All animal experiments adhered to the Chungnam National University policies regarding the care and use of animals. Male Sprague–Dawley rats (60~80 g) were housed under a 12/12-h light/dark schedule. Rats were randomly divided into two groups: euhydrated (EU) and chronic dehydarated (DE) animals. The EU group was allowed free access to normal tap water, whereas DE rats exposed by 2% saline for 7 days. All rats had access to food water ad libitum except for the DE periods throughout the experiments. Plasma osmolality was measured by freezing-point depression (Fiske Associates, Norwood, MA, USA) prior to sacrifice.

Hypothalamic slices (300 µm) were obtained as previously described [17]. Brian slices containing the SON were cut using a vibroslicer (Leica VT 100s, Leica, Bensheim, Germany) and placed in a holding chamber containing standard artificial cerebrospinal fluid (aCSF) until use. Standard aCSF consists of 126 mM NaCl, 26 mM NaHCO₃, 5 mM KCl, 2.4 mM NaH₂PO₄, 2.4 mM CaCl₂, 1.2 mM MgCl₂ and 10 mM glucose, pH 7.3~7.4. The medium was saturated with 95% O₂ and 5% CO₂.

Patch-clamp recordings were obtained using an Axopatch200-B amplifier (Axon Instruments, Foster City, CA, USA). For voltage-clamp recordings, a low Mg²⁺ aCSF (20 µM MgCl₂) was used to facilitate NMDAR-mediated currents at a holding potential (V_holding) of –70 mV. Periods of 180 sec of synaptic activity were analyzed using the Minianalysis 6.0.3 program (Synaptosoft Inc., Decatur, GA). In some cases, recordings were also obtained at a V_holding of –40 mV or –20 mV in normal standard aCSF. Currents were recorded in the presence of picrotoxin (100 µM) to inhibit ionotropic GABA receptors, if not mentioned. Current output was filtered at 2 kHz and digitized at 10 kHz (Digidata 1322A, Axon Instruments) in conjunction with pClamp 9.2 software. Patch pipettes (3~5 MΩ) were filled with a solution containing 140 mM K-gluconate, 10 mM KCl, 10 mM HEPES, 0.5 mM CaCl₂, 5 mM EGTA and 5 mM Mg²⁺ATP, pH 7.3.

Persistent activation of AMPARs and NMDARs was defined as the difference of holding current (I_holding) before and after application of the receptor agonists: the glutamate AMPA/kainate receptor antagonists, NBQX (2,3-dihydroxy-6-nitro-7-sulfamoyl-benzo[f]quinoxaline-2,3-dione) or DNQX (5,7-dintroquinoxaline-2,3-dione, 10 µM), and the NMDA receptor antagonist AP5 (D,L-2-amino-5-phosphonopentanoic acid, 100 µM), respectively [171819]. The I_holding was measured in 50 ms epochs of traces lacking PSCs using Mini Analysis (Synaptosoft, Decatur, GA, USA).

Series resistance was monitored throughout the experiment. Neurons with series resistance changes >15% during experiments were not included in the analysis.

Numerical data are presented as the mean±standard error of the mean (SEM). Statistical significance of the data was determined using independent or paired Student's t-test as needed.

To determine whether I_NMDA generated by eNMDARs alters AMPARs function to change the response to its agonists, glutamate was bath applied with a known concentration of the agonist (10 µM) in low Mg²⁺ aCSF. Glutamate caused a steady-state inward shift in I_holding (I_GLU) at V_holding of –70 mV, which was blocked by the sequential application of NBQX and NBQX+AP5 in SON MNCs (Fig. 1A and 1B). Interestingly, glutamate failed to change I_holding in the presence of AP5 in SON MNCs (Fig. 1A and 1B), suggesting that glutamate induced persistent tonic activation of AMPARs in an NMDAR-activation-dependent manner in SON MNCs. In agreement with this, glutamate induced NBQX-sensitive currents in normal aCSF when V_holding was depolarized to –40 mV or –20 mV, activating NMDARs. While glutamate caused minimal changes in I_holding (p>0.2, n=5) at V_holding of –70 mV, it induced NBQX-sensitive I_GLU at THE depolarized V_holding (–40 mV, 16.6±2.9 pA, n=7; –20 mV, 14.6±4.9 pA, n=7) (Fig. 1C and 1D). Furthermore, I_GLU was completely inhibited by pretreatment with AP5 at all tested potentials in SON MNCs (data not shown). In a subset of experiments, we further confirmed that I_GLU was blocked by the additional application of AP5 (Fig. 1C) inset.

To determine whether NMDAR activation caused an alteration in the steady state sensitivity of the AMPA receptor-activated channels to its agonists, AMPA was bath-applied in low-Mg²⁺ aCSF and normal aCSF. Application of AMPA (1 µM) induced a reproducible steady-state inward current (I_AMPA) in the SON MNCs, which was blocked by an AMPAR antagonist, DNQX or NBQX (Fig. 2). We recorded and compared I_AMPA amplitudes in the absence and presence of the NMDAR antagonist.

Although there is a tendency for I_AMPA amplitude to increase with repeated AMPA applications, pretreatment with AP5 consistently attenuated I_AMPA amplitude in the first and second AMPA applications (Fig. 2A). The inhibitory effects of AP5 on the first and second I_AMPA amplitudes were tested in the same number of neurons and combined to compare with the control values (Fig. 2B). In the presence of AP5, I_AMPA amplitude was decreased to 60.3±4.06% of that in the absence of the antagonist in low-Mg²⁺ aCSF (control: 134.8±17.0 pA; AP5: 77.8±8.5 pA; n=12, p<0.001) (Fig. 2A and 2B), while pretreatment with AP5 failed to affect I_AMPA amplitudes in normal aCSF (control: 91.7±10.9 pA; AP5: 86.9±11.5 pA; n=10, p>0.15). I_AMPA facilitation by I_NMDA was also evident in normal aCSF at depolarized V_holding activating NMDARs (Fig. 2C and 2D). In agreement with I_AMPA facilitation by I_NMDA in low-Mg²⁺ aCSF, AP5 reduced I_AMPA amplitudes from 92.0±11.2 pA to 65.3±11.7 pA (n= 12, p<0.001) at V_holding of –40 mV in normal aCSF. Pretreatment with AP5 also significantly reduced I_AMPA at V_holding of –20 mV (control: 65.2±7.0 pA; AP5: 39.3±5.1 pA; n=10, p<0.001) (Fig. 2D). In a subset of experiments, AP5 inhibition of I_AMPA was tested in the absence of picrotoxin (Fig. 2C, inset). Pretreatment with AP5 efficiently reduced I_AMPA in the absence of picrotoxin at V_holding of –20 mV in normal aCSF (control: 55.8±6.7 pA; AP5: 35.3±5.3 pA; n=4, p<0.01).

These results suggested that NMDAR activation enhanced the steady-state activation of AMPARs, altering the neuronal response to the receptor agonists in SON MNCs.

In the next experiments, we investigated whether eNMDARs that generate tonic I_NMDA contribute to I_AMPA facilitation in SON MNCs. To isolate eNMDAR function, we adopted an eNMDAR selective antagonist, memantine [2122]. To investigate the role of eNMDARs in I_AMPA facilitation, we recorded and compared I_AMPA amplitude in the absence and presence of memantine in low-Mg²⁺ aCSF (Fig. 3). Memantine induced an outward shift in I_holding (13.2±3.8 pA, n=6), which is comparable to tonic I_NMDA uncovered by AP5 (12.1±2.5 pA, n=7, p>0.7) in low-Mg²⁺ aCSF.

Pretreatment with memantine reversibly and significantly reduced I_AMPA amplitudes from 152.3±24.4 pA to 114.8±24.0 pA (n=8) in low-Mg²⁺ aCSF (p<0.05) (Fig. 3A and 3B), while it failed to affect I_AMPA in normal aCSF (control: 108.8±12.7 pA; memantine: 105.0±13.2 pA; n=6). The inhibition rate of I_AMPA by pretreatment with memantine (65.8±5.4% of control, n=8) was comparable to that of AP5 (60.3±4.06% of control, n=12, p>0.4) in low-Mg²⁺ aCSF. Pretreatment with memantine also significantly reduced I_AMPA at V_holding of –40 mV in normal aCSF (control: 59.2±6.1 pA; memantine: 42.3±5.1 pA; n=6, p<0.01) (Fig. 3B).

These results suggested that eNMDAR activation facilitates I_AMPA in SON MNCs.

To investigate the functional significance of I_AMPA facilitation by tonic I_NMDA in SON MNCs, we examined and compared the I_AMPA potentiation by I_NMDA in euhydrated (EU) and chronic dehydrated (DE) rats (Fig. 4). Chronic dehydration with a 7-day salt loading (2% NaCl) protocol increased plasma osmolarity from 306.7±2.3 mOsm (n=4) to 378.5±7.1 mOsm (n=5). Consistent with a previous report [23], DE significantly increased EPSC frequency in SON MNCs (EU: 1.81±0.35 Hz, n=6; DE: 4.09±0.62 Hz, n=7).

Bath application of AMPA induced reproducible I_AMPA in DE SON MNCs as in EU SON MNCs. Pretreatment with AP5 significantly reduced I_AMPA in low-Mg²⁺ aCSF at V_holding of –70 mV (control: 68.6±15.0; AP5: 45.9±12.8 pA; n=8, p<0.05) and in normal aCSF at V_holding of –40 mV (control: 70.1±10.4 Pa; AP5: 47.9±8.6 pA; n=8, p< 0.001) in DE SON MNCs. Although I_AMPA amplitudes were slightly lower in DE than in EU neurons, the difference did not reach statistical significance (p>0.2 in both cases). Furthermore, pretreatment with AP5 inhibited I_AMPA at similar rates in EU and DE SON MNCs (Fig 4), suggesting that I_AMPA potentiation by I_NMDA was preserved in DE SON MNCs.

Although receptors are often considered extrasynaptic if they are located more than 100 nm from the postsynaptic density, the precise delineation of the synaptic-extrasynaptic receptors seems to be specific to the parameter under considerations [24]. In terms of electrophysiological activities, synaptic NMDARs are defined as receptors recruited in response to spontaneous glutamate release, generating EPSCs, or during low-frequency afferent stimulation (less than 0.05 Hz), while eNMDARs correspond to those not activated during such conditions. In this sense, there is a general consensus that eNMDARs are responsible for the persistent tonic excitatory current in SON MNCs [625], while their synaptic counterparts mediate conventional EPSCs. In the present study, I_holding shift by a NMDAR antagonist, AP5, supported the presence of a persistent inward current with basal glutamate release mediated by NMDARs in SON MNCs. Combined with the fact that memantine failed to affect the basic properties of glutamate EPSCs in SON MNCs [6], our results showed that memantine, a selective eNMDAR blocker, mimicked the AP5-induced I_holding shift, which confirmed the notion that eNMDARs generate tonic I_NMDA in the neurons. The subunit composition of the eNMDAR mediating tonic I_NMDA in various brain regions has not been well clarified, while eNMDARs containing the NR2B subunit have been known to be partly responsible for tonic I_NMDA [6]. Future studies are warranted to delineate the subunit composition of the eNMDARs mediating I_AMPA facilitation in SON MNCs.

Our results showing that I_AMPA was inhibited by AP5 or extracellular Mg²⁺ in low-Mg²⁺ aCSF suggested that activated NMDARs facilitated AMPARs function in SON MNCs. Furthermore, similar inhibitory effects of AP5 and memantine on I_AMPA facilitation suggested that eNMDARs facilitated AMPAR function in the neurons. Combined with the fact that I_GLU amplitudes were dependent on AP5 (Fig. 1), these results suggested that eNMDAR activity modulates the steady-state sensitivity of AMPA receptors to their agonists in SON MNCs. Given that binding glutamate/AMPA to the AMPA receptor results in Na⁺ influx, which causes depolarization of the membrane, our results showing that AP5 inhibited I_AMPA/I_GLU could indicate that AP5 inhibited eNMDAR activated by the agonists. However, our results showing that AMPARs antagonists completely abolished I_AMPA in both the absence and presence of NMDAR antagonists (Fig. 2) argued against this possibility. It is not likely that a bath application of AMPA recruited additional NMDARs generating I_NMDA in our recording conditions. This idea was further supported by the finding that I_AMPA was insensitive to the following application of AP5 (data not shown).

In general, there has been a consensus that extrasynaptic glutamate receptors generate slow, persistent tonic currents, while their synaptic counterparts mediate AMPAR- or NMDAR-EPSCs. In the present study, it is not clear whether extrasynaptic AMPARs (eAMPARs) generate I_AMPA, and, if so, what portion of the currents is mediated by eAMPARs in SON MNCs. However, it is interesting to note that eAMPARs are highly mobile and move rapidly between the plasma membrane and the intracellular compartments by exocytosis and endocytosis, and diffuse laterally to and from synaptic sites [26272829]. Such continuous AMPAR exchanges between synapses and different cellular pools ensures a dynamic fit of synaptic AMPAR numbers in synaptic plasticity, including long-term potentiation (LTP). A prevailing two-step model is that NMDAR-dependent LTP is mediated by surface insertion and synaptic delivery of AMPARs, in which AMPARs deliver to the extrasynaptic sites first, and their synaptic targeting requires synaptic NMDAR activation that likely triggers the signal transduction cascade necessary to anchor AMPARs in the synapse. Given that activated eNMDARs cause Ca²⁺ influx [9], it is reasonable to assume that eNMDARs regulated AMPARs trafficking in the present study, resulting in I_AMPA facilitation. As a gliotransmitter, ATP has been known to activate purinergic receptors in PVN MNCs, promoting the insertion of AMPARs at the surface and, strengthening the excitatory synapses [14]. However, it is noteworthy that NMDAR-induced CaM-KII activation alters AMPARs channel properties in both a Ca²⁺-dependent and a Ca²⁺-independent manner [3031]. It is also possible that I_NMDA facilitated AMPAR-mediate currents via increased single channel conductance and/or open channel probability of AMPAR activated channels in SON MNCs. Future studies are warranted to delineate the cellular mechanisms of eNMDAR-mediated I_AMPA facilitation in SON MNCs.

A peculiar property of the SON is that it undergoes anatomical remodeling under certain physiological conditions, such as lactation and chronic dehydration. This remodeling includes a reduction in astrocytic coverage of neurons [32] and is associated with increased extracellular levels of glutamate in the SON [33]. The increased extracellular level of glutamate in the nucleus is also in agreement with enhanced EPSC frequency [23] and increased numbers of excitatory synaptic contacts in SON MNCs. Combined with our results showing that I_AMPA facilitation by I_NMDA is conserved in DE SON MNCs, these results suggested that I_AMPA facilitation by I_NMDA could contribute to enhanced neuronal activity and hormone release from the neurons during the osmotic challenge.

Spontaneous glutamate release from astrocytes synchronizes neuronal activity via eNMDAR activation [3435], and increases the frequency of AMPAR-mediated EPSCs via the activation of metabotropic glutamate receptors facilitating presynaptic release in CA1 pyramidal neurons [36] and kainite receptors in hippocampal interneurons [37], respectively. Combined with the fact that Ca²⁺ influx triggers the signal transduction cascade necessary for anchoring AMPARs in synapses [2638], our results showing that eNMDAR activation potentiated AMPAR-mediated steady-state currents in SON MNCs are in line with astrocytic glutamate release altering AMPARs, resulting in strengthening of excitatory synapses in SON MNCs.

In summary, our results support the view that tonic I_NMDA generated by activated eNMDARs facilitates AMPA receptor function in SON MNCs, which gives the gliotransmitter glutamate the ability to regulate postsynaptic efficacy during normal and physiological challenges, including chronic dehydration.