Pathogens Penetrating the Central Nervous System: Infection Pathways and the Cellular and Molecular Mechanisms of Invasion

E. coli. E. coli translocates from the upper compartment to the lower compartment in an HBMEC Transwell model without increasing cellular permeability, thus supporting a transcellular mechanism of penetration (163). Electron microscopy studies have demonstrated that E. coli invades HBMECs through a zipper-like, receptor-mediated endocytosis mechanism (164). Intracellular E. coli E44 bacteria were observed in membrane-bound vesicles within HBMECs at 30 min postinfection; by 45 min, the bacteria had translocated to the basolateral side (164). Both early (early endosome antigen 1 and transferrin receptor) and late (Ras-related protein 7 and lysosome-associated membrane protein 1) endosomal markers were recruited to vacuoles containing E. coli K1. However, the lysosomal hydrolytic enzyme cathespin D did not accumulate within the vacuoles, demonstrating that lysosomal fusion was prevented. This was found to occur in a capsule-dependent manner (165).

Several E. coli proteins contribute to the adhesion to and invasion of HBMECs (Table 1). Attachment of E. coli to HBMECs induces host cytoskeletal rearrangements and actin condensation below adherent bacteria (164). This is associated with a CNF-1-mediated recruitment of focal adhesion kinase and the cytoskeletal protein paxillin to the 67-kDa laminin receptor, which forms clusters that colocalize with adherent E. coli (150). Interactions between E. coli and HBMECs result in tyrosine phosphorylation of focal adhesion kinase and paxillin. The activity of focal adhesion kinase and its autophosphorylation site, tyrosine 397, were shown to be essential for E. coli invasion of HBMECs (166). Furthermore, the Src kinase-dependent activation of phosphatidylinositol 3-kinase and its interaction with focal adhesion kinase were required for E. coli K1 invasion and host cytoskeletal rearrangement (167, 168). Sukumaran and Prasadarao demonstrated that the association of phosphatidylinositol 3-kinase with focal adhesion kinase resulted in the downstream activation of protein kinase Cα (PKCα) in an Omp-A-dependent manner (169). Activated protein kinase Cα was recruited to the E. coli entry site, where it interacted with its substrate (myristoylated alanine-rich C kinase substrate), leading to an accumulation of actin (169). The activated protein kinase Cα also interacted with caveolin-1 colocalized with condensed actin underneath the bacterial entry site to form caveolae, which are plasma membrane invaginations that are involved in endocytosis and signal transduction (170). CNF-1 (via the 37/67-kDa laminin receptor) and type 1 fimbrial adhesin (FimH; via CD48) have also been shown to induce cytoskeletal rearrangements through activation of the GTPase RhoA (149, 171, 172).

In an independent mechanism, E. coli OmpA and new lipoprotein 1 (Nlp1) promote the activation of host cytosolic phospholipase A₂ (173, 174). The activation of cytosolic phospholipase A₂ generated arachidonic acid metabolites, induced host actin cytoskeletal rearrangements, and was essential for E. coli K1 invasion of HBMECs (173, 175 – 177). OmpA and invasion of brain endothelial cell protein A (IbeA) also trigger the phosphorylation of STAT3, which results in the activation of Rac1, a Rho family GTPase that regulates host cytoskeletal rearrangements (178). Notably, OmpA is highly conserved between Gram-negative bacteria and may play a role in CNS invasion by those whose mechanisms of entry are unknown. Similarly, a B. pseudomallei toxin, Burkholderia lethal factor 1 (BLF-1), is structurally similar to E. coli CNF-1 and possesses a number of conserved residues that, in CNF-1, promote deamidation of glutamine and result in the activation of RhoA and subsequent host cytoskeletal rearrangements (179). Thus, OmpA and BLF-1 may be attractive targets for future studies of the mechanisms of B. pseudomallei CNS invasion.

N. meningitidis.The attachment of N. meningitidis to HBMECs is mediated by type IV pili. This initial attachment is inhibited by high cerebral microcirculation shear stress levels; however, once attached, N. meningitidis proliferates and is able to resist high blood velocities by forming microcolonies with strong bacterium-bacterium interactions (124) and bacterium-host interactions (122). CD46 was previously identified as the receptor for type IV pili (180); however, this was not corroborated by additional studies, and it is likely that the type IV pilus attaches to host cells independently of CD46 (181). The type IV pilus protein PilQ, in addition to the outer membrane porin PorA, was demonstrated more recently to bind to the 37/67-kDa laminin receptor on HBMECs (151). Furthermore, studies with human bronchial epithelial cells show that the meningococcal type IV pilus binds to the platelet activating factor (PAF) receptor and that synergy between a pilin-linked glycan and phosphorylcholine decorating moieties is required for pili to efficiently engage the receptor (182).

N. meningitidis is thought to invade HBMECs primarily by paracellular mechanisms (see below); however, in vitro studies have demonstrated that small numbers of both encapsulated and nonencapsulated meningococci are internalized by HUVECs and HBMECs (183 – 185). Following initial bacterial attachment, the meningococcal minor pilin protein, PilV, triggers a cellular response in which endothelial cells form protrusions around N. meningitidis (122). These cellular projections further protect the bacterial microcolony from the mechanical stresses associated with high-velocity blood flow (122) and also lead to the engulfment and internalization of N. meningitidis (186). The host signaling events that contribute to these interactions have been elucidated. Following type IV pilus-mediated attachment, N. meningitidis activates the host cell β2-adrenoceptor, leading to the translocation of β-arrestin to the site of bacterial attachment at the plasma membrane (187). Several host proteins, including ezrin and moesin, which regulate the cortical cytoskeleton through F-actin-binding sites, are then recruited to the site of bacterial attachment and accumulate in honeycombed molecular complexes, referred to as cortical plaques, underneath the meningococcal microcolonies (186, 188, 189). The translocation of β-arrestin also results in the docking and activation of the protein tyrosine kinase Src, which phosphorylates cortactin (187, 188). Cortical actin is then polymerized in a Rho GTPase- and Cdc42-dependent manner, which leads to the formation of cell membrane protrusions (186). In human bone marrow endothelial cells, the recruitment of cortactin and formation of membrane protrusions were also shown to be dependent on the activation of a phosphoinositide-3-kinase/Rac1 signaling pathway by lipooligosaccharide (190).

Once internalized, the membrane-bound vesicles containing meningococci associate with transferrin receptor and lysosome-associated membrane protein 1 endosomal markers (185). It is likely that lysosomal fusion is subverted, as live meningococci have been shown to translocate from the basolateral side to the apical side in a Transwell BCSFB invasion model (140). In vitro studies have demonstrated that capsular polysaccharide inhibits the invasion of N. meningitidis into HBMECs (191) and human choroid plexus epithelial cells (140) but is essential for survival and replication within the intracellular niche (185).

For nonencapsulated meningococcal strains, the invasion of HBMECs is dependent upon outer membrane protein C (Opc). Opc binds to the human serum factor fibronectin, which acts as a bridging molecule and anchors the bacterium to the α₅β₁ integrin receptor of HBMECs (191). This interaction results in the activation of mitogen-activated protein kinase (c-Jun N-terminal kinase 1 [JNK1], JNK2, and p38) and protein tyrosine kinase signal transduction pathways (192). Pretreatment of cells with specific JNK1, JNK2, and protein tyrosine kinase inhibitors significantly reduced the internalization of N. meningitidis by HBMECs, without affecting bacterial adherence. Blocking of the α₅β₁ integrin receptor of HBMECs also decreased JNK activation in the presence of N. meningitidis. Furthermore, the use of Opc-deficient mutant strains demonstrated that JNK signaling, but not p38 signaling, was mediated by Opc expression (192). A more recent study identified that Opc preferentially binds to activated vitronectin within human serum and that this interaction promotes N. meningitidis invasion of HBMECs (155). In vitro, the enhanced invasion of nonencapsulated N. meningitidis strains is thought to be due to the unmasking of Opc; however, the role of Opc in vivo has yet to be determined.

S. pneumoniae. S. pneumoniae is internalized by human and rat brain microvascular endothelial cells via receptor-mediated endocytosis. This uptake does not involve the formation of cellular membrane protrusions as discussed for N. meningitidis (193). The adhesion of pneumococci to HBMECs is mediated by interactions between CbpA and the 37/67-kDa laminin receptor (151) and between neuraminidase A (NanA) and an unknown cellular receptor(s) (194). Uchiyama et al. demonstrated that NanA, a surface-anchored sialidase, was necessary and sufficient to promote pneumococcal adhesion to and invasion of HBMECs (195). Following initial attachment, the laminin G-like domain of NanA initiates chemokine signaling and inflammatory activation of endothelial cells (194). The activation of endothelial cells results in the increased expression of PAF receptor—the receptor for pneumococcal phosphorylcholine, which is a component of the cell wall and acts as a molecular mimic of the chemokine PAF (196).

In HBMECs, S. pneumoniae within vacuoles may (i) transit through the cell to the basolateral surface, (ii) recycle back to the apical surface, or (iii) be killed within the vacuole (193). Opaque S. pneumoniae variants, which express more capsular polysaccharide and less phosphorylcholine than transparent variants, were efficiently killed, suggesting that vacuoles containing these strains were targeted for lysosomal fusion. In contrast, transparent pneumococci may undergo some recycling back to the apical surface; however, by 5 h postinfection in a Transwell system, the majority of bacteria had translocated to the basolateral chamber (193). The invasion of HBMECs by transparent S. pneumoniae depends on phosphorylcholine, the PAF receptor, and β-arrestin, which targets G-protein-coupled receptors (such as PAF receptor) to clathrin-coated vesicles (193, 197). S. pneumoniae induces the translocation of β-arrestin from the cytosol to the plasma membrane, where it colocalizes with the PAF receptor. This interaction also stimulates mitogen-activated protein kinase signaling, which is required for pneumococcal uptake (197). Vacuoles containing S. pneumoniae colocalized with early and late endosomal markers; however, increased expression of β-arrestin subverted these vacuoles from lysosomal trafficking and promoted the transcytosis of viable bacteria. The small number of pneumococcus-containing vacuoles that underwent recycling to the apical surface demonstrated colocalization with Ras-related protein 11, which regulates endosome recycling. These findings suggest that the pneumococcus-induced interactions between PAF receptor and β-arrestin contribute to the transcellular penetration of HBMECs by S. pneumoniae (197).

Group B streptococcus.Several group B streptococcal proteins promote adhesion to and invasion of HBMECs (Table 1). The hypervirulent group B streptococcus adhesin (HvgA) was identified as a sequence type 17 (ST-17)-specific virulence factor which is anchored to the group B streptococcal cell wall by sortase A (198). ST-17 is a clonal complex belonging to the group B streptococcus capsular serotype III and is responsible for >80% of neonatal meningitis cases. Therefore, it was proposed that HvgA expression may contribute to the hypervirulence of ST-17. In vitro studies using an isogenic hvgA mutant demonstrated that HvgA was required for efficient group B streptococcal adhesion to HUVECs, primary rodent choroid plexus epithelial and brain microvascular endothelial cells, and immortalized HBMECs. Furthermore, in a murine model of hematogenous meningitis, HvgA was required for CNS invasion, although it had no effect on the levels of bacteremia (198).

A common characteristic of several other group B streptococcal adhesins/invasins is the ability to bind components of the HBMEC extracellular matrix, such as laminin, collagen, and fibronectin (109, 157, 159, 199). The group B streptococcal pilus tip adhesin (PilA) and the recently identified streptococcal fibronectin-binding protein A (SfbA) bind immobilized collagen and fibronectin, respectively. The collagen and fibronectin then associate with HBMECs via integrins, and these interactions facilitate the entry of group B streptococcus (109, 199). Similarly, the group B streptococcal serine-rich repeat 1 (Srr-1) glycoprotein binds the fibrinogen Aα chain within human blood through a dock, lock, and latch mechanism (200, 201). During this interaction, fibronectin docks between two IgG-like folds (N2 and N3 domains) of the binding region of the adhesin, which initiates a conformational change whereby a flexible region of the N3 domain latches to form a β-strand and completes a β-sheet within the N2 domain, effectively locking the ligand in place (200). Seo et al. demonstrated that deletion of the latch-like domain of the C-terminal end of the Srr-1 fibrinogen-binding region significantly reduced group B streptococcal adhesion to HBMECs (201). The fibrinogen-binding protein (FbsA) also mediates group B streptococcal adherence to HBMECs, by binding immobilized fibrinogen (156), although it is unknown if this interaction also occurs via a dock, latch, and lock mechanism.

Electron microscopy studies have demonstrated that in vitro, group B streptococcus associates closely with the cell membrane of HBMECs and is enveloped by microvillous structures (202). Group B streptococci are internalized within membrane-bound vesicles and translocate from the apical side to the basolateral side of HBMECs without a marked change in transendothelial electrical resistance. Serotype III group B streptococci were shown to invade HBMECs more efficiently than representative strains of serotypes Ia, Ib, II, and V; however, the serotype III capsule itself did not facilitate group B streptococcal invasion (202). Internalized group B streptococci did not undergo significant replication and survived within HBMECs for up to 20 h (202). Intracellular survival of group B streptococcus may be promoted by the CiaR-CiaH two-component regulatory system, which regulates several genes associated with stress tolerance and the subversion of host defenses (203). A serotype III strain that was deficient in the CiaR response regulator demonstrated adhesion and invasion levels similar to those of the wild-type group B streptococcus in HBMECs but was associated with a significant decrease in intracellular survival. Broth inhibition assays demonstrated that CiaR conferred resistance to antimicrobial peptides, lysozyme, and reactive oxygen species (203).

The host signal transduction pathways that are involved in the uptake of group B streptococcus by HBMECs are similar to those discussed for E. coli. In vitro, the trimolecular interactions between PilA, collagen, and α₂β₁ integrins on HBMECs stimulate the phosphorylation of focal adhesion kinase at tyrosine 397, which then induces host cytoskeletal rearrangements and the uptake of group B streptococci via the recruitment and activation of phosphatidylinositol 3-kinase and paxillin (109, 204). It has also been demonstrated that group B streptococcus invasion occurs via cytoskeletal rearrangements that are induced following the activation of RhoA and Rac1 (205). Furthermore, group B streptococcus induces serine 505 phosphorylation of host cytosolic phospholipase A₂, which leads to the release of arachidonic acid metabolites, including cysteinyl leukotrienes (206). In vitro pharmacological inhibition studies demonstrated that the activation of cytosolic phospholipase A₂ was required for efficient group B streptococcal invasion of HBMECs. In addition, brain colonization by group B streptococcus was significantly reduced in mice deficient in cytosolic phospholipase A₂ compared to wild-type animals in a model of hematogenous meningitis. Following cytosolic phospholipase A₂ activation and the release of cysteinyl leukotrienes, downstream signaling activates protein kinase Cα, which is involved in the regulation of actin cytoskeletal rearrangements (206).

S. suis.The zoonotic pathogen S. suis adheres to but does not invade HBMECs in vitro (207). In contrast, invasion of porcine brain microvascular endothelial cells has been observed for this swine pathogen. Inhibition studies demonstrated that S. suis invasion of porcine brain microvascular endothelial cells required actin filaments but not microtubular cytoskeletal elements or active bacterial RNA or protein synthesis (208). Electron microscopy studies showed that S. suis adhered to porcine brain microvascular endothelial cells at 5 min postinfection, and at 2 h postinfection, the streptococci were in close contact with the cells, within invaginated cell membrane structures and underneath the cell surface, behind the cell membrane. Unlike the case of group B streptococci, following internalization within membrane-bound vacuoles, the number of viable intracellular S. suis organisms steadily decreased over time, and viable cocci were not detected after 7 h (208). However, similar to the case with S. pneumoniae and group B streptococcus, the capsular polysaccharide of S. suis partially inhibited interactions with brain microvascular endothelial cells (208).

The S. suis proteins and virulence factors that mediate invasion of porcine brain microvascular endothelial cells remain largely unknown. Pretreatment of porcine brain microvascular endothelial cells with S. suis lipoteichoic acid reduced but did not abolish the adhesion and invasion of S. suis, in a dose-dependent manner, suggesting that lipoteichoic acid may play a role in S. suis interactions with porcine brain microvascular endothelial cells (209). Vanier et al. screened a transposon mutagenesis library of S. suis and identified several genes that may contribute to the invasion of porcine brain microvascular endothelial cells (210). These genes were characterized as encoding proteins belonging to the following groups: surface proteins, transport/binding proteins, regulatory functions, metabolism, amino acid synthesis, protein synthesis, and hypothetical proteins. However, the roles of these genes in S. suis invasion of porcine brain microvascular endothelial cells have not been confirmed using isogenic mutants.

As discussed above, the choroid plexus may be the site of entry to the CNS for S. suis. In order to study the interactions between S. suis and choroid plexus epithelial cells, Tenenbaum et al. developed an inverted Transwell model in which the translocation of S. suis from the basolateral, “blood” side to the apical, “CSF” side could be investigated (211). Following S. suis infection of the basolateral side, the transepithelial electrical resistance and paracellular permeability remained constant for up to 4 h. S. suis invaded porcine choroid plexus epithelial cells from the basolateral side in the inverted Transwell system; however, invasion was rare when S. suis was inoculated into the apical chamber in the standard Transwell model. These findings were also replicated in a human choroid plexus epithelial cell inverted Transwell model (140). This suggested that S. suis may interact with components of the extracellular matrix that are accessible on the basolateral side. The translocation of S. suis through porcine brain microvascular endothelial cells was inhibited when cells were treated with a phosphatidylinositol 3-kinase inhibitor, demonstrating that phosphatidylinositol 3-kinase may be involved in the uptake of S. suis (211).

L. monocytogenes and B. pseudomallei. L. monocytogenes can enter the CNS by the Trojan horse mechanism (see below) or by transcellular penetration of HBMECs. Using HUVECs and HBMECs, it was demonstrated that L. monocytogenes can invade endothelial cells directly or via cell-to-cell spread from adherent, infected mononuclear phagocytes (212, 213). In HUVECs, L. monocytogenes attaches to the cellular surface and induces membrane ruffling, which leads to internalization (214). In HBMECs, L. monocytogenes invasion is preceded by intimate interactions between the bacterium and microvilli on the cell surface (215). The invasion of L. monocytogenes into HBMECs requires actin microfilaments (213), but unlike the case for other meningeal pathogens, it does not involve the activation of phosphatidylinositol 3-kinase (213) or cytosolic phospholipase A₂ (175).

Once internalized, L. monocytogenes degrades the phagosome via the activity of a pore-forming toxin (listeriolysin) and two phospholipases (PlcA and PlcB), and it escapes into the cytoplasm by using actin-based motility, which is driven by the actin assembly-inducing protein, ActA. Following escape into the cytoplasm, L. monocytogenes can replicate within HUVECs and HBMECs and use actin-based motility to spread to adjacent cells (212, 213), thus avoiding an extracellular lifestyle. However, L. monocytogenes can be isolated from the CSF (216), suggesting that translocation through the cellular barriers of the CNS also occurs.

B. pseudomallei has an intracellular lifestyle similar to that of L. monocytogenes: the type III secretion system of B. pseudomallei (particularly the BopE and BopA effector proteins and the Bsa translocation apparatus) is required for escape from the vacuole and the subversion of autophagy (217 – 222). In an epithelial cell line or a murine macrophage-like cell line, cell-to-cell spread then occurs by cell fusion, with the formation of multinucleated giant cells, in a type VI secretion-dependent process. BimA, which is necessary for actin-mediated motility, and Fla2, which is thought to mediate intracellular motility, are also required and are considered to mediate cell-to-cell contact prior to cell fusion (223 – 227). To our knowledge, the ability of B. pseudomallei to directly invade HBMECs has not been investigated, but it would be reasonable to hypothesize that the intracellular lifestyle of B. pseudomallei may contribute to the pathogenesis of CNS invasion. Interestingly, a recent study investigated the variable virulence factors of B. pseudomallei associated with melioidosis in Australia and reported that B. pseudomallei strains harboring a bimA allele that shares 95% homology with B. mallei bimA (bimA _Bm) (228) were significantly associated with neurological melioidosis (229). Patients that were infected with B. pseudomallei harboring the bimA _Bm allele were found to be 14 times more likely to present with neurological involvement than patients infected with strains harboring the B. pseudomallei bimA variant (bimA _Bp) (229), indicating a role for actin-mediated motility in either transgression of the BBB/BCSFB or transit to the brain via the olfactory or trigeminal nerve pathways (see below).

Several studies have shown that internalin B (InlB) is required for L. monocytogenes invasion of HUVECs (212 – 214) and HBMECs (213, 215). Greiffenberg et al. demonstrated that although L. monocytogenes deficient in InlB could adhere to HBMECs at levels comparable to those of the wild-type strain, the invasive capability of the InlB mutant was reduced >100-fold (215). In contrast, one study reported that neither InlA nor InlB was required for L. monocytogenes invasion into HUVECs (230). It was suggested that these inconsistent findings may be attributable to differences in experimental conditions, especially involving the addition of normal human serum, which markedly affects InlB-mediated invasion due to the presence of anti-Listeria antibodies (231). In an inverted Transwell model of human choroid plexus epithelial cells, the cellular receptors for InlA and InlB (E-cadherin and met receptor tyrosine kinase, respectively) were expressed on the basolateral, “blood” side but not the apical, “CSF” side (232). In this model, L. monocytogenes invaded human choroid plexus epithelial cells exclusively from the basolateral side, and both InlA and InlB were required for efficient invasion.

Other L. monocytogenes virulence factors that may play a role in the attachment to and invasion of HBMECs include Vip and the autolysin IspC. Vip is a surface-expressed protein that is absent from nonpathogenic Listeria and binds to gp96 on host cells (160). A vip allelic replacement mutant was shown to be significantly less invasive than wild-type L. monocytogenes in Caco-2 cells and L2071 mouse fibroblasts but was not investigated in HBMECs. However, in intravenously inoculated mice, the Δvip strain was attenuated for virulence and associated with a significant decrease in bacterial loads within the brain (160). Similarly, IspC was also shown to contribute to L. monocytogenes invasion of the brain in a mouse model of hematogenous meningitis (233). Interestingly, the deletion of IspC did not affect the ability of L. monocytogenes to attach to and invade HBMECs in vitro. In contrast, the ΔispC strain demonstrated a significant reduction in attachment and invasion in sheep choroid plexus epithelial cells (compared to wild-type L. monocytogenes), and purified IspC was capable of binding to these cells. Proteomic analyses revealed that IspC regulates the expression of other virulence factors, including ActA, InlC2, and a flagellin homologue (FlaA); therefore, the phenotype observed for the ΔispC strain may be a result of the deletion of IspC in combination with a reduction in the expression of these factors (233).

M. tuberculosis. M. tuberculosis translocates from the apical chamber to the basolateral chamber in a Transwell model of HBMECs (118). In vitro, M. tuberculosis interacts with microvillus-like protrusions on HBMECs and is observed intracellularly but not paracellularly. Antibiotic protection assays demonstrated that M. tuberculosis invaded HBMECs. Furthermore, M. tuberculosis colocalized with actin on the cell surface. Treatment of HBMECs with cytochalasin D significantly decreased M. tuberculosis invasion, suggesting that actin polymerization is required for internalization (118).

The mechanisms and host-pathogen interactions involved in CNS invasion by M. tuberculosis remain incompletely understood. In an in vitro BBB model consisting of bovine brain microvascular endothelial cells (BBMECs) cocultured with rat astrocytes, the addition of purified recombinant heparin-binding hemagglutinin adhesin (rHBHA) induced actin filament rearrangement (234). This was shown to be dependent on the ability of rHBHA to bind to heparan sulfate glycosaminoglycans on the surfaces of BBMECs. The role of HBHA in brain microvascular endothelial cell invasion by M. tuberculosis has not been investigated in the context of live bacteria. Furthermore, gene expression profiling of M. tuberculosis associated with HBMECs in vitro (versus nonadherent bacteria) did not identify HBHA as a gene that was significantly upregulated during infection (118). Jain et al. demonstrated that 33 genes were upregulated >8-fold in M. tuberculosis associated with HBMECs. Transposon mutants were created to determine if these upregulated genes were associated with HBMEC invasion; of the 33 tested genes, 4 (Rv0980c, Rv0987, Rv0989c, and Rv1801) were required for efficient invasion and/or intracellular survival (118).

In vivo screening of transposon mutant libraries has also been employed to identify potential M. tuberculosis virulence factors that are associated with CNS invasion. Be et al. inoculated pools of M. tuberculosis mutants intravenously into BALB/c mice and determined that mutants deficient in the expression of 5 genes (Rv0311, Rv0805, Rv0931c, Rv0986, and MT3280) were significantly attenuated in the brain but not the lungs. In vitro studies confirmed that these mutants invaded HBMECs at significantly reduced levels compared to that of a negative-control transposon mutant (235). Rv0931c (also known as pknD) encodes a serine-threonine protein kinase and was also identified as a key microbial factor required for brain infection in a guinea pig model of hematogenous meningitis, from a library of 398 mutants (158). When studied independently in a mouse model of infection, the pknD mutant was associated with significant reductions in bacterial loads within the brain but not the lungs. In vitro, pknD was required for efficient invasion and intracellular survival within HBMECs; this phenotype was not observed in HUVECs, A549 lung epithelial cells, or J447 macrophages. Microspheres coated with pknD adhered to HBMECs at significantly higher levels than those of microspheres coated with bovine serum albumin, and it was proposed that pknD binds to HBMECs via interactions with laminin within the extracellular matrix (158).

E. coli.Several E. coli virulence factors promote the destruction of tight junctions between brain microvascular endothelial cells. In addition to triggering the internalization of E. coli K1 into HBMECs, the binding of OmpA to its gp96 homologue receptor stimulates inducible nitric oxide synthase and nitric oxide production, which enhances the generation of cyclic GMP (237). This increase in cyclic GMP activates protein kinase Cα (237), which leads to a Ras GTPase-activating-like protein (IQGAP1)-mediated dissociation of β-catenin from vascular endothelial cadherin at adherens junctions (238). Immunofluorescence experiments demonstrated that vascular endothelial cadherin was redistributed from the intercellular junctions to the sites of bacterial attachment (239). This resulted in a significant increase in HBMEC permeability to horseradish peroxidase and a parallel decrease in transendothelial electrical resistance (239). Furthermore, the activation of host cytosolic phospholipase A₂ by E. coli OmpA and Nlp1 may play a role in reducing the integrity of the BBB. Using cocultures of primary BBMECs and primary bovine retinal pericytes seeded onto Transwell membranes, Salmeri et al. demonstrated that the arachidonic acid released following cytosolic phospholipase A₂ activation acts as a substrate for cyclooxygenase to produce prostaglandins, which trigger the synthesis of vascular endothelial growth factor by brain endothelial cells (119). This was associated with a significant decrease in transendothelial electrical resistance and a dramatic detachment of pericytes from the Transwell membrane. These effects were not observed when the pericyte vascular endothelial growth factor receptor 1 was blocked using a polyclonal antibody. These findings suggest that the vascular endothelial growth factor released by brain endothelial cells may bind to the vascular endothelial growth factor receptor 1 on adjacent pericytes and trigger ablation of the pericytes from the basal membrane, potentially opening a paracellular route (119).

In vitro studies using the mouse brain microvascular endothelial cell line Bend.3 have demonstrated a role for lipopolysaccharide (LPS; from E. coli O55:B5) in inducing BBB permeability. Similar to OmpA, LPS was shown to activate protein kinase C isoforms, which led to the activation of RhoA and the stimulation of downstream NF-κB signaling and myosin light chain phosphorylation (240, 241). These signaling events were associated with a decrease in the expression of the tight junction proteins claudin-5 and zonula occludens-1 and a reduction in transendothelial electrical resistance. LPS also induced changes in F-actin organization, leading to paracellular gaps and stress fiber formation (240). In Shiga toxin-producing E. coli associated with hemolytic-uremic syndrome and CNS involvement, both Shiga toxin 1 and Shiga toxin 2 induce cytotoxic effects that lead to a decrease in BBB integrity and to cellular injury in vivo and in vitro (242 – 245). LPS may mediate these effects (242, 245).

H. influenzae.Quagliarello et al. first demonstrated that H. influenzae causes an increase in BBB permeability and tight junction disruption in adult Wistar rats following intracisternal inoculation of approximately 10⁶ CFU (236). Using this intracisternal inoculation model, the role of H. influenzae type B lipooligosaccharide (LOS) was investigated. Purified LOS and outer membrane vesicles containing non-cell-associated LOS induced a dose-dependent increase in BBB permeability to ¹²⁵I-bovine serum albumin, as well as CSF pleocytosis (246, 247). In leukopenic rats, no increases in BBB permeability and CSF leukocyte counts were observed following LOS or outer membrane vesicle challenge, suggesting that the decrease in BBB integrity is mediated by the intact host leukocyte response (246, 247). In contrast, Tunkel et al. demonstrated that H. influenzae type B LOS significantly increased the permeability of rat brain microvascular endothelial cells to ¹²⁵I-albumin in the absence of host inflammatory cells in vitro (248). Furthermore, in the absence of monocytic cells, purified LOS was cytotoxic to BBMECs. However, in vivo studies have shown that nitric oxide production within the CSF is required for LOS-mediated pleocytosis and increased BBB permeability (249) and that host PAF synergizes with H. influenzae LOS to augment these events (250).

Peptidoglycan within the H. influenzae cell wall may also contribute to BBB/BCSFB dysfunction in vivo. Intracisternally inoculated peptidoglycan induced an increase in BBB permeability to ¹²⁵I-albumin (251), as well as CSF leukocytosis (252), although the CSF leukocytosis was attenuated compared to that in animals challenged with LOS (252). It was subsequently hypothesized that H. influenzae cell wall components, such as peptidoglycan, activate brain microvascular endothelial cells and induce the separation of intercellular junctions. LOS may then augment this response by activating the recruited leukocytes to produce proinflammatory cytokines that further contribute to BBB breakdown (252); however, the mechanisms by which this may occur have not been elucidated. The H. influenzae type B porin may also contribute to these events. The inoculation of purified porin into the fourth cerebral ventricle of rats induced the expression of interleukin-1α (IL-1α), tumor necrosis factor alpha (TNF-α), and macrophage inflammatory protein 2, which preceded an increase in BBB permeability to serum proteins and leukocytes and an increased brain water content (108).

N. meningitidis.As discussed above, N. meningitidis-induced β2-adrenoceptor/β-arrestin signaling in HBMECs results in the formation of cortical plaques and membrane protrusions. In parallel, the type IV pilus-mediated translocation of β-arrestin to the cortical plaque also acts as a molecular scaffold for additional N. meningitidis-induced signaling events that open the paracellular route (187). These events include the recruitment of the Par3-Par6-PKCζ polarity complex to the cortical plaque, in a Cdc42-dependent manner, and the accumulation of adherens junction (vascular endothelial cadherin, p120-catenin, and β-catenin) and tight junction (zonula occludens-1, zonula occludens-2, and claudin-5) proteins underneath the site of bacterial adhesion (253). Labeling of vascular endothelial cadherin demonstated that N. meningitidis delocalized this protein from the adherens junctions between HBMECs and caused it to be redistributed to the cortical plaque. This depletion of junctional proteins was associated with an increase in barrier permeability to Lucifer yellow and the formation of gaps between cells, thus opening a paracellular route of entry (253). Interestingly, the recruitment of junctional components by N. meningitidis occurs only in endothelial cells, as it was demonstrated that meningococci do not activate the β2-adrenoceptor/β-arrestin signaling pathway in epithelial cells (254). Therefore, different host-pathogen interactions may occur at the BBB and BCSFB.

In an inflammation-mediated process, N. meningitidis stimulates HBMECs to produce matrix metalloproteinase-8, which was shown to cleave occludin and cause it to dissociate from tight junctions in vitro (255). The cleavage of occludin was prevented in the presence of matrix metalloproteinase-8 inhibitors and in cells transfected with matrix metalloproteinase-8-specific small interfering RNA (siRNA). Inhibition of matrix metalloproteinase-8 activity also significantly decreased the N. meningitidis-induced permeability of HBMECs to 40-kDa fluorescein isothiocyanate (FITC)-dextran at 24 h. Furthermore, the inhibition of matrix metalloproteinase-8 prevented the caspase-independent detachment of HBMECs from a Matrigel matrix on a solid support following N. meningitidis infection (255).

Streptococcus spp.Group B streptococcus, S. pneumoniae, and S. suis invade brain microvascular endothelial cells via transcellular mechanisms; however, the associated production of proinflammatory cytokines, the activity of pore-forming toxins, and the induction of plasmin activity may also lead to a concurrent weakening of the BBB, potentially opening a paracellular route of entry. In vivo and in vitro studies have shown that group B streptococci, pneumococci, and S. suis induce the production of a range of cytokines and chemokines, including IL-1β, IL-6, IL-8, IL-10, TNF-α, macrophage chemoattractant protein 1 (MCP-1), granulocyte-macrophage colony-stimulating factor (GM-CSF), Gro-α, and Gro-β, in brain microvascular endothelial cells (109, 256 – 258) and brain tissue (94, 259, 260). The stimulation of HBMECs with cytokines (including TNF-α, IL-1β, IL-6, and IL-17F) leads to cytoskeletal rearrangements and a redistribution of tight junction and adherens junction proteins, resulting in a decrease in barrier integrity (261 – 267). Therefore, the host inflammatory response to group B streptococcus, S. pneumoniae, and S. suis meningitis may contribute to the disruption of tight junctions. In HBMECs, the group B streptococcal PilA-mediated activation of focal adhesion kinase induces bacterial uptake but also triggers a parallel mitogen-activated protein kinase/extracellular signal-regulated kinase pathway, which leads to the secretion of IL-8 (109). Furthermore, the production of the IL-8 homologue (KC) in mice intravenously inoculated with group B streptococci resulted in the recruitment of neutrophils and an increase in BBB permeability to Evans blue and FITC-albumin (109). It is likely that disruption of the BBB/BCSFB by host inflammatory factors represents a common physiological event that occurs during bacterial CNS invasion.

Using a separate mechanism, group B streptococcus, S. pneumoniae, and S. suis may hijack the host plasminogen system to cause cellular injury and disrupt tight junctions between HBMECs. Plasminogen is found in plasma and extracellular fluids, and upon activation is converted to plasmin, which degrades the extracellular matrix and may upregulate matrix metalloproteinases that degrade tight junctions (268). In vitro, plasminogen binds to streptococcal glyceraldehyde-3-phosphate dehydrogenase (269 – 271). Enolase and choline-binding protein E have also been identified as additional pneumococcal plasminogen-binding proteins (272, 273). Once plasminogen is bound to its binding protein, it becomes activated and is subsequently converted to plasmin by exogenous tissue plasminogen activator and urokinase (269, 270, 274). Plasmin bound to the surfaces of group B streptococci enhanced the ability of bacteria to adhere to and invade HBMECs and induced a significant decrease in transendothelial electrical resistance (275). Furthermore, streptococcus-bound plasmin contributed to cellular injury, characterized by human brain microvascular cell detachment and lactate dehydrogenase release (275). In the human vascular endothelial cell line EaHy, plasmin bound to the pneumococcal cell surface cleaved vascular endothelial cadherin from adherens junctions and increased bacterial translocation across endothelial barriers (276). Taken together, group B streptococcus, S. pneumoniae, and S. suis may bind host plasminogen and hijack the proteolytic activity of plasmin to potentially open a paracellular route between cells. Plasminogen has also been shown to bind to N. meningitidis (277, 278), suggesting that this mechanism may be exploited by several common meningeal pathogens.

Group B streptococcus, S. pneumoniae, and S. suis produce pore-forming hemolysins, named β-hemolysin/cytolysin, pneumolysin, and suilysin, respectively. In HBMECs, these hemolysins are cytotoxic and promote lactate dehydrogenase release, the loss of cytoplasmic density, discontinuity of cytoplasmic membranes, clumping of nuclear chromatin, dilation of the endoplasmic reticulum, cell rounding, and detachment (97, 202, 207). These morphological changes most likely result in the formation of paracellular gaps. The effects of pneumolysin have also been investigated in rat astrocytes; the addition of pneumolysin to monolayers of primary astrocytes resulted in cell shrinkage and the subsequent separation of cells from each other (107). In rat brain slices, exposure to pneumolysin led to astrocytic process retraction and reorganization of astrocytes within the glia limitans. These pneumolysin-induced changes were associated with increased interstitial fluid retention and BBB permeability, which facilitated the penetration of macromolecules and bacteria into brain slices (107). These data suggest that in addition to mediating cytotoxic effects in brain microvascular endothelial cells, bacterial hemolysins may also induce morphological changes in other components of the neurovascular unit.

S. pneumoniae.Early studies by Rake demonstrated that S. pneumoniae rapidly enters the olfactory bulb in Swiss mice following intranasal inoculation (329). Remarkably, pneumococci were isolated from the olfactory bulbs of the brain at 1 min postinfection. At this time point, bacteria were not isolated from the blood, effectively excluding a hematogenous route of CNS invasion. At 2 min postinfection, pneumococci were observed between the sustentacular cells of the olfactory epithelium, within the perineural space of the olfactory nerve, and within the subarachnoid space (329). Similar findings were reported by van Ginkel et al., who demonstrated that intranasally delivered S. pneumoniae (strain EF3030) could be recovered from nasal washes, the olfactory nerve and epithelium, and the olfactory bulb at 24 h postinfection in CBA/CAHN/xid mice (330). These sites remained colonized throughout the duration of the experiment, until day 39 postinfection; however, this did not lead to extensive brain infection. In contrast, nonencapsulated pneumococci were unable to persist within the nasal cavity and olfactory epithelium and did not invade the olfactory bulb within the brain. Briles et al. confirmed that the pneumococcal capsular polysaccharide is an important mediator of olfactory bulb infection in CBA/CAHN/xid mice, as only opaque variants were isolated from the olfactory bulb following intranasal inoculation (331). In these studies, intranasally delivered S. pneumoniae did not lead to bacteremia, suggesting a role for direct brain invasion from the nasal cavity. S. pneumoniae was also recovered from the trigeminal ganglia, further suggesting that the trigeminal nerve bundles may also be a potential route to the brain for this bacterial pathogen (330). However, it must be noted that these studies utilized CBA/CAHN/xid mice, which possess a mutation in Bruton's tyrosine kinase gene and thus do not respond to thymus-independent type II antigens (332). These mice also fail to respond to capsular polysaccharide; therefore, pneumococcal infection in this mouse strain is unlikely to accurately model human disease.

Incubation of S. pneumoniae with gangliosides prior to intranasal infection resulted in reduced colonization of the nasal mucosa, olfactory system, and brain (330). Phosphorylcholine residues on the pneumococcal cell wall have been shown to bind to gangliosides (333), suggesting that gangliosides may be an important target for S. pneumoniae attachment in the neuroepithelium. In vitro studies have also demonstrated that S. pneumoniae can invade olfactory ensheathing cells via mannose receptor-mediated endocytosis (334); however, the role of olfactory ensheathing cells in pneumococcal infection and the exact mode of transport of S. pneumoniae along the olfactory and/or trigeminal nerves have not yet been studied.

N. meningitidis.In cases of meningococcal meningitis, >60% of patients develop meningitis without septic shock (335), suggesting the possibility that N. meningitidis may also penetrate the CNS without prior bacteremia. In a mouse model of intranasal N. meningitidis infection, 20% of mice developed lethal meningitis; however, no bacteria were isolated from the blood (324). At 3 days postinfection, lesions and polymorphonuclear cells were observed within the olfactory epithelium. Furthermore, N. meningitidis infection was associated with significant damage to the olfactory epithelium and reduced the thickness of discrete, noncontinuous sections of epithelium by more than 50% compared to the case in controls. Immunofluorescence studies demonstrated N. meningitidis within the olfactory epithelium, basement membrane, and lamina propria, along the olfactory nerves in the cribriform plate, and within the nerve fiber layer of the olfactory bulb. Meningococci were also isolated from the CSF. Due to the absence of bacteremia, it was suggested that N. meningitidis may travel from the nasal cavity to the meninges and subarachnoid space via the olfactory nerves (324). In rhesus macaques, commensal neisseria bacteria (RM Neisseria) were shown to be transmitted between animals and naturally colonized the epithelium covering the cribriform plate (336), suggesting that migration of neisseria bacteria along the olfactory pathway is not a phenomenon that is observed only in animals experimentally inoculated with pathogenic N. meningitidis. In the African “meningitis belt,” the onset of meningococcal meningitis epidemics coincides with harsh, dry harmattan winds. It has been hypothesized that these harsh environmental conditions may irritate the mucus membranes (such as the olfactory mucosa) and enable N. meningitidis to penetrate the epithelium and invade the CNS (337), although this has not yet been investigated.

Sjölinder and Jonsson demonstrated that the expression of the junctional protein N-cadherin in the olfactory epithelium and lamina propria was significantly reduced in mice intranasally infected with N. meningitidis (324). These data suggest that N. meningitidis may invade the olfactory epithelium via the destruction or rearrangement of cellular junctions between sustentacular cells and olfactory sensory neurons, although additional studies are required to confirm these findings. The mechanisms by which N. meningitidis may travel along the olfactory nerve pathway to invade the nerve fiber layer of the olfactory bulb, the meninges, and CSF are also unknown. The meninges and the subarachnoid space extend over the cribriform plate and further into the olfactory foramen, where the olfactory nerves pass through the cribriform plate (338, 339). Bacteria traveling along the olfactory pathway may potentially invade the meninges after traversing the cribriform plate, and subsequently enter the CSF. However, there is a positive pressure from the brain to the nasal lymphatics due to the drainage of CSF through the cribriform plate, and it is unknown how N. meningitidis could travel against this pressure gradient.

B. pseudomallei.The olfactory pathway has been identified as a route of CNS invasion by B. pseudomallei. Owen et al. used a capsule-deficient strain, which fails to survive within the blood, to demonstrate that bacteria were present in the olfactory epithelium and olfactory bulb of BALB/c mice at 24 h postinfection, in the absence of bacteremia. It was also shown that when colonization occurred in only one side of the nasal cavity, wild-type B. pseudomallei was detected in the olfactory epithelium and olfactory bulb of the infected side only. Combined, these data suggest that B. pseudomallei is able to directly invade the brain via the olfactory pathway, without blood-borne infection (105). These findings also demonstrate that the presence of the polysaccharide capsule is not a prerequisite for CNS invasion via the olfactory pathway, presumably in contrast to systemic invasion via the BBB/BCSFB. However, it might be the case that an absence of capsule does diminish translocation via the olfactory or trigeminal nerves, since it is not possible to definitively compare with the presence of capsule, due to hematogenous spread in the latter case. Figure 3D and E illustrate B. pseudomallei invasion of the nasal cavity, olfactory epithelium, and ventral nerve fiber layer of the olfactory bulb in a mouse model of acute melioidosis. It may be noted that mice, like most mammals, are naturally susceptible to melioidosis.

The mechanisms by which B. pseudomallei migrates along the olfactory nerves were recently investigated (326). In BALB/c mice, we demonstrated that intranasally delivered B. pseudomallei was associated with widespread crenellation of the olfactory epithelium and loss of the neuron cell bodies. This neuronal loss led to the degeneration of olfactory axons within 24 to 48 h, and the axon-devoid, hollow olfactory nerve fascicles, surrounded by olfactory ensheathing cells, provided an open conduit for bacterial passage from the nasal cavity through the cribriform plate and into the nerve fiber layer of the olfactory bulb. No bacteria were observed within the supporting sustentacular cells of the olfactory epithelium, nor within the Bowman's glands (326). By migrating within the nerve sheath, B. pseudomallei would be shielded from the flow of CSF out of the brain. This may also explain why B. pseudomallei is often not isolated from the CSF in cases of neurological melioidosis (56). During our initial studies in mice, we also demonstrated that B. pseudomallei was present in the brain stem, increasing in number for up to 48 h postinfection, indicating, though not definitively demonstrating, that bacteria may enter via the trigeminal nerve after intranasal inoculation (105). We have now confirmed that B. pseudomallei penetrates the intact respiratory epithelium in mice and migrates along Schwann cell-encased trigeminal nerve bundles to the cranial cavity (326). Figure 4B to D demonstrate B. pseudomallei localized within the branches of the trigeminal nerve. As noted above, clinical evidence points to a role for BimA, and hence actin-mediated motility, in neurological melioidosis (229); assuming inhalational acquisition to be associated with such cases, it remains to be determined where in the olfactory/trigeminal pathways such motility is utilized.

Although there is currently no direct evidence of B. pseudomallei penetration of the brain via the olfactory and/or trigeminal pathway in humans, two important considerations provide support for CNS invasion from the nasal mucosa. First, B. pseudomallei can colonize the nasal cavity in humans (340), and this may represent a potential reservoir of infection. Our unpublished data also demonstrate that following the inhalation of aerosolized B. pseudomallei, considered a major route of infection in humans, bacteria can be recovered in large numbers from the mouse olfactory epithelium. Second, for human cases of primary neurological melioidosis, Currie et al. reported that bacteremia was observed in only 3 of 14 patients (21%) (57). In addition to our animal studies, these data provide evidence that B. pseudomallei can invade the CNS by a nonhematogenous route. Furthermore, the trigeminal route of CNS invasion may explain the brain stem-related neurological presentations of melioidosis patients, without requiring frank encephalitis.

L. monocytogenes.In ruminants naturally infected with L. monocytogenes, bacteria migrate along several cranial nerves (including the trigeminal nerve) to the brain stem and cause encephalitis (341). Similar findings are observed in human cases of L. monocytogenes brain stem encephalitis, where inflammatory lesions may be observed in the nuclei and tracts of cranial nerves V, VII, IX, X, and XII, innervating the oropharynx (342). Of the cranial nerves, the trigeminal nerve pathway is thought to represent a primary route by which L. monocytogenes may invade the brain stem from the oropharynx (343, 344). It has been proposed that L. monocytogenes may bind to E-cadherin on Schwann cells surrounding trigeminal nerves and subsequently invade the trigeminal axons by receptor-independent cell-to-cell spread (345). However, binding of L. monocytogenes to E-cadherin-expressing Schwann cells has not yet been demonstrated.

In animals spontaneously infected and experimentally infected with L. monocytogenes, bacteria were observed within the axons of the trigeminal nerve, suggesting that these bacteria may use intra-axonal spread to migrate to the CNS (343, 344, 346). In a mouse model of brain stem encephalitis, L. monocytogenes was inoculated into the snouts of immunodeficient mice and was observed within the cytoplasm of trigeminal nerve cell bodies at 7 days postinfection (344). Following this, bacteria were then observed within the brain stem but not in other regions of the brain. In this model, gamma interferon (IFN-γ) release by natural killer cells was required for efficient control of L. monocytogenes neuroinvasion via the trigeminal route (344). In vivo and in vitro studies have demonstrated that colony-stimulating factor 1-dependent cells (including macrophages and dendritic cells) facilitate the neuronal spread of L. monocytogenes to the CNS (347, 348). Following replication and escape from the phagosome in colony-stimulating factor 1-dependent cells, L. monocytogenes may use actin-based motility to propel itself into adjacent nonphagocytic cells, such as axon terminals. Indeed, the presence of L. monocytogenes with actin tails has been demonstrated in vivo in the axons of sheep trigeminal nerves (346) and in vitro in mouse hippocampal neurons and hypothalamic neurons (349). Dramsi et al. showed that L. monocytogenes spreads from J774 macrophages to neurons in vitro, in a process that is dependent on actin polymerization by the bacterium (348), suggesting that the formation of actin tails is critical for neuronal spread in listeriosis.