Analysis of experimental data showed that IFN-$\alpha$2b had a significant effect on the function of AMPA receptors (AMPARs) after 2 min of perfusion of the vestibular apparatus with cytokine solution (Friedman’s test p 0.001) (Fig. 1).

Preliminary 2-minute perfusion of synaptic area with IFN-$\alpha$2b (10 ng/mL) increased the frequency of background activity, the dynamics of which varied widely from experiment to experiment. Sequential application of AMPA (2 µM) against the cytokine perfusion resulted in a biphasic positive-negative change in the frequency of pulse activity (Fig. 1B). In most experiments, the pattern of response to AMPA application was visually altered by co-perfusion with IFN-$\alpha$2b. However, the ratio of the maximum of impulse activity of AMPA evoked response to the background level altered by IFN-$\alpha$2b was 156.5 $\pm$ 8.0% (hereinafter: mean $\pm$ Standard Error) and was not significantly different compared to test presentation (157.5 $\pm$ 5.5%) (Paired-Samples t-test p $>$ 0.1; Wilcoxon signed-rank test p $>$ 0.1; n = 20). As a rule, the level of background activity was not restored after 15 min of washing the vestibular apparatus in normal solution.

However, 15 min after cessation of the combined action of IFN-$\alpha$2b and AMPA and subsequent washing the synaptic area in normal solution, the ratio of the maximum response rate to AMPA application to the prior background was 180.1 $\pm$ 12.4% and was significantly different compared to the test value (157.5%) (Paired-Samples t-test p = 0.018; Wilcoxon signed-rank test p = 0.015; n = 20) and compared to the ratio of the magnitude of AMPA induced response on the cytokine background (156.5 $\pm$ 8.0%) (Paired-Samples t-test p = 0.005; Wilcoxon signed-rank test p = 0.002; n = 20) (Fig. 1A).

Brief 30–40 s perfusion of the synaptic area with solutions of AMPA and IFN-$\alpha$2b produced a significant change in AMPAR evoked responses relative to the preceding background (Friedman test p = 0.045; n = 9) (Fig. 2).

Importantly, as with the 2-minute cytokine pre-application, the AMPA-evoked response to simultaneous presentation of AMPA and IFN-$\alpha$2b (214.5 $\pm$ 18.8%) did not alter the ratio of the maximum frequency of pulse activity to the preceding background activity level compared to first test presentation of the agonist (218.5 $\pm$ 23.1%) (Wilcoxon signed-rank test p = 0.721). But 15 min after cessation of co-perfusion and wash the vestibular apparatus in normal solution, the ratio of the maximum frequency of pulse activity relative to the preceding background only tended to increase compared with test presentation of AMPA (Wilcoxon signed-rank test p = 0.086; n = 9) and significantly increased compared with simultaneous perfusion of cytokine and AMPA 254.5 $\pm$ 20.9% (Wilcoxon signed-rank test p = 0.038; n = 9) (Fig. 2A).

Comparison of the results of the above two series suggests the potentiating delayed effect of IFN-$\alpha$2b on AMPARs for both prolonged and short-term cytokine exposure.

Brief simultaneous perfusion of the synaptic area with solutions of kainate (2 µM) and IFN (10 ng/mL) increased significantly the ratio of the maximum frequency of pulse activity relative to the preceding background (216.5 $\pm$ 25.3%) compared to the magnitude of the test response (187.3 $\pm$ 15.6%) (Wilcoxon signed-rank test p = 0.016; n = 11) and also potentiated kainate evoked response after 15 min of washing in normal solution (225.0 $\pm$ 30.4%) (Wilcoxon signed-rank test p = 0.05; n = 11) (Fig. 3).

To answer the question about the effect of IFN $\alpha$2b (10 ng/mL) on NMDA receptors (NMDARs), the magnitude of NMDA (50 mM) induced responses was compared before IFN exposure, after 2 minutes of perfusion of the synaptic area with cytokine solution, and 15 minutes after cessation of co-perfusion and washing of all drugs in normal solution (Fig. 4).

The ratio of the maximum frequency of pulse activity to the preceding background activity level differed significantly between NMDA and IFN-$\alpha$2b applications (Paired-Samples t-test p = 0.006; Wilcoxon signed-rank test p = 0.003; n = 11). Although the response pattern to NMDA application may have been visually different from the response after a two-minute perfusion of the synaptic area with cytokine solution, there were no significant differences between the magnitude of response before (219.2 $\pm$ 28.2% ) and after a 2-minute exposure to IFN-$\alpha$2b (198.4 $\pm$ 18.5%) (Paired-Samples t-test p = 0.222; n = 11). The magnitude of the response after 15-minute washing of the vestibular epithelium in normal solution (226.1 $\pm$ 26.6%) did not differ from the test value (Paired-Samples t-test p = 0.52; n = 11) and from the magnitude of the response to NMDA application after 2 min exposure to the cytokine (Paired-Samples t-test p = 0.16; n = 11).

Analysis of the obtained results allows us to conclude that interferon IFN-$\alpha$2b has a different effect on NMDA and non-NMDA receptors. Under these experimental conditions, the cytokine activated AMPARs and kainate receptors (KARs), but had no effect on NMDARs. The nature of the potentiating effect of IFN-$\alpha$2b on AMPARs and KARs had its own peculiarities.

As follows from the results of this work, IFN exerted a delayed potentiating effect on AMPA- and kainate-evoked responses (Figs. 1,2,3). To answer the question of whether IFN-$\alpha$2b can modulate the functional interaction between non-NMDA and NMDA receptors and thus influence the excitation process of the glutamatergic synapse, the effect of AMPA on the NMDA-induced responses was compared in the absence and presence of IFN-$\alpha$2b in the same vestibular apparatus (Fig. 5).

For this purpose, in the first part of the experiment, comparison of the NMDA-evoked responses before (NMDA (1) in Fig. 5A) and after AMPA exposure (NMDA (2) in Fig. 5A) allowed us to assess the effect of AMPARs activation on NMDARs. In the second part of the experiment, the vestibular apparatus was perfused for 2 min with IFN-$\alpha$2b solution (10 ng/mL), and then AMPA was applied against the background of continued perfusion with cytokine. After a 15-minute wash of the vestibular epithelium in normal solution, NMDA was again applied twice at 15-minute intervals. Comparison of responses to NMDA application before and after activation of AMPARs by cytokine would allow us to confirm or refute the hypothesis about possible modulation of functional interaction between AMPA and NMDA receptors.

Since the analysis of data distribution in some series of experiments revealed significant differences from normal, the Friedman test, Wilcoxon paired comparisons and in addition ANOVA RM analysis of variance for dependent variables were applied. The ratio of maximum frequency of impulse activity relative to the preceding background during NMDA application (50 mM) was found to increase 15 min after AMPA application (2 µM) (Wilcoxon signed-rank test p = 0.093; Paired-Samples t-test p = 0.043; n = 10), which implies minimal delayed potentiation of NMDARs by AMPARs 15 min after cessation of AMPA application and washing of the synaptic area in normal solution. Subsequent perfusion of the synaptic area with IFN and IFN+AMPA solutions had a significant effect on NMDA responses (ANOVA RM F(2.18) = 5.019, p = 0.019, $\eta$² = 0.358).

15 min after cessation of co-perfusion of the synaptic area with IFN and IFN+AMPA solutions the ratio of maximum response frequency to the preceding background of NMDA responses was 190.5 $\pm$ 18.6% (NMDA (3) in Fig. 5A), which was significantly different from the test NMDA-induced response after AMPA application (NMDA (2) in Fig. 5A) (ANOVA, p = 0.045; n = 10; Wilcoxon signed-rank test p = 0.022; n = 10). In 30 min wash in normal solution the ratio of NMDA-evoked response to resting activity equaled 201.7 $\pm$ 17.6% (NMDA (4) in Fig. 5A), that also significantly differed from the NMDA-induced response after AMPA exposure (NMDA (2) in Fig. 5A) (ANOVA p = 0.028; n = 10; Wilcoxon signed-rank test p = 0.009; n = 10).

Moreover, there was no significant difference between NMDA-induced responses 15 and 30 min after cessation of IFN + AMPA application (Wilcoxon signed-rank test p = 0.309; n = 10), suggesting a prolonged delayed stimulatory effect of interferon-activated AMPARs on NMDARs.

Thus, our data suggest that IFN-$\alpha$2b itself does not alter NMDA responses but exerts a delayed activating effect on AMPARs, which in turn potentiate NMDARs.

There is now no doubt that the nervous and immune systems are in continuous interaction. Under normal conditions, the release of inflammatory mediators usually represents an adaptive and regulated brain response to immune signals. When the immune challenge becomes prolonged and/or uncontrolled, the subsequent inflammatory response leads to maladaptive synaptic plasticity and brain disorders [2]. A growing body of evidence indicates that some of the immune system proteins, in addition to their immunological function, play a critical role in the formation, function and modulation of synaptic contacts [16, 17, 18]. This group includes proteins of innate immunity (Dscam, pentraxins, C1q and C3 compliments), the major histocompatibility complex class I (MHCI) family of proteins, and also pro-inflammatory cytokines (TNF$\alpha$, IL-6) [19, 20, 21, 22, 23]. Numerous studies show that both immune and neural functions of these proteins may be equally important for different cellular and systemic functions. It is known that pro-inflammatory cytokines can be synthesised in the normally functioning brain, actively penetrate the blood-brain barrier during inflammatory processes [24, 25], and have multidirectional modulatory effects on CNS functions [26]. It was shown, using mutational analysis, that some of the polymodal functions of IFNs, in particular immunoregulatory and neuroregulatory effects, were triggered by distinct domains of the cytokine molecule contacting different receptors [27, 28].

IFNs are known to have anti-inflammatory and immunomodulatory effects against viral and malignant cells [29]. In vivo IFNs are produced in the CNS by neurons, astrocytes and microglia, as well as immune cells: macrophages, monocytes, T-lymphocytes—at a low physiological level [30, 31].

IFNs, belonging to the group of mediators of innate and acquired immunity [1], are divided into three types according to the types of membrane receptors to which the cytokine binds [32].

Signalling from the interferon type I receptor (IFNRI) and activation of the Janus kinase 1/signal transducers and activators of transcription (JAK/STAT) pathway is a hallmark of viral control and host defence [4]. In addition to their protective function, IFNs, in a dose-dependent manner, exert multidirectional effects on synaptic processes.

Modulatory and toxic effects of different types of IFNs have been described in the CNS. Clinical and experimental data indicate that low doses of IFNs don’t cause severe side effects. Neurotoxic effects of IFNs in the glutamatergic system at the neuronal level are associated with activation JAK/STAT (Janus kinase 1/signal transducers and activators of transcription) cascade, and with direct and indirect impact on the various subtypes of iGluRs [33, 34]. IFNs alter the dendritic tree pattern, the number and nature of synapses, specifically decrease the levels of presynaptic protein vesicular glutamate transporter and postsynaptic density protein-95 [35].

The analysis of the mechanisms of IFNs influence on synaptic transmission is based mainly on the data obtained for CNS structures. The modulatory effects of IFN differ significantly in different CNS structures. IFN not only has both excitatory and inhibitory effects in various structures of the CNS, but also causes multidirectional effects on neurons of the same structure [33, 35, 36]. IFN-$\alpha$ suppresses excitatory synaptic transmission in spinal cord neurons, reducing the frequency of spontaneous excitatory postsynaptic currents [37], suppresses neuronal activity in the hypothalamus but activates it in the amygdala, hippocampus and in cortical neurons [38, 39]. IFN-$\beta$ significantly reduced the amplitude of long-term potentiation (LTP). Importantly, at the cellular level, IFN dose-dependently influenced subthreshold membrane response by raising the membrane resistance and the membrane time constant [39].

Data of clinical and experimental physiology provide numerous evidences of ototoxic effect of IFN on inner ear functions [40, 41, 42, 43, 44]. One of the causes of vestibular disorders may be impaired synaptic transmission in vestibular organs, which is reflected in changes in the nature and ratio of spontaneous and evoked activity in synaptic structures. This results in the balance problems, from minor sensations of instability to severe vertigo attacks. Our studies demonstrate for the first time that IFN-$\alpha$2b affects glutamatergic synaptic transmission in the posterior semicircular crista [13]. In this work, we continue and refine previous preliminary studies [45] and demonstrate different functional interactions of IFN with NMDARs, KARs, and AMPARs, as they are the receptors that trigger the most rapid excitatory synaptic processes whose hyperactivation may lead to excitotoxicity.

In our experiments, interferon increased the level of background activity of afferent fibres in contact with the epithelium of the posterior semicircular canal [13]. Despite the fact that the background activity of afferent fibres of the vestibular epithelium is due to activation of AMPARs [46], it seems unlikely that the observed increase in the level of background activity may be directly due to changes in AMPAR activity, since neither simultaneous nor sequential application of cytokine with AMPA changed the ratio of maximal activity to the preceding background.

The increase in the resting activity level of afferent fibres may be explained by direct interaction of cytokine with IFNRI, so far not identified in the vestibular epithelium, or with opiate receptors, to which cytokine binds specifically [47, 48]. In the CNS IFNs inhibit neuronal activity of the spinal cord, reduce pain responses by acting on µ-opiate receptors [37, 49].

In vestibular organs, opiate receptors, which are represented by µ- and $\kappa$-subtypes, are an important membrane structure involved in the cross-talk between the nervous and immune systems [50]. Earlier we had shown for the first time that the endogenous neutrophil antibiotic defensin, isolated from human and rabbit neutrophils, modulated glutamatergic transmission via opiate receptors [8], suggesting the mechanism other than traditional immunological ones. The possible modulatory effect of IFN on different opiate receptor subtypes in the vestibular epithelium may be the subject of a future separate study.

The increase in the level of background activity of afferent fibres upon IFN influence may also occur indirectly through an increase in the glutamate content in the synaptic cleft, which is carried out through various mechanisms. For example, IFN may change the properties of the basal membrane of the hair cell, which entails a change in the amount of spontaneously released mediator. It has been shown that in the CNS IFN changes membrane properties at the subthreshold level, increasing the resistance and time constant of the membrane [39].

Another mechanism affecting the level of background activity of afferent fibres in the vestibular epithelium under the IFN influence may be the change in the level of the astrocytic glutamate-aspartate transporter (GLAST), which clears the synaptic cleft of excessive neurotransmitter and thus ensures accurate and rapid transmission of information from the hair cell [51]. According to cited auther, IFN decreases the level of GLAST. The opposite effect, an increase in GLAST levels, was observed when interferon receptor (IFNAR) was inhibited [52].

Glutamate release from the basal membrane of the hair cell can also be modulated indirectly, through a possible depression of the efferent system. In amphibians, efferent fibres are known to contact directly with type II hair cells [53], supporting the complex and fine tuning of afferent glutamatergic transmission by balancing excitatory and inhibitory influences with the participation of cholinergic, dopaminergic and opioid systems [54]. It is known that acetylcholine (ACh) can have different effects on the level of background activity in the sacculus, utriculus and semicircular canals depending on the type of ACh receptors it affects [54, 55, 56]. In the semicircular canals, ACh increases background activity level by acting on muscarinic ACh receptors [57].

In the efferent synapses in the cochlea and in vestibular organs, ACh is co-localised with dopamine (DA), which is tonically released from efferent fibres and has a neuroprotective effect on glutamatergic transmission through activation of D1 and D2 dopamine receptors [58, 59, 60, 61]. DA has been shown to reduce the level of background afferent fibre activity and responses elicited by the application of glutamate and agonists of iGluRs and mGluRs. It is conceivable that suppression of DA release by IFN or inhibition of postsynaptic DA receptor activity may lead to increased level of background activity in afferent fibres.

As indicated previously, all subtypes of iGluRs have been identified in the vestibular apparatus. Under our experimental conditions, IFN had different effects on NMDARs, AMPARs, and KARs.

In accordance with current evidence KARs are tetramers associated of different combinations of the GluK1-GluK5 subunits. In CNS KARs are involved in the synaptic plasticity process, including long-term potentiation (LTP) and long-term depression (LTD), at presynaptic and postsynaptic levels using ionotropic and metabotropic mechanisms [62, 63].

GluK1-3 has been shown to form a functional homomeric channel. High-affinity GluK4-GluK5 subunits, characterised by a slow deactivation rate, do not function independently but must co-assemble with GluK1-GluK3 subunits for channel function. In contrast, receptors containing the GluK3 subunit, unlike other AMPARs/KARs, desensitise more rapidly at low agonist concentrations, rendering them insensitive to spillover from neighbouring synapses.

The functional role of different KARs in synaptic plasticity (including binding specificity, ion channel permeability and channel block) depends on the composition of the subunits and on the auxiliary subunits (neuropilin- and tolloid-like 1) Neto1 and (neuropilin- and tolloid-like 2) Neto2, which influence their synaptic trafficking [64, 65, 66, 67]. Modulation of MAP (mitogen-activated protein kinase) activity has been shown to be associated with a change in ion channel conformation from a ligand-bound fully open state to a fully closed state [68, 69, 70].

It is accepted that synaptic efficiency depends on the number of receptors on the synaptic membrane, which is regulated by the processes of endocytosis, recycling, exocytosis and lateral diffusion of receptors, including the phosphorylation process [71]. The combination of the above mechanisms provides a fine regulation of synaptic plasticity in various CNS structures at the pre- and postsynaptic level.

In contrast to CNS vestibular KARs are poorly studied. The first direct information about postsynaptic KARs in the vestibular epithelium was obtained in electrophysiological experiments on afferent synapses of the posterior semicircular canal of the axolotl (Ambystoma tigrinum). Block of synaptic transmission induced by a high-Mg²⁺ and low-Ca²⁺ solution did not alter the responses of the glutamate agonists AMPA and kainate, suggesting a postsynaptic influence of kainate and AMPA [72].

Immunoblotting of proteins from inner ear cell homogenate revealed 5 bands associated with KAR subunits with different molecular masses around 48,000 M_r. Immunocytochemical staining with polyclonal and monoclonal antibodies was detected on dendrites of afferent fibres, but not on hair cells or efferent fibres [73]. Immunoreactivity to kainate GluR5&6 and KA1&2 subunits was also detected on membranes of afferent fibres [74]. In vestibular ganglia culture, currents induced by kainate application were inhibited by the AMPA-kainate receptor antagonist CNQX (6-cyano-7-nitroquinoxaline-2,3-dione) and trimetazidine, suggesting postsynaptic localisation of KARs [75]. KARs associated with G protein activation were not detected in vestibular epithelium.

Importantly, in our experiments IFN-$\alpha$2b significantly increased the kainate-evoked response during cytokine application, and further potentiates the response 15 min after rinsing with normal solution, compared to the test value. The different temporal development of early (during simultaneous cytokine application) and late (in 15 min) responses to kainate exposure suggest the involvement of different mechanisms of modulation of KARs by IFN. The first increase in pulse activity observed during short-term simultaneous IFN application may presumably be related to modulation of ion channel function (probability or/and duration of opening). It is the increase in this first response during comparative simultaneous application that distinguishes the responses of KARs and AMPARs. The increased delayed response to kainate exposure recorded after 15-minute washout of the drug is presumably due to an increase in the number of KARs in the synaptic zone.

Since, as mentioned above, KARs have not been identified on hair cells, it can be argued that in our experiments IFN modulated kainate evoked responses at the postsynaptic level.

Although the structure of AMPARs and KARs is different, pharmacological tools that reliably distinguish between AMPARs and KARs have not been developed sufficiently [69]. Taking into account the fact that kainate is partially an agonist of AMPARs, it can be assumed that the delayed potentiation of KARs may be due not only to activation of KARs themselves, but also to activation, at least in part, of AMPARs. At the same time, we cannot exclude the possibility that IFN triggers common hypothetical mechanisms that lead to increased trafficking of both AMPARs and KARs to the synaptic membrane surface.

The background activity of afferent fibres was found to be due to glutamate release from hair cells and activation of postsynaptic AMPARs [46]. It was logically expected that one of the possible mechanisms for the increase in the background activity of afferent fibres upon cytokine exposure would be a change in the activity of AMPARs. However, this hypothesis was not confirmed by experimental data, because the ratio of the maximum frequency of pulse activity during AMPA application to the preceding level of background activity did not change during both short-time and long-time acute IFN applications.

At the same time, a delayed potentiating effect of the cytokine on AMPA-evoked responses has been observed 15 min after cessation of all exposures and washing vestibular epithelium in normal solution with both short-term and long-term IFN exposure. This indicates that the delayed potentiating effect of cytokine on AMPA-evoked responses is independent of the duration of IFN exposure. This suggests that IFN acts as a trigger of the delayed potentiating response of AMPARs.

The current literature on the effect of type I IFN$\beta$-1a on AMPARs are extremely limited, and the experiments were performed over different time frames, which makes correct comparative analysis of the data difficult.

Similar data on the effect of IFN$\beta$-1a on AMPARs function were obtained on striatum slices, where IFN$\beta$-1a had no effect on pharmacologically isolated AMPARs in the presence of the specific NMDAR antagonist amino-phosphono-valeriat (APV). Under these experimental conditions, IFN$\beta$-1a did not alter excitatory postsynaptic currents (EPSC) amplitude compared to control [36]. However, the authors did not investigate the delayed effect of the cytokine as we had performed on vestibular epithelium.

On the other hand, IFNs may exert toxic effects on CNS neurons via AMPARs. On mouse cortical neurons IFN-$\mathrm{{\rm Y}}$ was shown to cause an increase in Ca²⁺ entry into the cell, a decrease in Adenosine triphosphate (ATP) concentration, and an increase in nitric oxid (NO) concentration. IFN-$\mathrm{{\rm Y}}$ induced the formation of a unique membrane receptor complex consisting of type II interferon receptors (IFNR $\mathrm{{\rm Y}}$) and the Ca²⁺-permeable GluR1 subunit of AMPA receptors. It was shown that IFN-$\mathrm{{\rm Y}}$ phosphorylated GluR1 subunit and increased the formation of dendritic beads, which was eliminated by specific antagonists of AMPAR CNQX (6-cyano-7-nitroquinoxaline-2,3-dione) and specific antibodies to IFNRs [34].

Comparison of our results with the few current data [34, 76] demonstrates the principal possibility of interaction of different types of IFNs with AMPARs. The modulatory effect of type I and type II IFNs on AMPARs can be manifested in wide time intervals (from minutes to days) and trigger neuroimmunomodulatory and neurotoxic processes. The study of the mechanisms of influence of different types of IFNs on AMPARs is of special interest and requires close investigation, since hyperactivation of AMPARs may cause excitotoxicity, destruction of glutamatergic synapse.

According to the literature, AMPARs are tetramers consisting of dimers composed of GluA1/GluA2 and GluA2/GluA3 subunits [77, 78, 79, 80]. The function of AMPARs depend on subunit composition and on different auxiliary proteins (associate with the AMPAR subunits during their assembly). Each of AMPAR subunit has a large N-terminal extracellular domain (NTD), highly conserved extracellular ligand-binding domain (LBD), Transmembrane domains (TMD) 1–4 and C-terminal intracellular domain (CTD). It is the TMD2 that forms the pore-lining region and responsible for the channel properties of the receptor. If TMD2 of GluA2 is subject to nucleotide editing (conversion of glutamine (Q) to an arginine (R) at position 607)—the charge of the cannel pore changes from neutral to negative resulting in to formation of Ca²⁺-Impermeable AMPAR (CI-AMPAR). Ca²⁺-Permeable AMPARs (CP-AMPARs)—those lacking an edited GluA2 subunit—have a higher single-channel conductance than CI-AMPAR containing calcium-impermeable receptors. It is accepted that change of subunit composition or increase of GluA1:GluA2 ratio in synaptic zone may serve as a molecular base of short term plasticity.

We propose that interferon causes an increase in the number of CP-AMPARs in the synaptic zone either by the nucleotide editing of GluA2 at position 607 (R/Q), follow by a change in the channel charge, or by increasing the traffic of CP-AMPARs using auxiliary proteins.

Quantity of AMPARs in the synapse is in dynamic equilibrium between the synaptic and extrasynaptic membranes and the intracellular space through the mechanisms of AMPA exocytosis, lateral diffusion and clathrin-dependent endocytosis [81, 82].

The increase or decrease in the number of AMPARs on the synaptic surface is a key mechanism of synaptic plasticity revealed during LTP and LTD, and under functional stress. For example, it has been shown that accelerated AMPAR incorporation is observed during LTP [83]. In contrast, a decrease in AMPARs is associated with LTD or with the removal of desensitised AMPARs from the synaptic zone [82, 84, 85].

The process of AMPAR recycling has been identified in the central and peripheral nervous system: in the synapses of ascending fibres of Purkinje cells [86], in the cortex [87], in the hippocampus [82], and in structures of the inner ear [88]. Specifically, suppression of AMPAR endocytosis in cochlear neuronal cultures upon exposure to a glutamate agonist and acoustic loading results in excitotoxic responses to acoustic stimuli that were normally not excitotoxic. These included vacuolization in the nerve terminals and spiral ganglion, as well as irreversible threshold shifts. These data suggest that endocytosis of AMPAR plays an important protective role during excitotoxicity [88].

The process of dynamic equilibrium between synaptic, extrasynaptic and intracellular space can be modulated by endogenous ligands: insulin [89], stress hormone [90], and the pro-inflammatory cytokine TNF$\alpha$, which is released during damage and inflammation [91].

The effect of TNF$\alpha$ on AMPAR trafficking was studied in a spinal cord injury model. Western blotting with antibodies to AMPA1 and AMPA2 subunits revealed a significant increase in the GluR1:GluR2 ratio in the plasma membrane, indicating an increase in the proportion of receptors devoid of GluR2 in TNF$\alpha$-treated spinal cord samples. TNF$\alpha$ nanoinjections have been shown to induce a rapid dose-dependent increase in the number of AMPARs lacking GluR2 in the plasma membrane of neurons at both extrasynaptic and synaptic sites. TNF$\alpha$ sequestration reduced both AMPAR delivery to the membrane and acute neuronal excitotoxicity. These results suggest that spinal cord injury causing an increase in TNF$\alpha$ levels alters AMPAR transport [91].

In hippocampal and cortical cultures, glial TNF$\alpha$ has been shown to induce a rapid (15 min) increase in neuronal, surface localised synaptic AMPARs, presumably from intracellular depots. It is generally believed that an excess of surface localised AMPARs may predispose a neuron to excitotoxicity [92].

It was shown that in hippocampal neurons TNF$\alpha$ promoted exocytosis of AMPAR2 subunit with the participation of phosphotodilinositol-3 kinase mechanism. TNF$\alpha$ promoted endocytosis and reduced the number of gamma-Aminobutyric acid (GABA) receptors, which suppressed inhibitory processes [23]. Taking together, the data speaks in favor that TNF$\alpha$ may regulate membrane homeostasis.

There are no direct data indicating a change in AMPARs trafficking to the synaptic area upon exposure to IFN. We hypothesise that changes in the number of synaptic membrane receptors under the influence of proinflammatory cytokines is a common mechanism for the central and peripheral nervous systems, and that in our experiments IFN-$\alpha$2b is involved in the process of restructure AMPARs and trafficking of CP-AMPAR into the synaptic area, thereby increasing delayed AMPARs activity.

The study of the mechanisms of IFN influence on the synaptic recycling of AMPARs is a necessary step in the development of methods to prevent excitotoxicity and nerve cell death in inner ear structures.

Activation of AMPARs is accompanied by the entry of Ca²⁺ into the cell and subsequent membrane depolarization. In our experimental conditions, IFN-$\alpha$2b, although it changed the NMDA responses pattern, had no significant effect on them. Based on the above, we assumed that possible delayed membrane depolarization caused by IFN-$\alpha$2b action on AMPARs could change the function of NMDARs, which were not active in the absence of depolarization.

The results of our work indicate that different subtypes of iGluRs are affected by IFN to varying degrees. Analysis of literature data and our results indicate that IFN has a complex effect on glutamatergic synaptic transmission in the structures of the inner ear. It is known that AMPARs and NMDARs are co-localized on the hair cell and on the postsynaptic membrane.

Summarizing the data obtained, we suggest that IFN leads to a delayed activation of AMPARs, which causes long-term membrane depolarisation and activation of NMDARs (presumably by removing Mg²⁺ from the cationic channel of NMDAR and subsequently induced an influx of Ca²⁺ in cell) (Fig. 6).

At the presynaptic level, this is accompanied by activation of the glutamatergic ribbon synapse and abundant release of glutamate from the hair cells. At the postsynaptic level, it is accompanied by activation of postsynaptic glutamate receptors. In the vestibular epithelium, it is AMPARs that trigger the excitation process, determine the background activity of the afferent synapse, which maintains current excitability of the postsynaptic membrane and together with other GluRs provide instant transmission of information from the hair cell.

It is known that interferon disrupts the function of the glutamate transporter, leading to excess glutamate in the synaptic cleft, resulted in hyperactivation of GluRs and disruption of the balance of positive and negative process. In accordance of current evidence appropriate NMDAR activation is essential for neuronal survival and physiological functions, excessive activation contributes to pathological changes including cell death. It was suggested that the degree of excitotoxicity depends on the magnitude and duration of synaptic and extra-synaptic NMDAR co-activation [93]. While low-dose NMDA preferentially activated synaptic NMDAR, higher doses progressively activated increasing amount of extra-synaptic NMDAR along with synaptic NMDAR and triggered cell death program.

The delayed IFN-induced increase in AMPAR increase NMDA responses and changes cross talk of NMDAR and non-NMDA receptors.