GABA/glutamate co-release controls habenula output and is modified by antidepressant treatment

See allHide authors and affiliations

Science  19 Sep 2014:
Vol. 345, Issue 6203, pp. 1494-1498
DOI: 10.1126/science.1250469

A pathway that controls our mood

A brain area called the lateral habenula is involved in negative motivation and may thus play a role in depression. Shabel et al. investigated synaptic transmission in a brain pathway to the lateral habenula that transmits disappointment signals. Surprisingly, they found the simultaneous release of two antagonistic substances, glutamate and GABA, from individual nerve cells. In an animal model of depression, GABA release was reduced in this pathway. A widely used antidepressant drug, however, increased GABA co-release. These results reveal an unusual synaptic mechanism that affects lateral habenula activity. This mechanism may be instrumental for regulating the emotional impact of disappointment.

Science, this issue p. 1494


The lateral habenula (LHb), a key regulator of monoaminergic brain regions, is activated by negatively valenced events. Its hyperactivity is associated with depression. Although enhanced excitatory input to the LHb has been linked to depression, little is known about inhibitory transmission. We discovered that γ-aminobutyric acid (GABA) is co-released with its functional opponent, glutamate, from long-range basal ganglia inputs (which signal negative events) to limit LHb activity in rodents. At this synapse, the balance of GABA/glutamate signaling is shifted toward reduced GABA in a model of depression and increased GABA by antidepressant treatment. GABA and glutamate co-release therefore controls LHb activity, and regulation of this form of transmission may be important for determining the effect of negative life events on mood and behavior.

Excitatory transmission in the brain is generally balanced by feed-forward and feedback inhibitory actions of γ-aminobutyric acid–releasing (GABAergic) interneurons (1). The LHb largely lacks GABAergic interneurons (2), and thus other mechanisms must operate to control hyperactivity. One potential mechanism is for input neurons to release both excitatory (e.g., glutamate) and inhibitory (e.g., GABA) transmitter [(38), but see (7, 9, 10)]; balance could be achieved by regulating the relative amounts of transmitters released (11).

Hyperactivity in the LHb may contribute to depression (1219), possibly by overprocessing (20) negatively valenced events (21). Although enhanced excitatory input to the LHb contributes to depression-like behaviors in rodents (14, 15), little is known about LHb inhibitory transmission. Here, we examine the nature of the inhibitory input to the LHb from the basal ganglia, as well as its regulation in conditions related to depression. This pathway is of particular interest because it is active during negatively valenced events (22), which are thought to play a causal role in depression (23).

To examine the effects of GABAergic input on LHb output, we infected in vivo the basal ganglia output region, the entopeduncular nucleus (EP), of rodents with an adeno-associated virus (AAV) expressing channelrhodopsin-2 (ChR2) (24). Two to three weeks later, brain slices were prepared, and whole-cell responses were obtained from neurons in the LHb (Fig. 1A). Trains of brief light pulses produced a few synaptically driven action potentials, indicating predominant excitation (25); bath application of the GABA-A receptor (-R) antagonist picrotoxin increased the number of action potentials (Fig. 1, B and C), indicating that the EP excitatory drive was balanced by inhibition. In voltage-clamp mode recordings, light-evoked synaptic responses displayed AMPA-R, GABA-A-R, and N-methyl-d-aspartate receptor (NMDA-R) components (Fig. 1, D and E). In other brain slice experiments (Fig. 1F), pairs of light pulses elicited EP-LHb transmission and action potentials; the ratio of GABA-A-R to AMPA-R–mediated EP-LHb synaptic transmission was inversely related to the generation of action potentials.

Fig. 1 Mixed excitatory and inhibitory transmission from EP input controls persistent LHb activity.

(A) Diagram of experimental protocol. Microscopic images depict example injection of AAV-ChR2-YFP into WT rat; left, yellow fluorescent protein (YFP) fluorescence; right, bright-field illumination. (B) Trains of synaptic responses evoked by light pulses (50 Hz, blue dots) before (top) and after (bottom) bath application of picrotoxin recorded in whole-cell current clamp mode. Scale bars, 25 mV, 50 ms. (C) Summary of average responses for recording in (B). Similar results were obtained in three cells. (D) Whole-cell recordings in voltage-clamp mode of light-evoked EP-LHb transmission at different holding potentials (as indicated) and after application of drugs (as indicated). Scale bars, 100 pA, 10 ms. (E) Current-voltage plots for different EP-LHb synaptic conductances normalized to AMPA-R–mediated response at –60 mV (N = 9 cells). (F) Cell-attached (top) and subsequent whole-cell (middle) recordings at +20 mV (GABA, gray) and –50 mV (AMPA, black) holding potentials from two cells (i and ii) during synaptic responses evoked by pairs of light pulses (as indicated). Below, plot of change in firing rate (measured in cell-attached mode) versus GABA-A-R–mediated/AMPA-R–mediated synaptic response (measured in voltage-clamp mode; N = 24 cells); values for cells i and ii indicated by gray symbols. Scale bars, 400 pA, 50 ms. Error bars throughout indicate mean ± SEM.

Because the LHb contains few inhibitory neurons (2), we tested whether the GABA-A response was due to long-range monosynaptic EP to LHb transmission. Light-evoked synaptic responses, measured at –15 mV holding potential (to allow simultaneous measurement of AMPA-R and GABA-A-R–mediated currents), displayed inward followed by outward currents (Fig. 2A). An AMPA-R antagonist (NBQX; 2,3-dihydroxy-6-nitro-7-sulfamoyl-benzo[f]quinoxaline-2,3-dione), blocked the inward current, leaving intact a picrotoxin-sensitive outward current (Fig. 2A and fig. S1), indicating monosynaptic EP to LHb inhibition (see also fig. S2). In some whole-cell recordings of LHb neurons voltage-clamped at –15 mV, we observed isolated spontaneous responses (in the absence of tetrodotoxin; i.e., action potentials not blocked) that were inward immediately followed by outward currents, indicating co-release of glutamate and GABA from individual axons onto the recorded cell (Fig. 2B and fig. S3). Co-release from some individual vesicles could also be detected. In four of six experiments monitoring spontaneous miniature excitatory postsynaptic currents (in tetrodotoxin), biphasic events occurred more often than could be accounted for by independent inward and outward responses (Fig. 2C and fig. S4) (P < 0.05 for four experiments) (26).

Fig. 2 Co-release of GABA and glutamate from EP inputs to LHb: Electrophysiological, optogenetic, and mouse genetic evidence.

(A) Whole-cell recording at –15 mV holding potential of light-evoked EP-LHb transmission after bath application of indicated drugs. Scale bars, 60 pA, 10 ms. (B) Spontaneous responses from whole-cell recordings at –15 mV holding potential from three different LHb neurons (i, ii, and iii) in adult WT mice. Lower trace (iii′) is the same as above (iii) at higher temporal resolution. Scale bars, 50 pA, 200 ms (upper), 10 ms (lower). Biphasic events accounted for 2/3 of all large (>15 pA) outward events in these cells (79, 56, and 67%; large, biphasic events were found in 3 of 12 recordings). (C) Spontaneous miniature biphasic events (*) from whole-cell recordings at –40 mV with 0.5 mM chloride internal solution. Scale bars, 10 pA, 20 ms. (D) Whole-cell recording from Vglut2-cre mouse injected with AAV-DIO-ChR2-YFP at –15 mV holding potential of light-evoked EP-LHb transmission with indicated drugs in bath. Both inward and outward currents found in 17 of 17 cells. Scale bars, 100 pA, 10 ms. (E) Plot of effect of light-evoked EP-LHb transmission from Vglut2+ axons of adult mice delivered at 50 Hz (left) or 2 Hz (right) on normalized LHb neuronal firing rate before and after bath application of picrotoxin (*P < 0.05). (F) Same as (D) for GAD67-cre mice. Total of 11 of 13 cells, inward/outward; 2 of 13 cells, outward only. Scale bars, 40 pA, 10 ms. (G) Cell-attached recording (average of 10 trials) from adult GAD67-cre mouse LHb neuron; light delivery indicated. Scale bars, 5 mV, 20 ms.

To investigate specifically whether EP neurons projecting to the LHb co-release glutamate and GABA, we first took a genetic approach. In Vglut2-Cre mice, only glutamatergic neurons expressing the vesicular glutamate transporter Vglut2 express Cre recombinase (27). We injected their EP with an AAV that expresses ChR2 only in cells expressing Cre (AAV-DIO-ChR2) (28) (figs. S5 and S6) and prepared brain slices 2 to 3 weeks later. As in wild-type (WT) animals, light-evoked synaptic responses, measured at –15 mV holding potential, displayed a NBQX-sensitive inward current followed by a NBQX-insensitive, picrotoxin-sensitive outward current (Fig. 2D), indicating that EP-LHb glutamatergic neurons release both glutamate and GABA. In cell-attached recordings, picrotoxin increased spikes evoked by 50 Hz, but not 2 Hz, stimulation of Vglut2-expressing EP axons (Fig. 2E and figs. S7 and S8) (NBQX blocked evoked spikes) (26), indicating that co-release of GABA is important for limiting LHb output during high-frequency EP input; the greater effect at higher frequency is likely due to the slower GABA-A-R kinetics, permitting greater summation of GABA-A-R currents at 50 Hz.

In complementary experiments, we used GAD-67-Cre mice, in which only GABAergic neurons expressing glutamic acid decarboxylase type 67 express Cre recombinase (2931). We injected their EP with AAV-DIO-ChR2 (figs. S5 and S6) and prepared brain slices 2 to 3 weeks later. As in WT and Vglut2-Cre mice, light-evoked synaptic responses, measured at –15 mV holding potential, displayed a NBQX-sensitive inward current followed by a NBQX-insensitive, picrotoxin-sensitive outward current (Fig. 2F), indicating that EP-LHb GABAergic neurons release both glutamate and GABA. In cell-attached recordings, EP-LHb transmission evoked spikes, indicating that EP GABAergic neurons produce excitatory responses in LHb (Fig. 2G).

Co-release from EP to LHb was further supported by the following observations: (i) the trial-to-trial variability in evoked EP-LHb transmission recorded at –15 mV displayed significantly greater covariation of early inward and late outward currents than would be expected from independent release of glutamate and GABA (fig. S9); (ii) evoked EP-LHb GABA-A-R- and AMPA-R–mediated transmission displayed similar presynaptic modulation by bath application of serotonin (25) (fig. S10); (iii) evoked EP-LHb transmission recorded from different LHb neurons displayed similar GABA-A-R- and AMPA-R–mediated paired-pulse response ratios (fig. S11); and (iv) GABA-A-R- and AMPA-R–mediated transmission displayed similar response latencies and jitter (fig. S12).

We next examined immunohistochemically the synapses made by EP neurons to the LHb. In adult mice, the EP was injected with an AAV expressing green fluorescent protein (AAV-GFP) (fig. S6). Subsequently, 400-nm thin sections of the LHb were prepared, immunolabeled for GFP, glutamic acid decarboxylase (GAD), and Vglut2, and imaged with confocal microscopy. GFP-expressing structures displayed GAD and Vglut2 (Fig. 3A and fig. S13), indicating GABA and glutamate in individual EP-LHb terminals. A control region displayed little GAD and Vglut2 colabeling (Fig. 3B). Quantification of labeling (26) (fig. S14) in structures originating from the EP indicated that ~50% of Vglut2-positive structures showed GAD and ~ 80% that were GAD-positive had Vglut2; outside the LHb, Vglut2 showed little overlap with GAD (Fig. 3, C and D, and fig. S15). Thus, in EP-LHb structures, ~50% had both GAD and Vglut2, ~10% had only GAD, and ~40% had only Vglut2. Consistent with co-release, glutamate receptors and GABA receptors displayed overlap in LHb (fig. S16).

Fig. 3 Co-localization of GABA-producing enzyme (GAD) and vesicular glutamate transporter (Vglut2) at EP neuron terminals in LHb.

(A) Confocal fluorescent images of LHb sections from mouse injected with AAV-GFP into EP labeled for indicated proteins. Thick arrows indicate some of the sites of Vglut2, GAD, and GFP colocalization. Thin arrow indicates Vglut2 and GAD colocalization (with no GFP). Asterisk indicates GFP-expressing structure showing only Vglut2 expression. Scale bar, 3 μm. (B) Confocal fluorescent images of sections from LHb (top) and outside LHb (bottom) labeled for indicated proteins. Arrows indicate structures displaying co-labeling for GAD and Vglut2. Scale bar, 3 μm. (C) Distribution of GAD fluorescence in pixels with Vglut2 intensity above Vglut2 threshold [dashed line in (D); (26)], for images in LHb (for pixels displaying GFP, black line) and outside LHb (gray line). (D) Distribution of Vglut2 fluorescence in pixels with GAD intensity above GAD threshold (dashed line in C), as indicated. [(C) (D)] EP-LHb, N = 21 images, 3 mice; outside LHb, N = 5 images, 3 mice. (E) Electron micrograph showing a GFP-containing region in the LHb. Gold particles in the mitochondria were not analyzed. Scale bars, 200 nm. See figs. S17 and S18 for larger images of same region. (F) (Left) 18-nm gold labels Vglut2 (1); 12-nm gold labels GABA (2). (Middle) 18-nm gold labels Vglut2 (1); 6-nm gold labels GFP (3). (Right) 12-nm gold labels Vglut2 (1); 6-nm gold labels GABA (2). Scale bars, 30 nm. (G) Density of gold label pairs for distance intervals on x axis for a single experiment. Black line, Vglut2-GABA distance; gray line, GABA-GFP; dotted gray line, Vglut2-GFP. N = 72 Vglut2 (18 nm) labels, 486 GABA (12 nm) labels, 416 GFP (6 nm) labels. Arrow indicates peak for Vglut2-GABA counts at 20 to 30 nm. (H) Increased enrichment (26) of Vglut2-GABA (black lines; three experiments) but not GABA-GFP (gray lines; two experiments) or Vglut2-GFP (dotted gray; two experiments) labels at 15- to 30-nm separation. N = 1624, 2556, and 1558 total Vglut2, GABA, and GFP labels, respectively. *Values > 4 SD from the mean.

To examine anatomically whether GABA and glutamate can be released from individual vesicles at EP-LHb synapses, we used immunogold electron microscopy (EM). Two to three weeks after injecting adult mice with AAV-GFP in the EP, LHb sections were processed for EM and immunolabeled for GFP, Vglut2, and GABA with secondary antibodies linked to gold of three different sizes (26). Structures originating from the EP (Fig. 3E and figs. S17 and S18) showed gold particles of three clearly different sizes (Fig. 3F and fig. S17). The frequency of distances between Vglut2 and GABA gold labels displayed a peak at ~20 to 30 nm (Fig. 3G), consistent with colocalization of Vglut2 and GABA at individual vesicles (32). The distribution of GFP-Vglut2 and GFP-GABA distances was flat, indicating random distribution of particles. The number of gold particles less than 30 nm apart, compared with the number of gold particles between 30 and 100 nm apart, was significantly higher for Vglut2-GABA (34 of 248) compared with Vglut2-GFP (4 of 92; P < 0.03, χ2-test) and GABA-GFP (64 of 999; P < 0.001, χ2-test). Increased enrichment of Vglut2-GABA gold pairs at less than 30 nm was obtained from three separate experiments using different antibodies and size gold (Fig. 3, F and H) (26). Vglut2 and the GABA vesicular transporter (VGAT), which are responsible for glutamatergic and GABAergic transmission, respectively (fig. S19), also displayed enrichment at less than 30 nm (fig. S20).

Activity of the LHb is elevated in animal models of human depression (12, 14, 33). Chronic serotonin-based antidepressant [e.g., citalopram (34)] treatment reduces LHb activity (35) and normalizes behaviors modeling human depression (36). We thus examined EP-LHb transmission as above in mice treated with citolapram or saline for 2 weeks. The ratio of GABA-A-R- to AMPA-R–mediated light-evoked synaptic responses was significantly higher in the citalopram- compared to saline-treated group (Fig. 4, A and C, and fig. S21). This ratio was also reduced in cLH rats, a rat model of human depression (37), compared with control animals (Fig. 4, B and D, and fig. S21). No differences in paired-pulse ratio, a measure of presynaptic release probability, were observed (figs. S21 and S22). These changes were matched immunohistochemically: In animals treated with citalopram, the ratio of GAD to Vglut2 was larger in EP-LHb boutons (Fig. 4E and fig. S21) due to increased GAD labeling (Fig. 4, G and I, and fig. S21) whereas in cLH animals, this ratio was reduced (Fig. 4F and fig. S21) because of decreased GAD labeling (Fig. 4, H and K, and fig. S21). Little change was seen in Vglut2 labeling (Fig. 4, G, H, J, and L, and fig. S21). Consistent with these results, chronic antidepressant treatment normalized the GAD to Vglut2 ratio in cLH animals (fig. S23).

Fig. 4 Altered co-release of GABA in conditions related to depression.

(A) Sample whole-cell recordings of GABA-A-R- (black) and AMPA-R–mediated (gray) light-evoked EP-LHb responses from animals treated with saline or citalopram (10 mg per kg of weight, intraperitoneally). Scale bars, 200 pA, 20 ms. (B) Sample recordings, as in (A), in WT rats and cLH rats. Scale bars, 200 pA, 20 ms. (C) Bar graph of GABA-A-R–mediated/AMPA-R–mediated response at EP-LHb synapses for indicated groups normalized to average of saline ratios (N = 19 cells from four saline-treated mice and 24 cells from four citalopram-treated mice). (D) Bar graph of responses, as in (C), for WT and cLH rats normalized to average of WT ratios (N = 41 cells from nine WT rats and 23 cells from nine cLH rats). (E) Ratio of average background subtracted pixel intensities for GAD and Vglut2 in terminals originating from EP (26) for indicated groups (N = 24 images from three saline-treated mice, 21 images from three citalopram-treated mice). (F) Ratios, as in (E), for WT and cLH rats (N = 32 images from seven WT rats, 21 images from three cLH rats). (G) Example confocal images from a saline and a citalopram-treated mouse. Scale bar, 3 μm. (H) Example confocal images from a WT rat and a cLH rat. Scale bar, 3 μm. (I) GAD fluorescence in pixels displaying Vglut2 and GFP labels in tissue from animals treated with citalopram or saline (percentage of Vglut2+, GFP+ pixels showing GAD fluorescence: saline, 54 ± 5%, 21 images; citalopram, 79 ± 2%, 24 images; P < 10−4). (J) Vglut2 fluorescence in pixels displaying GAD and GFP labels in tissue from animals treated with citalopram or saline (percentage of GAD+, GFP+ pixels showing Vglut2 fluorescence: saline, 80 ± 3%, 21 images; citalopram, 79 ± 3%, 24 images; P = 0.8). (K) GAD fluorescence, as in (I), for images from WT and cLH rats (WT, 50 ± 3%, 32 images; cLH, 19 ± 3%, 21 images; P < 10−7). (L) Vglut2 fluorescence, as in (J), for images from WT and cLH rats (WT, 61 ± 3%, 32 images; cLH, 61 ± 3%, 21 images; P = 0.9). *P < 0.002.

The LHb plays important roles in reward computations (21, 22), is activated during negatively valenced events (21) via its basal ganglia inputs (22), and its hyperexcitation may contribute to human depression (16, 38). We found that co-release of GABA and glutamate occurs in the LHb and that the opponent actions of these neurotransmitters control the level of LHb activity by basal ganglia inputs. Our studies identify an unusual form of inhibition controlling the activity of this key brain region.

There is physiological evidence for co-release of GABA and glutamate (3) in the auditory system (5) and CA3 region of the hippocampus (4, 68), where it may contribute to refinement of neuronal connectivity during development (6, 39) and be induced by seizures in the adult hippocampus (11); GABA and glutamate co-release has not been universally accepted (10). Here, we show using a combination of electrophysiological, optogenetic, mouse genetic, and fluorescent and EM immunohistochemical techniques that in the adult rodent brain a substantial fraction of EP inputs to the LHb simultaneously release both glutamate and GABA and that these neurotransmitters have opposite effects on post-synaptic spike output. Co-release of glutamate and GABA can permit very local control, at the level of individual presynaptic terminals, in the balance of excitation and inhibition, a key mechanism controlling neuronal excitability (1).

The LHb receives long-range excitatory and inhibitory inputs from several sources (4042). It remains to be determined whether other inputs to the LHb co-release glutamate and GABA and the extent to which co-release of glutamate and GABA is used to regulate excitability in the rest of the nervous system (3). Co-expression of Vglut2 and GAD mRNA has been found in EP neurons that project to the motor thalamus, suggesting that co-release of glutamate and GABA may also occur in this pathway (43).

Given the proposed role of increased neuronal activity in the LHb in mood disorders (16), we examined the ratio of GABA to glutamate released at the EP-LHb synapse in conditions related to depression. In animals chronically treated with citalopram, a drug used to treat depression, evidence indicated enhanced GABA at EP-LHb synapses. In line with these data, EP-LHb synapses in cLH rats, which model aspects of human depression, showed evidence of reduced GABA, although additional changes (e.g., in receptor number) cannot be ruled out. Our results identify co-release of GABA and glutamate as a regulator of LHb output and potential determinant of the impact of negatively valenced events on mood and behavior.

Supplementary Materials

Materials and Methods

Figs. S1 to S23

References (44, 45)

References and Notes

  1. Materials and methods are available as supplementary material on Science Online.
  2. Acknowledgments: We thank H. Monyer and O. Kiehn for transgenic mice, R. Edwards for Vglut2 antibodies and DNA, K. Deisseroth for ChR2 DNA, J. Lin and R. Tsien for oChiEf DNA, T. Meerloo for electron microscopy assistance, and V. Joseph for histology assistance. We thank N. Spitzer and J. Issaacson and the Malinow laboratory for helpful suggestions. This work was supported by an NIH grant to R.M. awarded to the University of California, San Diego (UCSD) Neuroscience Core (NS047101). C.D.P was supported by a postdoctoral award by the Instituts de Recherche en Santé du Canada. All primary electrophysiological and immunohistochemical data are archived in the Center for Neural Circuits and Behavior at UCSD.
View Abstract

Navigate This Article