Almost half and one third of the astrocytic profiles show positive immunostaining for CB₁-R and DGLα, respectively, in the superficial spinal dorsal horn^19,26. Extending the scope of the previous studies, first we investigated whether astrocytes that carry CB₁-Rs also express DGLα.

First, we obtained sections of lamina I-II of the rat spinal dorsal horn and carried out triple immunostainings for glial fibrillary acidic protein (GFAP; a marker for astrocytes), CB₁-Rs and DGLα. We observed strong immunolabeling for all three molecules. Investigating 155 GFAP-immunoreactive (GFAP-IR) profiles, we found that 46.3 ± 2.5% of GFAP-IR profiles displayed CB₁-IR puncta and 37.5 ± 2.0% displayed DGLα-IR puncta. Most interestingly, 75.3 ± 2.7% of CB₁-R positive GFAP-IR profiles displayed immunostaining also for DGLα, whereas 91.7 ± 1.3% of DGLα positive GFAP-IR profiles showed immunostaining also for CB₁-R. This indicates a high degree of co-localization between CB₁-R and DGLα. Moreover, CB₁-R-IR and DGLα-IR spots were quite close together (Fig. 1a–d) - the closest DGLα-IR spot was never located more than 12 µm from a CB₁-R-IR spots. DGLα-IR spots showed a unimodal distribution around CB₁-R-IR spots, with a peak at 4 µm (Fig. 1e).

The activation of CB₁-Rs on hippocampal astrocytes leads to PLC-dependent mobilization of Ca²⁺ from cytoplasmic stores^15–17,24, resulting in an increase in [Ca²⁺]_i in a 10–20 µm long segment of astrocytic processes²⁷. Because an increase in [Ca²⁺]_i can lead to the stimulation of DGLα and the production of 2-AG^20,22,28, the close co-localization of CB₁-R and DGLα raises the possibility that spinal astrocytes might produce 2-AG following CB₁-R activation.

As a first step in investigating this possibility, we tested whether activation of CB₁-Rs on spinal astrocytes evokes an increase in [Ca²⁺]_i. To distinguish between neurons and astrocytes, we observed both electrophysiological current activities and the dynamics of calcium events in the recorded cells. Slices cut from the lumbar spinal cord of 10- to 15-day-old mice were loaded with the calcium indicator Oregon Green BAPTA-1-AM. Spontaneous electrical activity was recorded simultaneously with calcium imaging in individual cells in the superficial spinal dorsal horn, using extracellular loose patch recording. Activity in the recorded cells showed two different types of kinetics. Some cells presented rapid, short-duration calcium events, always preceded by a fast current spike in simultaneous loose-patch recordings (Fig. 2a–c). After blocking voltage gated sodium channels with 1 µM tetrodotoxin (TTX), the calcium events as well as the current spikes vanished (Fig. 2a), indicating that the recorded cells were neurons. Other cells presented slow calcium transients that were not coupled with current spikes (Fig. 2d). Since slow intracellular calcium transients in other areas of the central nervous system are characteristic to glial cells²⁹, we predicted that these cells were astrocytes. To test this, we filled some of the cells intracellularly with biocytin. The morphology of the labeled cells was characteristic of astrocytes (Fig. 2e,f), confirming our prediction.

Next, we investigated the effect of the CB₁-R agonist WIN 55,212-2 (WIN) on the [Ca²⁺]_i of cells that showed spontaneous slow calcium transients (assumed to be astrocytes). To block action potential firing and associated neurotransmitter release in the spinal cord slices, thus minimizing possible indirect effects evoked by the activation of neuronal CB₁-Rs, we supplemented the extracellular solution with 1 µM TTX. After recording spontaneous calcium events from individual cells in this condition, we applied 1 µM WIN in the presence of TTX. The application of WIN substantially increased both the frequency and amplitude of the calcium transients (Fig. 2g). To compare the calcium events recorded before and after the application of WIN, we calculated the area under the curve (AUC) for the entire control period and the entire drug application period for 73 cells recorded in slices obtained from three mice. Then we determined the average AUC per minute of recording and found that the average AUC during the control period was 29.0 ± 3.8, which increased significantly to 36.1 ± 4.1 (p = 0.0029) after the application of WIN (Fig. 2i).

To verify that the enhanced calcium events represented by the increase in AUC values were the result of CB₁-R activation, we carried out negative control experiments on slices cut from the lumbar spinal cord of 10–15-day-old CB₁-R knock out mice. Under identical experimental conditions as those used for recordings from wild-type animals, the application of WIN did not change the frequency and amplitude of the calcium transients in the absence of CB₁-Rs (Fig. 2h). The AUC values calculated for 40 cells recorded in slices obtained from two knock out mice were 20.0 ± 2.9 before and 22.7 ± 3.6 after the application of WIN (p = 0.8725; Fig. 2j).

The results clearly demonstrate that the activation of CB₁-Rs leads to an increase in [Ca²⁺]_i in astrocytes within the superficial spinal dorsal horn.

Next, we intended to test whether the CB₁-R-evoked increase in [Ca²⁺]_i has any effect on astroglial 2-AG mobilization. However, measuring 2-AG released by astrocytes in the central nervous system is far from trivial. Therefore, we generated primary cultures of spinal astrocytes from 7–9 day-old rats and mice (Fig. 3a), and conducted the rest of the experiment in a cell culture environment.

To show that cultured astrocytes express CB₁-R and DGLα similarly to spinal astrocytes in vivo, we carried out triple immunostaining for GFAP, CB₁-Rs and DGLα, and evaluated the immunolabeling similarly to the in situ conditions. We investigated 180 and 167 GFAP-IR astrocytes in cell cultures cultivated from rat and wild type mouse spinal cord, respectively.

In the rat cultures, 29.9 ± 21.9% of GFAP-IR cells displayed CB₁-IR puncta and 28.9 ± 1.8% displayed DGLα-IR puncta. The overlap between CB₁–R and DGLα immunostaining was remarkably high; 81.2 ± 2.2% of CB₁-R positive GFAP-IR cells displayed immunostaining also for DGLα, whereas 84.1 ± 1.7% of DGLα positive GFAP-IR cells showed immunostaining also for CB₁-R. CB₁-R-IR and DGLα-IR spots were very close to each other (Fig. 3c–k)- the closest DGLα-IR spot was never located more than 5 µm from a CB₁-R-IR spots. DGLα-IR spots showed a unimodal distribution around CB₁-R-IR spots, with a peak between 1–2 µm (Fig. 3b).

Results obtained from the wild type mouse cultures were practically identical to those collected from rat cultures in terms of co-localization and distribution pattern. In this case, 29.4 ± 1.3% and 29.3 ± 2.0% of GFAP-IR cells displayed CB₁-R-IR and DGLα-IR, respectively; and 86.4 ± 1.2% of CB₁-R positive GFAP-IR cells displayed immunostaining also for DGLα, whereas 86.7 ± 2.1% of DGLα positive GFAP-IR cells showed immunostaining also for CB₁-R. The distances between the CB₁-R-IR puncta and the closest DGLα-IR spots showed here also a unimodal distribution with a peak at 2 µm and a maximum value of 5 µm.

In cultures obtained from spinal cords of CB₁-R knock out mice, GFAP-IR astrocytes showed a strong immunolabeling for DGLα, as expected (Fig. 4c,d), but they were all negative for CB₁-Rs (Fig. 4b,d). In these cultures, we investigated 122 GFAP-IR astrocytes from which 20.4 ± 2.7% displayed immunostained puncta for DGLα.

Thus, fulfilling our expectations, in a substantial proportion of cultured rat and mouse spinal astrocytes, similarly to astrocytes in the superficial spinal dorsal horn, immunostaining for CB₁-R and DGLα showed a high degree of spatial co-localization.

The activation of CB₁-Rs increased [Ca²⁺]_i in astrocytes within the superficial spinal dorsal horn. Therefore, we next verified whether the activation of CB₁-Rs on cultured astrocytes also results in a transient increase in [Ca²⁺]_i.

First, we carried out whole cell calcium measurements on cultured rat astrocytes treated with CB₁-R agonists. Specifically, cells were loaded with the calcium sensitive dye Fura-2-AM, and a CB₁-R agonist (anandamide, 2-AG or WIN) was applied directly to the recorded cells. Fura-2 was alternately excited with a monochromatic light of 340 and 380 nm, and the emitted fluorescent signal was continuously monitored. The resting [Ca²⁺]_i calculated as the average of [Ca²⁺]_i of all recorded cells before CB₁-R agonists treatment, was 72 ± 3 nM. Astrocytes responded to the application of CB₁-R agonists with slow calcium transients, of which the full width at half maximum was 135.6 ± 29.0, 151.8 ± 19.5 and 74.1 ± 6.5 seconds, in response to the application of 2-AG, WIN and anandamide, respectively (Fig. 5a–c). Although all applied agonists evoked responses in some of the cultured astrocytes, different agonists induced changes in [Ca²⁺]_i to very different extents: anandamide was the most efficacious activator of CB₁-R, increasing the [Ca²⁺]_i by 237 ± 38 nM (Fig. 5c,d), whereas 2-AG and WIN raised the [Ca²⁺]i by only 63 ± 10 nM and 94 ± 25 nM, respectively (Fig. 5a,b,d; p = 0.4808, 2-AG vs WIN; p = 0.0354, anadamide vs WIN; and p = 0.0007 anandamide vs 2-AG,).

To verify that the increase in [Ca²⁺]_i in the cultured rat astrocytes was the result of CB₁-R activation, and to compare the efficacy of CB₁-R agonists on rat and mouse astrocytes, we carried out another set of experiments on astrocytes cultured from rat, wild type and CB₁-R knock out mice. We did not test the effect of 2-AG in these experiments, because the effect of 2-AG on [Ca²⁺]_i turned out to be very similar to that induced by WIN in the earlier experiment, and WIN is thermodynamically stable whereas 2-AG is prone to isomerization over time³⁰. Cells were loaded with the calcium indicator Fluo-8-AM and the fluorescence intensity within the cell body of individual astrocytes was then measured with a confocal laser scanning microscope, both at rest and after the application of the CB₁-R agonists WIN and anandamide. The experiments were run for 350–450 seconds and images were taken at one frame per second during the recording. After the effect of the applied CB₁-R agonist completely vanished, ATP was added to the cultures and its effect on [Ca²⁺]_i was also recorded. Only those cells that exhibited Ca²⁺ transients in response to ATP were considered alive and included in the analysis.

In rat astrocyte cultures, 4.3 ± 0.5% and 4.9 ± 1.2% of all living cells responded with calcium transients to WIN and anandamide, respectively (Fig. 5e–g). The amplitudes of the evoked calcium signals were 22.3 ± 5.5% (for WIN) and 29.9 ± 14.48% (for anandamide) of those induced by ATP (Fig. 5f,g,j, Table 1). In cultures obtained from wild type mice, 8.3 ± 0.7% and 18.3 ± 1.2% of living astrocytes responded with a normalized F_n value of 21.61 ± 4.1% and 27.4 ± 4.7% to WIN and anandamide, respectively (Fig. 5h,j, Table 1). The cannabinoid-evoked responses in rat and mouse astrocytes were very similar (p = 0.20 for WIN in rat vs WIN in mouse and p = 0,72 for anandamide in rat vs anadamide in mouse; Fig. 5j, Table 1).

In cultures obtained from CB₁-Rs knock out mice, we observed a remarkable reduction both in the proportion of ATP sensitive cells responding to the applied drugs and in the amplitude of the evoked transients (Fig. 5i, Table 1). Only 2.5 ± 0.2% and 5.3 ± 0.3% of the living cells responded to the application of WIN and anandamide with a normalized F_n value of 8.28 ± 1.01% and 19.28 ± 2.14%, respectively (Fig. 5i,j, Table 1). Comparing these values to those obtained in wild type mice, the amplitudes of calcium transients evoked by WIN were significantly lower in astrocytes cultured from CB₁-R knock out mice (p = 0.0118; wild type vs CB₁-R knock out mice; Fig. 5j, Table 1). We also observed a slight reduction in the case of anandamide application, but it was not significant (p = 0.6654).

Although nearly 30% of cultured astrocytes were positively stained for CB₁-Rs (presented in the previous section), much lower proportions of them responded to the application of CB₁-R agonists WIN and anandamide with Ca²⁺ transients. These seemingly contradictory findings may be explained by the fact that we measured [Ca²⁺]_i within the cell bodies of astrocytes, whereas the evoked calcium transients, like the CB₁-Rs may be primarily confined to astrocytic processes. Reinforcing this notion, Di Castro et al.²⁷ reported intense synaptic activity-driven local Ca²⁺ activity in short segments of processes of mature astrocytes in the dentate gyrus.

To test whether the application of CB₁-R agonists evoke local Ca²⁺ transients in spinal astrocytes similar to that reported by Di Castro et al.²⁷, we carried out calcium imaging on cultured mouse spinal astrocytes with a spinning disc confocal system equipped with a high sensitivity sCMOS camera. After loading the cells with the calcium sensitive dye Fura-8-AM and recording the spontaneous activity of astrocytes, we applied WIN to the bath solution. Similar to the previous observations, WIN application occasionally evoked calcium transients in cell bodies. We recorded calcium transients in 12 (8.9%) of the 134 investigated cell bodies. However, we also observed for the first time in spinal astrocytes that substantial elevations in [Ca²⁺]_i also occurred in short segments of glial process (Fig. 6a–d). These local Ca²⁺ signals appeared in large numbers (26.6 ± 6.4 active segments in a 700 × 350 µm large field of view; Fig. 6b,c).

To check the viability of astrocytes, two minutes after the application of WIN, ATP was added to the cultures and its effect on [Ca²⁺]_i was also recorded (Fig. 6c–e). WIN evoked activity was evaluated only for those microdomains of astrocytic processes that exhibited Ca²⁺ transients in response to the application of ATP. Finally, we evaluated WIN evoked calcium transients in 80 segments of astrocytic processes (Fig. 6e). To compare the calcium events recorded before and after the application of WIN, we calculated the AUC for the control period and for the drug application period, then normalized the values to the ATP responses. We found that the average AUC of the calcium signals during the control period was 5.4 ± 0.6% of the ATP response, which increased to 39.7 ± 1.7% (p = 8.81 × 10^–31) after the application of WIN (Fig. 6g).

In some cases, prior to the WIN application, cells were pretreated with the inverse CB₁-R agonist AM251, which completely abolished the effects of WIN (Fig. 6f,g). Following AM251 pretreatment, WIN application had no effect on the basal activity of the cells; the average AUC value calculated from recordings of 109 segments of astrocytic processes was 3.4 ± 0.3% of the ATP response.

Thus, we verified for the first time, that the activation of CB₁-R evokes local Ca²⁺ transients in short segments of processes of spinal astrocytes.

We next explored whether the elevated [Ca²⁺]_i in astrocytic processes was sufficient to activate DGLα, resulting in the production and release of 2-AG by cultured spinal astrocytes.

Cultured spinal astrocytes isolated from wild-type mice were stimulated by adding WIN into the culture medium. To evoke a CB₁-R independent general increase in [Ca²⁺]_i, other cultures were treated with the calcium ionophore A23187 which elevates [Ca²⁺]_i by increasing the calcium permeability of cell membranes. In addition, two other sets of experiments were run, in which astrocytes were pre-incubated with the inverse CB₁-R agonist AM251 (to block CB₁-R activation) or with the selective chelator of intracellular Ca²⁺ BAPTA-1-AM (to prevent calcium transients evoked by CB₁-R activation) before the administration of WIN. After the stimulations, the cells were scraped and the mixture of the cells and the medium was collected. Lipids were extracted from the cell suspensions and 2-AG, anandamide, and two other anandamide-like N‐acylethanolamides, oleoylethanolamide (OEA) and palmitoylethanolamide (PEA) were pre-purified and quantified by isotope dilution-liquid chromatography-atmospheric pressure chemical ionization-mass spectrometry (LC-APCI-MS).

The LC-APCI-MS data showed that the cultured mouse spinal astrocytes produced 32.1 ± 4.1 pmol/mg 2-AG under basal conditions (Fig. 7a), in accordance with earlier publications²⁰. The application of WIN significantly increased the level of 2-AG to 60.7 ± 7.6 pmol/mg (p = 0.023) (Fig. 7a). The WIN evoked response can be regarded as substantial, because even the calcium ionophore A23187 increased the resting level of 2-AG to only 91.3 ± 8.8 pmol/mg (Fig. 7a). Although A23187 had a greater impact on 2-AG production than WIN, the effects of the two compounds were not significantly different (p = 0.052). On the other hand, we found only 32.9 ± 2.3 pmol/mg (p = 0.96 vs basal) and 36.4 ± 5.3 pmol/mg (p = 0.71 vs basal) 2-AG in samples pretreated with AM251 or BAPTA-1 AM, respectively, before the application of WIN (Fig. 7a). These data demonstrate that the CB₁-R inverse agonist AM251 and the calcium chelator BAPTA-1-AM completely prevented WIN from inducing 2-AG synthesis. These results indicate that WIN acts on CB₁-Rs and subsequently increases [Ca²⁺]_i, in micro-domains of astrocytic processes. The increase in [Ca²⁺]_i, activate DGLα resulting in an increased synthesis of 2-AG.

In addition to 2-AG, the cultured spinal astrocytes also produced 1.2 ± 0.1 pmol/mg anandamide, 1.2 ± 1.3 pmol/mg OEA and 6.0 ± 5.4 pmol/mg PEA under basal conditions (Fig. 7b–d). In contrast to 2-AG, however, the administration of either WIN or A23187 did not alter the levels of these compounds (Fig. 7b–d), indicating that their production was unaffected by activation of CB₁-Rs and subsequent elevation of [Ca²⁺]_I.

Here, we provided evidence that activation of CB₁-Rs on spinal astrocytes, as in the hippocampus^16,17 and in other brain areas^31–37, results in a substantial increase in [Ca²⁺]_i. Importantly, although we applied endocannabinoids anandamide and 2-AG at concentrations that can activate receptors other than CB₁-R (e.g. GPR 55, TRPV1, CB₂), the control experiments verified that our results were not considerably biased by these off-target effects. Indeed, the proportion of astrocytes responding to WIN or anandamide in cultures obtained from CB₁-R knock out animals was less than one-third of those in culture from wild type mice, suggesting that the applied agonists exerted their effect mostly through CB₁-Rs. Some astrocytes isolated from CB₁-R knock out animals, however, did respond to WIN and anandamide. Although these calcium transients may represent spontaneous cellular activity coincided with the application of the drugs, we cannot exclude the possibility that some spinal astrocytes may express low levels of additional cannabinoid-sensitive receptors such as GPR55, TRPV1 or CB₂. The investigation of the roles of these receptors was beyond the scope of our present study, but one has to be aware that the expression of these receptors on astrocytes would substantially diversify the cannabinoid-mediated signaling pathways between neurons and astrocytes.

It is important to note that only a fraction of spinal astrocytes were sensitive to cannabinoids. Although it was previously shown that approximately 50% of astrocytic profiles in the superficial spinal dorsal horn of rats display immunoreactivity for CB₁-Rs¹⁹, here we recovered CB₁-Rs on only approximately 30% of cultured spinal astrocytes. Because we isolated astrocytes for culturing from the whole spinal cord, the difference in the previous and present data may indicate that CB₁-Rs occur more abundant in astrocytes within the superficial spinal dorsal horn than in other areas of the spinal cord. The distribution of immunoreactivity for CB₁-R within the spinal gray matter reinforces this notion¹⁹. Thus, previous and our present results suggest that endocannabinoid mediated neuron-astrocyte interactions may play a more substantial role in spinal pain processing than in other functions of the spinal cord, given the known importance of the superficial spinal cord in pain processing.

In the hippocampus, endocannabinoids released by neurons activate CB₁-Rs on astrocytes, which leads to the mobilization of Ca²⁺ from internal stores^16,17. A specific consequence of increased [Ca²⁺]_i in astrocytes is the release of gliotransmitters, such as glutamate, D-serine and ATP^2,12,38,39. In addition, an increase in [Ca²⁺]_i evoked by ATP or endothelin^20,22,23 can also activate the PLC-DGLα cascade, resulting in the production of 2-AG^20–22,28. However, 2-AG production evoked by CB₁-Rs activation has never been demonstrated. Filling up this gap in our knowledge, here we reported for the first time in the literature that the activation of CB₁-Rs evokes an increase in [Ca²⁺]_i which induces 2-AG mobilization in spinal astrocytes. The astrocytic 2-AG-mediated signaling mechanism may represent an important form of astrocyte-to-neuron communication, because, depending on the spatial distribution and actual activation pattern of excitatory and inhibitory synapses expressing CB₁-Rs, the release of 2-AG from astrocytes can either depress or enhance the net activity of local neural circuits within a given spatial compartment^39,40. Hence, astrocytes as targets and sources of endocannabinoids may substantially contribute to endocannabinoid-dependent plasticity.

The endocannabinoid signaling apparatus is activated by neural activities¹⁶. In the superficial spinal dorsal horn, in particular, neural activity is primarily initiated by glutamatergic nociceptive primary afferents, which activate spinal neurons. In the case of repetitive stimulation, postsynaptic neurons can release 2-AG from perisynaptic membrane compartments^26,41. The released 2-AG binds to presynaptic CB₁-Rs resulting in a transient depression of synaptic activity. A fraction of the 2-AG, however, can diffuse out from the site of release to a distance of 20 µm¹⁷ and activate CB₁-Rs on adjacent neurons and astrocytes^{16,20,22,23,42}. The activation of CB₁-Rs on astrocytes then leads to PLC-dependent Ca²⁺ mobilization in astrocytic processes^{16,24,43–45}. Considering the possibility that the incoming primary afferents may evoke complex, spatially and temporally distributed patterns of synaptic activity in the spinal dorsal horn, several astrocytic micro-compartments may respond with Ca²⁺ transients reflecting the original pattern of synaptic activity. The increased [Ca²⁺]_i in the activated compartments may activate PLC and DGLα, resulting in further release of 2-AG from astrocytes. This may act locally but may also diffuse out and act on neural CB₁-Rs, thereby affecting the functional properties of spinal neurons remote from the termination field of the originally activated nociceptive primary afferents⁴⁶. Thus, because of the involvement of astrocytes, the primary afferent-evoked original activity, initially limited to a relatively small area, may ultimately influence the functional properties of excitatory and inhibitory synapses in a much wider field. This mechanism may play a substantial role in the tonic tuning of neural excitability, the spike timing precision of neurons, and eventually the modulation of complex network functions¹⁵, such as pain processing in neural assemblies of the superficial spinal dorsal horn. Although the various mechanisms of actions underlying the endocannabinoid modulation of spinal nociceptive functions remain to be fully elucidated, CB₁-R-mediated neuron-astrocyte-neuron bidirectional communication should be considered when interpreting the effects of (endo)cannabinoids on spinal pain processing.

Experiments were carried out on adult and new born (7–9 day old) rats (Wistar-Kyoto, 250–300 g, Gödöllő, Hungary), C57BL6 wild type and CB₁-R knock out mice (kind gift of Andreas Zimmer, Bonn, Germany). All animal study protocols were approved by the Animal Care and Protection Committee at the University of Debrecen, and were in accordance with the European Community Council Directives.

To validate experimental findings obtained from wild type animals it was essential to conduct identical parallel experiments on wild type and CB₁-R knock out animals. For this reason, the use of mice was crucial for the present project. On the other hand, however, several earlier studies addressing the problem of spinal endocannabinoid mediated neuron-glia communication were performed in rats. Because we would have liked to compare our present results also to the earlier rat studies, we conducted the most critical experiments also in rats.

For immunocytochemical detection of CB₁-Rs and DGLα, the animals were deeply anesthetized with sodium pentobarbital (50 mg/kg, i.p.) and transcardially perfused with Tyrode’s solution (oxygenated with a mixture of 95% O₂, 5% CO₂), followed by a fixative containing 4% paraformaldehyde dissolved in 0.1 M phosphate buffer (PB, pH 7.4). After transcardial fixation, the L3-L5 segments of the spinal cord were removed, post-fixed in their original fixative for 1–4 hours, and immersed into 10% and 20% sucrose dissolved in 0.1 M PB until they sank. To enhance reagent penetration, the removed spinal cord was freeze-thawed in liquid nitrogen. Transverse sections (50 µm thick) were cut on a vibratome and washed thoroughly in 0.1 M PB.

Primary cell cultures of spinal astrocytes were generated from rats, wild-type mice and CB₁-R knockout mice (all 7–9 days old). After decapitation, the spinal cord was removed and placed into ice-cold dissecting buffer (136 mM NaCl, 5.2 mM KCl, 0.64 mM Na₂HPO₄, 0.22 mM KH₂PO₄, 16.6 mM glucose, 22 mM sucrose, 10 mM HEPES supplemented with 0.06 U/ml penicillin and 0.06 U/ml streptomycin), in which the meninges were carefully removed. The spinal cords were incubated in a solution containing 0.025 g/ml bovine trypsin (catalog no.: T4799, Sigma, St Louis, USA) for 30 minutes at 37 °C, then transferred into a tissue culture medium (minimum essential medium, catalog no.: 21090-022, Life Technologies, New York, USA) supplemented with 10% fetal bovine serum (catalog no.: F2442, Sigma, St. Louis, USA) for 5 minutes at room temperature. Tissue pieces were gently suspended by a Pasteur pipette. The suspension was filtered on a nylon mesh and centrifuged for 10 minutes at 800 rpm. Isolated cells with a density of 6 × 10⁵/ml were placed onto tissue culture dishes coated with 0.3 mg/ml poly-L-lysine or onto coverslips that were put into wells of tissue culture plates. Non-adherent cells were removed on the second day of culture. Adherent cells were cultured for 10–12 days in a CO₂ incubator (CO₂ concentration: 5%, humidity: 95%) at 37 °C. The tissue culture medium was changed every second day. The purity of the cultures was regularly controlled with staining cell nuclei with DAPI and immunocytochemical detection of markers specific for astrocytes (GFAP), microglial cells (CD11b) and neurons (NeuN). Only those cultures were used for the experiments in which the purity of astrocytes, the proportion of GFAP-IR cells, was more than 80% (a generally accepted purity for primary cultures). Some cells (2–8%) showed positive staining for NeuN, and CD11b immunoreactive cells were occasionally also recovered. Most of the non-GFAP-IR cells, however, turned out to be negative for both NeuN and CD11b.

Triple immunostaining protocols were performed to study the GFAP, CD11b and NeuN immunoreactivity of cultured cells as well as the co-localization of GFAP, CB₁-R and DGLα immunoreactivity in astrocytes within the spinal cord and in cell cultures obtained from rats as well as from wild type and CB₁-R knock out mice.

The specificity of the primary antibody against CB₁-R and DGLα has been extensively characterized previously in various immunostaining protocols and Western blots and also in knock out animals^19,26,47,48.

Single and short series of 1-µm-thick optical sections (15 optical sections with an overlap of 0.5 µm) were scanned from the spinal dorsal horn and cell cultures, using an Olympus FV1000 confocal microscope with a 60x oil-immersion lens (NA: 1.4). The confocal settings (laser power, confocal aperture and gain) were identical for all methods, and care was taken to ensure that no pixels corresponding to puncta immunostained for GFAP, CB₁-R and DGL-α were saturated. The scanned images were processed by Adobe Photoshop CS5 software.

The co-localization of CB₁-R and DGLα within the confines of GFAP immunoreactive astrocytic profiles was quantitatively analyzed in triple-stained sections and cell cultures.

For spinal cord sections, a 10 × 10 standard square grid in which the edge-length of the unit square was 4 µm was placed onto the regions of single confocal images corresponding to laminae I-II of the superficial spinal dorsal horn. The following anatomical features were used to ensure proper placement of the grid: (a) the border between the dorsal column and the dorsal horn, which was easily identified on the basis of the intensity of immunostaining. (b) the border between laminae II and III, which was approximated on the basis of previous observations^49,50. It has been demonstrated in ultrastructural studies that there are almost no myelinated axons in lamina II, whereas they are abundant in lamina III. Thus, the border between laminae II and III can be defined quite precisely in ultrastructural studies, and the thickness of laminae I-II can be measured. Therefore, immunoreactivities and co-localizations were investigated in the most superficial 150 µm thick zone of the dorsal horn that has previously been identified as the layer of the gray matter corresponding to laminae I and II in segments L3-L5 of the spinal dorsal horn.

Profiles that showed immunoreactivity for GFAP and was located on the edges of the standard grid were counted in the medial and lateral compartments of laminae I and II. The selected profiles were then examined to determine whether they were also immunoreactive for CB₁-R or DGLα. CB₁-R and DGLα immunolabeled puncta were always smaller than the area immunostained for GFAP. In addition, since the CB₁-R and DGLα antibodies used in the present study were raised against the intracellular domain of CB₁-R or DGL-α, immunolabeled puncta were expected to be located within the confines of the area immunostained for the markers. Thus, to define the co-localization values we counted only those CB₁-R or DGLα immunolabeled puncta that were located within the confines of the areas immunostained for GFAP. The co-localization was analyzed in three animals. The quantitative measurement was carried out in three sections that were randomly selected from each animal. Thus, the calculations for quantitative figures were based on the investigation of nine sections.

For cell cultures, the quantitative measurement was carried out on three independent cultures. We took five images of single optical sections from each triple stained culture in such a way that the images did not show any overlap with each other. Thus, the calculation of quantitative figures was based on the investigation of 15 images (regions of interest, ROI). We identified all GFAP immunoreactive astrocytes within the 15 ROIs. The processes of the selected astrocytes were then examined to determine whether they were also immunoreactive for CB₁-R or DGLα.

The distances between CB₁-R-IR and the closest DGLα-IR puncta were measured on short series of 1 µm thick optical sections both in the spinal cord and cell cultures. The sampling procedure was identical to that described above for the single optical sections. The z-stack images were imported into Imaris software (Imaris version 7.7.2., Bitplane) and GFAP immunostained astrocytic processes were segmented out manually with the Imaris Surface module (http://www.bitplane.com/imaris/imaris;nif-0000-00314). The segmented processes were then exported as surface objects within Imaris to run a distance transformation on the outside as well as inside of the surface objects to select the CB₁-R-IR and DGLα-IR spots associated with the surface of the segmented process. The distance transformation was done with the “distance transformation” XTension of the Imaris XT module (http://bitplane.com/imaris/imarisxt;nif-0000-00314). The CB₁-R-IR and DGLα-IR puncta were segmented with Imaris Spots retraction feature. The Spot objects on the surface of processes were then selected by using a filter on the resulting channel from the distance transformation. Finally, the distances between the CB₁-R-IR puncta and the closest DGLα-IR spots were measured.

Cultured cells were incubated in a solution of 10 μM Fura-2 AM (catalog no.: F-1201, Life Technologies, Budapest, Hungary) for 1 hour at 37 °C. The cultures were transferred into aCSF for 30 minutes at room temperature and placed onto the stage of an inverted fluorescent microscope (Diaphot; Nikon, Tokyo, Japan). The cultured cells within the recording chamber were continuously perfused with oxygenated aCSF while the CB₁-R agonists anandamide (catalog no.: 1339, Tocris Bioscience, Bristol, UK), 2-AG (catalog no.: 62160, Cayman Chemical, Ann Arbor, USA) or WIN (catalog no.: 1038, Tocris Bioscience, Bristol, UK) in 10 μM concentration were directly applied onto the cells through a capillary tube (Perfusion Pencil™; AutoMate Scientific, San Francisco, CA, USA) with the aid of a local perfusion system (Valve Bank™ 8 version 2.0, AutoMate Scientific). The agonist was carefully washed out, then 180 µM ATP (catalog no.: A7699, Sigma, St Louis, USA) was puffed onto the recorded cells to check their viability.

The investigated cells were illuminated with a light beam for which the wavelength was switched between 340 and 380 nm by a dual-wavelength monochromator (Deltascan, Photon Technology International, New Brunswick, NJ, USA), while the emission was monitored at 510 nm using a photomultiplier. [Ca²⁺]_i values (R) were calculated from the ratio of the recorded fluorescence intensities (R = F₃₄₀/F₃₈₀). The resting and the evoked [Ca²⁺]_i values were calculated and plotted against time.

Primary cell cultures of wild-type mouse spinal astrocytes were generated as described above. On days 10–12 in culture, the culture medium was replaced with 500 µL Hank’s Balanced Salt Solution (HBSS). Cells were maintained in HBSS for 30 minutes and then treated with 10 µM WIN or the ionophore A23187 (5 µM) for 2.5 minutes. In two other sets of experiments, cells were treated with WIN following a preincubation of the astrocytes with 5 µM AM251 or 5 µM BAPTA-1-AM. The stimulation was stopped by adding 500 µL ice-cold methanol. Cells were scraped, and an additional 500 µL of methanol was pipetted into the wells. Thus, we collected a sample solution of 1500 µL from each well, containing both the cells and supernatant. For the quantification of 2-AG under basal conditions, cells were maintained in 500 µL HBSS for 32.5 min. The treatment with WIN and ATP was omitted, but the application of methanol was identical to the sample preparation steps mentioned above. All samples were stored at −80 °C for liquid chromatography – mass spectrometry.

Lipids were extracted from cell suspensions and 2-AG, anandamide, palmitoylethanolamide (PEA) and oleoylethanolamide (OEA) pre-purified and quantified using LC-APCI-MS as described previously⁵¹. First, cells were Dounce-homogenized and extracted with chloroform/methanol/Tris-HCl 50 mM pH 7.5 (2:1:1, v/v) containing internal deuterated standards (5 pmol) for anandamide, 2-AG, PEA and OEA quantification by isotope dilution ([²H]₈ AEA, [²H]₅ 2-AG, [²H]₄ PEA, [²H]₄ OEA (Cayman Chemicals, MI, USA). The lipid-containing organic phase was dried down, weighed and pre-purified by open-bed chromatography on silica gel. Fractions were obtained by eluting the column with 99:1, 90:10 and 50:50 (v/v) chloroform/methanol. The 90:10 fraction was used for anandamide, 2-AG, PEA and OEA quantification by LC-APCI-MS, as previously described and using selected ion monitoring at M + 1 values for the four compounds and their deuterated homologues, as described previously⁵².

Before the statistical analysis, outliers were identified (Q1-3IQR or Q3 + 3IQR, where Q is quartile and IQR is interquartile range) in each data set and were excluded from further analysis. Sample sizes were not predetermined; however, post hoc power tests were performed to verify that they are adequate for quantifying the experimental data. Box-and-whisker plots were generated, and all statistical analyses were performed using Origin Pro 8.0 software. Data sets were compared with the Wilcoxon signed rank test (Fig. 2i,j) or the two-tailed nonparametric Mann-Whitney U test (Figs 5d,j and 6g, 7a). Equal variances between data sets were assumed. Differences were considered significant when p < 0.05. In the text, data sets are presented as mean ± SEM.

All relevant data are available from the authors.

The authors declare no competing interests.

All relevant data are available from the authors.