GluN2 Subunit-Dependent Redox Modulation of NMDA Receptor Activation by Homocysteine

Homocysteine (HCY) molecule combines distinct pharmacological properties as an agonist of N-methyl-d-aspartate receptors (NMDARs) and a reducing agent. Whereas NMDAR activation by HCY was elucidated, whether the redox modulation contributes to its action is unclear. Here, using patch-clamp recording and imaging of intracellular Ca2+, we study dithiothreitol (DTT) effects on currents and Ca2+ responses activated by HCY through native NMDARs and recombinant diheteromeric GluN1/2A, GluN1/2B, and GluN1/2C receptors. Within a wide range (1–800 μM) of [HCY]s, the concentration–activation relationships for recombinant NMDARs revealed a biphasicness. The high-affinity component obtained between 1 and 100 µM [HCY]s corresponding to the NMDAR activation was not affected by 1 mM DTT. The low-affinity phase observed at [HCY]s above 200 μM probably originated from thiol-dependent redox modulation of NMDARs. The reduction of NMDAR disulfide bonds by either 1 mM DTT or 1 mM HCY decreased GluN1/2A currents activated by HCY. In contrast, HCY-elicited GluN1/2B currents were enhanced due to the remarkable weakening of GluN1/2B desensitization. In fact, cleaving NMDAR disulfide bonds in neurons reversed the HCY-induced Ca2+ accumulation, making it dependent on GluN2B- rather than GluN2A-containing NMDARs. Thus, estimated concentrations for the HCY redox effects exceed those in the plasma during intermediate hyperhomocysteinemia but may occur during severe hyperhomocysteinemia.


Introduction
Homocysteine (HCY) is a thiol-containing amino acid derived from methionine metabolism as a byproduct. The deficit of B 6 and B 12 vitamins [1] or mutation in the methylenetetrahydrofolate reductase gene [2] lead to HCY accumulation in the blood plasma and cerebrospinal fluid, which can exceed 100 µM in severe hyperhomocysteinemia [3]. Elevated HCY concentrations contribute to neuronal death in many neurodegenerative diseases like stroke, Alzheimer's disease, Parkinson's disease, etc. [4,5]. In vitro, HCY promotes apoptosis in cortical [6,7] and cerebellar [8,9] neurons.

Patch-Clamp Recordings
Whole-cell currents from cultured neurons or HEK293 cells were recorded using a MultiClamp 700B patch-clamp amplifier with Digidata 1440A controlled by pClamp v10.2 software (Molecular Devices). Recordings were low-pass-filtered at 100 Hz. Perfusing solution exchange was performed by BPS-8 fast-perfusion system (ALA Science Inc., Farmingdale, NY, USA) with the tip of the manifold placed 100-200 µm from the recorded cell. The external bathing solution contained (in mM): 140 NaCl; 2.8 KCl; 1.0 CaCl 2 ; 10 HEPES, at pH 7.2-7.4, 310 mOsm. The pipette solution contained (in mM): 120 CsF, 10 CsCl, 10 EGTA, and 10 HEPES, 300 mOsm, pH 7.4. Patch-pipettes of 4−6 MΩ were pulled from Sutter BF150-89-10 capillaries. Experiments were performed at 23-25 • C. Both neurons and HEK293 cells were voltage-clamped at −70 mV (holding voltage, V h ). The liquid junction potential was~12 mV between the Na + -containing bathing solution and the Cs + -containing pipette solution. Wherever V h is shown, this value is indicated without a correction for the liquid junction potential. To activate NMDARs, both N-methyl-d-aspartate (NMDA) or l-homocysteine (HCY) were always co-applied with 30 µM glycine as a co-agonist.

Calcium Imaging
For loading with Fluo-8, fluorescent Ca 2+ -sensitive dye cortical neurons were incubated in the basic media containing 2 µM Fluo-8 acetoxymethyl ester (Fluo-8 AM) at room temperature for 60 min. The cells were washed out from dye by 20 min incubation in pure basic solution, and coverslips were transferred to a Leica TCS SP5 MP inverted microscope (Leica Microsystems, GmbH, Wetzlar, Germany). The imaging bath was permanently perfused at a flow rate of 1.2 mL/min with the same basic solution as used for patch-clamp recording. Local media exchange was achieved using the fast local-perfusion system, which allowed the rapid application of various compounds. HCY (100 µM) was applied together with 30 µM glycine. Fluo-8 fluorescence was excited with a 488 nm laser and detected at the 510-560 nm spectral range with~2 s sampling interval (frame 1024 × 1024 pixels, 20× objective). The fluorescence intensities were measured in regions of interest (ROIs) chosen inside of individual neuronal bodies. The intensity of fluorescence in ROIs in the absence of NMDAR agonists was taken as 1.

Data Analysis
Quantitative data are expressed as mean ± SEM. GraphPad Prism software was used for statistical analysis. Student's two-tailed t-test was used to compare groups of measurements. The number of experiments is indicated by 'n'. The level of statistical significance was set to p < 0.05. Data marked by *, **, ***, or **** are significantly different with p < 0.05, p < 0.01, p < 0.001, or p < 0.0001 correspondently. Approximations
To improve a statistical comparison of DTT effects and eliminate HEK293 cell variability concerning NMDAR expressions, in each pair of NMDA and HCY applications, the steady-state amplitudes of NMDAR currents activated by 100 µM NMDA were used as a reference current, presumably corresponding to saturated receptor activation. We, therefore, normalized the steady-state amplitude achieved during the HCY application (I HCY ) by the steady-state amplitude achieved during the NMDA application (I NMDA ). This allowed presenting HCY-elicited currents as a fraction of NMDA-activated current corresponding to saturated receptor activation ( Figure 1C). The ratio (I HCY /I NMDA ) values exhibited less dispersion and strengthen the statistical significance of DTT effects on HCY-elicited currents for both GluN2A (n = 11, p = 0.00008) and GluN2B (n = 14, p = 0.0017) containing NMDARs ( Figure 1C). We further studied HCY-activated currents within a wide range of HCY concentrations from 1 μM to 800 μM before and after 1 mM DTT treatment. Sequential stepwise HCY applications with an increment concentration caused currents with subsequently increased amplitudes for all GluN1/2A (Figure 2A  We further studied HCY-activated currents within a wide range of HCY concentrations from 1 µM to 800 µM before and after 1 mM DTT treatment. Sequential stepwise HCY applications with an increment concentration caused currents with subsequently increased amplitudes for all GluN1/2A (Figure 2A In general, for all three types of NMDARs, the dependence of the amplitude of currents on HCY concentration appeared to be biphasic under control conditions. Previously, we demonstrated that activation of these receptors within the range of HCY concentrations from 1 to 200 μM reaches saturation and could be well fit with the Hill equation [19,25]. Here, in experiments on GluN1/2A receptors using up to 800 μM HCY, we observed biphasic concentration dependence of amplitudes of HCY-activated currents, in which an initial receptor activation with EC50 < 10 μM [19] was followed by the saturation at 50-100 μM and further additional current increase at [HCY]s ≥ 200 μM. As a result, the current concentration dependence was better fitted with the biphasic Hill equation than a monophasic one ( Figure 2A). Notably, the amplitudes of currents did not reach saturation even at [HCY] = 800 μM. Similar biphasic concentration dependence of currents was found for GluN1/2B and GluN1/2C NMDARs unless the initial portion of curves was characterized by EC50s of about 50 μM [19,25] that caused the break in the curves to be less obvious. It should be noted that the second branch of curves was found at similar [HCY]s for all three types of NMDARs, suggesting similar mechanisms of HCY action. Because HCY molecules contain a thiol group, we assume that In general, for all three types of NMDARs, the dependence of the amplitude of currents on HCY concentration appeared to be biphasic under control conditions. Previously, we demonstrated that activation of these receptors within the range of HCY concentrations from 1 to 200 µM reaches saturation and could be well fit with the Hill equation [19,25]. Here, in experiments on GluN1/2A receptors using up to 800 µM HCY, we observed biphasic concentration dependence of amplitudes of HCY-activated currents, in which an initial receptor activation with EC 50 < 10 µM [19] was followed by the saturation at 50-100 µM and further additional current increase at [HCY]s ≥ 200 µM. As a result, the current concentration dependence was better fitted with the biphasic Hill equation than a monophasic one ( Figure 2A). Notably, the amplitudes of currents did not reach saturation even at [HCY] = 800 µM. Similar biphasic concentration dependence of currents was found for GluN1/2B and GluN1/2C NMDARs unless the initial portion of curves was characterized by EC 50 s of about 50 µM [19,25] that caused the break in the curves to be less obvious. It should be noted that the second branch of curves was found at similar [HCY]s for all three types of NMDARs, suggesting similar mechanisms of HCY action. Because HCY molecules contain a thiol group, we assume that possible self-potentiation of HCY-activated currents by HCY interaction with NMDAR redox sites may contribute to this reaction. We did not test [HCY]s higher than 800 µM in this type of experiments because they substantially exceeded physiological values observed in hyperhomocysteinemia.
To evaluate how the redox status of NMDA receptors would affect the HCY action, similar measurements to those described above were performed after treatments of HEK293 cells with 1 mM DTT for 90 s, which presumably should cause reductions of a majority of NMDAR disulfide bonds. In contrast to the DTT effects on NMDA-activated currents, the pretreatment with DTT caused a decrease of currents activated by HCY through GluN1/2A receptors (Figure 2A). In agreement to its action on NMDA-activated currents, DTT increased HCY-activated currents through GluN1/2B ( Figure 2B) and GluN1/2C ( Figure 2C) receptors. The DTT effects depended on [HCY] since any influence on currents activated by HCY below 100 µM was not found. In addition, even at 800 µM HCY, the enhancement of currents through GluN1/2C receptors was still rather moderate ( Figure 2C).
The difference between currents obtained in the same [HCY]s before and after DTT treatment ( Our experiments demonstrated that the DTT effects are most pronounced in GluN1/2B receptors, which, as known, are desensitized when activated by HCY [19]. The DTT pretreatment increases HCY-activated currents through GluN1/2B receptors in a much greater extent than NMDA-activated currents ( Figure 1A), so that the amplitudes of HCY-and NMDA-evoked currents became similar ( Figure 1C). This allows us to suggest that the cleavage of disulfide bonds could abolish GluN1/2B receptor desensitization, and redox regulation could contribute to the HCY effects at a concentration above 100 µM.

Redox Modulation of HCY-Induced Desensitization of Native NMDARs
In order to obtain more evidence in favor of our assumption that the HCY redox action may abolish the GluN1/2B receptor desensitization, experiments were performed on native receptors of cortical neurons. It is well established that cortical neurons express GluN1, GluN2A, and GluN2B subunits, which form diheteromeric and triheteromeric NMDARs [38][39][40][41]. The co-existence of a mixture of different NMDARs in the plasma membrane of cortical neurons predicts a more complex effect of DTT than the effects found in pure receptor populations. Indeed, NMDAR currents activated by HCY in neurons which are expected to be transferred by GluN2A-containing receptors because of the desensitization of GluN2B-containing NMDARs [19] and might be depressed by the DTT pretreatment (Figure 2A), demonstrated a considerable increase of amplitudes in a wide range of [HCY]s ( Figure 3A), whereas the DTT effect on NMDA-activated currents agreed to those found on recombinant receptors ( Figure 2). In addition, as for the recombinant receptors, the concentration activation curves for HCY in neurons were biphasic, and DTT increased currents, suggesting a major contribution of GluN2B-containing NMDARs in their amplitudes ( Figure 3B). Furthermore, in neurons, 100 µM HCY-elicited currents were characterized by >50% desensitization ( Figure 4A). The DTT pretreatment abolished the desensitization of HCY-activated currents ( Figures 3A and 4A) (p = 0.00007, n = 8, Student's t-test), but did not influence the desensitization of NMDA-elicited currents ( Figure 4A,B). These experiments support our conclusion that changes of GluN1/2B receptor redox status either by the DTT pretreatment or during the action of HCY concentrations ≥ 200 µM may prevent their HCY ligand-dependent desensitization.

The Contribution of GluN2A and GluN2B-Containing NMDARs to HCY Redox Effects in Neurons
At this point, it seems appropriate to use GluN2 subtype-specific inhibitors to study the contribution of GluN2A and GluN2B subunits to the redox regulation of HCY effects in neurons. In control neurons, the steady-state currents evoked by HCY were not inhibited by GluN2B antagonist ifenprodil (10 μM) ( Figure 5A). Pretreatment of neurons with 1 mM DTT for 90 s caused a significant increase (about 2-fold) of HCY-elicited steady-state currents (n = 8, p = 0.0002). After DTT treatment, the HCY-activated currents were partially inhibited by ifenprodil (n = 8, p = 0.0022) ( Figure 5B). An

The Contribution of GluN2A and GluN2B-Containing NMDARs to HCY Redox Effects in Neurons
At this point, it seems appropriate to use GluN2 subtype-specific inhibitors to study the contribution of GluN2A and GluN2B subunits to the redox regulation of HCY effects in neurons. In control neurons, the steady-state currents evoked by HCY were not inhibited by GluN2B antagonist ifenprodil (10 μM) ( Figure 5A). Pretreatment of neurons with 1 mM DTT for 90 s caused a significant increase (about 2-fold) of HCY-elicited steady-state currents (n = 8, p = 0.0002). After DTT treatment, the HCY-activated currents were partially inhibited by ifenprodil (n = 8, p = 0.0022) ( Figure 5B). An

The Contribution of GluN2A and GluN2B-Containing NMDARs to HCY Redox Effects in Neurons
At this point, it seems appropriate to use GluN2 subtype-specific inhibitors to study the contribution of GluN2A and GluN2B subunits to the redox regulation of HCY effects in neurons. In control neurons, the steady-state currents evoked by HCY were not inhibited by GluN2B antagonist ifenprodil (10 µM) ( Figure 5A). Pretreatment of neurons with 1 mM DTT for 90 s caused a significant increase (about 2-fold) of HCY-elicited steady-state currents (n = 8, p = 0.0002). After DTT treatment, the HCY-activated currents were partially inhibited by ifenprodil (n = 8, p = 0.0022) ( Figure 5B). An appearance of ifenprodil-sensitive component of HCY-induced currents after the treatment with DTT could be explained by the increased contribution of GluN2B-containing NMDARs to HCY-activated currents.
NMDARs caused a pronounced (of about 50%) inhibition of currents activated by 1000 μM HCY ( Figure 5C,D). Therefore, the fraction of NMDAR currents transferred by GluN2B-containing receptors considerably increases when currents are activated by 1000 μM HCY in comparison with currents activated by 100 μM HCY. This observation supports our conclusion that the cleavage of disulfide bonds both by 1 mM DTT and 1 mM HCY can induce similar gain of GluN2B-dependent HCY-activated currents, which occurs due to abolishing agonist-dependent desensitization of GluN2B-containing NMDARs.
We further focused on the contribution of GluN2A-containing NMDARs to HCY-elicited currents, since these receptors provide a major contribution to HCY-induced neurotoxicity. Zinc (200 nM), used as a specific GluN2A-subunit antagonist [42,43], caused a remarkable (n = 11, p = 0.0001) inhibition of current elicited by 100 μM HCY ( Figure 6A,B). However, after DTT treatment, zinc inhibition of the currents was weakened ( Figure 6B). This observation favors the suggestion that DTT potentiates HCY-induced response of GluN2B-, but not GluN2A-containing NMDARs. To verify that HCY itself can induce DTT-like effects in NMDARs, we compared ifenprodil action on currents activated by either 100 or 1000 µM HCY ( Figure 5C). Whereas ifenprodil did not inhibit currents elicited by 100 µM HCY ( Figure 5A,C), this GluN2B-selective antagonist of NMDARs caused a pronounced (of about 50%) inhibition of currents activated by 1000 µM HCY ( Figure 5C,D). Therefore, the fraction of NMDAR currents transferred by GluN2B-containing receptors considerably increases when currents are activated by 1000 µM HCY in comparison with currents activated by 100 µM HCY. This observation supports our conclusion that the cleavage of disulfide bonds both by 1 mM DTT and 1 mM HCY can induce similar gain of GluN2B-dependent HCY-activated currents, which occurs due to abolishing agonist-dependent desensitization of GluN2B-containing NMDARs.
We further focused on the contribution of GluN2A-containing NMDARs to HCY-elicited currents, since these receptors provide a major contribution to HCY-induced neurotoxicity. Zinc (200 nM), used as a specific GluN2A-subunit antagonist [42,43], caused a remarkable (n = 11, p = 0.0001) inhibition of current elicited by 100 µM HCY ( Figure 6A,B). However, after DTT treatment, zinc inhibition of the currents was weakened ( Figure 6B). This observation favors the suggestion that DTT potentiates HCY-induced response of GluN2B-, but not GluN2A-containing NMDARs.
The cytosolic-free Ca 2+ increase by HCY governs its neurotoxicity. To confirm our electrophysiological data, we further performed optical imaging of intracellular Ca 2+ to evaluate the DTT effects on a population of neurons. In control cortical neurons, the Ca 2+ responses to 100 μM HCY were effectively suppressed by Zn 2+ (n = 4, p = 0.02). (Figure 6C,D), which specifically inhibit GluN2A-containing NMDARs. This suggests that under control conditions, the HCY-induced Ca 2+ entry was predominantly GluN2A-subunit-dependent. DTT pretreatment completely prevented Zn 2+ inhibition of HCY-induced Ca 2+ responses, which speaks in favor of the GluN2B-mediated Ca 2+ entry after the NMDAR redox modification. This observation coincides well with the conclusion that DTT suppresses HCY-induced GluN1/GluN2A and enhances GluN1/GluN2B-mediated transmembrane currents.  The cytosolic-free Ca 2+ increase by HCY governs its neurotoxicity.
To confirm our electrophysiological data, we further performed optical imaging of intracellular Ca 2+ to evaluate the DTT effects on a population of neurons. In control cortical neurons, the Ca 2+ responses to 100 µM HCY were effectively suppressed by Zn 2+ (n = 4, p = 0.02). (Figure 6C,D), which specifically inhibit GluN2A-containing NMDARs. This suggests that under control conditions, the HCY-induced Ca 2+ entry was predominantly GluN2A-subunit-dependent. DTT pretreatment completely prevented Zn 2+ inhibition of HCY-induced Ca 2+ responses, which speaks in favor of the GluN2B-mediated Ca 2+ entry after the NMDAR redox modification. This observation coincides well with the conclusion that DTT suppresses HCY-induced GluN1/GluN2A and enhances GluN1/GluN2B-mediated transmembrane currents.

Discussion
We demonstrated here that the cleavage of disulfide bonds by DTT modulates NMDA-and HCY-elicited NMDAR currents in different ways. DTT enhanced amplitudes of NMDA-activated currents for GluN1/2A, GluN1/2B, and GluN1/2C receptors and did not affect the NMDAR desensitization. In HCY-elicited currents, DTT augmented currents through GluN1/2B and GluN1/2C NMDARs, while GluN1/2A receptor currents were even reduced. Under control conditions, HCY-evoked GluN1/2B currents exhibited pronounced desensitization [19], which almost disappeared after cleaving disulfide bonds with 1 mM DTT or 1 mM HCY. In contrast, HCY-elicited GluN1/2A currents decreased after DTT treatment. Therefore, the redox status of NMDARs may govern the contribution of GluN2A and GluN2B subunit-containing receptors to HCY-caused neurotoxicity, which is an unexpected conclusion. This observation can provide some clues as to why treatment with reducing reagents like hydrogen sulfide (H 2 S) rescues NMDAR expressing neuroblastoma cells [44], neurons [45], and prevents behavioral alterations [46] in hyperhomocysteinemia.

Molecular Basis of Redox Agent Effect on NMDARs
The structural basis of redox action on NMDARs is well established. The observed diverse HCY effects on NMDARs of different subunit compositions may originate from a difference in molecular locations of redox sites in the structure of GluN1/2A and GluN1/2B receptors [29,31]. The most functionally essential pairs of cysteines 744-798 [31] and 780-726 [32] are located in the GluN1 subunits common for all NMDARs. It is known, however, that distinct conformational states of GluN1 in the structure of GluN1/2A and GluN1/2B receptors alter the access of reducing agents to disulfide bonds [47]. For example, enhancement of glutamate-elicited GluN1/2A currents can be achieved by the cleavage of any of the following bonds: GluN1 780-726 [32], 744-798 [31], 79-308 [34], and GluN2A 87-320 [34]. In contradiction, only a reduction of 744-798 [31] augments GluN1/2B currents. As a result, in GluN1/2B receptors, redox agents enlarge the frequency of openings, while in GluN1/2A receptors, these thiol-containing compounds also increase the open time of receptors [47], presumably due to interaction with additional redox sites existing in the GluN2A subunit of NMDARs.
The thiol-containing agents acting from the outside of cells can interact disulfides at the N-terminal domain of NMDARs. DTT represents a traditional, widely used reagent to study the redox regulation of NMDAR functional properties. In all NMDAR subunit compositions, DTT increases currents elicited by glutamate or NMDA [29,31]. Unlike DTT, endogenous reducing agents, including H 2 S, HCY, or glutathione, when causing the cleavage of disulfide bonds, stay bound to thiol groups of NMDARs. Hence, s-persulfidation (s-sulfhydration) with H 2 S results in the conversion of thiol groups to perthiols (persulfides), causing enhanced GluN1/2A and suppressed GluN1/2B charge transfer [33]. Conversely, DTT as a sulfhydryl-reducing agent diminishes H 2 S effects on NMDARs [33].
HCY can reduce disulfide bonds due to its higher acid dissociation constant for the thiol group as of cysteine [26,27]. Thus, HCY cleaves accessible cysteine disulfide bonds and binds to cysteine residues (s-homocysteinylation), altering protein molecular structure and/or function [28]. Elevated [HCY] in animal hyperhomocysteinemia may lead to s-homocysteinylation of most reactive free cysteine residues and intramolecular disulfide bonds of proteins [48].

HCY as NMDAR Agonist and Reducing Agent
We hypothesized that HCY at concentrations above 200 µM acts as the reducing agent for NMDARs. There are no quantitative data on the subject of NMDAR s-homocysteinylation. However, there are some clues concerning a reducing [HCY], since many other proteins reach saturated s-homocysteinylation at [HCY]~100-140 µM [27,49], which is in order of magnitude higher than EC 50 for GluN1/2A receptor activation [19]. In agreement, similar EC 50 values for the redox modulation by HCY (180 µM) of NMDARs were obtained in our experiments. We suppose that for GluN1/2A NMDARs, the biphasic concentration-activation relationship allows a pure distinction between the HCY agonist effect (EC 501 0 µM) and the HCY-positive redox modulation (EC 50 > 300 µM) of NMDARs. For other NMDAR subunit compositions (GluN1/2B and GluN1/2C), the EC 50 s for HCY as an agonist of NMDARs are about 50 µM, and EC 50 s for the HCY redox modulation are~180 µM. The short gap between EC 50 values of two distinct modes of HCY effects makes a biphasicness of the concentration-activation curves less articulated.
The modification of NMDAR redox status by DTT caused remarkable modulation of HCY-activated currents only when [HCY]s were >100-200 µM, and the statistical significance of DTT effects tended to increase with the rise of [HCY]s. It is clear that the efficacy of HCY as a reducing agent expectably augments with the [HCY] increase. It should be noted that the sign (negative or positive) of DTT effects on GluN1/2A and GluN1/2B or 2C receptors was the opposite. Whereas currents through GluN1/2A receptors in the second branch of the concentration-activation curve were decreased by DTT, presumably because of the saturating cleavage of disulfide bonds before the HCY action, the currents through GluN1/2B receptors considerably increased so that the concentration-activation curve approached a monophasicness. Consequently, this pronounced increase of currents through GluN1/2B receptors activated by HCY made them be of similar amplitudes to currents activated by NMDA, suggesting that the DTT prevented the HCY desensitization of these receptors. The DTT potentiation of currents through GluN1/2C receptors was moderate and observed at the largest [HCY] used in these experiments.

HCY Redox Modulation of NMDARs Is GluN2 Subunit-Specific
To understand how redox regulation by HCY may contribute to the functioning of a mixture of NMDARs of different subunit compositions, pharmacological experiments with the usage of GluN2B and GluN2A selective antagonists, which are ifenprodil and Zn 2+ , correspondently, were performed on native NMDARs of cortical neurons. We demonstrated that currents activated by 100 µM HCY at the steady state could be suppressed by GluN2A-selective, but not GluN2B-selective antagonist, that exhibits a lack of contribution of GluN2B-containing receptors to the currents, presumably because of their desensitization. Currents activated by 1000 µM HCY at the steady state, however, were effectively inhibited by the GluN2B-selective antagonist. The remarkable increase of an involvement of GluN2B-containing NMDARs to the ion transfer during HCY-activated responses most likely occurred due to an absence of GluN1/2B receptor desensitization abolished by self-redox regulation by 1000 µM HCY. In fact, the elimination of HCY-induced desensitization of GluN2B-containing NMDARs was achieved by redox modulation with both 1 mM DTT and 1 mM HCY. Presumably, the transition of GluN2B-containing receptors to the desensitized state after their activation by low [HCY]s (of about 100 µM) captures the HCY molecules on the binding sites, since the further recovery conformational changes are prohibited. The cleavage of disulfide bonds either by 1 mM DTT or HCY makes the NMDAR molecule more flexible, which allows the recovery from desensitization.
The structural basis for GluN2 subunit-specific redox regulation of currents activated by HCY is not yet fully understood. Unlike DTT, the reducing agent H 2 S shows GluN2 subunit-specific modulation of NMDARs [33]. Probably, the effects of both s-persulfidation and s-homocysteinylation depend on the difference of redox-reactive sites between GluN1/2A and GluN1/2B receptors [31,32,34]. In addition, Bolton et al. [50] described that in the presence of low (about 200 nM) glycine, HCY potentiates glutamate-activated NMDAR currents interacting as an agonist with the glycine binding sites. This effect disappeared in the presence of 10 µM glycine. Therefore, it seems unlikely that it could contribute to our observations because the saturating concentration of glycine (30 µM) was used in our experiments.
It is now commonly accepted that GluN2A-dependent accumulation of free intracellular Ca 2+ is responsible for HCY-induced neurotoxicity [51,52]. We have found that cleaving NMDAR disulfide bonds forces HCY-induced Ca 2+ accumulation to depend on GluN2B, but not GluN2A subunit-containing NMDARs. The approximate concentration range for HCY redox effects strongly exceeds normal and mild hyperhomocysteinemia levels but could be achieved in severe hyperhomocysteinemia ([HCY] > 200 µM).

Conclusions
The thiol-containing agents like HCY can cleave disulfide bonds at redox-sensitive sites of NMDARs, and the efficacy of HCY as a reducing agent augments with the [HCY] increase. The biphasic concentration-activation relationship of NMDAR currents allows the distinction between the HCY agonist effect and the HCY-positive redox modulation. An estimated concentration for HCY redox effects exceeds mild hyperhomocysteinemia levels but could be achieved in severe hyperhomocysteinemia (>200 µM). We demonstrated that the cleavage of disulfide bonds in the NMDAR molecule enhances HCY-activated transmembrane currents due to reduced desensitization of GluN1/2B receptors. Consequently, cleaving NMDAR disulfide bonds makes HCY-induced Ca 2+ accumulation dependent on GluN2B, rather than GluN2A subunit-containing NMDARs.

Conflicts of Interest:
The authors declare no conflict of interest.