Implementation of combined antiretroviral therapy (cART) has significantly reduced the mortality associated with human immunodeficiency virus (HIV) infection and increased the life expectancy of people living with HIV (PLWH) (Günthard et al., 2016). However, the long-term use of these drugs has been reported to induce neurological and psychiatric adverse effects with variable frequency and severity (Abers et al., 2014; Bertrand et al., 2021). In particular, HIV-associated neurocognitive disorders (HAND) have become a major complication of HIV and great health concern (Animut et al., 2019; Wang et al., 2020). Depending on severity, HAND has been categorized in three stages: HIV-associated dementia (HAD), asymptomatic neurocognitive impairment (ANI), and minor neurocognitive disorder (MND). Although the incidence of HAD, characterized by severe cognitive impairment and motor dysfunction has remarkably declined since the implementation of cART, the milder forms such as ANI and MND continue to occur and are even on the rise (Elbirt et al., 2015; Clifford, 2017).

The blood-brain barrier (BBB) is the major physiological barrier that separates the brain parenchyma from the systemic circulation and plays a central role in maintaining central nervous system (CNS) homeostasis (Xu et al., 2019; Kadry et al., 2020). This barrier is mainly composed of microvascular endothelial cells (EC) that are continuously sealed by intercellular tight junction proteins (TJs) and adherent junction complexes (AJ) which strictly restrict the paracellular diffusion of small and larger molecules (Varatharaj & Galea, 2017; Bors & Erdö, 2019; Xu et al., 2019). TJ proteins primarily include: occludin (OCLN) and claudin-5 (CLDN-5) as transmembrane proteins, and zonula occludens -1 (ZO-1) as accessory protein for anchoring TJ protein complex to the actin cytoskeleton (Fanning et al., 2002; Greene et al., 2019). Pericytes and astrocytes foot processes which are embedded in the basement membrane of the EC cells are important in maintaining BBB function by providing structural and metabolic support to the microvasculature (Abbott et al., 2006; Armulik et al., 2010). In addition to the physical characteristics, the BBB displays very effective biochemical barrier properties by expressing multiple drug efflux transporters and metabolic enzymes (Dauchy et al., 2008). Previous data demonstrated that the brain microvascular endothelial cells express a variety of membrane transporters belonging to the ATP-binding cassette (ABC) and Solute Carrier Families (SLC). In particular, the efflux transporters, P-glycoprotein (P-gp) and Breast Cancer Resistant protein (BCRP) are robustly expressed at the apical membrane of brain capillary endothelial cells as “gatekeepers” of the BBB by preventing harmful substances from entering the brain parenchyma from the blood circulation (Lee et al., 2001; Bendayan et al., 2002). As glucose is the primary metabolic fuel for the mammalian brain, a continuous and highly regulated supply is critical to maintaining normal brain function (Shah et al., 2012). Glucose uptake into the brain is primarily mediated by the facilitative glucose transporter Glut-1 expressed by brain microvascular endothelial cells, and its regulation is essential for normal CNS function (Devraj et al., 2011; Koepsell, 2020). Taken together, an intact BBB is essential to maintain normal CNS function and its dysregulation is known to be associated with several neurodevelopmental and neurodegenerative diseases, including Alzheimer’s disease and Parkinson’s disease (Erickson & Banks, 2013).

BBB disruption occurs early during HIV infection and is not significantly improved by the implementation of cART, as Rahimy et al. reported an elevated ratio of albumin level in the cerebrospinal fluid (CSF) to that in serum among PLWH receiving antiretroviral drugs (ARVs) (Louboutin & Strayer, 2012; Rahimy et al., 2017; Bertrand et al., 2019). An increased BBB permeability is a critical contributor to HAND pathogenesis as its disruption facilitates the CNS infiltration of free virions, infected or uninfected macrophages and leukocytes as well as ARVs from the periphery, resulting in an escalated susceptibility of inflammatory assault and toxicity in the brain (Atluri et al., 2015; Chaganti et al., 2019).

Based on molecular mechanisms and resistance profiles, ARVs are classified in six major pharmacological classes: nucleoside reverse transcriptase inhibitors (NRTI); non-nucleoside reverse transcriptase inhibitor (NNRTI); integrase strand transfer inhibitor (INSTI), protease inhibitor (PI); fusion inhibitors and coreceptor antagonist (Arts & Hazuda, 2012; Pau & George, 2014). Current regimen for HIV treatment generally consists of two NRTIs administered in combination with a third active ARV drug from one of three drug classes: INSTIs, NNRTI, or PIs with a pharmacokinetic enhancer (Hirsch, 2008). Although side effects are inevitable for every drug, it is crucial to ensure that the anti-HIV drugs do not aggravate any pathological conditions caused by HIV infection. However, a cohort study on HIV positive stable patients demonstrated that the discontinuation of ARVs significantly improved neurocognitive function, with no neurocognitive improvements observed with cART re-implementation among those who had developed neurocognitive impairment (Robertson et al., 2010). Although limited, emerging clinical evidence suggests that ARVs-mediated toxicity in the CNS may play a critical role in the pathogenesis of HAND (Bertrand et al., 2021).

Efavirenz (EFV), a NNRTI, has long been considered a first-line therapy owing to its potency and efficacy in viral suppression and immune function restoration (Bengtson et al., 2018). However, its clinical use has been discouraged in most developed countries as a result of various neurological and neuropsychiatric adverse reactions, including insomnia, dizziness, depression and psychosis (Gutiérrez et al., 2005; van de Wijer et al., 2016; Qin et al., 2020). While the underlying mechanisms for this toxicity are not fully understood, previous literature suggests its potential in impairing BBB by inducing endoplasmic reticulum (ER) stress, mitochondrial and autophagy dysfunction (Apostolova et al., 2015).

Dolutegravir (DTG), an INSTI, is the current World Health Organization (WHO) recommended preferred first- and second-line treatment for PLWH (Estill & Bertisch, 2020). It is commonly prescribed in combination with a two-drug backbone of tenofovir alafenamide and emtricitabine or abacavir and lamivudine (Estill & Bertisch, 2020). However, the reported rate of DTG discontinuation due to neuropsychiatric adverse events was significantly higher than other INSTIs-based cART, reaching as high as 6% within 12 months of treatment in the general population and even higher (18%) amongst women and older patients (Hoffmann et al., 2017). With a positive correlation between DTG plasma-trough concentrations and CNS side effects reported in a Japanese population, the safety of using DTG has gained emerging concern, especially amongst PLWH with pre-existing neuropsychiatric disorders (Cailhol et al., 2017; Yagura et al., 2017; Yombi, 2018). Bictegravir (BTG), a recently approved second-generation INSTI that was structurally derived from DTG is primarily recommended as a first-line cART alternative (Stellbrink et al., 2020). The most common side effects reported with the use of BTG include nausea and headache (Hill et al., 2018). Although limited, Hoffmann et al. reported a comparable rate of neuropsychiatric effects between BTG and DTG, suggesting a potential pharmacological class effect of INSTI (Hoffmann et al., 2021).

While studies are suggesting a CNS toxicologic potential of INSTI, the underlying mechanism is poorly understood. The present study aimed to investigate the effect of DTG and BTG in inducing structural and metabolic dysfunction of the BBB using human and mouse BBB model systems in vitro, ex vivo and in vivo.

Despite the implementation of cART which has successfully transformed HIV-1 infection from a deadly disease to a treatable chronic condition, HIV remains incurable with the life-long burden of ARVs intake (Elbirt et al., 2015). HAND, manifested by the alteration of cognitive, behavioural, motor and autonomous functions, is one of the major health complications among PLWH on cART (Clifford, 2017). HAND has been mainly attributed to limited viral suppression due to the ineffective penetration of ARVs into the brain (Atluri et al., 2015; Saylor et al., 2016). However, the full neuropathogenesis of HAND is not completely understood. Penetration-Effectiveness (CPE) rank, developed by pharmacokinetics’ characteristics of ARVs, is commonly used to estimate the efficacy of ARVs in the CSF (Letendre et al., 2008; Mukerji et al., 2018). Although ARVs with good CNS penetration, as indicated by a high CPE score, are generally more effective in controlling CSF viral replication, these ARVs are associated with poorer neurocognitive performance (Caniglia et al., 2014; Beardsley & Le, 2015; Mukerji et al., 2018). Taking consideration of such evidence, we hypothesized that the use of ARVs is associated with CNS toxicity, particularly by inducing BBB dysfunction. Herein, we applied in vitro experiments using cultures of human brain microvessel endothelial cells (hCMEC/D3) and primary cultures of mouse brain microvascular endothelial cells as well as ex vivo experiments using isolated mouse brain capillaries and in vivo mouse experiments to investigate whether ARVs induce BBB dysfunction that may ultimately contribute to the development of HAND.

INSTI based cART are the current preferred first line clinical regimen in most clinical settings. DTG, an INSTI with the highest CPE score (4) is associated with various neuropsychiatric side effects and a relative high discontinuation rate in the clinic (Cailhol et al., 2017; Povar-Echeverría et al., 2021). Therefore, we focused on investigating the potential DTG-mediated toxicity in various BBB models. To further address whether a pharmacological class effect could occur, BTG, a second generation INSTI that was structurally derived from DTG, was also examined in the current study. EFV is a NNRTI which has been removed from the first line regimens in the United States but continues to be used as a cost-effective option in many developing countries (Bengtson et al., 2018). EFV was widely implicated in mild-to-severe neuropsychiatric adverse events and neurocognitive impairment (Gutiérrez et al., 2005; Qin et al., 2020). A considerable number of in vitro and in vivo experiments have underlined its high risk in CNS toxicity such as neurotoxicity and BBB dysfunction (Apostolova et al., 2015). Therefore, EFV was used as a positive control in our study. The concentrations of ARVs applied in this study were carefully chosen to correspond to the therapeutically relevant plasma concentrations reported in human subjects (Villani et al., 1999; Elliot et al., 2016; Rigo-Bonnin et al., 2020).

Brain endothelial cells play a central role in the formation of BBB and TJs constitute the most important junctional complex in regulating the paracellular diffusion of small molecules. Our results demonstrated that all three ARVs (EFV, DTG, BTG) at therapeutically relevant plasma concentrations downregulated the mRNA expression of at least one TJ protein in our in vitro human and mouse BBB models, and ex vivo, in isolated mouse brain capillaries. Our gene expression data using EFV, corroborates previous findings by others demonstrating EFV can alter CLDN5/Cldn5 expression and increase microvessel endothelial permeability in human and mouse BBB models (Bertrand et al., 2016). In addition, we showed that acute EFV exposure downregulated gene expression of two most relevant drug efflux transporters, ABCB1 and ABCG2 in vitro human and mouse BBB cell models and ex vivo mouse brain capillaries. While EFV is neither considered a substrate nor an inhibitor of P-gp (Dirson et al., 2006), studies in rats have shown that it is a substrate and inhibitor of BCRP in gastrointestinal epithelial cells (Peroni et al., 2011). In the current study, we found EFV exerted a downregulatory effect on ABCG2 mRNA expression at the BBB. Inconsistency of gene expression of glucose transporter and proinflammatory cytokines was observed in our BBB models exposed to EFV. We observed a downregulation of SLC2A1 mRNA expression in human BBB cells and mouse brain capillaries, but upregulation in primary mouse BBB cells. Furthermore, EFV mediated robust upregulation of mRNA expression of proinflammatory cytokines (IL6, IL8, IL1β) and oxidative stress marker (NOS2) in the human BBB cells, but only induced Il6 mRNA expression in primary mouse cells and Il1β in isolated mouse brain capillaries. Species differences, incubation time points and model-specific response could play a role in the observed inconsistency.

In comparison to EFV, our gene expression data illustrates a greater potential of DTG in disrupting TJs in primary cultures of mouse microvascular endothelial cells, reflected by the downregulation of the mRNA expression of three TJ proteins by up to 80%. Similar to EFV, DTG robustly induced gene expression of proinflammatory cytokines (IL6, IL8, IL1β) and oxidative stress marker (NOS2) in hCMEC/D3 cells, but with only Il6 being robustly induced in primary cultures of mouse brain microvascular endothelial cells, and Il1β being mildly upregulated in mouse brain capillaries. Our results are consistent with data of a previous study performed by Ma et al., who reported that DTG significantly downregulated TJP1, CLDN5 and JAM-2 gene expression in bovine brain microvascular endothelial cells, and increased mRNA levels of proinflammatory cytokines TNFα and IL1β in the cell culture supernatants (Ma et al., 2020). However, the DTG dose (25 μM/∼10,480 ng/mL) used in that study exceeded the clinically relevant plasma concentrations observed in human (up to ∼5,000 ng/mL) and may explain why certain changes were not observed in our hands. In addition, we observed a robust downregulation of the efflux drug transporters, Abcb1a and Abcg2 mRNA expression in the primary cultures of mouse brain microvascular endothelial cells following DTG treatment. Previous studies reported DTG being a substrate and inhibitor of P-gp and BCRP in vitro, in intestinal epithelial cells (Cottrell et al., 2013; Reese et al., 2013). Although such inhibitory effects were not observed at clinically relevant concentrations (Reese et al., 2013), potential drug-drug interactions between DTG and co-administered medications that are substrates or inhibitors of P-gp or BCRP remain to be investigated. Furthermore, although DTG pharmacokinetics were not significantly altered in the clinic by the co-administration of lopinavir/ritonavir which are known P-gp and BCRP inhibitors (Song et al., 2011), an in vitro study using bovine BBB model showed a potential synergistic toxic effect in impairing BBB integrity with the administration of DTG in combination with the P-gp inhibitor sertraline, a selective serotonin reuptake inhibitor, that is commonly used as antidepressant therapy (Ma et al., 2020). Considering depression and other psychiatric disorders are common co-morbidities among PLWH, drug-drug interactions involving drug efflux transporters remain a concern, as any increased penetration of ARVs or psychiatric medication can potentially result in elevated microglial activation and neurotoxicity that may lead to cognitive impairment (Wu et al., 2017).

In comparison to DTG, which robustly altered TJ proteins in primary cultures of mouse brain microvascular endothelial cells, acute exposure to BTG, had a minor effect on mRNA expression of Tjp1, Ocln, Cldn5, and proinflammatory cytokines (Il6, Il1β, Nos2). We additionally observed a BTG-mediated decrease in ABCB1 and SLC2A1 mRNA expression in hCMEC/D3 cells, a downregulated Abcg2 gene expression in mouse cells and an upregulated gene expression of Abcb1a in isolated mouse brain capillaries. BTG has been reported to be a substrate of P-gp and BCRP, however, the clinical relevance of these transporters interactions remains unknown (Deeks, 2018). To our knowledge, we are the first research group to report any effect of BTG on TJ proteins, membrane-associated transporters, and inflammatory markers at the BBB in vitro, ex vivo and in vivo.

In brief, our in vitro and ex vivo data illustrate that the acute exposure of EFV and DTG significantly, disrupted TJ proteins and induced inflammatory response in in vitro and ex vivo BBB models. Unexpectedly, DTG elicited a stronger effect than EFV, which was considered a positive control for its well-known induced CNS toxicity.

Herein, we also conducted animal studies to further investigate whether the observed in vitro changes have in vivo implications. Considering that most ARVs are taken in the long term, we treated animals for a prolonged time period to mimic chronic exposure to ARVs before experimental assessment. To mimic the oral administration of ARVs in most clinical settings and to avoid the detrimental effects of repetitive oral gavage in mice, we implemented an alternative method of oral administration by formulating ARVs into PB pellets. Considering the duration of treatment and species-specific differences in pharmacokinetics, doses of ARVs were carefully selected to yield human therapeutic plasma concentrations in mice (Osborne et al., 2020; Mohan et al., 2022).

Our NaF permeability assay demonstrated that 14-day chronic treatment with EFV or DTG, in mice, significantly increased the permeability of the BBB in vivo, reflected by a ∼40% and ∼70% increase in brain NaF level, respectively, suggesting a greater potential of DTG in compromising BBB integrity than EFV. In contrast, BTG treatment did not induce a significant effect on BBB permeability in these animals. Our mRNA expression data on isolated brain capillaries collected from mice treated with DTG or BTG for 14 days demonstrated a dysregulation of TJ gene expression. DTG elicited a stronger effect than BTG in downregulating both Tjp1 and Ocln, potentially explaining the greater effect of DTG in increasing BBB permeability we observed from the NaF assay. Chronic exposure to DTG or BTG for 14 days significantly decreased the expression of drug efflux transporters, and DTG upregulated Il1β gene expression, suggesting their potential in disrupting BBB functionality in addition to its integrity. Notably, our positive control EFV data was consistent with a previous study performed by Bertrand et al., in which a ∼30% increase in BBB permeability was observed in mice treated with EFV for 14 days, along with a robust decrease in Cldn5 protein expression in isolated mouse brain capillaries (Bertrand et al., 2016). Interestingly, our data showed that chronic exposure to DTG or BTG disrupted BBB primarily by targeting ZO-1 and OCLN, suggesting a potential differential mechanism from EFV-mediated BBB disruption which predominantly targeted CLDN5.

In general, our data from acute exposure of ARVs in vitro and ex vivo BBB models provided similar responses in vivo, following a prolonged exposure time. However, the ARVs-mediated changes of glucose transporter expression were only observed in vitro, and further studies need to be performed to validate these findings in vivo. Our in vivo NaF assay data confirmed a more deleterious effect of DTG than EFV in increasing BBB permeability and our gene data revealed a similar effect of DTG and BTG in disrupting the physical and physiological function of BBB by altering TJ proteins and membrane associated transporters, suggesting a pharmacological class effect of INSTIs. However, the deleterious effect on BBB permeability was more robust for DTG than BTG treatment, indicating a potential safer toxicological profile of BTG in the CNS. A very limited number of studies on BTG-induced CNS toxicity are currently available except for a few clinical reports. Hoffmann et al. concluded that BTG-based regimen leads to a comparable neuropsychiatric adverse event-related discontinuation rate but some favourable patient-reported outcomes in randomized clinical trials compared with DTG-based regimen (Hoffmann et al., 2021). Another study which enrolled 24 HIV + individuals with HIV-related CNS impairment on BTG-based cART indicated a significantly higher BTG concentration in cerebrospinal fluid among patients aged over 51 years old (Gelé et al., 2021), suggesting a potentially elevated risk of BTG-mediated toxicity among the aged population. To date, to the best of our knowledge, we are the first laboratory to report the effect of BTG in the CNS using in vitro, ex vivo and in vivo BBB models.

Our study presents some limitations. The CNS concentration of ARVs was not quantified in the mouse model due to the limited volume of cerebrospinal fluid. Although the ARV doses were chosen to yield a human equivalent concentration in the plasma based on previous studies in rodents (Kala et al., 2018; Mohan et al., 2021; Mohan et al., 2022), complementary pharmacokinetic studies would be valuable to further validate the findings.

In summary, our studies demonstrated the potential of ARVs in altering the functionality of the BBB by inducing inflammation and drug/nutrient transporter changes in addition to structural impairment. A compromised BBB facilitates viral entry of free virions and infected monocyte-macrophages, leading to enhanced cerebral viral infection (Atluri et al., 2015; Elbirt et al., 2015; Clifford, 2017). The activation of neighbouring resident microglia and astrocytes in response to viral infection is often followed by a robust secretion of proinflammatory cytokines and neuronal toxins, which as a result, leads to severe brain damage (Tavazzi et al., 2014). In addition, the inflammatory response may also facilitate a positive feedback loop, aggravating the BBB impairment by enhancing cellular trafficking, solute permeability and altering signal transduction cascades such as NF-κB, prostaglandin E2 synthesizing-enzymes, cyclooxygenase-2 and microsomal prostaglandin E synthase 1 (Wilhelms et al., 2014; Galea, 2021). Furthermore, a compromised BBB also facilitates the brain entry of ARVs, resulting in elevated drug exposure and risk of toxicity in the brain parenchyma which can cause progressive neuronal damage (Apostolova et al., 2015). Taken together, the BBB disruption, associated inflammation and drug toxicity can have an additive or synergistic negative effect on cognitive function among PLWH, and may contribute to the high frequency of short-term neuropsychiatric adverse events associated with cART (Gutiérrez et al., 2005; Yombi, 2018; Hoffmann et al., 2021), and in a long term, high incidence of ANI and MND observed in the clinic (Elbirt et al., 2015; Clifford, 2017). By studying the first line INSTIs and EFV, our study revealed a comparable potency of DTG and EFV but not BTG in inducing inflammation and disrupting integrity and functionality of the BBB. These findings suggest a safer CNS toxicological profile of BTG compared to DTG and EFV. Further studies are required to understand the underlying toxicological mechanisms exerted by these drugs in order to minimize the risk of BBB-mediated CNS complications.