Control of aversion by glycine-gated GluN1/GluN3A NMDA receptors in the adult medial habenula

See allHide authors and affiliations

Science  11 Oct 2019:
Vol. 366, Issue 6462, pp. 250-254
DOI: 10.1126/science.aax1522

An inhibitor causes neuronal excitation

Glycine is thought to be primarily an inhibitory neurotransmitter. However, it also acts as a coagonist on excitatory N-methyl-D-aspartate (NMDA) receptors. Otsu et al. examined the function of the NMDA receptor subunit combination GluN1/GluN3A in the medial habenula (MHb) of adult mice. This NMDA receptor subunit combination in MHb neurons is activated by glycine released from astrocytes. Activation of GluN1/GluN3A NMDA receptors causes depolarization and increased spiking of MHb neurons. Reducing GluN3A receptor subunit levels in the MHb blocks conditioned place aversion.

Science, this issue p. 250


The unconventional N-methyl-d-aspartate (NMDA) receptor subunits GluN3A and GluN3B can, when associated with the other glycine-binding subunit GluN1, generate excitatory conductances purely activated by glycine. However, functional GluN1/GluN3 receptors have not been identified in native adult tissues. We discovered that GluN1/GluN3A receptors are operational in neurons of the mouse adult medial habenula (MHb), an epithalamic area controlling aversive physiological states. In the absence of glycinergic neuronal specializations in the MHb, glial cells tuned neuronal activity via GluN1/GluN3A receptors. Reducing GluN1/GluN3A receptor levels in the MHb prevented place-aversion conditioning. Our study extends the physiological and behavioral implications of glycine by demonstrating its control of negatively valued emotional associations via excitatory glycinergic NMDA receptors.

Glycine is a major inhibitory neurotransmitter of the central nervous system, which acts via anion-permeable receptors. It is also a well-characterized coagonist of excitatory N-methyl-d-aspartate receptors (NMDARs) via GluN1 subunits (1). In addition, glycine binds the unconventional NMDAR subunits GluN3A and GluN3B, which generate atypical glutamate-activated triheteromeric NMDARs with GluN1 and GluN2 subunits (2), as well as glycine-gated diheteromeric excitatory complexes with GluN1 (3). GluN3B is appreciably expressed only in caudal areas (4). GluN3A expression is broader but assumed to be limited to early development (2). Mainly examined in recombinant systems (5), native GluN1/GluN3A receptors (GluN1/GluN3ARs) have been identified in the juvenile hippocampus (6) but never in adult neurons.

We found strong immunohistochemical expression of GluN3A subunits in the ventral subdivision of the medial habenula (MHb) of adult mice (Fig. 1A) (7), an area that mediates aversive behaviors (812). We investigated GluN3A-immunostained sections with electron microscopy (EM) to identify the ultrastructural localization of the subunit. Frequently in juxtaposition with glial structures, all the 3,3-diaminobenzidine (DAB)–labeled profiles were identified as postsynaptic dendrites and somata (n = 92) (Fig. 1B). Supported by further confocal microscopy results (fig. S1), these data demonstrated that GluN3A subunits were abundantly expressed in adult MHb neurons.

Fig. 1 Functional GluN1/GluN3ARs in the adult MHb.

Both (A) confocal and (B) electron microscopy demonstrated that GluN3A subunits are expressed in the ventral MHb. DAB EM staining [black deposits in (B)] was detected in dendrites (yellow), close to glial extensions (red), rarely in somas (black arrows), but never in axons (blue; white arrowheads highlight presynaptic specializations). Hipp, hippocampus; LHb, lateral habenula; 3rd V, third ventricule. (C) GluN1/GluN2/GluN3ARs are not functional in the MHb. No difference was found in the rectification of NMDAR currents in WT (black) and GluN3AKO (red) mice, elicited by extracellular stimulation (a) or pressure-ejected NMDA (b). (D) Glycine puffs increased firing rates in MHb cells from control and GluN2AKO mice, but not from GluN3AKO animals. (E and F) Similarly to heterologous GluN1/GluN3R currents, glycine induced outwardly rectifying, rapidly rising and inactivating currents in control and GluN2AKO, but not GluN3AKO, neurons. Data are illustrated as averages ± SEM.

We next examined whether GluN3A subunits formed operational GluN1/GluN2/GluN3ARs. These receptors show reduced rectification compared with GluN1/GluN2 NMDARs (2). We thus examined the current-voltage (I-V) curves of synaptic (13) and puff-evoked NMDAR currents in MHb neurons from wild-type (WT) and GluN3A knock-out (GluN3AKO) (14) mice, in which excitatory transmission to the MHb is not modified (fig. S2). No significant differences were found (Fig. 1C, a and b), suggesting the absence of functional receptors.

To investigate whether glycine-activated GluN1/GluN3ARs were operational, glycine (1 or 10 mM) was pressure-ejected while recording spontaneous firing activity (13) in the loose cell-attached configuration (LCA). In all tested WT neurons, glycine puffs (300 to 500 ms) potently increased firing in the presence of d-2-amino-5-phosphonovalerate (d-APV) and strychnine (Fig. 1D). This excitatory effect was not attributable to conventional NMDARs, because it was present in GluN2AKO mice (Fig. 1D), where GluN1/GluN2 NMDARs are nearly undetectable (13). The effect on firing activity was absent in GluN3AKO mice (Fig. 1D), supporting the presence of GluN1/GluN3ARs in the MHb.

Glycine puffs induced rapidly rising and inactivating inward currents in voltage clamp recordings of both WT and GluN2AKO ex vivo MHb neurons performed in the presence of d-APV and strychnine. The currents could not be elicited in GluN3AKO mice (Fig. 1E). As in recombinant systems, the glycine-evoked currents showed slight outward rectification (Fig. 1F).

We investigated the pharmacology of the glycine-induced currents. In recombinant systems, occupation of the higher-affinity GluN3A site activates the receptor, whereas glycine binding to GluN1 entrains rapid desensitization (5, 15). We recently discovered that the GluN1 antagonist [(1S)-1-[[(7-Bromo-1,2,3,4-tetrahydro-2,3-dioxo-5-quinoxalinyl)methyl]amino]ethyl]phosphonic acid (CGP78608) enhanced receptor responses by reducing desensitization (6). We found that CGP78608 (1 μM) potentiated and prolonged the responses in WT but not GluN3AKO mice (Fig. 2A). 6-Cyano-7-nitroquinoxaline-2,3-dione (CNQX), 5,7-Dichlorokinurenic acid (5,7-DCKA), and d-serine inhibit the currents produced by GluN1/GluN3Rs (3, 5, 6, 15). These drugs also potently inhibited the responses to glycine in MHb neurons (Fig. 2B). Direct applications of d-serine increased neuronal excitability (fig. S4) and elicited currents (Fig. 2C) to a lesser extent than glycine (3, 15). With the exception of zinc, which had no effects (16) (fig. S5), the pharmacological profile of the native currents thus mirrored that of recombinant GluN1/GluN3Rs.

Fig. 2 Properties of GluN1/GluN3Rs in the MHb.

(A) GluN1 subunit block with CGP78608 potentiated glycine-elicited currents. (B) Bath applications of CNQX (blue), 5,7-DCKA (green), and d-serine (purple) antagonized control and potentiated GluN1/GluN3AR currents. (C) Pressure-ejected d-serine had only partial agonist effects on GluN1/GluN3ARs. (D) Viral expression in the MHb of a GluN3A-targeting shRNA (AAV-shRNA-GluN3A-GFP and shRNA) led to reduction of GluN3A mRNA levels, and of glycine-induced currents with respect to scrambled RNA–expressing mice (scrAAV). Percentage values and absolute amplitudes are illustrated in the lower graphs. (E) GluN1/GluN3ARs are Ca2+ permeable. Changing extracellular Ca2+ concentrations led to a significant reversal potential shift [(a), highlighted on the right] in heterologous systems expressing GluN3A and point-mutated, glycine-insensitive GluN1 subunits. In the presence of synaptic receptor antagonists and Ca2+ channel blockers, two-photon imaging revealed glycine-elicited Ca2+ transients in ex vivo MHb neurons (b). ΔG/R, ratio between green fluorescence increases and basal red fluorescence. Stars represent statistically significant differences. Data are illustrated as averages ± SEM.

We injected the MHb with a short hairpin ribonucleic acid (shRNA)–expressing adeno-associated virus (AAV) targeting the GluN3A subunit (AAV5-shRNA-GluN3A) (17). The virus efficiently reduced both the messenger RNA levels of GluN3A in the MHb (Fig. 2D) and the amplitude of the glycine-induced currents compared with scrambled RNA (AAV5-scrRNA-GluN3A)–expressing mice (Fig. 2D).

The permeability of GluN1/GluN3ARs to calcium (Ca2+) remains uncertain (3, 18). We examined this question by first analyzing in human embryonic kidney (HEK) cells a GluN1/GluN3AR variant with potentiated glycine responses (GluN1-F484A/GluN3A) (5, 6). Increasing the extracellular Ca2+ concentration led to a shift of the reversal potential of the responses to glycine (Fig. 2E, a), which demonstrated significant Ca2+ permeability, although smaller than for WT GluN1/GluN2A receptors (fig. S6).

Using two-photon microscopy in the presence of synaptic and Ca2+ channel blockers, we could detect glycine-evoked Ca2+ transients originating from GluN1/GluN3ARs in ex vivo MHb cells (Fig. 2E, b). We concluded that native GluN1/GluN3A receptors were also permeable to Ca2+.

We then searched for physiological sources of glycine. We had previously found that the MHb was completely devoid of glycinergic neuronal specializations (fig.S1) (19). Thus, we tested whether glial cells could control extracellular glycine levels in mice expressing the excitatory DREADD (designer receptor exclusively activated by designer drugs) hM3Dq under the control of the promoter of the glial fibrillary protein (GFAP). In these mice, hM3Dq receptor (hM3DqR) activation triggers Ca2+ elevations specifically in glial cells (20). We confirmed that hM3DqR expression was exclusively glial (Fig. 3A and fig. S7).

Fig. 3 Glial activation in GFAP-hM3Dq mice increases neuronal excitability via GluN1/GluN3ARs.

(A) The specific expression of hM3Dq receptors in MHb glia was demonstrated by the colocalization of GFAP and GFP in aldehyde dehydrogenase-1 family member l1-green fluorescent protein (Aldh1l1-GFP) mice, and of hM3DqRs and GFP in GFAP-hM3Dq::Aldh1l1-GFP and GFAP-hM3Dq::S100β-GFP mice. (B) CNO applications increased spontaneous firing rates of MHb neurons. Representative traces preceding and following CNO are shown for specified experiments. The average time courses (C) and quantifications (D) of the CNO effects are depicted for all the experiments. The smaller increase of the firing rate in CGP78608 than in control likely derived from reduced affinity for glycine in the presence of the drug (6). DMSO veh., dimethyl sulfoxide vehicle. Stars represent statistically significant differences. Data are illustrated as averages ± SEM.

To determine whether GluN1/GluN3ARs mediated glia-neuron interactions in the MHb, we recorded ex vivo spontaneous firing activity in GFAP-hM3Dq mice (Fig. 3). Application of the hM3DqR agonist clozapine-N-oxide (CNO; 10 μM), produced rapid firing rate increases.Preincubation with 5,7-DCKA and CNQX strongly reduced the effect of CNO, suggesting a substantial contribution from GluN1/GluN3ARs (Fig. 3). CGP78608 application significantly augmented basal firing rates (fig. S3), indicating that ambient glycine may suffice to bind GluN3A. The potentiation of the firing rate triggered by CNO was instead reduced by CGP78608 preincubations.

Furthermore, the effect of CNO was significantly smaller in GFAP-hM3Dq mice injected with the AAV5-shRNA-GluN3A-GFP than with AAV5-scrRNA-GluN3A-GFP virus (GFP, green fluorescent protein) (Fig. 3). In contrast to GlyT2 (fig. S1), we found expression in MHb glia of the membrane glycine transporter GlyT1 (fig. S8), which can contribute to glial glycine accumulation (21). In the presence of the GlyT1-specific blocker N-[(3R)-3-([1,1'-Biphenyl]-4-yloxy)-3-(4-fluorophenyl)propyl]-N-methylglycine (ALX5407), CNO applications increased firing rates to greater values than in control (Fig. 3, C and D), likely because of greater buildup of extracellular glycine levels. Together, these experiments suggested that glial cells potentiated neuronal activity via activation of GluN1/GluN3ARs.

Finally, we examined whether GluN1/GluN3ARs in the adult MHb were behaviorally relevant in mice injected with either the AAV5-shRNA-GluN3A-GFP or the AAV5-scrRNA-GluN3A-GFP virus (fig. S9). No differences were detected in locomotor and exploratory activity in an open field arena (Fig. 4A). In contrast, the test mice spent significantly less time in the open arms of an elevated place maze, suggesting that reduced GluN1/GluN3AR levels in the MHb were mildly anxiogenic (Fig. 4, A and B). GluN1/GluN3ARs in the MHb did not modulate learning and memory capabilities, because test and control mice scored similarly in a novel object recognition test (Fig. 4C).

Fig. 4 MHb GluN1/GluN3ARs control place-aversion conditioning.

(A) No difference was detected between AAV-shRNA-GluN3A– and AAV-scrRNA-GluN3A–injected mice in the distance traveled in an open field and in the time spent in its center. (B) Knocking down MHb GluN3A expression was mildly anxiogenic as indicated by the shorter time spent by GluN3A-deficient mice in the open arms of an elevated plus maze. Object exploration times and cue- and context-induced freezing were similar in (C) novel object recognition and (D) fear conditioning tests, respectively. (E) After lithium conditioning, AAV-shRNA-GluN3A–injected mice developed no aversion for the malaise-associated compartment in contrast to scrRNA-expressing animals. Data are illustrated as averages ± SEM.

The MHb can contribute to the development of negatively valued emotional states. We thus examined the behavioral outcomes of a fear conditioning protocol (Fig. 4D). We found no difference between AAV5-shRNA– and AAV5-scrRNA–injected mice in the freezing time elicited either by the cued or the contextual stimulus (Fig. 4D).

Finally, we examined lithium-induced conditioned place aversion, a paradigm that relies on intact MHb function (22, 23). In contrast to AAV5-scrRNA animals, lithium-treated AAV5-shRNA mice did not develop aversion for the conditioned compartment (Fig. 4E). Overall, these results indicate that impaired GluN3A signaling in the MHb alters the capability of associating negatively valued external conditions with internal aversive states.

In addition to its ubiquitous role as an inhibitory neurotransmitter, glycine can exert excitatory actions via unconventional GluN1/GluN3A NMDARs. Our study unveils a functional role of this aspect of glycine physiology by demonstrating that full expression of GluN1/GluN3ARs in the MHb is mandatory for modifying the internal emotional landscape in response to specific environmental challenges.

Supplementary Materials

Materials and Methods

Figs. S1 to S9

Table S1

References (2439)

References and Notes

Acknowledgments: We thank M. Galante, R. Lambert, and N. Leresche for critically reading the manuscript; M. McNicholas and S. Kirchherr for assistance during behavioral experiments; and B. Mathieu at the IBENS Imaging Facility. N. Nakanishi and S. Lipton (Scintillon Institute) provided the GluN3AKO mice and M. Mishina (Tokyo University) provided the GluN2AKO line. The anti-GluN3A antibody was a gift from M. Watanabe (Hokkaido University). We thank C. Bellone (Geneva University) for the scr/shRNA-expressing viruses. Funding: This study was supported by fellowships (NKTH-)ANR-09-BLAN-0401 to S.D., M.A.D., and L.A., and ANR-17-CE16-0014 to P.P., S.D., B.L.K., and M.A.D.; by Emergence Sorbonne (S17JRSU003) and CNRS (PICS 7415) to M.A.D.; by the European Research Council (693021) to P.P.; by ERC-FRONTHAL (742595) to L.A.; by grants FK124434 and 2017-1.2.1-NKP-2017-00002 to F.M.; and by the National Institute of Drug Abuse (05010) and the National Institute on Alcohol Abuse and Alcoholism (16658) to B.L.K. C.V.R. was supported by FRM grant FDT20100919977. C.A. was supported by a Paris School of Neuroscience “Chair of Excellence” award and by a NARSAD Y.I. Award. Author contributions: M.A.D. designed and developed the project. Y.O. and M.A.D. performed the electrophysiology with K.P., S.A.G., T.B., and E.S. K.P., C.V.R., and C.A. performed confocal immunohistochemistry. F.M. performed electron microscopy under supervision of L.A. E.D. performed the behavioral tests under the supervision of B.L.K., with M.A.D. and K.P. T.G. performed the HEK experiments under supervision of P.P. S.D. contributed equipment. C.A. provided transgenic mice. M.A.D. wrote the manuscript, which was edited by all authors. Competing interests: The authors declare no competing interests. Data and materials availability: Data values and statistics are included as supplementary materials.

Stay Connected to Science

Navigate This Article