Two extreme Loss-of-Function GRIN2B-mutations are detrimental to tri-heteromeric NMDAR-function, but rescued by pregnanolone-sulfate

Mutations within various N‐methyl-D-aspartate receptor (NMDAR) subunits are tightly associated with severe pediatric neurodevelopmental disorders and encephalopathies (here denoted GRINopathies), for which there are no treatments. NMDARs are tetrameric receptors and can be found at the membrane of neurons in various compositions, namely in di- or tri-heteromeric forms. The GluN2B subunit appears very early in development and, therefore, prenatally this subunit is predominantly found within di-heteromeric receptors, exclusively composed of the GluN1 and GluN2B subunits. Postnatally, however, the GluN2A subunit undergoes rapid increase in expression, giving rise to the appearance of tri-heteromers containing the GluN1, GluN2A and GluN2B-subunits. The latter are emerging as the principal receptor-type postnatally. Despite more than a decade of research of numerous GRINopathies, not much is known regarding the effect of GRIN variants when these are assembled within tri-heteromers. Here, we have systematically examined how two de novo GRIN2B variants (G689C and G689S) affect the function of di- and tri-heteromers. We show that whereas a single mutated subunit readily instigates a dominant negative effect over glutamate affinity of tri-heteromers, it does not dominate other features of the receptor, notably potentiation by pregnanolone-sulfate (PS). This led us to explore PS as a potential treatment for these two severe loss-of-function (LoF) mutations in cultured neurons, in which case we indeed find that the neurosteroid rescues current amplitudes. Together, we present the first report to examine LoF GRIN2B mutations in the context of di- and tri-heteromeric receptors. We also provide the first demonstration of the positive outcome of the use of a GRIN2B-relevant potentiator in the context of tri-heteromers. Our results highlight the importance of examining how different mutations affect features in various receptor subtypes, as these could not have been deduced from observations performed on purely di-heteromers. Together, our study contributes to the ongoing efforts invested towards understanding the pathophysiology of GRINopathies as well as provides insights towards a potential treatment.

expressed GluN2B-wt-C1 with GluN2B-wt-C2 and find that these channels readily express at the membrane and yield specific glutamate-dependent currents (Fig. 1b, c), whereas the sole expression of just one subunit (e.g., GluN2B-wt-C2 with GluN1a) yields very small, albeit bona-fide, glutamatecurrents (denoted leak), but these were mainly noticeable when oocytes are exposed cells to high (saturating) glutamate concentrations (10 mM , Fig. 1b, c, but also see Suppl. 1 for more leak examples). We next proceeded to compare the apparent glutamate affinity (EC 50 ) of purely diheteromeric channels composed of GluN2B-G689C-C1/C2 and GluN2B-G689S-C1/C2-channels, to GluN2B-wt-C1/C2 receptors. We observed that GluN2B-G689C and GluN2B-G689S instigate a drastic reduction in EC 50 , namely ~2000-and >3000-fold reduction in EC 50 , respectively, on par with our previous observations with the non-tagged variants (Fig. 1d,  We then proceeded to examine the effect of single variants within a mixed di-heteromeric channel (i.e., GluN2B-wt-C1+GluN2B*-C2). Expectedly, we noted that mixed channels showed major reductions in EC 50 values, with GluN2B-G689S-containing channels displaying a more severe rightward shift (Fig. 2a, b, Table 1). Nevertheless, both mixed di-heteromeric channels showed ~four-fold higher (i.e., improved) affinities compared to purely di-heteromeric mutated channels ( Fig. 2b, solid vs. dashed plots, and see Table 1). These observations are somewhat reminiscent of our previous report, in which we have also observed slightly improved EC 50 values in oocytes injected with different ratios of wt and mutated subunits' mRNAs (see 9 ).
Thus, our current results demonstrate that a single variant within a mixed di-heteromeric channel is sufficient to engender a strong dominant negative effect over the entire channel, however the variant's affinity does not govern the apparent affinity of the entire complex (i.e., is not the limiting factor). Instead, the channel's EC 50 is adjusted by both subunit types (wt and mutant). Our results thereby elegantly demonstrate that robust channel opening (on a macro-current level) depends on the liganding of all subunits, strongly supporting previous observations [29][30][31] , as well as demonstrates the positive allosteric effects between subunits 32 .
We next examined the effect of the variants over the more common, if not the dominant, channel-subtype in the adult brain, namely tri-heteromeric channels composed of GluN2A and GluN2B-subunits 7 . GluN2A-wt tagged with the C1 tail, co-expressed with C2-tagged GluN2B variants, also exhibited severe LoF; yielding ~300 and ~600-fold reductions in EC 50 by GluN2B-G689C and 2B-G689S variants, respectively, in comparison to GluN2A/2B-wt tri-heteromeric channels (Fig. 3, Suppl   1a-b, and summary in Table 1). Together, our results demonstrate that GluN2B-variants induce severe reductions in glutamate potency when incorporated in either mixed di-or tri-heteromeric channels.

Mixed di-and tri-heteromeric channels containing the GluN2B-variants respond differently to
GluN2B-selective potentiators-To assess whether channel activity could be rescued pharmacologically, we turned to examine spermine, a GluN2B-selective potentiator 33 . Of note, we were reluctant because purely di-heteromeric channels (containing two copies of the mutated subunits) do not respond to the drug 9 . However, the binding site for spermine lies at the dimer interface between the GluN1 and GluN2B subunits (at the N-terminal domains), and it remains debated whether a single intact interface is sufficient for its potentiation 33 . Remarkably, we found that mixed di-heteromeric receptors, composed of GluN2Bwt-C1 assembled with GluN2B-G689C-C2 or GluN2B-G689S-C2, undergo robust potentiation by spermine, highly similar, although lower, to responses of wt receptors to the drug (Fig. 4a, b). Thus, GluN2B-G689C and GluN2B-G689S do not exert a dominant negative effect over spermine potentiation when assorted with a GluN2B-wt subunit. We also find that all GluN2A-wt-containing tri-heteromeric receptors were non-responsive to spermine (actually, slightly inhibited) under physiological pH and temperature (and -60 mV holding potential), regardless the identity of the GluN2B subunit (i.e., wt or mutated) ( Fig. 4a, b, Suppl. 1c). These results show that potentiation by spermine indeed requires only a single intact interface between GluN1-wt and GluN2B-wt subunits ( Table 1), however this does not hold true for tri-h eteromeric receptors. Thus, these results argue against the use of spermine as a possible treatment for GluN2B's LoF mutations, at least postnatally, during which period tri-heteromeric receptors are the more abundant form. In fact, the use of spermine may actually worsen the clinical phenotype by further inhibiting extant tr-heteromers.
We proceeded to explore the effect of the neurosteroid pregnenolone-sulfate (PS), a wellestablished positive allosteric modulator (PAM) of di-heterometric GluN2A-or GluN2B-containing receptors 34,35 . Notably, and to the best of our knowledge, the use of PS on tri-heteromeric channels has yet to be explored. We found that purely wt or mixed di-heteromeric receptors containing GluN2B-subunits exhibit similar potentiation (~200%) by 100 M PS (Fig. 5a, b, Suppl. 1d).
Unintuitively, all tri-heteromeric receptors (whether with a wt or a GluN2B variant) show similar potentiation ( Fig. 5b) and, perhaps the most striking, purely mutated di-heteromeric receptors show the greatest potentiation (G689S-~5-fold, G689C-~4-fold, Fig. 5b). Thus, these clearly demonstrate that the binding domain of PS (within the transmembrane 35 ) is not affected by the mutations in the LBD of the GluN2B subunits. Of note, the different extents of potentiation observed for the dicompared with tri-heteromers is suggested to stem from the diverging potentiation mechanisms of PS onto GluN2A and GluN2B subunits by PS 34 . Together, we show that PS is a powerful potentiator of purely mutated di-heteromers. We extend these findings towards tri-heteromers only to show, for the first time, that PS is a powerful potentiator of all channel types, regardless whether it includes a wt or mutated variant within the receptor complex. Thus, PS may serves to rescue the effects of the mutations by enhancing the currents of the receptors, as frequently suggested for di-heteromers [35][36][37][38] .
Lastly, we also examined the effect of Olanzapine, a derivative of clozapine, which has been previously suggested to act as a potential 'enhancer' of NMDARs [39][40][41] . Aside previous reports, additional motivation behind the examination of olanzapine is the fact that it is an FDA-approved anti-psychotic drug and therefore, if indeed active, would present novel opportunities to quickly obtain approval for treating GRINopathies at the clinic. Despite the latter, we found no evidence for any direct effect of three different physiologically-relevant concentrations of the drug over di-or triheteromeric receptors (Suppl. 2a-e), even though it did show its established effect (inhibition) over hERG channels (Suppl. 2f) 42 .
PS rescues NMDAR-current amplitudes in cultured hippocampal neurons-We next turned to examine the functional outcome of the use of PS in neurons. We transfected cultured hippocampal neurons only with the different mutated GluN2B-subunits, relying on the endogenous subunits to assemble with, and traffic them to the membrane 43 . We first noted that overexpression of the variants caused a significant reduction in the NMDAR-current (Fig. 6a, b-asterisk, and c), as previously shown 9 , and-importantly-application of 100 M PS increased the steady-state NMDAdependent currents in neurons expressing the G689C and G689S variants to similar amplitudes to control neurons (Fig.6b, c). We also observed that potentiation by PS was most prominent in neurons overexpressing GluN2B-G689S (~60%) (Fig.6d). Based on our oocytes recordings (see Fig. 5), this result may have suggested the over-abundance of purely di-heteromeric channels consisting of GluN2B-G689S variants at the membrane of neurons, although we ruled-this out by use of ifenprodil (a selective GluN2B-inhibitor 23,26 ), showing that these neurons enclose the same fraction of GluN2B containing receptors (Fig. 6e).
Collectively, our results demonstrate that the G689C-and G689S-variants reduce the total NMDAR-dependent currents, and instigate a very strong dominant negative effect over glutamate affinity of all forms of di-or tri-heteromeric receptors (Figs. 1-3). However, in the case of the use of spermine, only purely di-heteromeric channels containing the variants are non-responsive to the reagent, whereas mixed di-heteromers remain almost completely unperturbed (Fig. 4). The latter applies for PS as well, as all channel types are responsive to micromolar concentrations of the neurosteroid (Figs. 5, 6).

Discussion
The tendency of NMDAR-subunits to assemble into tri-heteromeric receptors (i.e., two GluN1subunits and two different GluN2 subunits, notably GluN2A and GluN2B) in the postnatal brain is widely gaining acceptance 3,21,44 . However, very little is known regarding the effect of single GRIN variants over the function of tri-heteromers, in particular at the synapse and their sensitivity (or irresponsiveness) to NMDAR-selective drugs (i.e., 15,16,24,45,46 ). This scenario limits our understanding of the heterozygous-nature of most GRINopathies, and is likely behind the very slow progression towards development of treatment to patients (e.g., 12 ).
Here, we make a small, albeit significant, step forward by examining the effect of two very extreme de novo analogous mutations in the context of di-and tri-heteromers, previously found in two pediatric patients (specifically, G689C and G689S 9 ). We first focus on the most detrimental feature of both mutations over receptor function, namely their ultrapotent (~2000-fold) reduction in glutamate affinity. We have previously shown that the mutated residue (G689) is precisely located at the opening of the LBD, which thereby affects the correct coordination and, consequently, binding of glutamate, yielding receptors with extremely high EC 50 values (~mM). Here, we show anew that a single GluN2B-variant, whether assembled with a GluN2B-wt subunit to form mixed di-heteromer or with a GluN2A-wt-subunit (i.e., tri-heteromer), strongly reduces the apparent glutamate affinity of the receptor, but the resulting affinities are less extreme than those obtained from purely diheteromeric receptors containing two GluN2B-variants (Figs. 1-3, and Table 1). This was somewhat surprising, as NMDARs opening requires all four subunits to be liganded (i.e., occupied by a ligand) which implies that the least affine subunit should have dominated the final affinity of the receptor 29 .
This can be explained by the well-established cooperativity observed between the subunits; each strengthening the affinity of the neighbors to their ligands 32 . Despite the latter, the negative effect of the two variants over mixed di-and tri-heteromeric receptors is strong enough so that their effectiveness is presumably below the expected glutamate concentrations that are readily obtained at the synaptic cleft, rendering both receptors and synapses oblivious (i.e., silent) to neurotransmission 9 . We would like to emphasize that this effect is not to be confused with cases of haploinsufficiency, for instance due to mutations that cause a drastic reduction in expression of synaptic receptors such as truncation mutations, even though they may share several functional similarities such as reduced currents at synspase 25,46,47 .
Another interesting, and counterintuitive finding is the effect of spermine, a GluN2Bselective potentiator. We have previously demonstrated that purely mutated di-heteromers are poorly responsive (or even inhibited in the case of the G689S variant) to spermine. However, when combined with a GluN2B-wt subunit, mixed di-heteromers regain their sensitivity to the reagent (Fig. consequently, spermine sensitivity, whereas glutamate affinity is governed by a single subunit (see above, Figs. 1, 2). Even more surprising, to which we and others have yet to find a plausible answer 21 , is the fact that wt tri-heteromers do not strongly respond to spermine despite the presence of an intact GluN2B-GluN1-interface. This likely stems from the dimer-of-dimer arrangement of the receptors, in which the transduction of the effect of spermine by the gating machinery requires coordinated action all dimers, and in tri-heteromers, there is one missing interface, which is not the case with the GluN2B variants. In fact, our findings further demonstrate that all tri-heteromers (whether with wt or variants) undergo slight, albeit significant, inhibition (-20%) by spermine (Fig. 4b). Our observations support the established consensus whereby spermine potentiation is absent in tri-heteromers (see review by Stroebel 7 ), however the genuine inhibition observed here is unique and stands in contrast to previous reports (of which there are only two reports 26,44 ). Closer scrutiny shows that the different effects seen heer may have resulted from very different experimental schemes, such as much higher acidity (pH =6.4), voltage (-30 mV) and lower glycine/spermine concentrations employed in previous reports; features which likely favor potentiation and/or mask inhibitions 48 . Regardless, these, show the challenge in trying to foresee how different variants may affect this highly complex and coordinated gating of NMDARs. Together, our results demonstrate that, despite the positive effect of spermine over mixed di-heteromers (which should be prevalent in prenatal stages), its use should be avoided postnatally as the largest population of receptors at synapses are likely tri-heteromers and these are readily inhibited by it ( Fig. 4b).
We then proceeded to examine the suitability of a potentiator from the neurosteroids family of molecules, explicitly pregnenolone-sulfate (PS). Of note, and to the best of our knowledge, PS has never been assessed with tri-heteromers. We now show that PS effectively potentiates all receptor types, namely purely or mixed di-heteromeric receptors assembled with the various GluN2B-variants and all tri-heteromers tested (Fig. 5). In fact, purely di-heteromeric receptors (bearing two mutated GluN2B-subunits) exhibit significantly improved potentiation by PS compared to other receptor subtypes (Fig. 5b). These results are supported by our recordings of NMDAR-dependent currents in primary cultured neurons overexpressing the different GluN2B variants (Fig. 6), not to mention complement reports showing the ability of PS to enhance short-and long-term potentiation (STP 49,50 and LTP 51-54 , respectively). Thus, and in spite of the observations above with spermine, the two GluN2B variants have no negative effect over the different receptors' responses to PS. Yet, it should be noted that PS may suffer from lack of specificity, as it may jointly perform as a negative allosteric modulator (NAM) of GluN2C and GluN2D-containing di-heteromers 55 (never tested on triheteromers) and of the GABA A -receptor 34 . The latter may promote susceptibility to epileptic seizures. Thus, owing to this the complex array of activities, whether PS could serve as a viable treatment for GRINopathies remains actively explored [32][33][34]54 . An alternative neurosteroid with similar potentiation capabilities towards GluN2B-receptorss is 24(S)-hydroxycholesterol (24-S), the major cholesterol metabolite in the brain 38 , to which a handful of synthetic analogues have been synthesized (e.g., SGE-201, SGE-301) 38 . Alternatively, there may be several means to increase the concentrations of 24(S) in the brain, for instance by promoting the activity of cytochrome P450 46A1 (CYP46A1) which converts cholesterol to 24-hydroxycholesterol, by FDA-approved drugs such as Efavirenz (an anti-retroviral compound) 56 . Despite this realization and potential benefit, we have yet to find a report that combines GRINopathies and Efavirenz, and we have yet to explore this strategy ourselves owing to lack of success in generating GRIN2B-G689C/S transgenic animals (see Suppl.

Text).
In summary, we examined how a single variant affects the function of di-and triheteromers. We show that while a single dysfunctional subunit readily instigates a dominant negative effect over glutamate affinity, it does not dominate the effect over allosteric features of the receptor, notably potentiation by spermine and PS. Our results highlight the importance in examining how different mutations affect features in various receptor subtypes, as these cannot be deduced from observations from purely di-hetereomeric receptors. Lastly, and perhaps the most important, we show the feasibility of using a neurosteriod to rescue glutamatergic neurotransmission. Together, our study contributes to the ongoing efforts invested into the understanding of the pathophysiology of GRINoparthies variants, and provide insights into potential treatments.

Dissociation, culturing, maintenance, and transfection of primary hippocampal neurons:
Cultures of hippocampal primary neurons were established as previously stated 58 . Briefly, extracted rat neonates (P0) hippocampi were dissociated and plated onto 12 mm poly-D-lysine (Sigma-Aldrich, Cat. #P6407)-treated glass coverslips. Cultures were then maintained in an enriched growth media and grown at 37°C and 5% CO 2 . Following five days in-vitro (DIV), growth medium was supplemented with 4 M cytosine-arabinoside (ARA-C) to suppress glia proliferation. At nine DIV, neurons were transfected using the calcium-phosphate method with 0.3 g DNA of eYFP and 2 g of GluN2B-G689C or GluN2B-G689S. Recordings were performed four to seven days past transfection.

Two Electrode Voltage Clamp recordings in Xenopus laevis oocytes
Two electrode voltage clamp (TEVC) recordings were carried out 24-72 hrs after mRNA injections, as previously described 9,27 . We use a commercial amplifier (Warner Instruments, USA) and Digitizer Glutamate dose response experiments were performed with glutamate concentrations ranging between 0.2 M and 10 mM (all containing 100 M glycine). In the case of hERG channel recordings, oocytes were clamped at -60 mV for 1 s, followed by a voltage jump to 20 mV for 4 seconds, and an additional voltage jump to -50 mV for 6 seconds and return to baseline voltage, as previously described 42 .