Abstract

   Background: Alterations of astrocyte function play a crucial role in
   neuroinflammatory diseases due to either the loss of their
   neuroprotective role or the gain of their toxic inflammatory
   properties. Accumulating evidence highlights that cannabinoids and
   cannabinoid receptor agonists, such as WIN55,212-2 (WIN), reduce
   inflammation in cellular and animal models. Thus, the endocannabinoid
   system has become an attractive target to attenuate chronic
   inflammation in neurodegenerative diseases. However, the mechanism of
   action of WIN in astrocytes remains poorly understood.

   Objective: We studied the immunosuppressive property of WIN by
   examining gene expression patterns that were modulated by WIN in
   reactive astrocytes.

   Materials and Methods: Transcriptomic analysis by RNA-seq was carried
   out using primary human astrocyte cultures stimulated by the
   proinflammatory cytokine interleukin 1 beta (IL1β) in the presence or
   absence of WIN. Real-time quantitative polymerase chain reaction
   analysis was conducted on selected transcripts to characterize the
   dose-response effects of WIN, and to test the effect of selective
   antagonists of cannabinoid receptor 1 (CB1) and peroxisome
   proliferator-activated receptors (PPAR).

   Results: Transcriptomic analysis showed that the IL1β-induced
   inflammatory response is robustly inhibited by WIN pretreatment. WIN
   treatment alone also induced substantial gene expression changes.
   Pathway analysis revealed that the anti-inflammatory properties of WIN
   were linked to the regulation of kinase pathways and gene targets of
   neuroprotective transcription factors, including PPAR and SMAD (mothers
   against decapentaplegic homolog). The inhibitory effect of WIN was
   dose-dependent, but it was not affected by selective antagonists of CB1
   or PPAR.

   Conclusions: This study suggests that targeting the endocannabinoid
   system may be a promising strategy to disrupt inflammatory pathways in
   reactive astrocytes. The anti-inflammatory activity of WIN is
   independent of CB1, suggesting that alternative receptors mediate the
   effects of WIN. These results provide mechanistic insights into the
   anti-inflammatory activity of WIN and highlight that astrocytes are a
   potential therapeutic target to ameliorate neuroinflammation in the
   brain.

   Keywords: immunosuppression, inflammation, neurobiology, synthetic
   cannabinoids

Introduction

   Neuroinflammation is a key component of neurodegenerative diseases such
   as Alzheimer's disease (AD), Parkinson's disease, and HIV-associated
   neurocognitive disorders (HAND).^[31]1,[32]2 Neuroinflammation is
   facilitated by astrogliosis, a process wherein astrocytes react to
   injuries in the central nervous system with an increase in
   proliferation and pronounced morphological changes.^[33]3 Reactive
   astrocytes also exhibit altered functional properties that affect
   blood–brain barrier permeability, extracellular levels of
   neurotransmitters such as glutamate, and availability of energy
   substrates to neurons.^[34]4–12 A variety of factors can elicit the
   activation of astrocytes, including those released by reactive
   microglia (e.g., cytokines), injured neurons (e.g., glutamate and
   reactive oxygen species), and pathogenic proteins from exogenous
   sources (e.g., viruses and bacteria).^[35]13,[36]14 Astrocyte
   reactivity triggers several intracellular signaling cascades that
   activate transcription factors (TFs) such as NF-κB,
   CCAAT-enhancer-binding proteins (CEBP), activator protein (AP) 1, and
   others.^[37]15–17 Upon activation, these TFs regulate the expression of
   genes involved in inflammation, mitochondrial function, and cell
   adhesion, all of which are altered in neurodegenerative
   diseases.^[38]2,[39]18 Reactive astrocytes express inflammatory genes,
   including cytokines such as interleukin 6 (IL6) and interferon, and
   complement proteins such as complement component 3 (C3), which are all
   implicated in the pathogenesis of neurodegenerative diseases.^[40]19–23
   C3 expression is implicated in inflammatory responses in AD and
   HAND.^[41]24–28 Astrocytes also express TFs that suppress inflammatory
   transcriptional responses such as peroxisome proliferator-activated
   receptors (PPAR)α and γ, which are implicated in neuroprotective
   mechanisms,^[42]29–32 or members of the SMAD family of TFs, which are
   known mediators of the transforming growth factor-β anti-inflammatory
   function.^[43]33 Due to these dynamic functions of astrocytes, they are
   increasingly recognized as potential therapeutic targets to mitigate
   chronic neuroinflammation in neurodegenerative diseases.^[44]34

   Multiple lines of evidence indicate that cannabinoids and cannabinoid
   receptor agonists exert anti-inflammatory properties,^[45]35–37 in
   part, by activating PPARα or γ.^[46]37–41 Specifically, WIN-55212-2
   (WIN) is an aminoalkylindole that acts as potent agonist of cannabinoid
   receptors 1 and 2 (CB1 and CB2) at concentrations in the nanomolar
   range and with 20-fold higher affinity for CB2.^[47]42 WIN has been
   shown to attenuate inflammatory responses in cellular or animal models
   of neuroinflammation.^[48]36,[49]37,[50]43–47 However, the effects of
   WIN on the regulation of global gene expression changes in reactive
   astrocytes have not been studied. The identification of genes and
   pathways modulated by WIN in reactive astrocytes may reveal a promising
   therapeutic strategy for the prevention and treatment of
   neurodegenerative diseases.

   In this study, we profiled the transcriptome of human primary astrocyte
   cultures after stimulation with recombinant IL1β in the presence or
   absence of WIN. Bioinformatic analysis showed that the IL1β-induced
   inflammatory response is robustly inhibited by WIN pretreatment.
   Pathway analysis suggested that the anti-inflammatory properties of WIN
   were linked to the regulation of kinase pathways and gene targets of
   neuroprotective TFs, including PPAR and SMAD. Using astrocytes from a
   different donor, we showed that the inhibitory effects of WIN were
   independent of CB1 or PPARα/γ receptors. Overall, these data support
   the therapeutic potential of targeting astrocytes to prevent or reverse
   inflammatory gene expression and reveal a potential mechanism, by which
   cannabinoids ameliorate inflammatory responses in astrocytes.

Materials and Methods

Primary human astrocyte cultures

   The cell model for astrocytes was approved by the University of
   California San Diego Human Research Protections Program and the
   National Institutes of Health as part of a grant actively funded by the
   National Institute for Mental Health. Astrocytes were isolated from
   fetal human brain tissue from elective terminated pregnancy between 12
   and 16 weeks of gestation, acquired from Advanced Bioscience Resources.
   Donors gave written informed consent for research-use of the cells and
   tissue. Tissue was fragmented and mechanically dissociated using a
   scalpel and washed three times with HBSS media (cat. no. 14175-095;
   Gibco) with 1 mM Glutamax (cat. no. 35050-061; Gibco), 20 μg/mL
   Gentamicin (cat. no. 15710-064; Gibco), and 5 mM HEPES (cat. no.
   15630-080; Gibco). The tissue was homogenized with the addition of
   15 mL of 0.25% trypsin EDTA (cat. no. 25200-056; Gibco) for 5 min in a
   37°C incubator. After 5 min, 1 mL of a trypsin inhibitor (cat. no.
   10109; Roche) and 24 mL of DMEM (cat. no. 11960-044; Gibco) with human
   serum (cat. no. 35-060-cl; Corning) was added. The mixture was then
   centrifuged for 5 min at 4°C to pellet the cells. The cells were
   resuspended in 5 mL of DMEM and strained with a 70 μM strainer (cat.
   no. 352350; Falcon). The cell suspension was underlaid with 7 mL of a
   solution of filtered 8% BSA in PBS and cells were centrifuged at 1×10^4
   rpm at 4°C for 10 min. The cells were resuspended in DMEM with human
   serum, 1 mM GlutaMAX, and 20 μg/mL gentamicin. Astrocytes were plated
   at a density of 1×10^7/T75 flask and cultured as adherent monolayers.
   After 1 week, the DMEM with human serum was replaced with DMEM with 10%
   fetal bovine serum (cat. no. 16000044; Gibco) and 1%
   penicillin/streptomycin (cat. no. 30-001-CI-1; Corning). Every 3 days,
   a half media exchange was performed. One donor line was used for the
   transcriptomic analysis and different passages of an additional donor
   line were used for the real-time quantitative polymerase chain reaction
   (RT-qPCR) experiments.

Treatments for RNA-seq experiment

   Primary astrocytes were treated with WIN (cat. no. 1038; R&D Systems)
   and IL1β (cat. no. rcyec-hil1b; InvivoGen) at concentrations of 10 μM
   and 10 ng/mL, respectively. DMSO was used as vehicle treatment. After
   treatment for 24 h with vehicle and WIN, IL1β was added for 6 h before
   RNA extraction. IL1β was used because of its widespread role in
   neuroinflammatory diseases.^[51]48–50 While many inflammatory cytokines
   are implicated in neurodegenerative diseases, IL1β is one of first and
   most robustly expressed cytokines in models for inflammation, including
   in vitro astrocyte cultures.^[52]51–54

RNA isolation and RT-qPCR

   RNA was extracted with RNeasy plus mini kit (cat. no. 74136; Qiagen)
   according to manufacturer's instructions. After cDNA synthesis, gene
   expression was determined using TaqMan gene expression assays using
   primers specific to IL6 (cat. no. hs00174131; TaqMan), C3 (cat. no.
   hs00163811; TaqMan), and ActB (cat. no. 1612290; Applied Biosystems).
   The PCRs were carried out using 2×Fast advanced master mix (cat. no.
   4444557; Thermo Fisher Scientific) at 48°C for 30 min and 95°C for
   10 min, followed by 40 cycles of 95°C for 15 sec and 60°C for 1 min.
   Fold-change of mRNA transcripts compared to vehicle-treated cells was
   calculated using the comparative CT method, as previously
   described.^[53]9,[54]55

WIN dose-response on IL1β-stimulated human astrocytes

   Astrocyte cultures were treated with WIN at 100 nM, 1 mM, 10 μM, and
   20 μM for 24 h and then treated with IL1β (10 ng/mL) for 6 h before RNA
   isolation for analysis by RT-qPCR. The concentration of IL1β used
   (10 ng/mL) is consistent with the relevant literature.^[55]56,[56]57
   Moreover, mouse models utilizing adenovirus-driven IL1β overexpression
   in the brain achieved expression levels of ∼10 ng/mg total protein 7
   days postinjections^[57]58 or a mean of 41 ng in the whole striatum 8
   days postinjections.^[58]59

Time-dependent effects of WIN on IL1β-stimulated human astrocytes

   To determine how time of exposure to WIN affects IL1β-induced changes
   in astrocyte gene expression, astrocytes were treated with WIN (20 μM)
   in one of three ways: (1) 24 h before IL1β, (2) same time as IL1β, or
   (3) 1 h after IL1β. IL1β was added at 10 ng/mL for 6 h.

Treatment of astrocytes with WIN and IL1β, and with PPARα, PPARγ, CB1, and
CB2 selective inhibitors

   To inhibit PPARα, PPARγ, and CB1, the following antagonists we used,
   respectively: GW9662 (cat. no. M6191; 10 μM; Sigma-Aldrich), GW6471
   (10 μM), and SR141617 (16 nM). Inhibitors were added to the cells for
   30 min followed by WIN (10 μM) for 24 h and then IL1β at 10 ng/mL for
   6 h before RNA isolation.

RNA-seq library preparation

   RNA was isolated as described above. RNA integrity was measured using
   an Agilent 2100 Bioanalyzer (Santa Clara, CA) with an RNA Integrity
   Number ≥8.5. cDNA libraries were prepared with 1 μg of starting total
   RNA using the Illumina TruSeq RNA Library Kit (Illumina, Inc., San
   Diego, CA). The libraries were amplified via 15 cycles of PCR and the
   amplified library was sequenced using an Illumina HiSeq 4000.

RNA-seq bioinformatics analysis

   Raw sequencing data were processed using Rosalind developed by OnRamp
   BioInformatics, Inc., (San Diego, CA). Reads were trimmed using
   cutadapt.^[59]60 Quality scores were assessed using FastQC. Reads were
   aligned to the Homo sapiens genome build hg19 using STAR.^[60]61
   Individual sample reads were quantified using HTseq^[61]62 and
   normalized via Relative Log Expression using DESeq2 R library.
   Sample-to-sample variation was assessed using a correlation matrix of
   Pearson's correlation coefficients ([62]Supplementary Fig. S1A).
   Downstream analysis was performed using raw counts and R packages
   supported by integrated Differential Expression and Pathway
   analysis.^[63]63 Exploratory data analysis was performed after data
   were transformed using the regularized-logarithm (rlog) transformation
   function of the DESeq2 package.^[64]64 Hierarchical clustering was
   computed by ranking all genes by standard deviation across all samples
   and the top 2000 genes were selected based on the distribution of
   variance ([65]Supplementary Fig. S1B). Principal component analysis
   (PCA) was performed using the matrix of rlog normalized read counts.
   The first and second principal components are shown to describe the
   largest variability in the dataset. Differential gene expression
   analysis was conducted using DESeq2 with false discovery rate (FDR)
   smaller than 0.01 and fold change (FC) larger than 2 as cutoffs. All
   comparisons were generated using the vehicle-treated samples as
   baseline. Volcano plots for each comparison show statistical
   significance (negative log of FDR) versus FC. Functional annotation of
   differentially expressed genes (DEGs) was done with enrichment analysis
   of biological pathways supplied by KEGG (Kyoto Encyclopedia of Genes
   and Genomes). For k-means clustering, we used the rlog transformed data
   and ranked the first 2000 genes by standard deviation. Based on the
   within-group sum of squares plot ([66]Supplementary Fig. 1C) as a
   reference, we chose k=4. Enrichment analysis is conducted for each
   cluster using Gene Ontology (GO) biological process terms. For pathway
   analysis, we used the FC values returned by DESeq2 and applied
   Parametric Analysis of Gene Set Enrichment using FDR <0.05 as
   cutoff.^[67]65 Gene sets associated to TFs were identified using the
   Transcriptional Regulatory Element Database.^[68]66

Statistical analysis

   RT-qPCR experiments using astrocyte cell cultures were performed in
   biological duplicates in astrocytes derived from an independent donor.
   Different passages were used to run different experiments in
   [69]Figures 5–7. The bars represent the mean of biological replicates.
   The error bars represent standard error of the mean. Data were analyzed
   by one-way analysis of variance (ANOVA) and post hoc analyses using
   Tukey's Test using GraphPad Prism. Differences in means were considered
   significant if p<0.05.

FIG. 5.

   [70]FIG. 5.
   [71]Open in a new tab

   WIN reduced IL1β-induced inflammatory gene expression in astrocytes in
   a dose-dependent manner. Fold-change of IL6 (A) or C3 (B) mRNA levels
   normalized to ACTB mRNA levels in total RNA isolated from human
   astrocytes. One-way ANOVA was conducted for the effect of treatment on
   IL6 [F (9, 10)=24.83, p<0.0001] and C3 [F (9, 10)=107.3, p<0.0001]. A
   post hoc Tukey's test was conducted; corrected p-values are shown
   (***p<0.001; ****p<0.0001 vs. vehicle; ^^^p<0.01, ^^^^p<0.001,
   ^^^^^p<0.0001 vs. IL1β-treated cells). ANOVA, analysis of variance.

FIG. 6.

   [72]FIG. 6.
   [73]Open in a new tab

   WIN treatment before, at the same time, and after IL1β treatment
   robustly blocked IL1β-induced inflammatory gene expression. (A–C)
   Fold-change of IL6 mRNA transcript levels normalized to ACTB mRNA
   levels in total RNA isolated from human astrocytes. One-way ANOVA was
   conducted for WIN 24 h before IL1β [F (3, 4)=377.6, p<0.0001], WIN at
   the same time as IL1β [F (3, 4)=91.26, p=0.0004], WIN after IL1β [F (3,
   4)=90.3, p=0.0004]. (D–F) Fold-change of C3 mRNA transcript levels
   normalized to ACTB mRNA levels in total RNA isolated from human
   astrocytes. One-way ANOVA was conducted for WIN 24 h before IL1β [F (3,
   4)=233.8, p<0.0001], WIN at the same time as IL1β [F (3, 4)=148.4,
   p=0.0001], WIN after IL1β [F (3, 4)=270.0, p<0.0001]. A post hoc
   Tukey's test was conducted; corrected p-values are shown (**p<0.01;
   ***p<0.001; ****p<0.0001 vs. vehicle, ^^^p<0.01, ^^^^p<0.001,
   ^^^^^p<0.0001 vs. IL1β-treated cells).

FIG. 7.

   [74]FIG. 7.
   [75]Open in a new tab

   PPARα, PPARγ, and CB1 antagonists did not block the inhibitory effect
   of WIN on the IL1β-induced inflammatory gene transcription. (A)
   Fold-change of IL6 mRNA transcript levels normalized to ACTB mRNA
   levels in total RNA isolated from human astrocytes. One-way ANOVA was
   conducted [F (12, 13)=55.96, p<0.0001]. (B) Fold-change of C3 mRNA
   transcript levels normalized to ACTB mRNA levels in total RNA isolated
   from human astrocytes. One-way ANOVA was conducted [F (12, 13)=90.84,
   p<0.0001]. A post hoc Tukey's test was conducted; corrected p-values
   are shown (***p<0.001, ****p<0.0001 vs. vehicle; n.s. vs. WIN+IL1β).
   CB1, cannabinoid receptor 1; PPAR, peroxisome proliferator-activated
   receptors.

Data availability

   Raw sequencing files and processed transcript counts generated in this
   study are deposited on the GEO at accession [76]GSE160092.

Results

RNA-seq analysis showed that substantial transcriptional changes were induced
by IL1β and WIN in primary astrocytes

   To evaluate the effect of WIN on the global transcriptional responses
   induced by IL1β in astrocytes, we performed transcriptomic analysis by
   RNA-seq. Astrocytes were treated with or without WIN (24 h) before
   stimulation with IL1β (6 h). Control RNA was extracted from
   vehicle-treated cultures. To assess variability among replicate
   samples, we used a Pearson's correlation matrix of normalized read
   counts, which showed that the variation was minimal with correlation
   coefficients (r) >0.9 ([77]Supplementary Fig. S1A). This result was
   confirmed by different clustering methods, including hierarchical
   clustering ([78]Fig. 1A), PCA ([79]Fig. 1B) and k-means clustering
   ([80]Fig. 3). To explore the overall variation in gene expression, we
   used hierarchical clustering and showed that both WIN and IL1β
   treatments induced substantial gene expression changes in astrocytes
   ([81]Fig. 1A). The PCA showed expected grouping among replicates and
   sample groups spread across the two PCs ([82]Fig. 1B). The PCA plot
   highlighted that there was a clear difference between WIN- and
   IL1β-treated samples. Overall, these results demonstrated that
   astrocytes showed a strong transcriptional response following IL1β and
   WIN treatments.

FIG. 1.

   [83]FIG. 1.
   [84]Open in a new tab

   Exploratory data analysis of transcriptomic data. (A) Hierarchical
   clustering of top 2000 genes showing that the variation between
   replicates was minimal and that substantial changes were induced by
   IL1β and WIN treatments. (B) PCA plot showing that the majority of
   variance is explained by PC1 (68%) and PC2 (21%). IL, interleukin; PCA,
   principal component analysis. Color images are available online.

FIG. 3.

   [85]FIG. 3.
   [86]Open in a new tab

   WIN diminished the IL1β-induced inflammatory response. (A) k-Means
   clustering visualizing 4 groups using the top 2000 differentially
   expressed genes. (B) The top GO biological process terms from each
   cluster is shown with associated p values. (C) For each cluster, the
   normalized expression values for three representative genes are plotted
   and error bars represent standard deviation values. GO, Gene Ontology.
   Color images are available online.

The IL1β-induced inflammatory transcriptional response in astrocytes was
ameliorated by the cannabinoid receptor agonist WIN

   Comparing IL1β, WIN, and IL1β+WIN-stimulated cells with vehicle-treated
   cells, we identified 1204, 3827, and 6046 genes differentially
   regulated genes, respectively ([87]Fig. 2A–C). To functionally annotate
   the up- and downregulated genes, we performed pathway enrichment
   analysis using KEGG pathways ([88]Fig. 2D–F). As expected, the
   IL1β-induced genes (n=914) were significantly enriched with pathways
   related to inflammatory response, such as TNF signaling,
   cytokine–cytokine receptor interaction, and NF-κB signaling ([89]Fig.
   2D). We did not identify pathways significantly enriched for the IL1β
   repressed genes, consistent with the small fraction of genes
   downregulated in this group (n=290). For WIN-regulated genes, the
   enriched pathways were distinct for up- and downregulated genes
   ([90]Fig. 2E). The genes repressed by WIN (n=1872) were involved in
   axon guidance and hippo signaling pathways. The genes induced by WIN
   (n=1955) were related in protein processing in endoplasmic reticulum.
   For genes regulated in the presence of both WIN and IL1β ([91]Fig. 2F),
   the upregulated genes (n=2797) were related to inflammation and protein
   processing in the endoplasmic reticulum, consistent with the pathways
   modulated by each individual treatment. The enrichment analysis of
   downregulated genes (n=3249) in presence of WIN and IL1β identified
   several kinase pathways, including PI3K-Akt signaling and Hippo
   signaling, suggesting that WIN activates specific signaling cascades
   and that the modulation of these pathways may be involved in the
   immunosuppressive action of WIN.

FIG. 2.

   [92]FIG. 2.
   [93]Open in a new tab

   IL1β and WIN treatments induced substantial changes in gene expression.
   Volcano plots visualizing the up- and downregulated genes by plotting
   the statistical significance (−log 10 FDR) versus the log2-fold change
   for pairwise comparisons between IL1β versus vehicle (A), WIN versus
   vehicle (B) and WIN+IL1β versus vehicle (C). (D–F) Enrichment KEGG
   pathways analysis of DEGs is shown for each pairwise comparison. The
   relationship among enriched KEGG pathways is visualized as a tree.
   DEGs, differentially expressed genes; FDR, false discovery rate; KEGG,
   Kyoto Encyclopedia of Genes and Genomes. Color images are available
   online.

   To gain insight into the molecular pathways underlying different
   patterns of gene expression in activated astrocytes that are exposed to
   WIN, we used k-means clustering, an unsupervised method for clustering
   genes into groups based on their expression pattern across all samples.
   [94]Figure 3 shows the gene clusters ([95]Fig. 3A), associated
   functional annotations ([96]Fig. 3B and [97]Table 1) based on the most
   enriched GO terms, and representative genes from each cluster ([98]Fig.
   3C). Clusters C and D were strongly enriched in genes related to
   inflammatory responses. These clusters of genes were upregulated in
   astrocytes stimulated with IL1β and downregulated in astrocytes
   pretreated with WIN. On the contrary, genes in clusters A and B were
   independent of IL1β and showed to be regulated by WIN. Cluster A
   included genes induced by WIN that were enriched in response to
   unfolded proteins. Cluster B contained genes repressed by WIN that were
   related to cell migration. These GO enrichment results confirmed the
   pathway enrichment analysis of DEGs in [99]Figure 2.

Table 1.

   Gene Ontology Terms Enriched for Different Clusters
   Cluster  Adj.p                          Pathways
   A       3.00E-17 Endoplasmic reticulum unfolded protein response
   A       1.93E-16 Response to endoplasmic reticulum stress
   A       1.20E-15 Response to unfolded protein
   A       1.20E-15 Cellular response to unfolded protein
   A       2.43E-15 Response to topologically incorrect protein
   A       2.48E-14 Cellular response to topologically incorrect protein
   A       5.93E-09 IRE1-mediated unfolded protein response
   A       8.19E-06 Carboxylic acid metabolic process
   A       9.96E-06 Oxoacid metabolic process
   A       1.29E-04 ER-nucleus signaling pathway
   A       1.35E-04 Organic substance transport
   A       1.62E-04 PERK-mediated unfolded protein response
   A       4.27E-04 Protein exit from endoplasmic reticulum
   A       4.27E-04 Golgi vesicle transport
   B       7.97E-31 Anatomical structure morphogenesis
   B       4.46E-28 Cell migration
   B       2.07E-27 Movement of cell or subcellular component
   B       2.04E-26 Regulation of developmental process
   B       2.65E-26 Cell motility
   B       4.25E-25 Locomotion
   B       2.67E-24 Tissue development
   B       8.43E-23 Cellular developmental process
   B       2.62E-22 Regulation of cell migration
   B       6.94E-22 Regulation of multicellular organismal process
   B       7.77E-22 Regulation of cell motility
   B       2.64E-21 Animal organ development
   B       3.38E-21 Cell differentiation
   B       3.73E-21 Regulation of signaling
   C       1.44E-24 Response to cytokine
   C       8.86E-24 Immune system process
   C       1.93E-23 Cellular response to chemical stimulus
   C       2.21E-23 Defense response
   C       5.56E-23 Response to organic substance
   C       5.58E-22 Response to virus
   C       5.58E-22 Cellular response to organic substance
   C       6.63E-22 Cellular response to cytokine stimulus
   C       5.98E-21 Response to stress
   C       1.89E-20 Regulation of multicellular organismal process
   C       5.45E-20 Response to external stimulus
   C       5.45E-20 Defense response to virus
   C       3.05E-19 Positive regulation of response to stimulus
   C       4.27E-19 Positive regulation of multicellular organismal process
   C       7.77E-19 Cell proliferation
   D       5.83E-51 Response to cytokine
   D       4.71E-49 Response to external biotic stimulus
   D       9.40E-48 Cytokine-mediated signaling pathway
   D       4.51E-46 Defense response
   D       1.20E-45 Cellular response to cytokine stimulus
   D       2.07E-41 Immune response
   D       1.27E-39 Immune system process
   D       2.37E-37 Response to external stimulus
   D       2.47E-32 Cellular response to organic substance
   D       4.94E-32 Response to organic substance
   D       2.22E-30 Multiorganism process
   D       1.06E-29 Response to stress
   D       1.59E-29 Inflammatory response
   [100]Open in a new tab

   Overall, this unsupervised analysis showed that the inflammatory
   response activated by IL1β was strongly ameliorated in the astrocytes
   pretreated with WIN ([101]Fig. 3B) and suggested that modulation of
   specific signaling pathways (e.g., PI3K-Akt) could mediate the
   repressive action of WIN. Moreover, these results showed that WIN
   induced substantial gene expression changes in resting astrocytes
   independently of IL1β stimulation.

TF enrichment analysis revealed TFs associated with inflammatory responses in
astrocytes

   To explore gene regulatory mechanisms underlying different patterns of
   gene expression, we performed pathway analysis to identify coherently
   altered upstream TFs associated to DEGs. This analysis confirmed that
   IL1β induced the activity of TFs known to regulate inflammatory genes
   induced by cytokines in immune cells, such as NF-κB, CEBP, and STAT
   ([102]Fig. 4A).^[103]15–17,[104]67 The analysis of genes regulated by
   WIN suggested that MYC was associated with upregulated genes and that
   E2F4 was associated with downregulated genes. Finally, we analyzed the
   genes regulated in presence of WIN and IL1β ([105]Fig. 4C). Among TFs
   associated with the action of WIN in presence of IL1β, PPAR and SMAD
   were of particular interest, as they have been linked to
   neuroprotective and anti-inflammatory pathways ([106]Fig.
   4C).^[107]8,[108]38,[109]68 Overall, this analysis revealed potential
   gene regulatory mechanisms underlying the anti-inflammatory properties
   of cannabinoid receptor agonists.

FIG. 4.

   [110]FIG. 4.
   [111]Open in a new tab

   TFs associated with inflammatory responses in astrocytes. Pathway
   analysis of DEGS was performed using the TRED database. Enriched TF are
   ranked by p-values (FDR <0.05 cutoff). Heatmaps are shown for pairwise
   comparisons between IL1β versus vehicle (A), WIN versus vehicle (B) and
   WIN+IL1β versus vehicle (C). TF, transcription factor; TRED,
   Transcriptional Regulatory Element Database. Color images are available
   online.

WIN inhibits the transcriptional activation of IL1β-induced inflammatory
genes

   To determine the dose-response curve of WIN-mediated inhibition of
   IL1β-induced genes, we used primary human astrocyte cultures that were
   exposed to increasing doses of WIN and then treated with vehicle or
   IL1β for 6 h. The expression of two inflammatory genes, IL6 and C3, was
   analyzed by RT-qPCR. IL1β robustly increased (∼55-fold) levels of IL6
   mRNA, while increasing doses of WIN had no effect on levels of IL6 mRNA
   ([112]Fig. 5A). WIN blocked IL1β-induced IL6 mRNA starting at 10 μM and
   no further reduction was seen with 20 μM WIN ([113]Fig. 5A). IL1β
   induced a ∼23-fold increase in C3 mRNA levels compared to vehicle and
   this induction was also blocked by 10 and 20 μM WIN by ∼35% and 55%,
   respectively ([114]Fig. 5B). These results showed that WIN exhibited a
   repressive effect on the transcriptional activation of inflammatory
   genes starting at concentration of 10 μM in vitro.

WIN treatment before, at the same time, and after IL1β treatment robustly
blocked IL1β-induced inflammatory gene expression

   To better understand how WIN affects inflammatory gene expression in
   astrocytes, cells were exposed to WIN 24 h before, at the same time,
   and 1 h after exposure to IL1β. After 6 h of IL1β treatment, total RNA
   was isolated and analyzed for IL6 and C3 mRNA levels. IL1β induced a
   robust increase in IL6 mRNA levels ([115]Fig. 6A–C). WIN treatment 24 h
   before IL1β reduced IL1β mRNA by over 90% ([116]Fig. 6A), while WIN
   treatment at the same time or 1 h after IL1β treatment resulted in a
   decrease of ∼75% compared to IL1β treated cells ([117]Fig. 6B, C). WIN
   treatment for 30 h induced a significant increase in C3 mRNA compared
   to vehicle-treated cells ([118]Fig. 6D). WIN-mediated effects on C3
   mRNA levels were not as robust as the effects on IL6 levels, with WIN
   mediating ∼30% reductions in IL1β-induced C3 mRNA levels when WIN was
   used 24 h before, at the same time, or after IL1β exposure ([119]Fig.
   6D–F). Collectively, these data suggest that the anti-inflammatory
   effects of WIN affect the expression of genes differently, likely
   depending upon the gene-specific regulatory mechanisms of
   transcription. The anti-inflammatory effects of WIN are, however,
   robust when used before, concomitant with, or after the inflammatory
   cytokine IL1β.

PPARα, PPARγ, and CB1 antagonists did not block the inhibitory effect of WIN
on the IL1β-induced inflammatory gene transcription

   To determine the involvement of PPAR and cannabinoid receptors, we
   sought to use selective antagonists against PPARα, PPARγ, and CB1
   receptors to modulate the inhibitory effects of WIN on the activation
   of IL6 and C3 mRNAs. We did not include CB2 due to undetectable or very
   low levels of expression of this gene in the primary astrocytic
   cultures ([120]Supplementary Fig. S1D).

   Astrocyte cultures were treated with or without WIN in the presence of
   GW6471 (10 μM, PPARα), GW9662 (10 μM, PPARγ), and SR141617 (16 nM,
   CB1). Treatment with IL1β induced a 20-fold increase in IL6 mRNA levels
   and this was reduced by ∼85% by WIN pretreatment. None of the selective
   inhibitors had a significant effect of reversing WIN-mediated
   inhibition ([121]Fig. 7A). The PPARγ inhibitor with IL1β showed modest
   increases in IL6 levels compared to IL1β alone, but the differences
   were not significant ([122]Fig. 7A). The lack of a modulatory effect of
   the PPARγ inhibitor is in contrast to results that we previously
   reported in astrocytes derived from a different donor, suggesting that
   the pathways mediating IL1β and WIN activity may vary by donor.^[123]8

   IL1β induced a ∼40-fold increase in C3 mRNA levels compared to
   vehicle-treated cells. WIN reduced the IL1β-induced changes by ∼50% and
   treatment with a PPARα and γ, and CB1 antagonists did not reverse the
   effects of WIN ([124]Fig. 7B). The selective antagonists had no effect
   on C3 mRNA levels in the presence of WIN, but they all increased the
   levels of IL1β-induced C3 mRNA compared to IL1β alone, although the
   differences did not reach significance ([125]Fig. 7B). Overall, these
   data suggest that the repressive effects of WIN did not depend on CB1
   or PPAR. However, the involvement of these pathways may vary by donor.

Discussion

   This work provides, for the first time, detailed transcriptomic
   analyses of IL1β-activated human astrocytes in the presence or absence
   of the cannabinoid receptor agonist WIN. The analysis of differentially
   expressed genes revealed the anti-inflammatory effect of WIN at a
   genome-wide level and identified specific signaling cascades regulated
   by IL1β and/or WIN. By using selective inhibitors and independent lines
   of human astrocytes, we showed that the anti-inflammatory effects of
   WIN are independent of canonical CB1 receptor.

   The robust induction of gene expression by IL1β is consistent with
   published findings showing that astrocytes are highly reactive to
   inflammatory cytokines and conditioned media from immunoactivated
   monocyte-derived macrophages and microglia.^[126]15,[127]69 Upon
   stimulation, astrocytes secrete inflammatory cytokines that act in
   autocrine and paracrine manners.^[128]70,[129]71 Reactive astrocytes
   also express adhesion molecules and chemokines that influence
   blood–brain barrier permeability and the passage of immune cells from
   the periphery into the brain.^[130]72 Unchecked, the inflammatory cycle
   may lead to neurodegeneration. Indeed, the transcriptomic changes
   presented here are consistent with these astrocytic properties and with
   molecular signatures of neuroinflammatory diseases such as AD and
   HAND.^[131]25,[132]73,[133]74 In particular, activation of
   transcriptional networks involving responses to viral infection,
   autoimmunity, and cancer shown in [134]Figure 2D are consistent with in
   vitro and in vivo studies on astrocytes in neurodegenerative
   diseases.^[135]2,[136]25,[137]75

   WIN has been studied extensively for its anti-inflammatory effects, but
   no studies have detailed the genome-wide effects of WIN on human
   astrocytes. WIN-mediated activation of the unfolded protein response
   corroborates previous studies showing that the cannabinoid receptor
   agonist stresses the function of the endoplasmic reticulum.^[138]76 The
   activation of PI3 kinase signaling in activated astrocytes treated with
   WIN is consistent with previous reports suggesting that PI3 kinase
   mediates a neuroprotective role of WIN.^[139]77

   Clustering analysis highlighted the suppressive effects of WIN on
   transcriptional responses induced by IL1β. These data are consistent
   with studies in animal models that show that cannabinoids can slow
   progression of neurodegenerative disease and behavioral deficits
   through reductions in inflammatory responses.^[140]78–80 Follow-up
   experiments are needed to determine if the immunosuppressive effect is
   a property of other cannabinoids. In future studies, it will be
   important to assess toxicity using in vitro and in vivo models to
   optimize the therapeutic use of WIN and other cannabinoids. However,
   the fact that WIN robustly reversed many of the IL1β-induced
   transcriptomic changes bodes well for therapeutic targeting of
   astrocytes to disrupt chronic inflammation in the brain as is prevalent
   in the neurodegenerative diseases AD and HAND.

   Bioinformatic analysis of the promoter of DEGs revealed that the
   consensus motif of PPAR and SMAD TFs are enriched in IL1β-induced genes
   that are inhibited by WIN. Previous studies have suggested that the
   anti-inflammatory effects of WIN and other cannabinoids may be
   independent of CB1 and CB2 receptors.^[141]81–83 Other studies,
   including our own, support a potential role for the PPAR family in
   mediating the anti-inflammatory effects of WIN and possibly other
   cannabinoids.^[142]8,[143]38,[144]40 However, using a new donor line,
   the immunosuppressive effect of WIN on gene transcription was not
   reversed by selective antagonists of PPARα and PPARγ, suggesting that
   the pathways mediating the effects of WIN on astrocytes may vary by
   donor. Follow-up experiments comparing different donor lines are needed
   to confirm whether genetic background is an important factor in
   determining the effects of WIN in astrocytes.

   This comprehensive dataset generated from primary human astrocytes
   should be viewed in the context of the following limitations. Any study
   of only a single brain cell in a culture is limited in its implications
   as astrocytes function in concert with other brain cell
   types.^[145]69,[146]84,[147]85 It will be important in future studies
   to compare the transcriptome of astrocytes from in vitro cultures with
   those isolated from diseased brain tissues with the goal of determining
   the degree to which in vitro reactive astrocytes reflect the condition
   in neurodegenerative diseases. Moreover, donor-specific effects should
   be addressed when using human tissues. Using two donor lines, we showed
   that the immunosuppressive effect of WIN was independent of the
   specific donor line; however, we did not reproduce the modulatory
   effect of PPARγ inhibitors in contrast to results previously published
   using a different astrocyte donor line. Further studies using
   astrocytes generated from multiple independent genetic backgrounds are
   needed to capture interindividual variabilities in the response to
   inflammatory stimuli and cannabinoid receptor agonists. These
   interindividual differences in anti-inflammatory effects highlight the
   potential of using personalized medicine to identify therapeutic
   strategies for neurodegenerative diseases. An additional limitation is
   the use of a single inflammatory cytokine, which is of limited
   relevance to the multifactorial aspect of neuroinflammatory conditions.
   Moreover, these data represent only a single snapshot of the
   transcriptome of astrocytes, which likely changes over time.

   In conclusion, our results suggest that reactive astrocytes may
   contribute to some of the molecular signatures associated with
   inflammation in neurodegenerative diseases, and that these signatures
   and therapeutic strategies to reverse them can be modeled in vitro by
   stimulating primary human astrocytes with the inflammatory cytokine
   IL1β. The global anti-inflammatory activity of WIN may explain recent
   studies that suggest cannabis is neuroprotective in AD and
   HAND.^[148]78,[149]86 Future studies examining the effects of different
   stimuli and combinations of varying proportions on astrocytes and other
   brain cells may lead to better modeling of and therapeutic testing for
   neurodegenerative diseases.

Supplementary Material

   Supplemental data
   [150]Supp_FigS1.docx^ (91.2KB, docx)

Acknowledgments