Male Wistar rats weighing 220-240 g were obtained from Institute for Cognitive Sciences studies (ICSS), Tehran, Iran. All animals were maintained five per Plexiglas cage in a room. The room was kept at a stable temperature (22 ± 2 °C), humidity (40-60 %) and a 12/12 light-dark cycle, lights on at 7:00 a.m. Rats were allowed to get acclimatized to the lab environment for 7 days before starting experiments. Also they had ad libitum access to food and water except during the experiments. All experimental schemes were carried out during the light phase between 8:00 and 17:00. We tried to reduce the number of animals used and their suffering. The experimental protocol in this study was approved by the Research and Ethics Committee of the School of Advanced Technologies in Medicine, Tehran University of Medical Sciences and were done in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals (NIH publications No. 80-23).

All animals were anesthetized by intraperitoneal injection of a ketamine and xylazine (100 and 10 mg/kg, respectively) mixture. After administration, rats were placed in a stereotaxic apparatus (Stoelting Co., Wood Dale, IL, USA). Two steel guide cannulae (22-guage) were bilaterally implanted into the CA1 region of the dorsal hippocampus (AP: -3.3 mm from bregma; ML: ± 2 mm from midline; DV: 2.8 mm from the skull surface) according to Paxinos and Watson's atlas (Paxinos and Watson, 2007[61]). The guide cannulae were secured by jeweler's screws, and the incision was closed with rapidly polymerizing dental acrylic cement. After surgery, all animals were allowed to recover for 7 days. This period was necessary for rats to getting clear from the anesthetic effects.

Lithium chloride (Daroopakhsh, Tehran, Iran) was dissolved in normal saline and administrated intraperitoneally (i.p.). The selective CB1r antagonist, AM251, and the selective CB1r agonist, ACPA, were stored at -14 °C. Doses of 1, 10 and 100 ng of AM251 and doses of 0.001, 0.01 and 1 µg of ACPA were dissolved in anhydrous ethanol and were diluted by normal saline to achieve their required doses (Khodayar et al., 2016[33]; Najar et al., 2015[54]). The bilateral microinjections of AM251 and ACPA into the CA1 region (intra-CA1) were done by using a 2 µl Hamilton syringe. This syringe was connected to an injection needle (27 guage) by a polyethylene tube. The injection needle was inserted 1 mm beyond the tip of the guide cannula. Infusions were delivered in volume of 0.5 µl during 60 s into each side.

Repeated measure analysis on spatial learning showed that for both escape latency and traveled distance, the main effect of dose [(F_3,28 = 32.83, P < 0.001) and (F_3,28 = 39.37, P < 0.001), respectively], the effect of trial [(F_3.87,108.44 = 409.32, P < 0.001) and (F_4.51,126.31 = 426.98, P < 0.001), respectively] and interaction effect [(F_11.62,108.44 = 3.55, P < 0.001) and (F_13.53,126.31 = 5.22, P < 0.001), respectively] were significant. Post hoc comparisons demonstrated that ACPA at all administered doses increased escape latency and traveled distance (trials 3 to 8). While, ACPA increased both only at doses of 0.01 and 1 µg/ rat (trial 2).

Additionally, repeated measure analysis on swimming speed showed that the main effect of dose (F_3,28 = 5.74, P < 0.01), the effect of trial (F_4.90,137.28 = 65.52, P < 0.001) and interaction effect (F_14.71,137.28 = 4.25, P < 0.001) were significant. Post hoc comparisons demonstrated that ACPA at all administered doses decreased swimming speed (trials 3 and 4). Additional analyses indicated that ACPA decreased swimming speed only at dose of 0.01 µg/rat (trial 5), ACPA decreased swimming speed only at doses of 0.001 and 0.01 µg/rat (trial 7), and finally that ACPA increased swimming speed only at doses of 0.01 and 1 µg/rat (trial 6) (Figure 1(Fig. 1), see also Supplementary data).

Repeated measure analysis on spatial learning showed that for both escape latency and traveled distance, the main effect of dose [(F_3,28 = 75.27, P < 0.001) and (F_3,28 = 75.13, P < 0.001), respectively], the effect of trial [(F_4.12,115.45 = 370.69, P < 0.001) and (F_4.34,121.56 = 328.35, P < 0.001), respectively] and interaction effect [(F_12.37,115.45 = 1.92, P < 0.05) and (F_13.02,121.56 = 2.38, P < 0.01), respectively] were significant. Post hoc comparisons demonstrated that AM251 at all administered doses increased escape latency and traveled distance (trials 3 to 8), and further that AM251 increased escape latency only at doses of 10 and 100 ng/rat (trial 2).

Also repeated measure analysis on swimming speed showed that the main effect of dose (F_3,28 = 19.14, P < 0.001), the effect of trial (F_4.73,132.44 = 38.51, P < 0.001) and interaction effect (F_14.19,132.44 = 3.20, P < 0.001) were significant. Post hoc comparisons demonstrated that AM251 at all administered doses decreased swimming speed (trials 3, 4 and 7). While, AM251 decreased swimming speed only at doses of 10 and 100 ng/rat (trial 2). Also AM251 decreased swimming speed only at dose of 10 ng/rat (trial 8). But AM251 increased swimming speed at doses of 1 and 100 ng/rat (trial 6) (Figure 2(Fig. 2), see also Supplementary data).

In all experimental groups, lithium was not associated with impairment on spatial learning and acquisition (Figure 3A, B(Fig. 3)); however, lithium was associated with impaired spatial memory retrieval (Figure 4(Fig. 4)), and it decreased swimming speed in some trials (Figure 3C(Fig. 3)). Lithium is used for long-term treatment of bipolar disorder, and has a powerful effect for the prevention of manic/depressive recurrences (Geddes et al., 2010[22]). The clear mechanism of lithium remains unknown (Misztal et al., 2017[52]), but lithium can inhibit GSK3-β by increasing phosphorylation on the key serine residue (Ser9) (Jope, 2003[30]; Klein and Melton, 1996[35]). GSK3-β inhibition induced by lithium is an important contribution to its mood stabilizing therapeutic effects (Jope, 2011[29]). Glycogen Synthase Kinase (GSK), is a kinase that phosphorylates and inactivates glycogen synthase (Embi et al., 1980[19]), and it has important roles in several cell functions (Jope and Johnson, 2004[31]). The activity of GSK3 is regulated by phosphorylation on a regulatory serine (serine-21 in GSK3-α and serine-9 in GSK3-β) and is inhibited by phosphorylation of these regulatory serines. Many of the GSK3-β targets are involved in pathways which are related to cell growth, apoptosis, learning and memory (Kaidanovich-Beilin and Woodgett, 2011[32]). Although lithium has many therapeutic effects on learning, memory and neurological disorders such as Down syndrome, Fragile X syndrome and Alzheimer's disease (AD) (Contestabile et al., 2013[10]; Liu and Smith, 2014[44]; Ly et al., 2013[45]; Nunes et al., 2007[56]), there are important side effects that must be considered including the inducement of dementia and other cognitive impairments (Dunn et al., 2005[17]; Weingartner et al., 1985[80]; Wingo et al., 2009[81]). Administration of lithium and other GSK3-β inhibitors may have impairment effect on memory and other cognitive performances (Stip et al., 2000[73]; Wingo et al., 2009[81]). However, the effect of GSK3-β inhibition on learning and memory is still inconsistent (Chew et al., 2015[8]). A previous study showed that GSK3-β was inhibited during Long-Term Potentiation (LTP), but was activated during Long-Term Depression (LTD) (Peineau et al., 2007[62]). Also, GSK3-β may be involved in memory reconsolidation in the hippocampus, because the genetic reduction and pharmacological inhibition of GSK3-β impaired reconsolidation of hippocampus-dependent place memory. However, the mechanism of memory formation and consolidation in the hippocampus may involve consonant activities in the other brain regions (Kimura et al., 2008[34]). On the other hand, glutamate, an abundant excitatory neurotransmitter in the brain, activates various receptors involved in synaptic plasticity, learning and memory (Collingridge and Bliss, 1995[9]). The current study showed that N-methyl-D-aspartate (NMDA) receptors have a role in spatial learning and memory, indicating that administration of D-AP5 (NMDA receptor antagonist) induced spatial memory impairment, while administration of NMDA facilitated spatial learning (Parsaei et al., 2016[60]). Chronic lithium treatment increases glutamate synaptosomal uptake and decreases hippocampal GluR1 expression in the hippocampus, which has beneficial effects on treatment of mood disorders (Du et al., 2004[16]; Hashimoto et al., 2002[25]). Also, it showed that lithium enhances VGLUT1 (necessary for glutamate uptake) mRNA expression in the cerebral cortex neurons (Moutsimilli et al., 2005[53]). These effects of lithium on the balance of glutamate and GSK3-β activity may lead to the impairment of spatial memory retrieval. Specifics of our experimental findings are outlined below.

Cannabinoid drugs, ACPA (CB1r agonist) and AM251 (CB1r antagonist), increased escape latency and traveled distance to the platform, suggesting a spatial learning impairment (Figure 1A, B(Fig. 1), Figure 2A, B(Fig. 2), respectively). In the probe test, ACPA impaired spatial memory retrieval, while AM251 did not impair (Figure 4(Fig. 4)). Also, both ACPA and AM251 decreased swimming speed, however, increased swimming speed in one trial (trial 6) (Figure 1C(Fig. 1) and Figure 2C(Fig. 2), respectively). Previous studies showed that CB1r agonists, such as ACPA, induce cognitive impairments in rodents (Kruk-Slomka and Biala, 2016[36]; Kruk-Slomka et al., 2016[37]; Pamplona and Takahashi, 2006[58]), But CB1r antagonists, such as AM251, often enhance the performance of rodents in different memory tasks (Kruk-Slomka and Biala, 2016[36]; Lichtman, 2000[41]; Takahashi et al., 2005[76]). The effect of cannabinoids on locomotor activity is inconsistent. In some researches, cannabinoids increased locomotor activity and in some researches decreased it (Chaperon and Thiebot, 1999[7]; Sulcova et al., 1998[74]). Acute administration of Δ9-tetrahydrocannabinol (Δ9-THC) or CB1r agonist impairs acquisition of memory in rodents (Da and Takahashi, 2002[11]; Lichtman et al., 1995[42]). Cannabinoids reduce glutamatergic synaptic transmission in different brain regions which have important roles in memory-related functions (Azad et al., 2003[4]; Fujiwara and Egashira, 2004[21]; Robbe et al., 2001[67]). Generally, activation of CB1r in the hippocampus, inhibits synaptic transmission (Hoffman and Lupica, 2000[26]). CB1r activation reduces LTP and inhibits the release of glutamate in the hippocampus (Sullivan, 2000[75]). Cannabinoids have influence on the function of NMDA receptors via various ways, such as reduction of releasing presynaptic glutamate into the synaptic cleft (Li et al., 2011[39]). However, it has been described that AM251 and other CB1r antagonists increased extracellular glutamate (Xi et al., 2006[82]). Cannabinoids have modulation effect on hippocampal glutamatergic activation, because releasing of intracellular calcium which have mediated by NMDA receptors, reduced by CB1r activation (Hampson et al., 2011[24]). CB1r activation reduces synaptic processing in the hippocampus, and also, reduces flexibility and adaptation of neurons for successful memory encoding and retrieval by decreasing the modulation of intracellular calcium releasing in critical circumstances (Deadwyler et al., 2007[12]; Deadwyler and Hampson, 2008[13]). NMDA receptors are essential for brain development, synaptic plasticity and direct effect on learning and memory consolidation (Dingledine et al., 1999[15]). However, it is notable that the reducing effect of cannabinoids on glutamate releasing may occur by other mechanisms unrelated to the CB1r (Sanchez-Blazquez et al., 2014[68]). Also, cannabinoids such as THC, inactivate GSK3-β. CB1r mediates this effect, because blocking this receptor by CB1r antagonists abolish the phosphorylation of GSK3-β (Ozaita et al., 2007[57]). Spatial memory impairment induced by cannabinoids may occur for their reducing effect on glutamate releasing, and also, for their inhibiting effect on GSK3-β.

Following the existing literature, we designed some experiments to elucidate the consequences of interaction between lithium and cannabinoids on spatial memory in MWM apparatus. The results showed that in the trial session, administration of lower dose of ACPA (0.001 µg/rat) after injection of lithium (0.5 mg/kg) increased both escape latency and traveled distance during spatial learning (trial 3). Also, administration of lower dose of AM251 (1 ng/rat) after injection of lithium (1 mg/kg) increased only escape latency during spatial learning (trial 6) (Figure 3A, B(Fig. 3)). In the probe test, both lower doses of cannabinoid drugs restored the spatial memory impairment effect of the higher dose of lithium (5 mg/kg), while potentiated the spatial memory impairment effect of lower doses of lithium (0.5 and 1 mg/kg) (Figure 5(Fig. 5)). Interaction between ACPA and lithium (5 mg/kg) increased swimming speed only in trial 3, and interaction between AM251 and lithium (1 mg/kg) increased swimming speed only in trial 8 (Figure 3C(Fig. 3)). As a result, the data showed that lower doses of ACPA (0.001 µg/rat) and AM251 (1 ng/rat), have dual effect on lithium-induced spatial memory impairment in rats. Previous reports demonstrated that cannabinoids may have dual effects in different experiments (Li et al., 2013[40]; Sarne et al., 2011[69]). For example, some experiments showed that cannabinoids decrease body temperature, inhibit nociception and suppress motor activity (Ameri, 1999[2]; Chaperon and Thiebot, 1999[7]). But nevertheless, many experiments showed opposite results: hyperalgesia, increasing motor activity and elevating body temperature (Sulcova et al., 1998[74]; Taylor and Fennessy, 1977[77]). Previous studies showed that acute administration of cannabinoids has protective effects in different models of acute brain injuries. However, chronic activation of CB receptors has neurotoxic consequences in heavy cannabis users (Arnone et al., 2008[3]; Ehrenreich et al., 1999[18]; Matochik et al., 2005[50]; McHale and Hunt, 2008[51]; Solowij et al., 2002[71]). Also, previous study showed that the CB1r agonist, ACPA, has dual effect on memory in step-down passive avoidance learning task (Nasehi et al., 2015[55]). CB1r peptide agonists RVD and VD, impair memory processing in mice, but in Amyloid-beta₁-₄₂-treated mice, RVD and VD improve Aβ_1-42-induced memory impairment (Zhang et al., 2016[83]). Another research showed that Cannabidiol (CBD) has dual effect on lithium-induced vomiting, indicating that lower doses of CBD suppressed lithium-induced vomiting, while higher doses enhanced it (Parker et al., 2004[59]). In this research, both cannabinoid drugs, ACPA and AM251, had similar effect on lithium-induced spatial memory impairment. The similar effect of CB receptor agonists and antagonists has been revealed in some experiments before. For instance, in the test of unconditioned fear, both selective CB1r agonist, ACEA, and selective CB1r antagonist, AM251, increased anxiety-like behavior (Simone et al., 2015[70]). Micromolar concentrations of WIN55212-2 (non-selective CB1r agonist), WIN55212-3 (selective CB2r antagonist) and AM251 (selective CB1r antagonist) inhibit the uptake of dopamine in striatal synaptosomes. Also, both WIN55212-2 and AM251, displace the binding of the cocaine analogue, [3H]WIN35,428, indicating a direct interaction with the transporter (Price et al., 2007[65]). In another research, results showed that AM251 failed to reverse WIN-induced inhibition of dopamine uptake in neocortical synaptosomes in rats (Steffens and Feuerstein, 2004[72]). In a related study, results showed that SR141716A (Rimonabant), the CB1r antagonist, has an inhibitory effect on forskolin-stimulated cyclic AMP accumulation. This effect is similar to the effect of cannabinoid agonists, suggests it may be acting as a low intrinsic activity agonist, rather than a pure antagonist at CB1r (Holland et al., 1999[27]).