The spinal dorsal horn (SDH) represents the primary relay station of pain pathway (Todd, 2010). Accordingly, maladaptive structural and functional plasticities in the superficial laminae of the spinal cord play a substantial role in the development and maintenance of persistent pain (Kuner, 2010; Luo et al., 2014). We have previously shown that DGL-α and CB1, the enzyme producing 2-AG and the receptor mediating its effect in the central nervous system, are expressed by neurons and glial cells in the SDH (Hegyi et al., 2009, 2012). COX-2 has also been demonstrated on spinal neurons, astrocytes and microglia (Ghilardi et al., 2004; Wang et al., 2011). As a first step of our investigation, we asked if peripheral inflammation alters the expression level and pattern of these molecules in SDH of mice.

In agreement with our previous results, DGL-α showed a strong immunostaining in the SDH with a heavily stained band in lamina I and much weaker staining in all deeper laminae (Hegyi et al., 2012). Following a treatment with Complete Freund’s Adjuvant (CFA), DGL-α showed a more dispersed staining, but its immunoreactivity in SDH altogether did not change significantly (p = 0.658; Figures 1A–C; Table 1).

Similar to our previous data, immunostaining for CB1 showed a twin-band in the SDH confined to lamina I and the inner segment of lamina II (Hegyi et al., 2009). On day 4 following CFA treatment, the characteristics of CB1 immunostaining did not change profoundly in SDH, but its intensity decreased by 21% (p = 0.007; Figures 1D–F; Table 1).

The SDH of naïve mice showed a relatively weak and diffuse COX-2 immunoreactivity, which appeared to be moderately heavier in the superficial laminae (Figure 1G). On day 4 of CFA-induced peripheral inflammation, the expression of COX-2 in SDH increased by 28% (p = 0.003; Figures 1H,I; Table 1). Some of the immunostained elements appeared to be cell bodies, but the majority of immunostaing showed a fine punctate pattern. These data suggest that, under inflammatory conditions, quantities of released 2-AG may not be altered profoundly. However, the probability that 2-AG is converted to prostaglandins instead of activating CB1 receptors is significantly increased.

Spinal astrocytes are extensively involved in cannabinoid signaling by releasing and responding to the endocannabinoid 2-AG (Hegyi et al., 2018; Duffy et al., 2021). Microglia has also been shown to express COX-2 (de Oliveira et al., 2008; Gong et al., 2008; Schlachetzki et al., 2010). The expression of COX-2 in astrocytes, however, seems much less straightforward in the literature. Astrocytes were found to be lacking COX-2 in previous publications (Anrather et al., 2011; Avraham et al., 2015), whereas others confirmed the expression of COX-2 by astrocytes (Wu et al., 2011; Kaizaki et al., 2013; Hsieh et al., 2018). To determine the possible role of spinal glia in the oxidative metabolism of 2-AG, we next investigated the expression of COX-2 in SDH astrocytes and microglia in naïve and CFA-treated mice.

Similar to single immunohistochemistry, double immunofluorescence revealed 3.32 ± 0.91 × 10⁴ pixels corresponding to COX-2 immunostaining in the SDH of naïve mice (Figures 2A,B,M; Supplementary Figures 1A,B,M), of which 12.41 ± 2.79% and 15.64 ± 3.02% co-localized with astrocytes (GFAP-positive profiles; Figures 2C,D,N; Supplementary Figures 1C,D,N) and microglia (Iba1-immunostained profiles; Figures 2E,F,N; Supplementary Figures 1E,F,N), respectively. Following a CFA treatment, the number of pixels for COX-2 staining increased by 240%, to 7.98 ± 1.67 × 10⁴ pixels (Figures 2G,H,M; Supplementary Figures 1G,H,M), of which 22.4 ± 3.77% and 28.91 ± 4.12% co-localized with GFAP (Figures 2I,J,N; Supplementary Figures 1I,J,N) and Iba1 (Figures 2K,L,N; Supplementary Figures 1K,L,N) immunostained elements, respectively. Based on these data, CFA treatment induced a 4-fold increase in COX-2 expression in spinal glia (astrocytes and microglia together, from ~0.92 × 10⁴ to ~4.23 × 10⁴ pixels) suggesting that, in inflammatory conditions, these cells may effectively participate in the conversion of 2-AG to prostaglandines.

Astrocyte and microglia profoundly change their expression phenotype and quickly undergo an inflammatory transition in response to an LPS treatment (Cunningham et al., 2019; Wendimu and Hooks, 2022; Lawrence et al., 2023). Therefore, as the subsequent experiments were carried out on primary spinal astrocyte-microglia co-cultures, neuroinflammatory response of glial cells was induced by an LPS stimulus.

To detect global, LPS-induced changes in the level of COX-2 expression in spinal glia, we performed a capillary electrophoresis immunoassay (WES) on control and LPS-treated astrocyte-microglia co-cultures. To demonstrate the time-course of the induction of COX-2 expression, LPS was applied for 30, 60 and 120 min, which increased COX-2 expression from 1.69 ± 0.12 AU to 7.27 ± 0.42 (p < 0.005), 10.3 ± 0.13 (p = 0.005) and 12.49 ± 0.46 AU (p = 0.005), respectively (Figures 3A,B). In the subsequent experiments, an LPS stimulus for 120 min was applied to induce the expression of COX-2.

To verify that both astrocytes and microglia participate in the increased expression of COX-2 seen in WES analysis, we immunostained control and LPS-treated cell cultures for GFAP and Iba1, as well as for COX-2. In control conditions, 0.88 ± 0.12 and 1.12 ± 0.27 COX-2 / μm² immunoreactive puncta were recovered on astrocytes (Figures 4A,B) and microglia (Figures 4C,D), respectively. LPS treatment for 2 h increased the number of COX-2 (immunreactive puncta to 4.88 ± 0.96 / μm², p < 0.001; Figures 4A,B) in astrocytes and (to 6.35 ± 1.04 / μm², p < 0.001; Figures 4C,D) in microglia.

Astrocytes and microglia express DGL-α (Hegyi et al., 2012) and release 2-AG (Carrier et al., 2004; Hegyi et al., 2018). Our data showed that spinal astrocytes and microglia express COX-2 and, based on earlier data, they both can release PGE2 (Zhang et al., 2009; Clasadonte et al., 2011; Xia and Zhu, 2011). Considering that oxidative metabolism of 2-AG is driven at least partly by COX-2 resulting in the synthesis of PGE2 (Alhouayek and Muccioli, 2014), next, using a metabolite ELISA essay, we tested if control and LPS-treated spinal glia convert 2-AG to PGE2.

Untreated astrocyte culture contained 10.71 ± 0.72 pg./mL PGE2 which increased to 14.11 ± 0.33 pg./mL (p < 0.001) following a treatment with LPS. We next administered 200 pmol 2-AG to the media, which in 5 min increased PGE2 concentration to 18.37 ± 0.66 pmol/mL in control (p < 0.001, vs. untreated basal values) and 21.84 ± 0.32 pmol/mL in LPS-treated cultures (p < 0.001, vs. LPS basal values). Pretreating the cultures with the COX-2 inhibitor nimesulide at 100 μM prior to the application of 2-AG decreased PGE2 levels to 9.47 ± 0.24 (p < 0.001, vs. 2-AG-treated cells) and 10.12 ± 0.44 (p < 0.001, vs. 2-AG-treated LPS-stimulated cells; Figure 5). These results confirm that the increased expression of COX-2 in reactive spinal glia allows an increased conversion of 2-AG to prostaglandines.

Astrocytes in SDH express CB1 (Hegyi et al., 2009) which mediate the effects of 2-AG and induces an increase in the intracellular Ca²⁺ concentration (Hegyi et al., 2018). Astrocytes also possess prostaglandin EP3 receptor (Takemiya and Yamagata, 2013) and respond to PGE2 with Ca²⁺ transients (Takemiya et al., 2011). Because the efficacy and potency of 2-AG and PGE2 can be different at inducing Ca²⁺ transients, conversion of 2-AG into PGE2 may increase astroglial calcium signaling following 2-AG mobilization. Thus, we next recorded the changes in intracellular Ca²⁺ concentration of astrocytes in untreated spinal astrocyte-microglia co-cultures following the application of 2-AG and PGE2 from 10 pM to 1 mM and illustrated the AUC of calcium signals as the function of the concentration of ligands to obtain standard concentration-response curves. To check the viability of cells, 180 μM ATP was applied at the end of the measurement which has been shown to induce a robust Ca²⁺ transient (Hegyi et al., 2018).

Astrocytes showed Ca²⁺ signals in response to both ligands, with a notable differences at equimolar concentrations regarding the pattern and AUC of the transients, normalized to the AUC of ATP-induced transients (Figure 6A; Table 2). The lowest concentration of 2-AG which induced Ca²⁺ transients was found to be in the nanomolar range, whereas the maximal response evoked by 2-AG was 89.98 ± 7.42% of that of ATP, with an EC50 value of 0.16 μM (Figure 6B). In agreement with our current calculation, EC50 of 2-AG previously was found to be 0.6 μM on CB1 transfected COS7 cells (Dócs et al., 2017). Increasing concentrations of 2-AG and PGE2 elevated the frequency of induced Ca²⁺ transients from 0 to 1.88 ± 0.23 / minute and from 2.01 ± 0.19 to 3.1 ± 0.49 / minute, respectively (Table 3).

PGE2 evoked Ca²⁺ transients at 10 pM, and its maximal effect induced at 1 uM was 114.67 ± 7.12% of ATP response. EC50 of PGE2 was calculated to be 2.98 nM, 55 times lower than that of 2-AG (Figure 6B). These data indicate that an increased expression of COX-2 in glial cells, through the conversion of 2-AG to PGE2, may indirectly lead to an exaggerated calcium signaling in spinal astrocytes.

Aberrant astroglial Ca²⁺ signals have been linked to certain neuropathological conditions (Dong et al., 2018; Shigetomi et al., 2019; Bancroft and Srinivasan, 2021), and they also play role in pain processing (Cho and Huh, 2020; Xu et al., 2021). Due to the capacity of spinal glia to convert 2-AG to PGE2, exposing these cells to 2-AG may lead to a PGE2-mediated alteration of astroglial Ca²⁺ signaling, an effect that may be exacerbated in neuroinflammatory astrocytes. Thus, we investigated spontaneous and 500 nM 2-AG-induced Ca²⁺ transients of astrocytes in control and LPS-treated astrocyte-microglia co-cultures.

The AUC, but not the frequency, of Ca²⁺ transients in control astrocytes was increased by 60% following the application of 2-AG (Figures 7A,C,I,J; Tables 4, 5). In reactive astrocytes, both AUC and frequency of spontaneous Ca²⁺ activity were elevated by 71 and 40%, respectively, which were further increased after the application of 2-AG by 106 and 84% (Figures 7B,D,I,J; Tables 4, 5).

Reactive spinal astrocytes responded to 2-AG with exaggerated Ca²⁺ signal, which may be related to an increased COX-2 activity and a consequent conversion of 2-AG to PGE2. To test this hypothesis, untreated and LPS-stimulated astrocyte cultures were pretreated with either the CB1 inverse agonist AM251 (5 μM) or the COX-2 inhibitor nimesulide (100 μM) prior to the application of 2-AG.

In resting astrocytes, prevention of CB1 mediated effects by AM251 resulted in a Ca²⁺ signaling similar to basal Ca²⁺ activities, that is, 2-AG this time had no effect on the intracellular Ca²⁺ concentration (Figures 7E,I,J; Table 4, 5). On the other hand, nimesulide pretreatment was unable to prevent 2-AG from inducing Ca²⁺ transients, and the evoked responses appeared to be practically identical to those without nimesulide (Figures 7G,I,J; Tables 4, 5).

Increased Ca²⁺ transients evoked by 2-AG in reactive astrocytes were only partially inhibited by AM251 (Figures 7F,I,J; Tables 4, 5). Nimesulide, however, almost completely abolished 2-AG-induced exaggerated Ca²⁺ transients (Figures 7H–J; Tables 4, 5). Plotting the AUC of Ca²⁺ signals as the function of their frequency revealed a clustering of responses of 2-AG-treated reactive astrocytes with or without AM251 pretreatment. Similarly, Ca²⁺ signals of resting astrocytes in response to 2-AG following AM251 pretreatment formed a cluster with spontaneous Ca²⁺ activities of unstimulated cells (Figure 8).

Experiments were carried out on six, 12–15 weeks old male mice (NMRI, 30–35 g). All animal study protocols were approved by the Animal Care and Protection Committee at the University of Debrecen (license number: 12/2018 DEMAB), and were in accordance with European Community Council Directives. Animals were deeply anesthetized with sodium pentobarbital (50 mg/kg, i.p.) and transcardially perfused with physiological solution, followed by a fixative containing 2% paraformaldehyde (for fluorescent double immunostaining and for peroxidase-based immunohistochemistry) dissolved in 0.1 M phosphate buffer (PB, pH 7.4). After transcardial fixation, the lumbar segments (L4-L5) of the spinal cord were removed, post-fixed in their original fixative for 4 h, and immersed in 20% sucrose dissolved in 0.1 M PB until they sank. Spinal cords were sectioned at 40 μm on a cryostat, and extensively washed in PBS solution.

Peripheral inflammation was induced by the administration of Complete Freund’s Adjuvant (CFA, containing 1 mg of Mycobacterium tuberculosis (H37Ra, ATCC 25177), heat killed and dried, 0.85 mL paraffin oil and 0.15 mL mannide monooleate per mL; Hylden et al., 1989). CFA was dissolved in physiological solution in a ratio of 1:1 and injected into the right hind paw of three mice. Based on previous studies, inflammation and mechanical hypersensitivity develops 4 days after CFA injection (Gajtkó et al., 2020). Therefore, animals were sacrificed on day 4 following the administration of CFA.

For single and double immunostaining, the following primary antibodies were used: mouse-anti-GFAP (1:2000, Synaptic Systems, Cat. No: 173011), guinea pig-anti-Iba1 (1:500, Synaptic Systems, Cat. No: 234013), goat-anti-DGL-α (1:200, Frontier Institute, Cat. No: MSFR101310), guinea pig anti-CB1 (diluted 1:200; Frontier Institute, Cat. No: CB1-GP-Af530-1) or rabbit-anti-COX-2 (1:200, Synaptic Systems, Cat. No: 12282). The specificity of primary antibodies has been extensively tested in earlier studies (Hegyi et al., 2009, 2012; Chopra et al., 2019).

The distribution of immunoreactivity for GFAP, DGL-α, CB1 and COX-2 in the SDH of naïve and CFA-treated adult NMRI mice was studied performing a single immunostaining protocol. Before applying antibodies, free-floating sections were first treated with 50% ethanol for 30 min followed by 10% normal goat or rabbit serum (Vector Labs) for 50 min. Sections were then incubated in mouse-anti-GFAP, guinea pig-anti-Iba1, goat-anti-DGL-α, guinea pig anti-CB1 or rabbit-anti-COX-2 for 48 h at 4°C, followed by a biotinylated goat-anti-rabbit, goat-anti-mouse, goat-anti-guinea pig or rabbit-anti-goat IgG (1:200; Vector Labs, Burlingame, CA, United States) for 4 h at 4°C. The sections were then transferred to an avidin biotinylated horseradish peroxidase complex (1:100, Vector Labs) for 1 h at room temperature. The immunoreaction was visualized with a 3,3′-diaminobenzidine (Sigma, St Louis, MO, United States) chromogen reaction. Antibodies were diluted in 10 mM Tris-phosphate-buffered saline (TPBS, pH 7.4) supplemented with 1% normal goat serum (Vector Labs). Sections were mounted on glass slides and covered with a neutral medium after dehydration.

Primary spinal astrocytes and microglia were cultured from 4 days old NMRI mice. Following a decapitation of animals and the removal of spinal cord, samples were transferred 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). After the removal of meninges, spinal cords were first incubated in a solution containing 0.025 g/mL bovine trypsin (catalog no.: T4799, Sigma, St Louis, United States) for 30 min at 37°C, followed by a Dulbecco’s Modified Eagle’s Medium containing high glucose (catalog no.: D6429, Sigma, St. Louis, United States) and 10% fetal bovine serum (catalog no.: F2442, Sigma, St. Louis, United States) to stop the action of trypsin. Samples were suspended, and the cell suspension was filtered on a nylon mesh. Thereafter, the isolated cells were transferred into 24-well plates coated with 0.3 mg/mL poly-L-lysine. For all experiments, cell density was 6 × 10⁵/ml. Cells were maintained for 7–12 days at 37°C in a CO2 incubator (CO2 concentration: 5%, humidity: 95%), and the medium was changed every second day. Contaminating neurons were identified by immunocytochemical staining against a marker specific for neurons (NeuN), whereas astrocytes and microglia were identified by GFAP and Iba1 immunoreactivity, respectively. Cultures with a purity of at least 80% (that is, the proportion of GFAP and Iba1 positive cells was more than 80%) were used for the experiments.

Astrocytes and microglia respond to LPS treatment with a profound change in their protein expression profile resulting in the formation of reactive astrocytes. Importantly, rodent astrocytes lack receptors and downstream signaling components required for LPS activation (TLR4 and MYD88), indicating that LPS results in the formation of reactive astrocytes through microglial activation (Liddelow et al., 2017). Therefore, 10–12 day old cell cultures were treated with 50 ng/mL LPS for 2 h (30 min, 1 and 2 h in Simple Western experiments) to induce reactive glia. Cells were kept in Dulbecco’s Modified Eagle’s Medium containing high glucose (catalog no.: D6429, Sigma, St. Louis, United States), enriched in 10% fetal bovine serum (catalog no.: F2442, Sigma, St. Louis, United States) in a CO2 incubator for the time of LPS stimulus.

Tissue sections obtained from naïve and CFA-treated adult NMRI mice, as well as control and LPS-stimulated cell cultures were incubated in a mixture of mouse-anti-GFAP or guinea pig-anti-Iba1 and rabbit-anti-COX-2. The sections and cultures were then treated with goat anti-rabbit IgG conjugated with Alexa Fluor 488 (1:1000, catalog no.: A11034, Invitrogen) and goat anti-mouse IgG conjugated with Alexa Fluor 555 (1:1000, catalog no.: A21422, Invitrogen) secondary antibodies. Prior to the antibody treatments, the samples were incubated in 10% normal goat serum (Vector Labs) for 50 min. For the dilution of antibodies, PBS (pH 7.4) containing 1% normal goat serum (Vector Labs) was used. Sections were mounted on glass slides. Cultures and sections were covered with VectaShield-DAPI (Vector Labs).

Image aquisition of single immunostained sections was carried out with an Olympus DP72 camera built on an Olympus BX53 microscope using a 20x dry objective lens (NA: 0.40). Grayscale intensity histogram profiles were taken from 3 randomly selected immunostained spinal dorsal horns of naïve and CFA-treated animals. Pixel intensities in laminae I-II, defined as a 100-μm-thick band of spinal dorsal horn from the border between white matter and gray matter, were measured in Fiji. Illustrations were prepared in Originpro 9 and Adobe Photoshop CS5.

Immunostained sections and cultures were investigated using an Olympus FV3000 confocal microscope with a 10x dry objective (NA: 0.25) or a 60x oil-immersion lens (NA: 1.4). Laser power, confocal aperture and gain were identical for all the samples to be compared. Single 1-μm-thick confocal optical sections or short image stacks with an overlap of 0.5 μm were recorded and further analyzed in Imaris and Fiji. Illustrations were prepared using Adobe Photoshop CS5 software.

The colocalization between immunostainings for COX-2 and GFAP or Iba1 was analyzed in double stained spinal cord sections in Fiji. Quantitative measurement was carried out in three randomly selected spinal dorsal horns of naïve and CFA-treated animals, ipsilateral to the treatment. Laminae I-II of spinal cord were defined as the most superficial 100-μm-thick band of the dorsal grey matter. Colocalization of immunostainings was identified in Fiji based on the overlapping pixels between the corresponding channels.

The colocalization of COX-2 and GFAP or Iba1 was analyzed in cell cultures in Imaris. The analysis of astrocyte cultures was performed on three independent cultures. All GFAP or Iba1 positive cells in five randomly selected fields of views from each culture were investigated for co-localization with COX-2-positive particles.

Samples from unstimulated and LPS-treated cell cultures were collected and lysed in a radioimmunoprecipitation (RIPA) lysis buffer containing Pierce protease inhibitor tablet (catalog no.: A32963, Thermo Fisher) and 1% NP-40. Prior to the collection of samples, cells were washed 2x in a medium lacking fetal bovine serum. The protein content of samples was measured using Pierce BCA protein assay kit (catalog no.: 23225, Thermo Fisher). All samples were denatured in the presence of 1X Fluorescent Master Mix (PS-FL01-8, ProteinSimple) for 5 min at 95°C based on the manufacturer’s instructions. Total protein fractions (1 μg/mL) were assayed using a ProteinSimple WES automated capillary-based electrophoresis instrument. A WES Separation Module protocol (ProteinSimple) was performed using a 12–230 kDa separation module (SM-W004, ProteinSimple) and an anti-rabbit detection module (DM-001, ProteinSimple). The COX-2 protein was detected using rabbit anti-COX2 (1:25, Synaptic Systems) primary antibody. Results were analyzed and the area under the curve (AUC) was calculated in Compass for SW software (v6.1.0).

PGE2 concentration in control and LPS-treated cell cultures was measured using a commercially available Prostaglandin E Metabolite ELISA competitive assay (Cayman Chemical, catalog no.: 514531), which converts prostaglandines to a single, stable derivative. Before performing the assay, samples were stored at −80°C. prior to the experiments, proteins were removed from the samples using an acetone precipitation. Thereafter, acetone was removed and supernatant was dried using a nitrogen evaporator device. Samples were then resuspended in ELISA buffer and assays were performed according to manufacturer’s recommendation. Results were evaluated in Cayman’s ELISA analysis tool.¹ The concentrations of samples were calculated using a linear regression fit.

Before calcium imaging experiments, astrocyte-microglia co-cultures were loaded with 1 μM Fluo-8-AM in the presence of 0.01% pluronic 127 at room temperature for 30 min. Ca²⁺ imaging was performed using a differential spinning disk (DSD2, Andor Technology) connected to an Andor Zyla 5.5 sCMOS camera. The imaging setup was built on an Olympus IX-81 inverse microscope. Using a 20× objective (NA: 0.40), images of 540 μm × 306 μm field of view, which contained around 15 to 40 cells were acquired at 10 frames per second in Andor iQ3 software. Astrocytes were distinguished from microglia based on their morphology, verified by a GFAP immunostaing following Ca2+ imaging experiments. Traces from cells negative for GFAP were discarded. Fluo-8-AM loaded astrocytes were excited at 488 nm and emission was detected at 520 nm. Acquisition parameters (illumination intensity, exposure time, readout time, frame rate) were identical for all measurements. Changes in fluorescence intensities were measured for 10 min over visually identified processes of astrocytes by drawing freehand ROIs around them. Changes in the intracellular calcium concentration were estimated as changes of the fluorescence signal over baseline (ΔF/F0, where F0 was the average initial fluorescence). Changes in fluorescence intensities of an astroglial process was considered a calcium signal (either spontaneous or treatment-induced) if ΔF/F0 was three times the standard deviation of the baseline for at least five consecutive images. Recorded data were analyzed in Microsoft Excel 365 (Microsoft). FFT filtering to reduce noise and calculation of area under the curve (AUC) were performed in Originpro 9 (Originlab, Northampton, MA, United States).

Statistical analysis was performed using Originpro 9 software. Data for Figure 1 were analyzed using the Kolmogorov–Smirnov test. Analyses of data for Figures 2, 4 were performed using the two-tailed non-parametric Mann–Whitney U test. Data for Figures 3, 5, 7 were analyzed using the two-way Anova with a Bonferroni correction. Cluster analysis for Figure 8 was performed using K-means cluster analysis. Differences were considered significant when the p level was <0.05. Data are presented as mean ± standard error of the mean (SEM) Differences were indicated in the figures as follows: * if p < 0.05; ** if p < 0.005; *** if p < 0.001.