At excitatory synapses between layer 4 (L4) spiny stellate neurons and layer 2/3 (L2/3) pyramidal neurons in rodent primary somatosensory cortex (barrel cortex), cannabinoid receptor signaling is required for the induction of spike-timing dependent long-term depression (tLTD) and the development of receptive fields^1–3.

In the central nervous system, the effects of endogenous and exogenous cannabinoids are mediated by Type 1 cannabinoid receptors (CB1Rs), the most abundant G-protein coupled receptors in the brain⁴. Canonical CB1R signaling suppresses synaptic release via retrograde transmission of endocannabinoids⁵. CB1R signaling is also involved in LTD induction in diverse brain areas including hippocampus, cortex, striatum, and cerebellum⁶. CB1Rs can thus reduce the likelihood of neurotransmitter release in both the short and long term, and these effects are usually mediated by presynaptic CB1Rs.

There are conflicting data regarding whether CB1Rs are expressed presynaptically at L4-L2/3 synapses^7,8. However, astrocytes also express CB1Rs, including in L2/3^9,10. Unlike neuronal CB1Rs, astrocyte CB1Rs tend to be G_q-coupled, such that their activation is linked to increases in intracellular calcium through IP₃R signaling¹¹. In hippocampus⁹ and layer 2/3 of somatosensory cortex^3,12, CB1R activation elicits calcium transients in astrocytes during exogenous agonist application or during patterns of activity that induce tLTD³. In L2/3, these calcium signals are thought to trigger astrocytic vesicular release of glutamate, which in turn is sensed by neuronal NMDARs. A more recent study¹⁰ found that CB1R activation increased extracellular glutamate both in cortical astrocyte cultures and also in vivo in medial prefrontal cortex, albeit on a much longer time scale.

If CB1R activation increases extracellular glutamate, this would have consequences for synapse and circuit function. Excessive extracellular glutamate impairs rapid neuronal signaling by desensitizing receptors¹³, and chronic glutamate exposure is considered the primary driver of neuronal excitotoxicity¹⁴, and has been hypothesized to contribute to a large number of pathologies¹⁵. Crucially, L2/3 pyramidal neurons in barrel cortex display NMDAR-mediated current at membrane potentials generally regarded as non-permissive in the presence of extracellular Mg²⁺ (V_m = -70 to -80 mV)^16,17. Given the sensitivity of L2/3 pyramidal neurons to glutamate, as well as the importance of preserving the ability to discriminate signal from noise in a sensory circuit, cannabinoid receptor-induced glutamate release would have important and heretofore unexplored consequences for synaptic transmission. We therefore sought to investigate extracellular glutamate in L2/3 of barrel cortex and its modulation by cannabinoid receptor activation.

To determine the basal extracellular glutamate concentration in L2/3 of mouse barrel cortex, we made whole-cell recordings from L2/3 pyramidal neurons in acute brain slices. Neurons were voltage-clamped at + 30 mV in physiological ACSF to relieve Mg²⁺ block of NMDAR channels only in the recorded cell. NMDAR currents were isolated by including in the superfusate TTX, NBQX, and picrotoxin. D-serine (10 µM) was present to ensure NMDAR glycine binding sites were saturated. In these conditions, application of the competitive NMDAR antagonist D-AP5 (50 µM; Fig. 1a) blocks a component of the standing current due to the activation of NMDARs by ambient extracellular glutamate. On average, D-AP5 reduced the standing current by 50.5 ± 8.6 pA (n = 13 cells from 10 mice, Fig. 1c).

The size of the NMDAR current depends on both the concentration of glutamate and the number of receptors present. Application of a known concentration of glutamate would provide information about the total number of NMDARs. However, transporters present in the slice prevent exogenously applied glutamate from reaching NMDARs^18,19. To overcome this problem, in each recording we measured the current evoked by 5 µM NMDA (Fig. 1b), a non-transported NMDAR agonist. Taking into account the different potency and efficacy of the two NMDAR agonists^19,20 allowed us to convert the D-AP5-sensitive current into a fraction of the maximal glutamate current. We used a previously derived glutamate dose-response curve¹⁹ to determine that this corresponds to 28.3 ± 3.3 nM glutamate (n = 13, Fig. 1). This is in close agreement with the concentration reported by NMDARs in hippocampus, nucleus accumbens, and medial nucleus of trapezoid body^19,21–23.

We next tested whether the synthetic CB1R agonists ACEA (Fig. 2a-c) or WIN (Fig. 2d,e) altered extracellular glutamate concentration compared to vehicle (ethanol or DMSO, respectively), with experimenter blind to condition. In a previous study³, the effect of CB1R activation on astrocytic calcium transients required over two minutes of agonist presentation, so to maximize the likelihood of detecting an effect of CB1R activation on glutamate, solution containing ACEA, WIN, or vehicle was applied for at least 2.5 min before blocking NMDAR current with D-AP5. ACEA (1 µM) had no effect on D-AP5-sensitive current or NMDA-evoked current (Fig. 2b). Accordingly, the extracellular glutamate concentration was the same in ACEA and vehicle (Fig. 2c). Similar results were obtained using WIN (5 µM; Fig. 2d,e). Because depolarization can trigger endocannabinoid synthesis and release²⁴, the lack of effect of CB1R agonists could reflect occlusion due to endocannabinoid activation of the receptors. In our experiments, postsynaptic calcium entry is limited by both inactivation (due to sustained depolarization of the neuron) and the presence of voltage-dependent calcium channel antagonists. To test whether CB1R-dependent glutamate release contributed to our basal extracellular glutamate estimates, we measured D-AP5 sensitive current in the presence of the CB1R antagonist AM251 (1 µM). AM251 was applied for at least three minutes before cells were clamped at + 30 mV. In these experiments, the extracellular glutamate concentration was 21.2 ± 2.6 nM (n = 8 cells from 3 mice; versus 28.3 ± 3.3 nM, P = 0.15, t-test) which indicates that CB1R activation does not contribute to extracellular glutamate measurements using this technique. Although there was no detectable increase in extracellular glutamate in the continued presence of CB1R agonists, one caveat is that glutamate uptake was intact during these experiments.

Given the proposed extrasynaptic origin of CB1R-mediated glutamate release³, transporters could buffer glutamate before NMDARs detect it. To address this possibility, we used the non-substrate glutamate transporter blocker DL-TBOA (TBOA) to inhibit glutamate uptake²⁵.

TBOA (100 µM) led to a reversible increase in standing current (Fig. 2h), reflecting the activation of NMDARs by accumulation of extracellular glutamate^21,22. To test whether CB1R activation further increases extracellular glutamate, we applied a second solution containing 100 µM TBOA and either WIN (5 µM) or vehicle, with experimenter blind to condition (Fig. 2h). Neither WIN (Fig. 2g) nor vehicle (Fig. 2h) changed the size of the current evoked by TBOA. To compare the two conditions, currents were normalized to the current measured in TBOA alone. The current evoked by TBOA in the presence of WIN was indistinguishable from the current evoked by TBOA and vehicle (Fig. 2i). Although the glutamate concentration was not directly measured in these experiments, the current measured in TBOA is unlikely to represent maximal activation of NMDARs. First, the current was roughly the same magnitude as the current elicited by 5 µM NMDA in other sets of experiments; this dose is well below the EC₅₀ (37.7 µM) for NMDA¹⁹. Second, in a subset of recordings, 1 µM glutamate was applied with TBOA at the end of the experiment, which increased NMDAR-mediated current relative to TBOA alone (WIN group: 461.4 ± 56.6 pA versus 328.7 ± 51.1 pA, n = 9 cells from 5 mice, P = 0.0003, paired t-test; vehicle group: 458.4 ± 58.8 pA versus 318.9 ± 48.2 pA, n = 10 cells from 6 mice, P = 0.009, paired t-test), indicating that NMDARs were not saturated. These experiments indicate that application of cannabinoid receptor agonist also does not lead to measurable increases in extracellular glutamate concentration within seconds, and that glutamate uptake did not mask the effect of CB1R activation.

To test whether this experimental approach is sensitive enough to detect a change in extracellular glutamate under conditions of increased vesicular release, we took advantage of the fact that hyperosmotic extracellular solution promotes vesicle release^26,27. Supplementing ACSF with 40 mM sucrose reversibly increased the frequency of miniature excitatory postsynaptic currents (mEPSCs) measured in L2/3 neurons (Fig. 3a, b). Sucrose increased mEPSC frequency (Fig. 3b, c; 7.7 ± 1.0 Hz versus 4.0 ± 0.7 Hz, n = 9 cells from 5 mice, P = 0.0001, paired t-test) and had no effect on mEPSC amplitude (sucrose: -12.9 ± 0.8 pA, baseline: -12.9 ± 0.8 pA, n = 9, P = 0.78, paired t-test).

To determine whether this modest increase in absolute mEPSC frequency was detectible using the standing current approach, we measured the D-AP5-sensitive current (Fig. 3d) before and after sucrose application (Fig. 3d). Finally, we measured the NMDAR-evoked current to normalize the D-AP5-sensitive currents and estimate the glutamate concentration. Although glutamate uptake was intact, the current blocked by D-AP5 was greater when preceded by sucrose application (Fig. 3e and 85.1 ± 6.1 pA versus 51.6 ± 5.4, n = 11 cells from 6 mice, P = 0.001, paired t-test). This corresponded to an increase in extracellular glutamate of 19 nM (Fig. 3f and 52.5 ± 7.2 nM versus 33.4 ± 3.9 nM; n = 10 cells from 6 mice, P = 0.009). Both the increase in mEPSC frequency and the increase in D-AP5-sensitive current were small in absolute terms, yet resolvable with our technique. These results demonstrate that NMDAR-mediated current is a sensitive measure of extracellular glutamate, and can report sub-micromolar changes attributable to increased vesicular release. We conclude that if CB1R activation leads to astrocytic glutamate release, it must be either be rare, i.e., occurring in only a small subset of astrocytes, or minimal, i.e., involve very little glutamate, or both.

The question remains, then, of whether CB1R-dependent tLTD requires astrocytic glutamate release. In L2/3, CB1R activation reliably increased astrocytic calcium transients; however, only a small subset of astrocytes contain the necessary machinery to package and release glutamate²⁸. Because tLTD is a robust phenomenon, this raises the question of whether the salient signal is something other than astrocyte-derived vesicular glutamate. In several preparations, ATP/adenosine^12,29,30 and d-serine³¹ have been identified as mediators of the downstream effects of astrocytic CB1R activation, yet in other cases, the soluble messenger(s) remain unknown³². Given that astrocyte calcium signals have been linked with an vast array of effects and effectors^32–34, it remains possible that L2/3 astrocytes regulate tLTD induction through a mechanism other than vesicular glutamate release.