7).

The EOS has long been implicated in stress-related behaviors and physiological responses. On the one hand, stress induces an increase in enkephalin, dynorphin, and β-endorphin release in stress-related brain regions (such as the amygdala, hypothalamus, locus coeruleus, nucleus accumbens, hippocampus, etc.). These endogenous opioids activate opioid receptors (μ, δ, κ) to modulate the hypothalamic–pituitary–adrenal axis, autonomic nervous system, and behavioral responses during stress (Bali et al., 2015; Valentino and Van Bockstaele, 2015). In particular, the μR-mediated opioid signaling pathway plays a key role in stress coping and has anti-stress effects, such as reducing the risk of developing depression and attenuating stress-induced anxiety (Bali et al., 2015; Valentino and Van Bockstaele, 2015). On the other hand, the μR signaling pathway has also been shown to be involved in various cognitive tasks, including hippocampus-dependent memory processing (Fichna et al., 2007; Kitanaka et al., 2015; Kibaly et al., 2016). Peripheral or intra-hippocampal administration of naloxone or naltrexone (both μR antagonists) blocks the deficit in memory retrieval induced by acute stress or glucocorticoids (Rashidy-Pour et al., 2004; Sajadi et al., 2007; Colgin et al., 2009; Cao et al., 2015; Liu et al., 2016), suggesting that opioid receptors participate in stress-related memory deficits. The present results provide a reasonable explanation for these behavioral observations: acute stress alters GABAergic inhibition of hippocampal CA1 pyramidal neurons by endogenously activating μRs, thereby altering excitatory synaptic plasticity and ultimately impacting memory retrieval.

GABAergic interneurons are essential for maintaining a functional balance between excitatory and inhibitory neuronal activity throughout the brain. Their network activities regulate activity-dependent excitatory synaptic plasticity (Meredith et al., 2003) and control the oscillations of the hippocampal network, both of which are thought to be essential for memory performance (Csicsvari et al., 2003). Neural disinhibition of the hippocampus caused by dysfunction of GABAergic neurons has been implicated in memory deficits and psychiatric disorders (Marin, 2012). The selective reduction of functional inhibitory synapses or GABAergic depletion in the dorsal hippocampus impairs spatial learning and memory (Reichel et al., 2014). The μR is a vital modulator of the function of GABAergic interneurons. In the hippocampus, μRs are exclusively expressed on GABAergic inhibitory interneurons (Drake and Milner, 2002). Endogenous or exogenous activation of μRs can inhibit the firing rate and neurotransmitter release of the interneurons. GABAergic interneuron-mediated feedforward and feedback inhibition are two fundamental modes of regulating the timing of action potential generation by controlling the temporal summation of excitatory inputs (Pouille and Scanziani, 2001, 2004). We found that EP stress reduced the feedforward (Figure 3) and feedback inhibition (Figure 4) of hippocampal CA1 pyramidal neurons via activation of μR_GABA on GABAergic neurons. The PS amplitude that reflects the recruitment of spiking neurons depends both on excitatory and inhibitory synaptic inputs. The previous study showed that blockade of GABAergic transmission with bicuculline caused a significant increase in the PS amplitude without affecting the fEPSP (Levkovitz and Segal, 1998). The present study revealed that EP stress altered the PS without affecting the basal fEPSP (Figure 2). These results suggest that the inhibitory input rather than the excitatory input to pyramidal neurons is impacted by EP stress, and such depression of inhibitory input disinhibits pyramidal neurons. The reduction in feedforward and feedback inhibition by activation of the μR_GABA located in the stratum oriens and the stratum lacunosum moleculare not only modulates hippocampal network oscillations (Lupica and Dunwiddie, 1991) but also alters glutamatergic synaptic plasticity (Wagner et al., 2001), both of which are associated with hippocampus-based memory encoding/retrieval (Colgin et al., 2009; Ruediger et al., 2011).

In brain slices, activating μRs exogenously using a selective agonist or endogenously by enhancing the release of opioids facilitates 1-Hz LFS-induced LTD at SC-CA1 synapses through modulation of feedforward inhibition (Wagner et al., 2001). Another report showed that exogenously activating μRs also facilitated a 3-Hz LFS-LTD at similar synapses, but this effect was independent of GABA-mediated inhibitory neurotransmission (Tian et al., 2015). Although both forms of the SC-CA1 LTD in vitro require NMDAR activation (Cui et al., 2013), the difference between these two studies raises questions about the functional site of μRs in LTD induction. For example, the activation of μRs in astrocytes cannot be excluded, and μRs are highly expressed on astrocytes in the CA1 hippocampus (Nam et al., 2018). However, the present results show that the facilitation of 3-Hz LFS-LTD in vivo depends on the activation of μR_GABA, which confirms that GABA-mediated inhibitory neurotransmission is required. Although the distinct mechanisms underlying the effects of GABAergic transmission on glutamatergic synaptic plasticity remain speculative, there are several possible ways to generate GABAergic transmission-mediated LTD facilitation: (1) the reduction in postsynaptic GABA_A receptor-mediated GABA response depolarizes pyramidal neurons and thus promotes LTD induction by lowering the threshold for NMDAR activation (removes shunting and hyperpolarizing effects) (Wagner and Alger, 1995); (2) the reduction in postsynaptic GABA_A receptor activation also enhances cAMP/PKA signaling, which subsequently leads to LTD (cAMP-mediated LTD) (Yu et al., 2003); (3) reduced activation of GABA_B receptors at presynaptic excitatory terminals, whose function is to inhibit synaptic transmission via G protein-mediated modulation of presynaptic Ca²⁺ channels, may also facilitate LTD induction. In addition to the CA1 area, μR-mediated facilitation of LTD induction were also reported at excitatory synapses in the dorsal striatum (Atwood et al., 2014) and paraventricular nucleus of the hypothalamus (Wamsteeker Cusulin et al., 2013), indicating μR-mediated long-term plasticity exists widely in emotion- and cognition-related brain regions thereby modulating the relevant behaviors.

Notably, exogenous activation of μRs facilitates LTD induction (Wagner et al., 2001) but suppresses LTP induction (Tian et al., 2015) in SC-CA1 synapses in a manner very similar to the effects of acute stress (Kim et al., 1996). In addition, acute stress exposure precludes the effect of μR activation on synaptic plasticity (Yang et al., 2004). These data indicate that the μ-opioid system is involved in mediating the effects of stress on synaptic plasticity. Our data demonstrating the absence of acute stress-facilitated LTD in μR_GABA-/- mice strongly supports that μRs on GABAergic interneurons play a crucial role in stress-related modulation of synaptic plasticity. Consistent with this finding, recent studies have shown that exposure to stress can alter the GABAergic function via changes in GABA release and the expression of specific GABA_A receptor subunits (Maggio and Segal, 2009); such changes interrupt the normal balance of inhibition and excitation in memory- and emotion-related circuits in the brain (Tzanoulinou et al., 2014; Lee et al., 2016; Hadad-Ophir et al., 2017). In addition to the spatial reference memory observed presently, the opioid receptor-mediated pathway is reportedly involved in stress-altered other hippocampus-dependent learning and memory, such as the inhibitory avoidance task (Rashidy-Pour et al., 2004) and recognition memory task (Liu et al., 2016), implying they might share the similar mechanism. The overall effects of this μR-mediated specific GABAergic pathway on the modulation of synaptic plasticity at excitatory synapses remain to be elucidated, including phasic versus tonic inhibitory currents, feedforward versus feedback inhibition, and other aspects. However, the present work provides novel insight into the cellular mechanisms underlying stress-induced memory deficits.

Z-QL and M-MS designed the research, drafted the manuscript, and supervised the project. K-MF performed the behavioral experiments, analyzed data, and prepared figures. L-JQ collected the electrophysiological data. NM and Y-ND performed in situ hybridization. Z-QQ and C-LW verified the genotype of mutant mice. WR and JH provided revisions to the overall the manuscript. All authors agree to be accountable for the content of the work.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.