Adolescent Δ⁹-THC exposure and astrocyte-specific genetic vulnerability converge on NF-κB-COX-2 signaling to impair memory in adulthood.

Abstract

Background:

Although several studies linked adolescent cannabis use to long-term cognitive dysfunction, there are negative reports too. The fact that not all users develop cognitive impairment suggests a genetic vulnerability to adverse effects of cannabis, which are attributed to action of delta-9-tetrahydrocannabinol (Δ⁹-THC), a cannabis constituent and partial agonist of brain cannabinoid receptor 1 (CNR1). As both neurons and glial cells express CNR1, genetic vulnerability could influence Δ⁹-THC-induced signaling in a cell type-specific manner.

Methods:

Here we use an animal model of inducible expression of dominant-negative Disrupted-In-Schizophrenia-1 (DN-DISC1) selectively in astrocytes to evaluate the molecular mechanisms whereby an astrocyte genetic vulnerability could interact with adolescent Δ⁹-THC exposure to impair recognition memory in adulthood.

Results:

Selective expression of DN-DISC1 in astrocytes and adolescent treatment with Δ⁹-THC synergistically affected recognition memory in adult mice. Similar deficits in recognition memory were observed following knockdown of endogenous Disc1 in hippocampal astrocytes in mice treated with Δ⁹-THC during adolescence. At the molecular level, DN-DISC1 and Δ⁹-THC synergistically activated the NF-kB-COX-2 pathway in astrocytes and decreased immunoreactivity of parvalbumin-positive pre-synaptic inhibitory boutons around pyramidal neurons of the hippocampal CA3 area. The cognitive abnormalities were prevented in DN-DISC1 mice exposed to Δ⁹-THC by simultaneous adolescent treatment with the COX-2 inhibitor, NS389.

Conclusions:

Our data demonstrate that individual vulnerability to cannabis can be exclusively mediated by astrocytes. Results of this work suggest that genetic predisposition within astrocytes can exaggerate Δ⁹-THC-produced cognitive impairments via convergent inflammatory signaling, suggesting possible targets for preventing adverse effects of cannabis within susceptible individuals.

One sentence summary:

Adolescent GxE in astrocytes impairs adulthood memory

Introduction

Cannabis is the most commonly used illicit drug of abuse in the United States (1). Although several studies have reported no long-term cognitive impairments after cannabis use, chronic cannabis exposure during adolescence has been associated with persistent deficits in some cognitive domains, including attention, memory, and processing speed (2–4). Cannabis sativa plant includes more than 400 different chemical constituents, of which about 70 are cannabinoids (5). Cannabis-induced adverse effects are mediated by delta-9-tetrahydrocannabinol (Δ⁹-THC), the principal psychoactive constituent of cannabis and a partial agonist of brain cannabinoid receptors (CNR1) (6). However, the mechanisms underlying Δ⁹-THC-induced long-lasting behavioral and cognitive abnormalities remain unknown.

Although neurons highly express CNR1, the role of glial CNR1 is being increasingly appreciated (7, 8). Two recent studies have shown that detrimental effects of Δ⁹-THC on learning and memory in mice are mediated by astrocyte CNR1 (9, 10), activation of NF-κB signaling and up-regulation of cyclooxygenase-2 (COX-2) that might lead to excessive glutamate release by astrocytes (11).

Notably, not all cannabis users demonstrate cognitive impairment suggesting a genetic vulnerability to adverse effects of cannabis (12–14). Similarly, preclinical studies have reported that mice carrying mutations in candidate genes for psychiatric disorders exhibit greater responses to adverse effects of Δ⁹-THC on memory (15–19). However, the underlying molecular mechanisms of how genetic mutations could moderate cognitive effects of cannabis remain unknown. Further, although astrocytes appear to play a major role in mediating effects of Δ⁹-THC on memory (9, 10), the molecular underpinning of how genetic risk factors could interact with Δ⁹-THC in a cell type specific manner to impair cognitive abilities have never been studied. In order to address these questions, we used an animal model of selective astrocyte expression of a rare, highly penetrant mutation, a dominant-negative form of Disrupted in Schizophrenia 1 (DISC1) (18, 20–28).

DISC1 is a gene disrupted by the balanced (1;11) (q42.1; q14.3) translocation, segregating in a Scottish family with several major psychiatric disorders (29, 30). Although the DISC1 locus has not been reported in recent genome-wide association studies (GWAS)(31), rare mutations of large effects contribute to behavioral and cognitive abnormalities (32) and have important roles in mechanistic studies (33, 34). It is in this context that we use a C-terminus truncated form of full-length DISC1 as a dominant-negative molecular tool (DN-DISC1). In this study, we sought to determine the molecular basis of gene-environment interaction (GxE) in astrocytes and elucidate how GxE interplay could shape up individual vulnerability to adverse cognitive effects of cannabis on cognitive abilities.

Materials and Methods

Animals

In order to evaluate the cell-specific role of astrocytes in gene environment interaction (GxE), mice expressing dominant negative form of DISC1 (DN-DISC1) in astrocytes were exposed to chronic Δ9-THC treatment (8 mg/kg, sc, daily) for three weeks from postnatal day (P) P30 and on. 21 days later, mice were assessed in a series of behavioral tests. All procedures were approved by the JHU ACUC.

Behavioral tests

The following tests were used: open field, spontaneous alternation, spatial recognition in Y maze, novel object recognition test, novel place recognition test and fear conditioning as previously described (24, 35–37).

AAV injections

AAV-Gfa-mir30-Disc1-EGFP or AAV-Gfa mir30-control-EGFP were injected in the CA2-CA3 areas of the hippocampus at P15–17.

Isolation of RNA and RNAseq analyses

Total RNA was purified from mouse hippocampus upon completion of Δ9-THC. RNAseq analysis was done as described in supplemental 1.

Biochemistry

Expression of Phospho-p65, Phospho-IκBα and COX-2 was assessed with standard Western blotting in primary astrocytes or hippocampal tissue samples. DISC1-IκBα binding was analyzed with co-immunoprecipitation as previously described (38).

Immunohistochemistry

GAD67-positive (GAD67⁺) presynaptic boutons within parvalbumin-positive (PV⁺) branches were evaluated on the surface of pyramidal neurons of the CA1 and CA3 areas of hippocampus.

Pharmacological treatment with the COX-2 inhibitor

Effects of COX-2 inhibition on cognitive deficits were assessed with the COX-2 inhibitor, NS398, (10mg/kg, daily sc).

Measurement of glutamate in hippocampal tissue and in culture medium

Glutamate concentration in the hippocampus or primary astrocyte culture were assayed using Kusakabe’s method (39) or glutamate assay, respectively.

For detailed information, please see Supplemental 1.

Results

Astrocyte DN-DISC1 and adolescent Δ⁹-THC impair memory in adult mice

Based on our prior studies (26, 28), we hypothesized that expression of DN-DISC1 in astrocytes (aDN-DISC1) would synergistically interact with adolescent Δ⁹-THC exposure to affect learning and memory in adult mice. In order to test this hypothesis, we treated control or aDN-DISC1 male and female mice with single daily injections of Δ⁹-THC (8mg/kg; SC) starting at postnatal day 30 (P30) for three weeks (14) to span mouse adolescence (P30–51) that corresponds to human adolescence from 12 to 19 years of age (40–43). Upon completion of treatment, the mice were left undisturbed for another three weeks before behavioral testing was commenced (Fig.1A).

Figure 1. Cognitive impairments in astrocyte DN-DISC1 mice.

A - Schematic diagram of the treatment protocol

In all graphs the Y-axes depict the preference (%); the X-axes depict the experimental groups: C-V – control mice treated with the vehicle (6 males and 16 females); C-T – control mice treated with Δ⁹-THC (10 males and 16 females); DISC1-V – aDN-DISC1 mice treated with vehicle (11 males and 7 females); DISC1-T - aDN-DISC1 mice treated with Δ⁹-THC (10 males and 8 females)

Male mice

B- Spatial recognition memory in the Y maze. Compared to other groups, aDN-DISC1 male mice treated with Δ⁹-THC exhibited the significantly decreased preference for the previously blocked arm. Two-way ANOVA of the preference data revealed a significant effect of DN-DISC1, F(1,33)=5.49, p=0.025 and the significant aDN-DISC1 by Δ⁹-THC interaction, F(1,33)=4.46, p = 0.045; Fisher LSD post-hoc test showed that aDN-DISC1-Δ⁹-THC mice were significantly different from both C-T and DISC1-V mice, ** p<0.01.

C- Novel object recognition test (NORT). Compared to other groups, aDN-DISC1 male mice treated with Δ⁹-THC exhibited the significantly decreased preference for the novel object. Two-way ANOVA of the preference data revealed a significant effect of DN-DISC1, F(1,33)=26.12, p<0.001; a significant effect of Δ⁹-THC, F(1,33)=9.26, p=0.005 and the significant aDN-DISC1 by Δ⁹-THC interaction, F(1,33)=14.48, p<0.001; Fisher LSD post-hoc test showed that DN-DISC1- Δ⁹-THC mice were different from both C-T and DISC1-V mice, ** p<0.01.

D- Novel place recognition test (NPRT). Compared to other groups, aDN-DISC1 male mice treated with Δ⁹-THC exhibited the significantly decreased preference for the novel place of one of two identical objects. Two-way ANOVA of the preference data revealed a significant effect of Δ⁹-THC, F(1,33)=7.89, p=0.008. Planned post-hoc tests showed that significantly decreased preference in aDN-DISC1 mice treated with Δ⁹-THC compared to vehicle-treated aDN-DISC1 mice (DISC1-V) but there was no difference in the preference between Δ⁹-THC –treated- aDN-DISC1 (DISC1-T) and control (C-T) mice (p=0.074). * p<0.05.

Female mice

E- Spatial recognition memory in the Y maze. Compared to other groups, aDN-DISC1 female mice treated with Δ⁹-THC exhibited the significantly decreased preference for the previously blocked arm. Two-way ANOVA of the preference data revealed a significant effect of Δ⁹-THC, F(1,43)=4.74, p=0.035 and the significant aDN-DISC1 by Δ⁹-THC interaction, F(1,43)=12.88, p<0.001; Fisher LSD post-hoc test showed that aDN-DISC1-Δ⁹-THC (DISC1-T) mice were different from vehicle-treated - aDN-DISC1-vehicle (DISC1-V) mice and Δ⁹-THC -treated control (C-T) mice, ** p<0.01.

F- NORT. No group differences were found.

G- NPRT. No group differences were found.

For male mice, compared to other groups, aDN-DISC1 mice treated with Δ⁹-THC exhibited synergistically impaired performance in the spatial recognition test in the Y maze (the significant aDN-DISC1 by Δ⁹-THC interaction, F(1,33)=4.46, p = 0.045) and the novel object recognition test (NORT) (the significant aDN-DISC1 by Δ⁹-THC interaction, F(1,33)=14.48, p<0.001), as well as significantly worse performance in the novel place recognition test (NPRT) compared to aDN-DISC1 mice treated with vehicle, p<0.05 (Fig.1B–D). For female mice, the aDN-DISC1-Δ⁹THC combination produced synergistic impairment in the spatial recognition test in the Y-maze only (the significant aDN-DISC1 by Δ9-THC interaction, F(1,43)=12.88, p<0.001) (Fig.1E–G). Memory deficits were not associated with group differences in general exploratory activity during the training phase for the above tests, distance traveled in the Y maze, or novelty-induced activity. No effects of Δ⁹-THC were found in the forced swim test, or context- or cue-dependent delay fear conditioning either (Fig.S1–S3 in Supplement 1). Thus, our results suggest that adolescent treatment with Δ⁹-THC and expression of DN-DISC1 in astrocytes synergistically affect recognition memory in adult mice. As the most robust behavioral effects were found in aDN-DISC1 male mice, we focused on male mice in all subsequent tests.

In order to examine whether the above synergistic effects were dependent on adolescent Δ⁹-THC exposure (39), adult control and aDN-DISC1 mice were treated using the same protocol. We found no significant effects of adult Δ⁹-THC exposure on recognition memory in aDN-DISC1 mice (Fig.S4 in Supplement 1).

Although expression of aDN-DISC1 in the brain reaches the maximum by late adolescence, there is expression of aDN-DSC1 during late gestation and early postnatal period (28) that coincides with astrocyte proliferation and maturation (44, 45). In order to evaluate a possible contribution of early expression of aDN-DISC1 to the cognitive phenotypes, we turned off expression of aDN-DISC1 using doxycycline-containing food beginning at P21 and on (Fig.S5A in Supplement 1). We observed no significant cognitive effects in any group (Fig.S5BD in Supplement 1, suggesting that developmental expression of aDN-DISC1 unlikely contributed to the cognitive abnormalities.

We next wondered whether a different psychoactive compound could also interact with aDN-DISC1 to impair recognition memory (46, 47). In order to evaluate this possibility, control and aDN-DISC1 adolescent male mice were treated with amphetamine (1 mg/kg, ip). No significant changes in the same memory tests were found in either group (Fig.S6 in Supplement 1).

We also evaluated whether expression of DN-DISC1 in neurons could also lead to the cognitive deficits after adolescent Δ⁹-THC treatment. We generated mice with neuronal expression of DN-DISC1 (nDN-DISC1) by crossing TRE-DN-DISC1 mice with CAMKII-tTA mice to express DN-DISC1 in forebrain neurons (23). Control and nDN-DISC1 male mice were treated with the same treatment prFig.S7 in Supplement 1), suggesting a cell type specific GxE to impair recognition memory.

The hippocampus is sufficient for mediating the major effects of interaction

The hippocampus plays the critical role in spatial recognition memory (48–50). The cognitive effects of exogenous cannabinoids have been linked to adverse effects on hippocampal neuronal circuits (51–54). In addition, we previously reported strong expression of astrocyte DN-DISC1 in the hippocampus compared to the frontal cortex (28). Thus, we evaluated the contribution of hippocampal aDN-DISC1 to memory deficits observed in Δ⁹-THC-treated mice. In order to address this question and alter expression of DISC1 in astrocytes with a different genetic tool, we engineered an adeno-associated viral (AAV) vector to knockdown (KD) endogenous Disc1 selectively in astrocytes, Gfa-GFP-mir30-Disc1, or control (scrambled) vector, Gfa-GFP-mir30-Ctrl (Fig.S8 in Supplement 1). Gfa-GFP-mir30-Disc1 AAV decreased expression of Disc1 in vivo (Fig.S8A) and transduced astrocytes only (Fig.S8B) (see Supplement 1).

Since prior studies have suggested that Δ⁹-THC exposure may affect CA1-CA3 circuits (9, 55, 56), control and mir30-Disc1 AAV were injected in the CA2-CA3 areas of the hippocampus at P16 to let the effects of KD take place by P30 when Δ⁹-THC treatment was commenced. Three weeks later behavioral testing was initiated (Fig.2A). Consistent with our earlier findings (Fig.1), compared to the Gfa-GFP-mir30-Ctrl AAV, Gfa-GFP-mir30-Disc1 AAV synergistically impaired spatial memory in the Y-maze test (the significant Disc1-KD by Δ⁹-THC interaction, F(1,22)=13.90, p=0.001), NORT (the significant the Disc1-KD by Δ⁹-THC interaction, F(1,22)=4.35, p=0.05) but not NPRT in mice treated with Δ⁹-THC during adolescence (Fig.2B–D). These effects were unlikely related to non-specific changes in locomotion or exploratory activity (Fig.S9 in Supplement 1). Upon completion of behavioral tests (P90), we confirmed that AAV transduction was present in astrocytes of the CA2-CA3 areas (Fig.S10 in Supplement 1). Thus, expression of aDN-DISC1 or Disc1 KD in hippocampal astrocytes synergistically exacerbated the adverse cognitive effects of adolescent Δ⁹-THC in adult mice.

Figure 2. Disc1 KD in hippocampal astrocytes.

A- Schematic diagram of the AAV vector injections and treatment protocol;

On all data panels: the Y-axes depict the preference (%); the X-axes depict the experimental groups: C-V – mice injected with Gfa-GFP-mir30-Ctrl AAV and treated with the vehicle (N=6); C-T - mice injected with Gfa-GFP-mir30-Ctrl AAV and treated with Δ⁹-THC (N=6); DISC1-V – mice injected with Gfa-GFP-mir30-Disc1 AAV and treated with vehicle (N=5); DISC1-T - mice injected with Gfa-GFP-mir30-Disc1 AAV and treated with Δ⁹-THC (N=6).

B- Spatial recognition memory in the Y maze. Compared to other groups, Disc1 AAV mice treated with Δ⁹-THC exhibited the significantly decreased preference for the previously blocked arm. Two-way ANOVA of the preference data revealed a significant effect of Disc1-KD, F(1,22)=5.77, p=0.027; a significant effect of Δ⁹-THC, F(1,22)=11.43, p=0.003, and the significant Disc1-KD by Δ⁹-THC interaction, F(1,22)=13.90, p=0.001. Fisher LSD post-hoc test showed that Disc1-KD- Δ⁹-THC mice were different from Δ⁹-THC-treated control mice (p<0.001) and Disc1-KD vehicle-treated mice (p<0.001); ** p<0.001.

C- Novel object recognition test (NORT). Compared to other groups, Disc1 AAV mice treated with Δ⁹-THC exhibited the significantly decreased preference for the novel object. Two-way ANOVA of the preference data revealed a significant effect of Disc1-KD, F(1,22)=11.20, p=0.003; and the borderline significance for the Δ⁹-THC by Disc1-KD, F(1,22)=4.35, p=0.051; Fisher LSD post-hoc test showed that Disc1-KD-THC mice were different from Δ⁹-THC-treated control mice (p<0.001) and Disc1-KD vehicle-treated mice (p=0.027); **p<0.01, * p<0.05.

D - Novel place recognition test (NPRT). No significant effects of Disc1-KD were found in NPRT.

Δ⁹-THC activates pro-inflammatory signaling in DN-DISC1 astrocytes

In order to gain an unbiased insight in the mechanisms of GxE, we performed RNA-seq analyses of hippocampal tissue samples derived from control and aDN-DISC1 mice treated with vehicle or Δ⁹-THC as above. Our analyses focused on the 56 genes differentially expressed in both the aDN-DISC1 and Δ⁹-THC conditions, but not in either condition alone (Fig.3A). This analysis revealed a significant up-regulation of genes involved in the inflammatory pathways, including NF-kB signaling (Fig.3B). Full lists of differentially expressed genes and pathways are presented in Tables S2 and S3 (in Supplement 2), respectively (the NCBI accession number is GSE116813). Together with prior studies demonstrating that both Δ⁹-THC and DISC1 influence NF-kB signaling (10), our results suggest that this inflammatory pathway may be a convergent target of DN-DISC1 and Δ⁹-THC in astrocytes.

Figure 3. RNA-seq identifies synergistic (GxE) genes, which are enriched in NF-kB signaling pathway.

A – In aDN-Disc1 mice exposed to Δ⁹-THC, there were 145 differentially expressed genes (FDR < 0.20; see circle at upper right). Over a third of the 145 genes (56, 38.6%; see red shaded portion of circle) were synergistic (GxE) genes since they were not found to be differentially expressed by aDN-Disc1 or THC alone. Two of the genes (Ddit4, Sgk1) were also differentially expressed by aDN-Disc1 or THC alone, and 87 genes were differentially expressed by aDN-Disc1 alone. The oval represents six genes differentially expressed after THC treatment: three were differentially expressed in wild type mice, and three in the aDNn-Disc1 mice of which two were in aDNA-Disc1 mice treated with THC.

B – GxE genes are enriched for membership in the NF-kB activation by viruses pathway (z-score = 2.00; p = 1.74E-03; BH-adjusted p-value [FDR] = 0.0253, 4 genes; all up-regulated: Akt2, Cd4, Itga5, and Prkch).

DISC1 regulates activation of NF-kB-COX-2 signaling in astrocytes

We focused on evaluation of expression of Ptgs2 (a.k.a. Cox-2) that encodes for a constitutively expressed and inducible enzyme, cyclooxygenase-2 (COX-2), that converts arachidonic acid to prostaglandins (57, 58). We chose to assess altered expression of Ptgs2 in our GxE model because Δ⁹-THC induces a robust increase in activity and expression of COX-2 in astrocytes and mediates synaptic and cognitive effects of Δ⁹-THC (10, 59). Since tissue astrocytes and cultured cells exhibit significant transcriptome differences (60), we assessed Ptgs2 expression in astrocytes acutely isolated from the hippocampus followed by Δ⁹-THC treatment. We found a synergistic up-regulation of Ptgs2 in astrocytes of aDN-DISC1 mice treated with Δ⁹-THC compared to other groups (the significant DN-DISC1 × Δ⁹-THC treatment interaction, F(1,8)=87.43, p=0.001) (Fig.4A).

Figure 4. Synergistic effects of DN-DISC1 and Δ⁹-THC on the NF-kB- COX-2 signaling in astrocytes.

A - DISC1 and Δ⁹-THC synergistically increased Ptgs2 (gene encoding for COX2 protein) expression in hippocamplal tissue astrocytes. The graph plots the individual data points and superimposes the mean and error bars. Each point represents an independent sample from a single animal assayed in triplicates. Two-way ANOVA revealed the significant effects of DN-DISC1 (F(1,8)=139.52, p<0.001), Δ⁹-THC (F(1,8)=59.15, p<0.001) and significant DN-DISC1 × Δ⁹-THC treatment interaction, F(1,8)=87.43, p=0.001. Fisher LSD post-hoc analysis showed that Ptgs2 expression in the DN-DISC1-Δ⁹-THC group was significantly greater compared to DN-DISC1-vehicle (DISC1-T vs. DISC1-V; p<0.001) or control- Δ⁹-THC group (DISC1-T vs. C-T; p<0.001); ** - denotes p<0.01.

B – Representative Western blot images and densitometric analysis showing that DISC1 and Δ⁹-THC treatment synergistically increased expression of COX-2 protein and phosphorylation of p65 (S536-p65/p65) in hippocampal astrocytes. Two-way ANOVA revealed the significant effects on COX-2 protein level (main effects: DISC1 (F(1,8)=18.76, p=0.003, Δ⁹-THC F(1,8)=18.79, p=0.002 and the significant DN-DISC1 × Δ⁹-THC interaction F(1,8) =15.52, p=0.004) and on phosphorylation of NF-κB p65 (main effects: DISC1 (F(1,8)=9.29, p=0.016, Δ⁹-THC F(1,8)=6.32, p=0.036 and the significant DN-DISC1 × Δ⁹-THC interaction: F(1,8)=6.03, p=0.040. Fisher LSD post-hoc analysis showed that protein level of COX-2 (protein encoded by Ptgs2 gene) in DN-DISC1-Δ⁹-THC group was significantly greater compared to control-Δ⁹-THC (DISC1-T vs. C-T; p<0.001) and DN-DISC1-Vehicle group (DISC1-T vs. DISC1-V; p<0.001). Fisher LSD post-hoc analysis showed protein level of Phospho-NF-κB p65 in DN-DISC1-Δ⁹-THC group was significantly greater compared to control-Δ⁹-THC (DISC1-T vs. C-T; p=0.005) or DN-DISC1-Vehicle group (DISC1-T vs. DISC1-V; p=0.008). No other significant differences were detected in expression of COX-2 or Phospho-NF-κB p65. Data are presented as the mean ± SEM; n=3 independent samples in each group; ** - denotes p<0.01.

C – Δ⁹-THC treatment of primary DN-DISC1 astrocytes up-regulated phosphorylation of IκBα and decreased IκBα expression (5 minutes interval) and increased phosphorylation of NF-κB p65 (one hour interval).

D- Protein interaction of DISC1 with phospho-IκBα (p32/36-IkBα) was assessed by coimmunoprecipitation (IP) in HEK293 cells. HA-tagged wild-type DISC1 interacts with endogenous phospho-IκBα. Overexpression of myc-tagged DN-DISC1 reduced wild-type DISC1- phospho-IkBα interaction. Input for each protein is presneted.

We then assessed the protein levels of COX-2 and phospho-NF-κB p65 in astrocytes isolated from the hippocampus one day after in vivo Δ⁹-THC or vehicle treatment. Consistent with mRNA data, we found a significant and synergistic increase in COX-2 level (the significant DN-DISC1 × Δ⁹-THC interaction, F(1,8) =15.52, p=0.004) and phosphorylation of NF-κB p65 compared to other groups (the significant DN-DISC1 × Δ⁹-THC interaction, F(1,8)=6.03, p=0.040) (Fig.4B).

To further evaluate activation of NF-κB p65 signaling in astrocyte, we measured the phosphorylation levels of NF-κB p65 and IκBα, an upstream signaling protein of NF-κB p65, in primary DN-DISC1 or control astrocytes treated with Δ⁹-THC (5 M for 5 min or 1 hour) or vehicle. Consistent with in vivo data, DN-DISC1 astrocytes treated with Δ⁹-THC had synergistically increased phosphorylation of NF-κB p65 (Δ⁹-THC treatment for an hour) and enhanced phosphorylation of IκBα (Δ⁹-THC treatment for 5 min), compared with other conditions (Fig.4C).

Given that phosphorylation of IκBα leads to its dissociation from p65 and degradation, an event that requires prior to nuclear translocation of the liberated p65 (61, 62), DISC1 may interact with phospho-IκBα for stabilization of cytoplasmic IκBα:p65 complex. Indeed, coimmunoprecipitation experiments confirmed protein interaction between wild-type DISC1 and phospho-IκBα. Importantly, this interaction was disrupted by over-expression of DN-DISC1 (Fig. 4D). Our results suggest that binding of DISC1 to phospho-IκBα may stabilize IκBα activity, supporting inhibitory action of IκBα on NF-kB. Expression of DN-DISC1 in astrocytes leads to decreased levels of endogenous DISC1 (26) that may facilitate degradation of phosphorylated-IκBα and activation of NF-kB as a result of its release from binding to IκBα.

As up-regulation of COX-2 could increase glutamate secretion by astrocytes (11), we examined the effects of Δ⁹-THC on levels of glutamate in the hippocampus and culture medium collected from primary astrocytes. Compared to other groups, there was a significant increase in glutamate levels in the hippocampus of aDN-DISC1 mice treated with Δ⁹-THC mice (p<0.05) (Fig.S11A in Supplement 1). In addition, we found a synergistically increased secretion of glutamate by primary DN-DISC1 astrocytes following stimulation of Δ⁹-THC (the significant aDN-DISC1 by Δ⁹-THC interaction, F(1,18)=6.74, p=0.02) (Fig.S11B in Supplement 1). Collectively, these results suggest that DN-DISC1 and Δ⁹-THC interact to activate the pro-inflammatory NF-kB-COX-2 signaling and increase secretion of glutamate by astrocytes.

Decreased GAD⁺PV⁺ immunoreactivity in the CA3 area of the hippocampus

Chronic exposure to Δ⁹-THC during adolescence was associated with decreased GAD67 expression in parvalbumin-positive (PV⁺) interneurons (63–67). Given enhanced vulnerability of GABA⁺PV⁺ neurons to adverse effects of chronic excitotoxicity that may be a result of increased secretion of glutamate by DN-DISC1 astrocytes (68–71), we assessed the integrity of pre-synaptic GAD⁺PV⁺ boutons in control and aDN-DISC1 mice. Since our AAV results strongly suggested that the CA areas of the hippocampus could be critically involved in producing deficient recognition memory, we focused on examination of pre-synaptic GAD⁺PV⁺ boutons in these areas (Fig.S12 in Supplement 1). Compared to aDN-DISC1 mice treated with vehicle or control mice treated with Δ⁹-THC, we found significantly decreased intensity of GAD⁺PV⁺ boutons on the surface of pyramidal neurons of the CA3 but not CA1 area in aDN-DISC1 mice treated with Δ⁹-THC (p<0.05) (Fig.5). No significant changes of GAD⁺PV⁺ boutons size or density were found (Fig.S13 in Supplement 1). Reduced immunoreactivity of GAD⁺PV⁺ boutons was not associated with a general decrease in GAD⁺ immunoreactivity (Fig.S14 in Supplement 1), suggesting PV⁺ GABA neurons could be selectively and synergistically affected in aDN-DISC1 mice treated with Δ⁹-THC during adolescence.

Figure 5. Astrocyte DN-DISC1 and Δ⁹-THC treatment synergistically decrease GAD⁺PV⁺ immunoreactivity in the CA3 area of the hippocampus.

On all data panels: the Y-axes depict the percentage of GAD boutons intensity in CA3 or CA1 PV neurons in relation to the level for the control-vehicle group; the X-axes depict the experimental groups (C-V – control mice treated with the vehicle; C-T – control mice treated with Δ⁹-THC; DISC1-V – aDN-DISC1 mice treated with vehicle; DISC1-T - aDN-DISC1 mice treated with Δ⁹-THC).

A- Representative images of GAD boutons intensity in pyramidal neurons of the CA3 area of the hippocampus; note decreased intensity in the DISC1+T group

B- Representative images of GAD boutons intensity in pyramidal neurons of the CA1 area of the hippocampus; note comparable intensity in all groups

C- Quantitative analyses of the intensity of GAD⁺PV⁺ presynaptic boutons around pyramidal neurons of the CA3 area of the hippocampus; n=3 sections per mouse, 3 mice per group; each data point represents one mouse. Two-way ANOVA of the intensity data revealed a significant effect of the group, F(1,8)=12.06, p=0.008 and significant effect of the Δ⁹-THC, F(1,8)=19.16, p=0.002; Fisher LSD post-hoc test showed that aDN-DISC1-Δ⁹-THC mice were significantly different from both C-T and DISC1-V mice, * p<0.05.

D- Quantitative analyses of the intensity of GAD⁺PV⁺ presynaptic boutons around pyramidal neurons of the CA1 area of the hippocampus. Two-way ANOVA of the intensity data revealed no significant effects of the DN-DISC1, F(1,8)=0.075, p=0.791, no significant effects of Δ⁹-THC, F(1,8)=2.129, p=0.183 and no DN-DISC1 by Δ⁹-THC treatment interaction, F(1,8)=0.347, p=0.572; N=3 sections per mouse, 3 mice per group; each data point represents one mouse.

Blockade of NF-kB-COX-2 activation prevents cognitive impairment

Since our results had suggested synergistic elevation of the pro-inflammatory NF-kBCOX-2 signaling in aDN-DISC1 treated with Δ⁹-THC, we hypothesized that inhibition of COX-2 activity concurrently with Δ⁹-THC injections during adolescence may prevent the development of cognitive impairment in adult mice. In order to test this prediction, we assessed the effects of COX-2 inhibition (NS398, 10mg/kg, daily S.C. injections 30 min before Δ⁹-THC injections) on different types of recognition memory in control and aDN-DISC1 treated with Δ⁹-THC at adolescence (Fig.6). We found that NS398 prevented the development of memory deficits in aDN-DISC1 mice treated with Δ⁹-THC (the significant aDN-DISC1 by NS398 interaction for Y maze, F(1,29)=5.03, p=0.033; for NORT, F(1,29)=8.99, p=0.006; and for NPRT, F(1,29)=19.81, p<0.001). No effects of NS398 were found on locomotor activity in mice in the Y maze (Fig.S15 in Supplement 1).

Figure 6. Rescuing the memory deficits with the COX-2 inhibitor.

A- Schematic diagram of the treatment protocol;

On all data panels: the Y-axes depict the preference (%); the X-axes depict the experimental groups: C-V/T– control mice treated with vehicle and Δ⁹-THC (N=7); C-N/T – control mice treated with the selective COX-2 inhibitor (NS398) and Δ⁹-THC (N=9); DISC1-V/T – aDN-DISC1 mice treated with vehicle and Δ⁹-THC (N=8); DISC1-N/T -aDN-DISC1 mice treated with the NS398 and Δ⁹-THC (N=9). NS398 (10 mg/kg SC injections) was administrated daily 30 minutes prior Δ⁹-THC injections (10 mg/kg SC injections).

B- Spatial recognition memory in the Y maze. Significantly decreased preference for the previously blocked arm in aDN-DISC1 mice treated with Δ⁹-THC was significantly restored by NS398 co-treatment. Two-way ANOVA of the preference data revealed a significant effect of aDN-DISC1, F(1,29)=4.68, p=0.039; no effect of NS398, F(1,29)=9.26, p=0.205 and the significant aDN-DISC1 by NS398 interaction, F(1,29)=5.03, p=0.033; Fisher LSD post-hoc test showed that aDN-DISC1 significantly reduced preference for the previously blocked arm in Δ⁹-THC-treated mice (DISC1-V/T vs. C-V/T, p =0.004) and NS398 co-treatment significantly increased this preference (DISC1-V/T vs. DISC1-N/T, p =0.007); ** - denotes p<0.01; N=7–9 per group.

C- Novel object recognition test (NORT). Significantly decreased preference for the novel object in aDN-DISC1 mice treated with Δ⁹-THC was significantly restored by NS398 co-treatment. Two-way ANOVA of the preference data revealed a significant effect of aDN-DISC1, F(1,29)=11.64, p=0.002; NS398, F(1,29)=8.61, p=0.006 and the significant aDN-DISC1 by NS398 interaction, F(1,29)=8.99, p=0.006; Fisher LSD post-hoc test showed that aDN-DISC1 significantly reduced preference for the novel object in Δ⁹-THC-treated mice (DISC1-V/T vs. C-V/T, p <0.001) and NS398 co-treatment significantly increased this preference (DISC1-V/T vs. DISC1-N/T, p<0.001); ** - denotes p<0.01; N=7–9 per group.

D- Novel place recognition test (NPRT). Significantly decreased preference for the novel place of one of two identical objects in aDN-DISC1 mice treated with Δ⁹-THC was significantly restored by NS398 co-treatment. Two-way ANOVA of the preference data revealed no effect of aDN-DISC1, F(1,29)=2.05, p=0.163; no effect of NS398, F(1,29)=2.89, p=0.100 and the significant aDN-DISC1 by NS398 interaction, F(1,29)=19.81, p<0.001; Fisher LSD post-hoc test showed that aDN-DISC1 significantly reduced preference for the novel object in Δ⁹-THC-treated mice (DISC1-V/T vs. C-V/T, p <0.001) and NS398 co-treatment significantly increased this preference (DISC1-V/T vs. DISC1-N/T, p=0.043); * - denotes p<0.05, ** - denotes p<0.01; N=7–9 per group.

We also evaluated whether the COX-2 inhibition would prevent elevated secretion of glutamate by primary DN-DISC1 astrocytes following stimulation with Δ⁹-THC. We found that NS398 reversed increased secretion of glutamate by primary DN-DISC1 astrocytes treated Δ⁹-THC (Fig.S16 in Supplement 1). These results suggest convergence of effects of DN-DISC1 and Δ⁹-THC on NF-kB-COX-2 signaling in astrocytes, leading to increased production of glutamate by astrocytes.

Discussion

We report that inducible expression of dominant-negative Disrupted-In-Schizophrenia-1 (DN-DISC1) in astrocytes but not neurons or knockdown of endogenous Disc1 in hippocampal astrocytes interact with adolescent Δ⁹-THC exposure to impair recognition memory in adult mice. The present findings suggest that DN-DISC1 and Δ⁹-THC synergistically activate the NF-kB-COX-2 pathway in astrocytes, leading to increased secretion of glutamate and decreased immunoreactivity of parvalbumin-positive pre-synaptic boutons around pyramidal neurons of the CA3 area of the hippocampus. Deficient recognition memory could be prevented with the COX-2 inhibitor. Our data demonstrate that astrocyte genetic risk factors can exacerbate cognitive effects of adolescent cannabis use and indicate a putative target for preventive treatment.

Adolescent not adult exposure to Δ⁹-THC was required for the development of deficient recognition memory in adult mice with expression of DN-DISC1 in astrocytes. These results are consistent with other pre-clinical reports on effects of adolescent exposure to cannabinoids and resulting cognitive impairments (3, 72–74). Lack of effects of Δ⁹-THC in aDN-DISC1 mice on fear conditioning is in line with the unaltered performance in the Morris water maze (75–77) or aversive memory tasks in mice following adolescent treatment with cannabinoids (2, 72, 78–80). This selectivity in cognitive effects of cannabinoids could be related to differential distribution of CNR1 in the neural circuits underlying various cognitive tasks.

The current work is congruent with human studies that demonstrate that cannabis use during adolescence could have lasting effects on cognition (2, 72, 80–82) that is likely related to continuing maturation of the brain in general (83–86) and cannabinoid receptors in particular (87–89). Our results are consistent with human studies that adolescent cannabis use tend to affect working memory (WM) in adulthood (90, 91), particularly spatial processing that is dependent on the integrity of the hippocampus (92, 93). While human studies suggest an association (94–96), animal models enable us to establish a causal relationship and neurobiological mechanisms. In this context, our study significantly extend the existing literature on effects of cannabinoids on spatial of WM as evaluated in rodents with spatial recognition tests (97).

Our data clearly demonstrate that expression of the same risk factor in different brain cells types produces differential neurobehavioral outcomes in mice treated with Δ⁹-THC. Astrocyte but not neuronal expression of DN-DISC1 interacts with adolescent Δ⁹-THC to lead to recognition memory impairment in adult mice. In contrast, neuronal expression of DN-DISC1 and Δ⁹-THC treatment seem to have greater effects on fear conditioning, consistent with our prior studies with constitutive DN-DISC1 model (18)

The effects on recognition memory in astrocyte DN-DISC1 mice are unlikely dependent on early developmental effects of DN-DISC1 as turning off expression of DN-DISC1 after P21 completely eliminates the cognitive effects observed in our model. This appears in line with our prior reports on differential effects of DN-DISC1 on various behaviors depending on time this risk factor was expressed in neurons or astrocytes (98, 99). In addition, our data with DOX manipulation suggests that expression of DN-DISC1 during adolescent exposure to Δ⁹-THC is critical for the cognitive effects observed in our model. However, one cannot completely rule out the potential effects of DOX itself on neuroinflammatory processes in astrocytes that may have contributed to the preventive effects of DOX treatment.

GFAP-tTA;DN-DISC1 model has some limitations that make identification of underlying neural circuits mechanisms challenging. The GFAP promoter is active in the hippocampus and subcortical regions (100), and in addition to astrocytes, it is active in progenitor cells of the dentate gyrus of the hippocampus and the olfactory bulbs (101–103). Our findings with the viral knockdown of Disc1 in the CA areas were designed to address these limitations and suggest that the bulk of cognitive effects observed in DN-DISC1 mice treated with Δ⁹-THC are related to altered expression of Disc1 in hippocampal astrocytes. Additionally, the similar outcomes of Disc1 KD and DN-DISC1 suggest that the observed behavioral outcomes are likely due to altered expression of endogenous Disc1 rather than “off-target” or so-called gain-of-function effects of DN-DISC1 (30).

Although previous research and our current work clearly indicate that adolescent cannabis exposure can produce long-lasting behavioral and cognitive problems, there has been no direct comparison made between cognitive effects of cannabis and other psychoactive drugs. Indeed, there are numerous reports on long-term effects of psychostimulants used during adolescence (104–107). We found that chronic treatment with amphetamine of DN-DISC1 mice did not replicate the phenotypes produced by Δ⁹-THC, suggesting some selectivity in behavioral outcomes of adolescence exposure to cannabinoids vs. psychostimulants. Future studies will need to perform a more comprehensive dose-dependent comparative analysis.

The majority of pre-clinical research on cannabis has focused on GABA or glutamatergic neurons (108–113). However, there is a growing appreciation that glial cells also contribute to the detrimental behavioral effects associated with cannabis (2–4) as glial cells also express CNR1 and other factors of the endocannabinoid system (114). A recent study has shown that deletion of Cnr1 in mouse astrocytes prevents acute effects of Δ⁹-THC on spatial working memory and long-term depression (LTD) at hippocampal CA3-CA1 synapses. Critically, abolition of the same receptor on GABA or glutamate neurons does not lead to the same rescue phenomenon, suggesting that deficits in working memory triggered by acute administration of Δ⁹-THC could be due to the activation of CNR1 signaling in astrocytes (9). In a further support of the major role of astrocytes in the mechanisms of cognitive impairment following Δ⁹-THC exposure, another study has demonstrated that chronic Δ⁹-THC triggers a sustained activation of COX-2 and increased production of prostaglandin E2 (PGE2) in the brain. The activation of this signaling mechanism is initiated via CNR1-coupled G protein βγ subunits (10). However, astrocytes also express CNR2 (115, 116). Thus, it is conceivable that at least some of the cognitive effects of THC may have been mediated by CNR2. Future research will address this critical question.

There are significant variations in response to cannabis among users, suggesting genetic disposition (78, 117–119). Consistent with human findings, pre-clinical studies with mouse models carrying mutations in Neuregulin 1 (NRG1), COMT or DISC1 have shown that the effects of Δ⁹-THC on adolescent or adult mutant mice can dramatically differ from those on control littermates (15, 120). However, the neurobiological and molecular underpinnings of how genetic variants could moderate effects of Δ⁹-THC remain poorly understood (16, 121, 122). Moreover, there are very few if any studies of the molecular mechanism of cell type-specific GxE that mediate adverse effects of environmental risk factors, including cannabis. We have studied the neurobiological mechanisms of gene-environment interaction relevant to major psychiatric conditions using a rare mutation of a neurodevelopmental risk factor, DN-DISC1(123) as an experimental genetic tool to identify the molecular mechanisms whereby DN-DISC1 in astrocytes influences the signaling pathways activated by Δ⁹-THC. Based on the results of an unbiased RNA-seq analysis and prior studies (9), we identified the neuroinflammatory signaling in astrocytes that appears to be a convergent target for DN-DISC1 and inflammatory factors up-regulated by Δ⁹-THC. Specifically, we found that DN-DISC1 and Δ⁹-THC synergistically activate NF-kB-COX-2 signaling that might lead to increased secretion of glutamate by astrocytes. In order to test this molecular hypothesis, we inhibited activation of COX-2 with the selective inhibitor and were able to prevent the development of cognitive deficits in aDN-DISC1 mice treated with Δ⁹-THC. We believe this pharmacological approach could be applied to other GxE rodent models with the goal to use COX-2 inhibitors to counteract and/or ameliorate psychosis-like behavioral alterations associated with neuroinflammatory conditions produced by several environmental factors, including chronic THC exposure during adolescence. This would be congruent with several studies that demonstrated that ad-on treatment with COX-2 inhibitors had some beneficial anti-psychotic and cognitive effects (124–129). However, given cell and regional heterogeneity of the hippocampus, future studies will need to validate the above molecular events using isolated tissue astrocytes from different areas of the hippocampus.

Consistent with prior rodent studies (108–113), we also found synergetic adverse effects of Δ⁹-THC on the integrity of GABA neurons. Our findings indicate that the intensity of PV⁺ pre-synaptic boutons around pyramidal neurons of the CA3 area are predominantly affected, suggesting that inhibitory influence of PV⁺ cells in the hippocampus could be compromised in aDN-DISC1 mice treated with Δ⁹-THC, potentially leading to altered excitatory-inhibitory balance underlying aspects of cognitive dysfunction. In addition, in line with the recent publication, decreased PV⁺ could also lead to abnormal long-term depression (LTD) at hippocampal CA3-CA1 synapses (10). Future studies will address these possibilities in detail.

In conclusion, our work for the first time demonstrates that a genetic predisposition and adolescent Δ⁹-THC exposure could synergistically produce a sustained activation of NF-kB-COX-2 signaling in astrocytes. This leads to elevated secretion of glutamate, reduced immunoreactivity of parvalbumin-positive pre-synaptic boutons around pyramidal neurons of the CA3 area of the hippocampus and deficient memory. The observed cognitive deficits can be prevented with the COX-2 inhibitor, suggesting future targets for therapeutic interventions.