The neurovascular unit (NVU) is an anatomical and functional unit that responds to the needs of neuronal energy demands by precisely regulating cerebral blood flow (Iadecola, 2017). The NVU is comprised of brain endothelial cells, astrocytes, mural cells (i.e., vascular smooth muscle cells and pericytes), microglia, neurons, and extracellular matrix components (Figure 1). The microvascular endothelial cells of the NVU comprise the blood-brain barrier (BBB), a physical and metabolic barrier important for maintaining central nervous system (CNS) homeostasis and protecting the brain from potentially harmful circulating substances. The metabolic component of the BBB consists of drug metabolizing enzymes that are capable of inactivating therapeutics and/or altering their ability to enter the CNS (Agundez et al., 2014). The BBB exhibits low permeability compared to peripheral blood vessels and only lipophilic molecules with molecular weights less than 500 Da, and few hydrogen bond donors and/or acceptors are able to passively diffuse across the BBB under physiological conditions (Banks, 2009; Banks and Greig, 2019). Nutrients such as glucose, amino acids, nucleosides, and certain neurotransmitters that are important for CNS function are able to enter the brain through carrier-mediated transport (Ohtsuki and Terasaki, 2007). Influx and efflux transporters, which extrude toxic metabolites from endothelial cells and limit CNS entry of xenobiotics and therapeutics from the bloodstream, are also a prominent feature of the BBB (Abdullahi et al., 2017). While receptors and transporters expressed at the endothelial plasma membrane selectively regulate brain entry of various solutes from the blood, the major factors limiting BBB permeability are its low rate of pinocytosis and tight junction (TJ) protein complexes which limit passage through the transcellular and paracellular routes, respectively (Reese and Karnovsky, 1967).

TJs at the BBB are protein complexes that form rows of extensive, overlapping occlusions between brain microvascular endothelial cells that greatly restrict paracellular diffusion of polar molecules and macromolecules into the CNS (Reese and Karnovsky, 1967; Brightman and Reese, 1969). The in vivo transendothelial electrical resistance (TEER) of brain microvessels is 1,500–2,000 Ω cm², which is considerably higher than peripheral microvessels that have a TEER of 3–30 Ω cm² (Crone and Olesen, 1982; Butt et al., 1990). Although previously considered to be static structures, the past few decades of research have revealed that TJs at the BBB are highly dynamic in nature and are dysregulated in many CNS disease states (Erdo et al., 2017; Reinhold and Rittner, 2017; Costea et al., 2019; Sweeney et al., 2019). While TJ dysregulation at the BBB can contribute to the pathogenesis of certain neurological disorders, understanding the mechanisms involved may provide an opportunity to improve CNS drug delivery through the paracellular pathway. Here, we provide an overview of the molecular composition of TJs at the BBB, disruption of TJs during CNS diseases, and strategies to manipulate TJs with the goal of improved drug delivery to the brain.

In the paracellular cleft, TJs form a physical seal between adjacent brain microvascular endothelial cells (Figure 2). In conjunction with active efflux transporters and the lack of fenestrations along the apical surface of the brain microvessel, TJ protein complexes contribute to physiological functioning of the BBB, including maintenance of a stable microenvironment within the CNS. TJ functional integrity at the BBB can be evaluated by measurement of extravasation of solutes (i.e., leak) that are not transport substrates and typically remain in the vascular lumen under physiological conditions. Such vascular markers cover a wide range of molecular weights and, if used appropriately, can enable a detailed evaluation of size selectivity of vascular “leak” in the setting of pathological stressors. Commonly used tracer molecules for the study of paracellular diffusion are presented in Table 1. Using transmission electron microscopy, BBB TJ protein complexes are visualized as a series of electron dense regions (i.e., “kissing points”) between apposing membranes of adjacent endothelial cells (Chiba et al., 2008). At the molecular level, TJs are comprised of transmembrane protein complexes protruding from neighboring cells. These transmembrane TJ proteins are linked to the cytoskeleton by intracellular TJ proteins, thereby forming complex networks that are highly responsive to pathological stimuli. Indeed, many different protein constituents are reported to be present at TJs in mammalian cerebral microvasculature (Haseloff et al., 2015; Reinhold and Rittner, 2017; Berndt et al., 2019). Below, we review the specific transmembrane and intracellular proteins that have a confirmed role in maintaining TJ functional integrity in health and disease.

Claudins are considered to be the principal proteins that establish the backbone of TJ and are critical determinants of paracellular “tightness” between adjacent BBB endothelial cells (Gunzel and Yu, 2013). The claudin family of proteins consists of 27 members that have been identified to date, and family members have an approximate molecular mass of ~26 kDa (Daneman et al., 2010; Mineta et al., 2011). They have a tissue-specific expression pattern (Hwang et al., 2014; Markov et al., 2015), which leads to different barrier properties in different tissues. In epithelial cells, claudin-3 and claudin-5 generally provide a sealing function (Amasheh et al., 2005; Milatz et al., 2010). At the BBB, claudin-5 has long been recognized as the dominant TJ protein (Greene et al., 2019). Depletion of claudin-5 in transgenic mice results in morphologically normal TJs at the BBB but also leads to detrimental brain defects and early neonatal death ~10 h after birth. Furthermore, brain microvasculature of claudin-5 knockout mice is highly permeable to molecules less than 800 Da (Nitta et al., 2003). In contrast, overexpression of claudin-5 in cultured brain microvessel endothelial cells results in elevated paracellular tightness, suggesting that expression levels of claudin-5 directly correlate with TJ function (Ohtsuki et al., 2007). Such notions have recently been challenged by the study of Schlingmann et al. (2016), which suggests that changes in TJ function are not simply the result of increased or decreased protein levels but rather arises from disruption in the overall balance of all TJ proteins (Schlingmann et al., 2016). Similar to claudin-5, expression of claudin-3 has also been reported in brain microvasculature (Ohtsuki et al., 2008; Ronaldson et al., 2009), where it may play a role in the physiological sealing function of the TJ (Winkler et al., 2020). Claudin-1 has been detected at the BBB at the messenger RNA (mRNA) level (Winkler et al., 2020); however, its expression at the BBB remains controversial in part due to possible antibody cross-reactivity between claudin-1 and claudin-3 (Bauer et al., 2014). Claudin-11 and claudin-12 have also been detected at the BBB, but claudin-12 is likely not important for barrier function (Castro Dias et al., 2019; Uchida et al., 2019).

Structurally, the claudins are tetraspan proteins with two extracellular loops (ECLs) and two cytoplasmic domains that consist of a short amino (N)-terminal sequence and a long carboxyl (C)-terminal sequence. The ECLs of claudins contribute to the formation of the TJ backbones by homo‐ or hetero-dimerization or oligomerization with neighboring claudins on the same membrane (cis-interaction) and/or with those on the apposing cell (trans-interaction; Piontek et al., 2008; Piehl et al., 2010). The ECLs are also responsible for determining charge‐ and size selectivity of the claudins (Piehl et al., 2010; Piontek et al., 2011). Therefore, interactions between claudin-5 with other claudins or with other transmembrane proteins may be necessary to properly restrict paracellular diffusion of molecules of varying charges and sizes. The C-terminus of claudins exhibits great diversity among members of the claudin family and ends with a PDZ binding motif for most claudins with the exception being claudin-12 (Piontek et al., 2011; Van Itallie and Anderson, 2018). PDZ-dependent interactions at the TJ play a prominent role in the organization and modulation of TJ proteins (Itoh et al., 1999; Poliak et al., 2002). The PDZ domain on the C-terminal tail of most claudins recruits cytoplasmic proteins, particularly zonula occludens (ZOs) proteins. Additionally, PDZ domains on the C-terminal of a claudin molecule can also promote interactions with other transmembrane TJ proteins including occludin and junction adhesion molecules (JAMs; Poliak et al., 2002). Notably, the claudin C-terminal domain contains multiple serine, threonine, and tyrosine residues (Hornbeck et al., 2015). Mass spectrometry (MS) data have revealed multiple phosphorylation sites in the C-terminal region of many claudins (Hornbeck et al., 2015; Van Itallie and Anderson, 2018). Although the regulatory role of claudin phosphorylation has not yet been directly studied at the brain microvascular endothelium, studies in other epithelial and endothelial tissues have shown roles in modulation of protein-protein interactions (Nomme et al., 2015), intracellular trafficking (Van Itallie et al., 2012; Twiss et al., 2013), and protein turnover mechanisms (Mandel et al., 2012; Hou et al., 2019). For example, Rho kinase (RhoK) has been demonstrated by Yamamoto et al. (2008) to phosphorylate claudin-5 at tyrosine 207 (T207) in mouse and human brain tissues. This observation was correlated with decreased TEER, suggesting that Rho-associated protein kinase (ROCK) signaling and associated claudin-5 phosphorylation may be a critical modulatory mechanism for TJ permeability at the BBB (Yamamoto et al., 2008).

There are three tight junction-associated marvel proteins (TAMPs) that have been identified at the mammalian BBB. These proteins are known as tricellulin, marvelD3, and occludin. Tricellulin plays a prominent role in regulating paracellular permeability to macromolecules (Krug et al., 2009; Haseloff et al., 2015) while marvelD3 appears to play a compensatory sealing role at the BBB (Raleigh et al., 2010). In contrast, occludin (molecular weight ~65 kDa), the first integral membrane protein identified at BBB TJs has been more extensively characterized. Occludin knockout mice exhibit structurally and functionally intact TJs, possibly due to compensation from other TAMPs; however, these mice display several histological abnormalities and postnatal growth retardation (Saitou et al., 2000). One study suggested that occludin’s physiological role in epithelial barriers is associated with epithelial differentiation rather than establishing barrier properties (Schulzke et al., 2005). Occludin has also been shown to be an important regulator of TJ assembly and functions (Hirase et al., 2001; Van Itallie et al., 2010; Doerfel and Huber, 2012; Buschmann et al., 2013). Additionally, occludin can assemble into higher-order oligomeric assemblies, which are essential to maintenance of TJ functional integrity (McCaffrey et al., 2008). Indeed, our laboratory has shown that disruption of occludin oligomeric structures results in BBB dysfunction characterized by an increase in paracellular permeability (Lochhead et al., 2010, 2012).

Occludin contains a tetraspanning marvel domain, two ECLs, and three cytoplasmic domains, including a short N-terminal domain, a long C-terminal domain, and an intracellular short turn. As has been demonstrated in a study with mutant occludin isoforms in MDCK cells, both ECLs together with at least one transmembrane domain serve important roles in selective permeability of TJs (Balda et al., 2000). The N-terminus of occludin contributes to maintenance of TJ integrity, as evidenced by the observation that deletion of the occludin N-terminal sequence leads to higher paracellular permeability and lower TEER (Bamforth et al., 1999). Lateral (cis-) homodimerization or oligomerization of occludin is mediated by the marvel motif, the cytoplasmic C-terminus, and the second extracellular loop (ECL2; Blasig et al., 2006; Yaffe et al., 2012). These domains are sensitive to changes in redox conditions due to the disulfide bonds between cysteine residues on ECL2 and the occludin C-terminal sequence (Walter et al., 2009; Bellmann et al., 2014). The C-terminus also interacts with ZO proteins (Li et al., 2005; Tash et al., 2012), an interaction that is essential for the trafficking of occludin to the TJ (Furuse et al., 1994). Similar to the claudins, the C-terminal domain of occludin is rich in potential phosphorylation sites and has been shown to interact with a wide range of kinases and phosphatases (Smales et al., 2003; McKenzie et al., 2006; Suzuki et al., 2009; Raleigh et al., 2011; Walsh et al., 2011). With few exceptions, in vitro experiments have repeatedly demonstrated that elevated phosphotyrosine and diminished phosphoserine/threonine are generally associated with disruptive barrier functions by interfering with occludin-ZO interactions and/or occludin trafficking (Kale et al., 2003; Elias et al., 2009; Suzuki et al., 2009; Han et al., 2010; Walsh et al., 2011). Interestingly, the phosphorylation status of the C-terminus of occludin is highly dependent on cellular localization of the protein itself; TJ-localized occludin is phosphorylated at a much higher level than those localized to the lateral membrane (Sakakibara et al., 1997), highlighting the importance of examining both cellular expression and localization when studying TJ proteins.

JAMs-1, -2, and -3 are members of the immunoglobulin superfamily that are highly enriched at TJs of the cerebral microvasculature. They are not required for TJ formation (Otani et al., 2019); however, complementary DNA (cDNA) transfection of JAMs in non-TJ forming cell cultures resulted in recruitment of occludin to regions of cell-cell contacts (Bazzoni et al., 2000). Such an observation suggests a potential role for JAMs in mediating occludin trafficking to the TJ. Additionally, loss of JAM protein expression and/or migration of JAM proteins away from the TJ directly lead to loss of BBB properties at the microvascular endothelium (Haarmann et al., 2010; Wang et al., 2014). JAMs consist of a single membrane-spanning domain, a short C-terminal tail, and an immunoglobulin G (IgG)-like extracellular domain. The extracellular domain of JAM proteins mediate homo‐ or hetero-dimerization of JAM family members in both trans‐ and cis-manners, contributing to adherence of adjacent cells at the TJs (Bazzoni et al., 2000). The cytoplasmic C-terminus contains multiple phosphorylation sites and ends with a PDZ binding motif which interacts with ZO-1 as well as other PDZ-containing proteins in the endothelial cells (Bazzoni et al., 2000; Ebnet et al., 2003). Binding affinity of JAMs to ZO-1 may be regulated through the phosphorylation of protein residues on the C-terminal sequence (Van Itallie and Anderson, 2018).

ZO-1 (225 kDa), ZO-2 (160 kDa), and ZO-3 (130 kDa) belong to the PDZ domain-containing protein and membrane-associated granulated kinase (MAGUK) family, which constitute the most prominent intracellular proteins at the TJs. ZO family members are essential for assembly of functional TJs by binding to and cross-linking a wide range of TJ proteins and tethering them to the actin cytoskeleton (Fanning et al., 1998; Shin and Margolis, 2006). The ZO proteins share considerable structural similarity except for the C-terminal sequence that is unique to each specific ZO protein (Bauer et al., 2014). Indeed, differences in these C-terminal protein sequences could explain functional differences of individual ZO family members. While ZO-1 and ZO-2 are considered indispensable for TJ assembly (Umeda et al., 2006), the specific role of ZO-3 at BBB TJs has not been not well-documented. Structurally, the ZO proteins are comprised of three PDZ domains, a Src homology 3 (SH3) domain, a guanylate kinase (GuK) domain, and a proline-rich domain. The first PDZ domain (PDZ1) interacts with the PDZ domain on the C-terminus of most of the claudin proteins (Itoh et al., 1999). The second PDZ domain (PDZ2) mediates ZO-1 heterodimerization with ZO-2 or ZO-3 (Itoh et al., 1999; Wittchen et al., 1999). Structural studies have suggested weak binding between PDZ3 and JAM-1 (Bazzoni et al., 2000). The SH3 domain is the most important element for binding of actin cytoskeleton and signaling molecules (Schmidt et al., 2004; McNeil et al., 2006). Actin cytoskeletal filaments also interact with a defined actin-binding region (ABR) at the C-terminus of the ZO-1 protein (Fanning et al., 2002). It is important to note that there are multiple organizational forms of actin that are regulated by actin-binding proteins which may affect dynamics of TJ proteins and barrier integrity (Tornavaca et al., 2015). GuK are kinases that normally catalyze ATP-dependent transformation of GMP into GDP; however, the GuK domain on the ZO subfamily of MAGUKs is likely enzymatically inactive due to a missing GMP binding-site and steric hindrance at the ATP-binding site (Lye et al., 2010; Zhu et al., 2012). Instead, the GuK domain of ZO proteins has been shown to bind occludin and PDZ motifs of the claudins (Fanning et al., 1998; Chattopadhyay et al., 2014).

BBB dysfunction is a pathological feature of several diseases affecting the CNS (Erdo et al., 2017; Reinhold and Rittner, 2017; Costea et al., 2019; Sweeney et al., 2019). BBB permeability alterations are often present before overt clinical symptoms, suggesting that BBB impairment may be a contributing factor to the pathogenesis of various neurological disorders (Figure 3). A number of studies have indicated that plasma proteins can be neurotoxic, which suggests that even if a compromised BBB does not cause these disorders it likely exacerbates them (Chodobski et al., 2011). In this section, we will review the neurological disorders most closely associated with impairments in BBB functional integrity with a focus on human data and in vivo studies.

Although the BBB plays an important protective role of the CNS, modulation of TJs may allow increased brain delivery of drugs through the paracellular route (Deli, 2009; Tscheik et al., 2013; Greene and Campbell, 2016). For example, we have shown that TJ alterations during PIP are associated with improved CNS delivery of the opioid analgesic codeine, which results in enhanced analgesia (Hau et al., 2004). Ideally, TJ alterations at the BBB would be transient in nature to allow enhanced drug permeation into the CNS without causing side effects or toxicity resulting from an open barrier. Pharmacological strategies to modulate TJs at the BBB must also exhibit low systemic toxicity as their delivery to brain endothelial cells would likely be through the bloodstream. Over the past several years, a number of different strategies have been tested to increase BBB paracellular permeability with the goal of improving drug delivery to the brain in vivo.

One strategy to increase brain delivery of drugs is intracarotid infusion of hyperosmotic concentrations of arabinose or mannitol. This technique causes reversible vasodilation and shrinkage of endothelial cells, which leads to a loosening of the TJs, an increase in paracellular diffusion, and bulk fluid flow across the BBB (Rapoport and Robinson, 1986; Rapoport, 2000; Dobrogowska and Vorbrodt, 2004). While this method has been used to enhance delivery of chemotherapeutics to brain tumors, limited efficacy and severe side effects such as brain edema and seizures have prevented its widespread use in the clinic (Rapoport, 2000; Marchi et al., 2007). In addition, its invasive nature and need for patients to be anesthetized during the procedure render the technique impractical for chronic administration of therapeutics (Marchi et al., 2007; Hersh et al., 2016).

Several different attempts to pharmacologically manipulate BBB permeability to increase brain delivery of drugs have also been attempted. BBB permeability to Evans blue was reversibly increased 15 min after intracarotid infusion of oleic acid in rats (Sztriha and Betz, 1991). Administration of oleic acid or linoleic acid to cats also caused BBB disruption as measured by MRI, but these effects were associated with brain edema, necrosis, and demyelination (Kim et al., 2005). Intravenous infusion of the bradykinin B₂ receptor agonist RMP-7 (Cereport®) increased lanthanum penetration through TJs of the BBB by loosening the junctional complexes (Sanovich et al., 1995). RMP-7 also showed enhanced uptake of chemotherapeutics in preclinical brain tumor models but was ineffective when used to enhance brain delivery of carboplatin to treat childhood brain tumors in a phase II trial (Emerich et al., 2001; Warren et al., 2006). Administration of alkylglycerols has also been shown to increase BBB permeability of a number of different drugs and tracers (Iannitti and Palmieri, 2010). Intracarotid administration of alkylglycerols induce a size-dependent permeability increase, suggesting that alkylglycerols increase BBB paracellular permeability by modulating TJs, but intracarotid administration is an invasive procedure, and alkylglycerol administration to increase CNS drug uptake has not been tested clinically (Erdlenbruch et al., 2003). A 12 kDa fragment of the ZO toxin (ΔG) from vibrio cholerae has been shown to increase BBB to (¹⁴C)-sucrose, likely by activating PKC (Fasano et al., 1995; Menon et al., 2005). These effects were modest without the addition of protease inhibitors, suggesting that ΔG is rapidly degraded in the bloodstream and likely not a good candidate to clinically improve drug delivery to the CNS (Menon et al., 2005).

Another strategy to transiently increase BBB paracellular permeability is by targeting specific components of the TJs with siRNA or novel peptidomimetic drugs. Campbell et al. (2008) showed that BBB paracellular permeability could be significantly increased up to 48 h after tail vein injection of siRNA targeting claudin-5 in mice (Campbell et al., 2008). This technique led to enhanced brain delivery of Gd-DTPA (742 Da), but not to FITC-dextran (4.4 kDa), showing a size-selective barrier alteration when targeting claudin-5 similar to what is seen in claudin-5 knockout mice. More recently, Dithmer et al. (2017) reported that newly engineered peptidomimetic drugs targeting claudin-5 could increase paracellular BBB permeability to various tracers across a wide molecular weight range such as lucifer yellow (457 Da), FITC-dextran (10 and 40 kDa), FITC-albumin (67 kDa), and tetramethylrhodamine-dextran (155 kDa) (Dithmer et al., 2017). These claudin-5 targeting peptides were also shown to increase CNS distribution of gadolinium 4 h following their administration, suggesting a novel approach for transient and reversal opening of the BBB for optimization of drug delivery (Dithmer et al., 2017).

Ultrasound in the presence of microbubbles is a technique that is increasingly being used to create a transient and focal opening of the BBB to deliver therapeutics to the CNS ranging in size from small molecules to antibodies, plasmids, and viral vectors. Focused ultrasound (FUS) bursts generated from an external source are able to utilize microbubbles as contrast agents that act as cavitation sites, which physically disrupt the BBB (Alonso, 2015; McMahon et al., 2019). Increased BBB permeability induced by ultrasound may involve transcytosis, opening of transendothelial channels, passage through injured endothelial cells, or opening of TJs (Sheikov et al., 2004). Studies in rats showed that FUS with microbubbles could increase BBB permeability to horseradish peroxidase and lanthanum. This BBB disruption was associated with ultrastructural alterations of claudin-5, occludin, and ZO-1, suggesting that FUS can increase BBB paracellular permeability by altering TJ proteins (Sheikov et al., 2008). Clinical trials utilizing this technique to improve CNS drug delivery are currently underway.

TJ protein complexes are the major determinants of paracellular permeability at the BBB. While brain microvascular endothelial TJs function to protect the CNS from potentially harmful substances in the blood, their protective role also represents a major obstacle to effectively deliver therapeutics for treatment of neurological diseases. Evidence is increasingly emerging that pathological alterations of TJs at the BBB may exacerbate or possibly initiate neurological dysfunction. A better understanding of how TJ expression, localization, and function is altered at the BBB during disease states may provide opportunities to protect the CNS from adverse consequences resulting from BBB disruption. Conversely, understanding the mechanisms involved in rapidly and transiently regulating TJ function at the BBB may allow improved delivery of therapeutics to the CNS through the paracellular route. Future research elucidating signaling pathways affecting TJ function at the BBB has the potential to both protect the CNS from a compromised barrier and more effectively treat neurological disorders by optimizing drug delivery.

JL and JY wrote the paper. PR and JL created the figures and tables. PR and TD edited the paper. All authors contributed to the article and approved the submitted version.