Investigation of Blood-Brain Barrier Disruption in an Animal Model of Mania Induced by d-Amphetamine

Background: Bipolar disorder (BD) is a recurrent chronic psychiatric disorder. Evidence indicates that many individuals with BD exhibit high serum levels of inammation and oxidative stress markers, which are further associated with mood symptoms and cognitive dysfunction. Due to the crosstalk between the periphery and central nervous system in BD, the disruption of the blood-brain barrier (BBB) has been proposed as a key mechanism of the pathophysiology of the disorder. This study aimed to investigate claudin-5 expression – a major protein of the BBB – in the brain of an animal model of mania induced by d-amphetamine (AMPH) and evaluate the effects of treatment with lithium. Results: AMPH-injected animals exhibited increased overall activity in the open eld test. In the serum, TBARS levels were augmented in the lithium-treated groups, regardless of AMPH injection, while TNFα was not detected. In the brain, TBARS and TNFα did not differ between groups but were positively and strongly correlated in the ST of AMPH-injected rats. Contrary to the primary hypothesis, AMPH and lithium injections did not affect brain claudin-5 protein levels. Conclusions: This is one of the rst attempts to investigate the effects of AMPH on BBB integrity. Although no evidence of BBB disruption was found in the current study, our results provide evidence and rationale for future research to elucidate the importance of such alteration in BD. Abbreviations: df, degrees of freedom; HIP, hippocampus; PFC, prefrontal cortex; ST, striatum; TBARS, thiobarbituric acid reactive substances; TNFα, tumor necrosis factor α. # AMPH+LI, n=5 these variables.


Background
The crosstalk between the periphery and the central nervous system (CNS) has implicated the blood-brain barrier (BBB) dysfunction in the pathophysiology of BD (Patel and Frey, 2015). The BBB consists of endothelial cells of the brain capillaries, astrocyte end-feet, and pericytes (Ballabh et al., 2004). The presence of tight and adherens junctions between the endothelial cells contributes to the tightness of the BBB and its selective function (Banks, 2015). As most blood-borne substances are not allowed to enter the brain under physiological conditions, BBB integrity is essential for CNS homeostasis (Segarra et al., 2021). Evidence indicates that in ammatory cytokines such as tumour necrosis factor (TNF) α and interferon (IFN) γ (Capaldo and Nusrat, 2009), and oxidative stress (Sajja et al., 2016) are detrimental and contribute to BBB disruption. Recently, extensive BBB leakage has been associated with a more severe and chronic course of BD (Kamintsky et al., 2019). Still, research on the potential involvement of the BBB in the pathophysiology of BD is very scant.
Moreover, it is suggested that lithium can prevent and reverse most of these alterations induced by AMPH and its analogues (Frey et al., 2006b; Macêdo et al., 2013;Valvassori et al., 2015). Considering that both in ammation and oxidative stress increase BBB permeability, we aimed to investigate if BBB disruption is observed in an AMPH-induced animal model of mania. Also, we sought to evaluate whether treatment with lithium reverses any BBB damage induced by AMPH. To date, only a few studies have evaluated, to some extent, BBB disruption in an animal model of mania (Valvassori et al., 2015) or other psychostimulants-induced models (Northrop and Yamamoto, 2012), while none has investigated a speci c marker, such as claudin-5.

Animals
Male Wistar rats (n=24, 2-month-old, 220-310 g) were purchased from Charles Rivers Laboratories (Massachusetts, USA). Animals were housed (n=2 per cage) at standard room temperature and 12-h inverse light/dark cycle, with free access to food and water. This study was approved by the institutional ethics committee from McMaster University (Animal Research Ethics Board, project #140828), and all experimental procedures were performed following national (Canadian Council on Animal Care, 2020) and international (National Research Council, US, 2011) ethical standards.

Open eld
On day 14, spontaneous locomotor behaviour was evaluated 2 h after the last injection of AMPH using the open eld. The apparatus consisted of a 60x40 cm open eld with 50-cm-high walls and divided into 12 equal rectangles. Each animal was placed in the center of the open eld and allowed to explore the apparatus for 5 min. All sessions were recorded using a webcam (C270 HD, Logitech), and a video tracking system was used to analyze the number of crossings, distance travelled (m), average speed (m/s) and time spent in the periphery (s).

Euthanasia and sample collection
After the behavioural task, rats were anesthetized with iso urane (4-5% with oxygen at 1-2 L/min) and decapitated, and the brain was removed as quickly as possible. The right and left prefrontal cortex (PFC), hippocampus (HIP) and striatum (ST) were dissected and snap frozen in dry ice. The troncular blood was also collected and centrifuged (2057g, 10 min, at room temperature). Finally, brain structures and serum were stored at -80°C until sample preparation.

Sample preparation
Right-and left-brain structures were prepared differently for biochemical analyses and western blot. Brain tissue was homogenized using phosphate-buffered saline (PBS) for biochemical assays or RIPA buffer 1x (20-188, Merck Millipore) for western blot at 1:4 (w/v). Both lysis buffers were prepared with EDTA-free protease inhibitor cocktail (11836170001, Roche) according to the manufacturer's instructions. Then, samples were centrifuged (10,000g, 5 min, 5°C), and supernatants were collected and kept at -80°C until further analysis.
2.6. Biochemical assays 2.6.1. Total protein The total protein was determined in the brain structures using DC Protein Assay Reagent (5000116, BioRad) according to the manufacturer's instructions. Brie y, samples were diluted 1:21 (v/v) in PBS and, after the addition of proper reagents, the total protein content in each sample was quanti ed in a spectrophotometer at 750 nm.

TNFα
TNFα levels were quanti ed using a sandwich ELISA kit (KRC3011, Invitrogen TM ) following the manufacturer's instructions. Serum samples were not diluted, while all brain structures homogenates were diluted at a 1:10 (v/v) ratio.

Thiobarbituric acid reactive substances (TBARS)
TBARS levels were detected and quanti ed using a colorimetric assay kit (KGE013, R&D Systems) according to the manufacturer's instructions. Serum samples were diluted at 1:2 (v/v) ratio, PFC and ST homogenates were diluted at 1:20 (v/v) and HIP at 1:20 (v/v) ratios.

Western blot
Brie y, homogenates samples of PFC, HIP, and ST were further diluted 1:1 (v/v) in 2x Laemmli sample buffer (1610737, BioRad Laboratories, Inc). After that, 20 µg of each sample was loaded in 4-20% precast and stain-free mini gels (4568094, BioRad Laboratories, Inc). A protein marker (1610373, BioRad Laboratories, Inc) was also loaded in the gels. Electrophoresis was performed at 100 V for 1 h 15 min. The protein transfer was performed using the Trans-Blot® Turbo™ Transfer System (BioRad Laboratories, Inc) to an LF-PVDF membrane (0.45 µm pore size, BioRad Laboratories, Inc) at 25 V for 7 min. The total protein and transfer were veri ed using the ChemiDoc™ Imaging Systems (BioRad Laboratories, Inc); no speci c staining was used. Blocking was performed using skim milk powder 5% (diluted in TBS-T) for 1h at room temperature. Primary (Claudin 5 Monoclonal Antibody, 35-2500, Thermo Fisher) and secondary (Peroxidase-A niPure Goat Anti-Mouse IgG (H+L), 115-035-003, Jackson Immuno Research) antibodies were diluted in TBS-T at 1:500 and 1:10,000, respectively. β-actin was the loading control (β-Actin Loading Control Monoclonal Antibody, MA5-15739, Thermo Fisher; same secondary antibody as previous), and antibodies were diluted in TBS-T at 1:10,000. Clarity Western ECL Substrate (1705060, BioRad Laboratories, Inc) was used for detection. Total protein was used for loading control as described previously (Taylor et al., 2013), and claudin-5 levels were normalized by the control group levels (i.e., SAL+SAL).

Statistical analysis
Shapiro-Wilk and Levene's tests were used to evaluate the normality of distribution and homogeneity of variance, respectively. Secondly, two-way ANOVA was performed, considering model (saline and AMPH) and treatment (saline and lithium) as independent factors, followed by Tukey posthoc analysis, if ANOVA is signi cant. Outliers were identi ed using the Grubbs' test (α=0.05) and excluded from the analysis (n=5). All p<0.05 were considered statistically signi cant.

AMPH induced hyperactivity
AMPH-injected rats presented overall hyperactivity in the open eld, which was indicated by an increased number of crossings, average speed, and distance travelled (Figure 1, A-C). Two-way ANOVA indicated a main effect of the model for all variables, with no main effects for treatment and interaction of factors (Table 1). Also, no differences were observed for time spent in the periphery ( Figure 1D, Table  1).

Effect of AMPH on serum and brain levels of TNFα and TBARS
In the serum, TNFα levels were below the limit of detection across all groups. In the brain regions analyzed, AMPH did not induce in ammation dependent on TNFα levels. Consequently, no effect of lithium was observed. Two-way ANOVA did not indicate main effects of model, treatment, and interaction for these variables (Figure 2A-D, Table 2).
TBARS levels in the serum and brain structures remained unchanged following AMPH injection ( Figure   2E-H). However, lithium treatment increased lipid peroxidation in the serum, which was indicated by treatment effect, but no main effects for model and interaction of factors were observed following twoway ANOVA ( Figure 2E, Table 2). In the brain, there were no main effects of model, treatment, and interaction (Table 2).
Interestingly, a signi cant strong positive correlation was observed between TNFα and TBARS levels in the ST of the AMPH+SAL (r=0.87, p<0.001) and AMPH+LI (r=0.96, p=0.002) groups. This correlation was weaker in the SAL+SAL group (r=0.80, p=0.056) and was negative in SAL+LI rats (r=-0.89, p=0.017)possibly due to the effect of lithium on TBARS levels.

Claudin-5 protein levels in the PFC, HIP and ST of AMPH-injected rats
Contrary to our hypothesis, the levels of claudin-5 in the brain remained unchanged following AMPH and lithium injection. Two-way ANOVA did not indicate main effects for the model, treatment, or interaction ( Figure 2I-L, Table 2).

Discussion
To our knowledge, this is the rst study that investigated BBB disruption in an animal model of mania induced by AMPH. Corroborating with previous studies, AMPH-injected rats exhibited hyperactivity, which was determined by an increased frequency of crossings and distance travelled. Although our model showed face validity, no changes in peripheral and CNS levels of TNFα and TBARS were observed following AMPH and lithium injection. Consequently, protein levels of claudin-5, the most enriched tight junction protein in the BBB, also remained unchanged in the brain regions analyzed.
Psychostimulant-induced animal model of mania, such as the AMPH model, is frequently used to investigate biological mechanisms and alterations that have already been described in BD (Kara and Einat, 2013; Sharma et al., 2016). Clinical studies often report an increase in in ammatory and oxidative stress parameters (Rowland et al., 2018), and most preclinical evidence is consistent with such ndings.
For instance, Valvassori et al. (2015) have shown that AMPH injections resulted in a pro-in ammatory effect. Injected rats presented with an augment of IL-4, IL-6, IL-10, and TNFα in the PFC, ST, and serum, which were restored to control levels following treatment with lithium. However, the authors did not nd signi cant alterations in the HIP and cerebrospinal uid (CSF). Other psychostimulant drugs, such as methylphenidate and methamphetamine, also seem to increase in ammatory markers in the HIP of rats (Beirami et al., 2017;Motaghinejad et al., 2017). However, no changes in TNFα levels were found in the serum and brain regions analyzed in our model. Still, no differences in in ammatory cytokines levelssuch as TNFα, IL-1β, and IL-10 -have been previously described in rodents injected with AMPH or its derivates (Bristot et al., 2019;Gubert et al., 2016).
Classically, AMPH is responsible for enhancing dopamine (DA) release by inhibiting its reuptake, promoting reverse transport of DA into the synaptic cleft independent of stimulus and releasing DA from synaptic vesicles in the cytoplasm (Calipari and Ferris, 2013). These are the primary mechanisms involved in AMPH-induced hyperactivity and its neurotoxic effects (Valvassori et al., 2021). If not stored in synaptic vesicles, cytoplasmatic DA has a highly autoxidative capacity that can impair mitochondrial function and increase oxidative stress resulting in cell death (Brown and Yamamoto, 2003;Yamamoto and Bankson, 2005). However, in the present study, lipid peroxidation was not augmented following AMPH injection, given by TBARS levels. TBARS levels were higher in the ST, a brain region with many dopaminergic projections, but no statistical difference between groups was found. Still, a strong positive correlation between TBARS and TNFα levels was observed in this same brain region in AMPH-injected groups, regardless of treatment with lithium. Overall, ndings regarding oxidative stress in AMPH models vary in the literature. More acute protocols or higher doses seem to be more likely to promote such Oxidative stress is an important promoter and product of the in ammatory response (Biswas, 2016), and it is supposed to underly AMPH-induced in ammation. As oxidative stress and in ammation are predictive of promoting BBB disruption and increasing its permeability (Patel and Frey, 2015), we hypothesized that AMPH would exert a deleterious effect in the BBB by downregulating claudin-5 expression. However, no signi cant changes were observed on claudin-5 levels in the PFC, ST and HIP of rats after AMPH and lithium injections. A previous study showing increased levels of in ammatory cytokines in the brain parenchyma but no alteration in the CSF suggested that AMPH-injected rats might not present with disrupted BBB (Valvassori et al., 2015). However, in vivo and in vitro studies have already reported that methamphetamine and other psychostimulant drugs can disturb BBB integrity (Kousik et al., 2012;Northrop and Yamamoto, 2012).
Besides in ammation and oxidative stress, associated mechanisms have been proposed to underlie BBB disruption. For instance, the activation of in ammatory pathways, such as nuclear factor kappa B (NFκB), can result in the ampli cation of a large array of genes involved in in ammation, including matrix metalloproteinases (MMPs) (Hurtado-Alvarado et al., 2016). MMPs are enzymes that degrade tight junctions, such as claudin-5, which are essential for maintaining BBB properties (Rempe et al., 2016). Methamphetamine has already been described to upregulate MMPs expression (Mizoguchi et al., 2007). Additionally, NFκB pathway activation following TNFα and IL-15 signalling may also be responsible for the downregulating the expression of tight junction proteins (e.g., claudin-2) in vitro (Stone et al., 2011). Therefore, it is relevant to investigate other pathways in BBB disruption and their potential association with BD and other psychiatric disorders.
Although BBB disruption has been implicated in the pathophysiology of psychiatric disorders (including BD; Greene et al., 2020), clinical and preclinical evidence is still scarce. A recent imaging study has shown that, among individuals with BD, only a sub-group exhibited an extensive BBB leakage that signi cantly differed from controls (Kamintsky et al., 2019). Interestingly, this sub-group of patients have a more chronic course of BD, with more severe symptoms of depression and anxiety. Other studies have evaluated biomarkers in the CSF and described higher levels of catecholamine and serotonin metabolites and in ammatory markers, such as IL-8, in individuals with BD (Isgren et al., 2015;Knorr et al., 2018). Furthermore, increased IL-8 levels in the CSF were associated with lithium treatment. Since its discovery, lithium has remained the rst-line therapeutic choice for BD treatment (Yatham et al., 2018), but its effects on in ammation have yet to be fully elucidated. During euthymia, individuals with BD treated with lithium exhibited increased levels of TNFα and IL-4 compared to unmedicated patients (Guloksuz et al., 2010), which has also been described in vitro (Liu et al., 2011); but there is also evidence to show otherwise (Fernandes et al., 2019;Knijff et al., 2007). It is worth mentioning that augmented peripheral TNFα levels were further associated with poor response to lithium treatment in BD (Guloksuz et al., 2012).
Despite its novelty, some limitations should be addressed in our study. First, we did not explore more dynamic markers of the interface integrity between the blood, brain, and CSF, such as Evans Blue, which could be helpful as a rst screening. Second, plasmatic levels of lithium were not assessed, but it has been described that therapeutic levels are reached following this protocol (Frey et al., 2006). Third, while only a few parameters were analyzed, their relevance to BD and the rationale for their role in the BBB disruption have been discussed. Although an ideal animal model for BD has not been developed, AMPH injection in rodents remains an established animal model of mania with good construct, face and predictive validity (Sharma et al., 2016). It should be noted that the latter has been questioned (Lan and Einat, 2019). Although this model may not mimic the vast complexity of BD pathophysiology, it would be advantageous to identify novel animal models that allow the evaluation of BBB disruption.
In summary, in one of the rst attempts to investigate the effects of AMPH on BBB integrity, we did not nd evidence that AMPH or lithium impact brain levels of claudin-5. It is only recently that clinical research has provided BBB disruption as a marker of progression in BD (Kamintsky et al., 2019). Still, our results provide evidence and rationale for future research to establish the best approach to model and better understand this relatively novel pathophysiological mechanism implicated in BD.    kDa. Two-way ANOVA, data expressed by mean±SEM.