Escherichia coli at the Blood-Brain Barrier Interface

E. coli is a gram-negative bacterium implicated in the majority of cases of neonatal meningitis. In vitro studies of infection of human brain microvascular endothelial cells (HBMECs) by E. coli K1, an encapsulated strain responsible for most cases of neonatal meningitis, have shown that E. coli K1 interacts with these cells in a unique manner involving both host- and pathogen-specific structures and signaling pathways. These interactions lead to alterations in the host actin cytoskeleton, membrane protrusion and ruffling around bacteria, and endocytosis of bacteria into membrane-bound vacuoles, where bacterial determinants act to prevent lysosome fusion and influence intracellular vacuole trafficking to achieve transcytotic passage.² Several bacterial determinants have been identified as part of the initial binding and invasion of HBMECs, including type 1 fimbriae, outer membrane protein A (OmpA), Ibe proteins, and cytotoxic necrotizing factor-1 (CNF-1).² Type 1 fimbriae are adhesins that bind to α-D-mannosides on the surface of host cells, thereby allowing binding and interaction of the bacterium with the host cell. Type 1 fimbriae have been shown to play an important role in the binding of E. coli K1 to HBMECs; deletion of fimH, the gene for the major adhesin protein of type 1 fimbriae, significantly decreases binding of E. coli K1 to HBMECs, a finding that is reversed by genetic complementation of fimH in deletion mutants.³ OmpA also facilitates binding of E. coli K1 to HBMECs via interaction with surface glycoproteins containing N-acetylglucosamine residues.⁴ In vivo studies using an experimental neonatal rat model of meningitis have shown that deletion mutants of ompA are impaired in their ability to enter the CNS in comparison to the parent K1 strain, and N-acetylglucosamine oligosaccharides are able to block penetration of the CNS by wild-type E. coli K1 in the same animal model.⁵

Once binding has occurred, the process of cellular invasion must take place for E. coli K1 to ultimately infect the CNS. Invasion of HBMECs is dependent on the host actin cytoskeleton; E. coli K1 invasion can be completely inhibited in vitro by using inhibitors of the actin cytoskeleton such as cytochalasin D.⁶ Entry of E. coli K1 into membrane-bound vacuoles in HBMECs involves the formation of membrane projections, described as zipper-like structures, and subsequent membrane ruffling before internalization (Fig. 39-1).⁷ These events are related to several important signal transduction pathways in the host cell known to be involved in the regulation of endocytosis, cell membrane interactions, and the actin cytoskeleton. In particular, E. coli K1 leads to activation of focal adhesion kinase (FAK) and phosphorylation of paxillin, a cytoskeletal protein that interacts with and regulates the actin cytoskeleton.⁸ The exact mechanisms underlying FAK activation by E. coli K1 are unknown but appear to play a role in HBMEC invasion because overexpression of a dominant-negative form of FAK significantly inhibits HBMEC invasion by E. coli K1.⁸

FIGURE 39-1 Transmission electron micrographs of Escherichia coli K1 strain RS 218 undergoing attachment (A) and internalization (B and C) into human brain microvascular endothelial cells. Scale bar, 1 µm.

(From Kim KS. Microbial translocation of the blood-brain barrier. Int J Parasitol. 2006;36:610.)

Another important signal pathway involved in E. coli K1 invasion of HBMECs is the family of phosphatidylinositol-3′-kinases (PI3Ks), which lie downstream of FAK activation. Pharmacologic inhibition of PI3Ks with LY-294002 significantly limits HBMEC invasion by E. coli K1.⁹ Akt/protein kinase B, a downstream effector of PI3K, is increased during E. coli K1 invasion of HBMECs.⁹ These observations indicate a role for the PI3K/Akt pathway in E. coli invasion of HBMECs, although the exact roles that these kinases play in the process have yet to be described.

Rho family guanosine triphosphatases (GTPases) have been shown to regulate a diverse array of processes that affect the actin cytoskeleton, cell motility, and cell-cell/cell-matrix interactions.¹⁰ E. coli strains associated with urinary tract infections and meningitis have been shown to produce CNF-1, an AB-type toxin with deaminase activity that targets this family of GTPases.¹¹ CNF-1 activates Rho GTPases and increases the in vitro uptake of bacteria by nonprofessional phagocytes such as epithelial and endothelial cells.¹¹ Deletion of the cnf1 gene from E. coli K1 significantly decreases the invasion of HBMECs, and this deletion is associated with decreased activation of RhoA and Cdc42, GTPases that are activated during E. coli K1 invasion of these cells.¹¹

Once E. coli K1 bacteria have been internalized in membrane-bound vacuoles within HBMECs, these organisms must survive the internal hostile milieu and avoid destruction in the lysosome to gain entry into the CNS. The K1 capsule appears to play a very important role in preventing the normal maturation of endosomes and fusion of vacuoles with the lysosome. K1 isogenic deletion mutants (K1⁻) have been shown to traffic through the endosomal system and colocalize with cathepsin D, thus confirming fusion of the lysosome with the vacuoles containing these bacteria.¹² The exact mechanisms underlying the K1 capsular effect on endosomal trafficking have yet to be elucidated, however.

In summary, a neuroinvasive strain of E. coli uses specific tools to gain entry into the CNS by binding/uptake into HBMECs and subsequent diversion of the normal protective trafficking of endosomal compartments to the lysosome (Fig. 39-2). These events allow E. coli K1 strains to cross the highly selective BBB and cause meningitis. The next section reviews the most recent information pertaining to crossing the BBB by GBS, gram-positive bacteria that remain a prominent cause of sepsis and meningitis in neonates and infants.

FIGURE 39-2 Summary diagram of Escherichia coli K1 binding and invasion of the blood-brain barrier. CNF-1, cytotoxic necrotizing factor-1; FAK, focal adhesion kinase; GTPase, guanosine triphosphatase; OmpA, outer membrane protein A; PI3K, phosphatidylinositol-3′-kinase.

Group B Streptococci: If You Can’t Beat Them, Join Them

GBS (i.e., Streptococcus agalactiae) are highly effective at colonizing the human female genitourinary tract and are responsible for significant morbidity and mortality in neonates and infants, predominantly in the form of pneumonia, sepsis, and meningitis. GBS use both similar and different strategies to gain entry into the CNS when compared with E. coli K1; these strategies are discussed here to highlight the diversity of virulence factors and pathogenic mechanisms used by neuroinvasive pathogens.

Like E. coli K1 strains, GBS express a polysaccharide capsule that serves many functions in the process of host invasion, including protection from opsonization, complement-mediated lysis, antibody-mediated clearance, and phagocytosis.¹³ Nine distinct GBS capsular serotypes have been identified, with serotype III strains dominating the clinical isolates associated with meningitis.¹³ Virulent strains of GBS have terminal sialic acid residues coating the surface of the polysaccharide capsule; GBS strains lacking these residues are less virulent than isogenic counterparts carrying these sialic acid groups.¹⁴ Many host glycoproteins have the same terminal sialic acid residues (α2 → α3 N-acetylneuraminic acid), so the presence of these residues on the GBS capsule may provide the bacterium with some protection from immune surveillance via molecular mimicry.¹³ These residues also inhibit activation of the alternative complement pathway, thus preventing opsonophagocytosis of GBS.¹³

Like E. coli K1, GBS must bind to the luminal surface of BMECs to initiate invasion of the CNS. Importantly, GBS are able to bind to several components of the extracellular matrix, including fibrinogen, laminin, and immobilized but insoluble fibronectin. ScpB, a C5 peptidase anchored in the GBS membrane, has been identified as a selective fibronectin-binding protein that differentially associates with bound fibronectin.¹³ FbsA is a surface-anchored protein that binds fibrinogen, and the gene fbsA is regulated by RogB, a transcriptional regulator that positively regulates several genes involved in binding to the extracellular matrix.¹³

Recently, GBS pili have been shown to play a role in GBS interactions with HBMECs, and targeted deletion of the gene for the pilus accessory surface protein PilA significantly reduces the ability of GBS to adhere to HBMECs, whereas deletion of pilB, the gene encoding the major pilus structural protein, does not affect adherence but significantly reduces HBMEC invasion by GBS.¹⁵

The dissociation of adherence and invasion observed with the pili mutants just described supports the hypothesis that these events, adherence and invasion, are mediated by distinct virulence factors expressed by GBS. Other invasion mediators that have been identified for GBS and HBMECs include a gene involved in modification of lipoteichoic acid (LTA), expression of a β-hemolysin/cytolysin, and factors that alter host cell signaling pathways. LTA is a major structural component of the surface of gram-positive bacteria and may mediate amphiphilic interactions with host cell membrane phospholipids. Allelic exchange of a glycosyltransferase homologue in GBS, iagA, reduces LTA anchoring in the GBS cell wall and significantly reduces HBMEC invasion in vitro and the development of meningitis in a murine in vivo model.¹⁶ β-Hemolysin/cytolysin is a pore-forming exotoxin expressed by GBS that is known to promote GBS invasion of HBMECs at sublytic concentrations and cause HBMEC cytolysis at higher concentrations.¹⁷ β-Hemolysin/cytolysin also induces HBMEC synthesis of interleukin-8 (IL-8), an extremely potent neutrophil chemoattractant, and promotes neutrophil transmigration across HBMEC monolayers.¹⁷ Mice hematogenously infected with a GBS mutant lacking this exotoxin had lower brain bacterial counts and lower mortality than did mice infected with the parent wild-type strain expressing this toxin.¹⁷

As with E. coli K1, GBS have been shown to modulate a number of host signaling pathways to facilitate passage across the BBB. In fact, the FAK/paxillin/PI3K pathway is a common target involved in both E. coli K1 and GBS invasion of HBMECs. Inhibition of FAK signaling via a dominant-negative form of FAK and pharmacologic inhibition of PI3K with LY-294002 both significantly inhibit GBS invasion of HBMECs.¹⁸ Another shared pathway involved in E. coli K1 and GBS invasion of HBMECs involves RhoA GTPase; RhoA levels are increased during GBS invasion of HBMECs, as are levels of Rac1.¹⁹ Inhibition of RhoA and Rac1 with a geranylgeranyl transferase I inhibitor, GGTI-288, and expression of dominant-negative forms of these GTPases both result in significantly reduced HBMEC invasion by GBS.¹⁹

In summary, GBS take advantage of cloaking, molecular mimicry, exotoxin synthesis, and usurpation of host signaling pathways to bypass the BBB and cause meningitis. The common theme of modulating FAK and Rho GTPase signaling pathways during HBMEC invasion by E. coli K1 or GBS may provide future therapeutic targets for the prevention or treatment of CNS infections caused by these pathogens.

Microglia: Ramón y Cajal’s “Third Element”

Santiago Ramón y Cajal (1852-1934) divided the histology of the brain into “elements,” including neurons (first element), astrocytes (second element), and a third category of non-neuronal, nonastrocytic cells that he termed the third element. This third element contained cells later identified as oligodendrocytes, as well as a group of small, highly branched cells that were morphologically distinct from other cells in the CNS. These cells were eventually termed microglia by Pio del Rio Hortega (1882-1945), who went on to further characterize these cells as a distinct entity in the brain parenchyma.²⁰ Like monocytes and macrophages, these cells are derived from the bone marrow and share many features of monocytes/macrophages with respect to immune modulation.²⁰

Microglia can exist in several different states, as defined by surface markers, morphology, migration status, and function. Resting microglia are small cells with few surface markers and prominent thin branches that are constantly reorganizing and sampling the microenvironment of the brain parenchyma.²¹ A large number of stimuli activate resting microglia, with the nature of the stimulus influencing the structural and functional changes that occur during activation. Cellular debris from CNS damage, particularly free adenosine triphosphate, activates microglia and induces a morphologic change from a small, ramified cell to an ameboid cell capable of phagocytosing such debris in a manner similar to macrophages.²¹ Microglia, like macrophages, express a large number of receptors associated with various mediators of the inflammatory response to tissue damage, pathogens, and immune stimuli. A major family of receptors expressed by microglia includes pathogen-associated molecular pattern (PAMP) receptors known as the Toll-like receptors (TLRs). Eleven TLRs (TLR1 to TLR11) have been identified to date in humans. Microglia express TLR1 to TLR9, which allows them to detect and respond to a huge array of PAMPs; LTA (TLR2), double-stranded RNA (TLR3), lipopolysaccharide (LPS; TLR4), flagellin (TLR5), single-stranded RNA (TLR7), and unmethylated CpG DNA (TLR9) are examples.²² Engagement of these receptors by cognate ligands triggers signal transduction of multiple intracellular pathways responsible for the inflammatory response. LPS-induced activation of TLR4 has been well characterized in microglia and leads to activation of nuclear factor κB (NF-κB), cytokine production (interferon-β [IFN-β]), tumor necrosis factor-α (TNF-α)], signal transducer and activator of transcription-1α (STAT-1α), production of reactive oxygen species (ROS), and production of nitric oxide (·NO, or NO).²³ Flagellin-mediated activation of TLR5 on microglia has been shown to upregulate expression of TLRs 1, 2, 4, and 5, as well as IL-6.²⁴ TLR3 activation in microglia has been demonstrated in response to HIV, and TLR9 activation in microglia leads to the production of multiple cytokines and chemokines (TNF-α, IL-1β, IL-6, IL-12, macrophage inflammatory protein-1α [MIP-1α], MIP-1β).^25,26

Activation of microglia by invading pathogens has many consequences, some beneficial to the host and some detrimental. Microglia, like other members of the monocyte lineage, are capable of NO synthesis and a respiratory burst, processes directed at producing oxidative damage to the offending pathogen. Microglia produce NO via inducible nitric oxide synthase (iNOS, NOS-2), which leads to the formation of peroxynitrite (ONOO⁻), a highly toxic product capable of damaging both host and pathogen.²⁷

Microglia also possess the metabolic machinery (reduced nicotinamide adenine dinucleotide phosphate oxidase) necessary for the generation of superoxide anion (·O₂⁻), an extremely reactive oxygen species that can damage nucleic acids, lipids, and proteins.²⁸ Pathogens expressing superoxide dismutase are able to neutralize this effective defense. Microglia also function as phagocytes in the CNS under both physiologic and pathophysiologic conditions. Microglia are the major scavengers of cell debris in the CNS and interact, in part, with apoptotic bodies expressing externalized phosphatidylserine.²⁹ The orphan receptor TREM-2 (triggering receptor expressed on myeloid cells-2) is important in transforming microglia into phagocytes, and activation of TREM-2 enhances phagocytosis while suppressing the production of proinflammatory cytokines, events that may be important for prevention of the autoimmune responses to the autoantigens present in apoptotic bodies.³⁰

Microglia also regulate the response of other CNS cells to injury or infection and recruit cells from outside the CNS via the production of a wide variety of cytokines, chemokines, and lipid mediators. Table 39-1 lists some of the cytokines and chemokines known to be generated by microglia in response to a number of activating stimuli.³¹ Microglia also produce factors that support glial and neuronal cells in their microenvironment, including nerve growth factor, NT-3, brain-derived neurotropic factor, glial-derived neurotropic factor, and basic fibroblast growth factor.³¹ Many potent lipid mediators are synthesized by microglia, including prostaglandins (D₂, E₂, F_2α, thromboxane B₂), leukotriene B₄, and platelet-activating factor.^32,33 Several of these lipid mediators serve an autocrine role; the EP2 receptor on microglia participates in the activation of microglia, and antagonism of this receptor may have a neuroprotective effect by preventing excessive microglial neurotoxicity.³⁴ The critical role of microglia in orchestrating the innate immune response has been established, and experimental evidence for this role is expanding rapidly; an extensive review of this evidence can be found elsewhere.^31,35,36

TABLE 39-1 Cytokines and Chemokines Produced by Microglial Cells

IL, interleukin; IP-10, interferon-γ–inducible protein-10; MCP-1, monocyte chemoattractant protein-1; M-CSF, macrophage colony-stimulating factor; MIP-1α, macrophage inflammatory protein-1α; RANTES, regulated on activation, normal T cell expressed and secreted; TGF-β, transforming growth factor-β; TNF-α, tumor necrosis factor-α.

Data from Hanisch UK. Microglia as a source and target of cytokines. Glia. 2002;40:140-155.

In addition to initiating the innate immune response to a variety of pathogens in the CNS, microglia also function as antigen-presenting cells and are able to prime CD4⁺ T cells to initiate the adaptive immune response. Various TLR ligands (LPS, peptidoglycan, polyinosinic-polycytidylic acid [poly-I:C], CpG DNA) and infection with Theiler’s murine encephalomyelitis virus result in increased expression of major histocompatibility class II complexes and costimulatory molecules on microglia, events that favor efficient antigen presentation and development of an adaptive immune response.²⁶ Cytomegalovirus (CMV), a gamma herpes virus capable of invading the human CNS, elicits CXCL10 (IFN-γ–inducible protein-10 [IP-10]) production from primary microglia but not from astrocytes (see later).³⁷ CXCL10 is an important chemokine involved in IFN-γ–induced T-cell recruitment, a process critical for control of CMV infection.³⁷ Remarkably, astrocytes infected with CMV produce the viral homologue of IL-10, UL111a, which suppresses the production of CXCL10 from activated microglia.³⁷ Thus, CMV is able to suppress T-cell recruitment in the CNS by subverting the production of an important antiviral chemokine by microglia.

Microglia, as potent regulators of the proinflammatory response to CNS injury and infection, are also able to downregulate these proinflammatory responses. Microglia express anti-inflammatory cytokines (IL-4, IL-10, IL-13, transforming growth factor-β [TGF-β]) and actively phagocytose apoptotic T cells after stimulation by IFN-β and IFN-γ.^31,38,39 Production of IL-4 and IL-13 ultimately triggers microglial apoptosis via autocrine receptor engagement, thereby providing a means of balancing inflammation in response to a specific CNS insult.⁴⁰

Microglia are clearly multifunctional cells that serve as central regulators of the innate and adaptive immune response in the CNS. However, these cells do not act in a vacuum and interact with both immune and nonimmune cells to coordinate the events surrounding CNS injury or invasion. One important cell type included in this response is the astrocyte. The next section briefly addresses the importance of astrocytes in the CNS during various stages of infection.

Astrocytes: Stellar Actors in Central Nervous System Immunopathogenesis

Astrocytes are resident glial cells derived from neuroectoderm and are often thought of as “nurse” cells for neurons in the brain parenchyma. These cells compose a large portion of the brain parenchyma and can be identified by their star-shaped morphology and expression of glial fibrillary acidic protein (GFAP). In addition to maintaining the BBB, astrocytes also participate in the immunologic processes that are associated with CNS injury and infection. The overt histologic response of astrocytes to CNS damage is termed reactive astrogliosis and is manifested as an increase in the amount of GFAP expressed by resident astrocytes at the site of insult. However, these observations are a small reflection of the complex responses of astrocytes to CNS injury or infection.

Like microglia, astrocytes express several TLRs, but to a more limited degree, which allows them to participate in the initial innate immune response to CNS invasion. The close proximity of astrocyte foot processes to BBB tight junctions and the huge surface area presented by the brain endothelium also highly favor early interaction between astrocytes and invading pathogens. TLRs expressed by astrocytes include 1, 3, 4, 5, and 9, with little or no TLRs 2, 6, 7, 8, or 10.^22,41 Among the expressed TLRs, TLR3 levels are the most prominent by quantitative real-time polymerase chain reaction (PCR).⁴¹ TLR3 ligation in astrocytes (by poly-I:C) triggers the production and release of IFN-β, CXCL10, and IL-6, as well as an increase in the expression of TLRs 1 to 5 and 9 mRNA by quantitative real-time PCR.⁴¹ Astrocytes are the main source of IL-6 production in the CNS; IL-6 is a multifunctional cytokine with diverse biologic activities, including neurotropism, protection of neurons from glutamate toxicity and ischemic injury, astrocyte proliferation, inflammation, and modulation of Fas/FasL expression in astrocytes.⁴² This latter effect on Fas/FasL expression may affect the general immunologic state of the CNS in that astrocytes normally express both elements of the Fas apoptosis apparatus yet do not succumb to this autocrine signaling.⁴³ The lack of astrocyte apoptosis despite Fas/FasL expression has been attributed to low-level expression of procaspase-8, whereas FasL expression by astrocytes is thought to provide a molecular barrier to circulating lymphocytes entering the CNS, cells that are highly sensitive to Fas/FasL-induced apoptosis.^44,45

In addition to the production of important immunologically active cytokines and chemokines, astrocytes also respond to chemokines through expression of chemokine receptors. Multiple chemokine receptors have been identified in astrocytes, and ligation of these receptors has many downstream effects on astrocyte function, including regulation of chemokine production and receptor expression (Table 39-2).⁴⁶ Of note, astrocytes express CXCR4, the receptor for stromal cell–derived factor-1α (SDF-1α), a chemokine with autocrine activity leading to influx of Ca²⁺ and chemostasis of astrocytes. Both CCR5 and CXCR4 serve as CD4-associated coreceptors for HIV. Although astrocytes lack CD4 receptors, CXCR4 binds to the gp120 glycoprotein of HIV, an event that leads to activation of the mitogen-activated protein kinases (MAPKs) extracellular signal–regulated kinase-1 (ERK-1) and ERK-2.⁴⁶ HIV also infects astrocytes via a CD4-independent mechanism, which results in a restricted (i.e., nonlytic, nonproductive) infection with incorporation of the HIV provirion into genomic DNA. Nonetheless, this restrictive infection alters astrocyte functions: increased expression of immunomodulatory molecules such as monocyte chemotactic protein-1 (MCP-1), complement factor 3 (C3), and iNOS and decreased glutamate uptake (see later).⁴⁷ The decreased glutamate uptake and increased iNOS expression by HIV-infected astrocytes probably contribute to the neurotoxicity observed with HIV infection of the CNS.

TABLE 39-2 Select Astrocyte Chemokine Receptors and Ligands and Astrocyte Responses to Receptor Activation

cAMP, cyclic adenosine monophosphate; ERK, extracellular signal–regulated kinase.

Data from Dorf ME, Berman MA, Tanabe S, et al. Astrocytes express functional chemokine receptors. J Neuroimmunol. 2000;111:109-121.

One of the major functions of astrocytes is uptake of glutamate at synaptic junctions in the CNS. Glutamate is an excitatory neurotransmitter that is highly neurotoxic, mainly via two mechanisms: (1) hyperactivation of neurons and (2) inhibition of cysteine uptake leading to oxidative damage to neurons through glutathione depletion. Astrocytes “soak up” glutamate in the CNS via excitatory amino acid transporters and rapidly convert glutamate to nontoxic glutamine by expression of the enzyme glutamine synthase.⁴⁸ This glutamine is then exported from astrocytes, taken back into neurons, and converted back to glutamate via a mitochondrial-based glutaminase for use as a neurotransmitter.⁴⁸ Pathogen-associated factors or inflammatory mediators released in the CNS during infection, or both, can alter the ability of astrocytes to control the CNS glutamate concentration and result in glutamate-induced neurotoxicity as a by-product of infection. For example, HIV gp120 has been shown to reduce astrocyte expression of glutamine synthase, and patients with HIV-associated dementia have higher brain glutamate levels than do nonaffected controls.^49,50 Glucocorticoid-induced expression of glutamine synthase is inhibited by IL-1β and TNF-α.⁵¹

Astrocytes also contribute to the balance of CNS protection/destruction during infection by secretion of matrix metalloproteinases (MMPs). MMPs are a family of proteases produced in a wide range of tissues, including the CNS, and are discussed in the next section, with particular attention on their roles in bacterial meningitis.

Matrix Metalloproteinases, “You Can’t Have Your Cake and Eat It Too!”

MMPs are a subgroup of the metzincin family of proteases that share a common Zn²⁺ binding site in their catalytic domain (…HExxHxxGxxH…, where x is any amino acid) and an associated methionine within a β turn, biochemical elements important for proteolytic activity.⁵² There are 24 members of the MMP group in mammals, and they can be further subdivided according to their domain structure (additional details can be found in the review by Yong⁵²). Most MMPs are secreted from cells, although some may be associated with cell surface molecules such as integrins (e.g., pro-MMP-2) or the hyaluronan receptor (active MMP-9) or are linked to the cell membrane via a glycophosphatidylinositol link.⁵² MMPs degrade a wide variety of substrates, including extracellular matrix components, receptors, growth factors, and adhesion molecules.⁵² As a consequence, MMPs must be tightly regulated to prevent extensive tissue destruction or unintended biologic sequelae. The first level of control involves the transcriptional regulation of MMP genes based on specific activation signals. MMPs may also undergo posttranslational modification and are highly compartmentalized within cells, thereby allowing further intracellular control of MMP activity.⁵² The second level of MMP regulation is based on secretion of these enzymes as zymogens, or proenzymes, which require additional cleavage to become active proteases. Finally, specific inhibitors, known as tissue inhibitors of metalloproteinases (TIMP-1 to TIMP-4), are expressed in tissues to counteract MMPs.⁵²

MMPs have been implicated as playing a role, either beneficial or detrimental, in many different diseases affecting the CNS, including infections, ischemic or traumatic injury, and autoimmune disorders.⁵² The cellular sources of these MMPs vary depending on the specific pathologic condition but generally include cellular elements of the CNS (microglia, astrocytes, neurons), endothelial cells, and infiltrating leukocytes, especially neutrophils and macrophages.⁵² Neutrophils, in particular, represent an important early source of MMP-9 because these cells are among the “first responders” of the innate immune response to injury or infection and contain preformed stores of this MMP.⁵² In general, MMPs can have complex effects on the inflammatory state of the CNS. The BBB is rich in potential substrates for MMPs, including type IV collagen, fibronectin, and laminin; MMP degradation of the BBB may favor transmigration of leukocytes, as well as the movement of macromolecules and water, thus contributing to brain edema. Injection of heat-killed Neisseria meningitidis results in BBB disruption and increased intracranial pressure (ICP), phenomena that are inhibited by the administration of batimastat, an MMP inhibitor.⁵³ MMPs also modify or degrade cytokines and chemokines relevant to this inflammatory state. For example, MMP-9 cleaves six amino acids from the N-terminal of IL-8 to produce a truncated form of IL-8 that is more potent with respect to neutrophil chemoattraction.⁵⁴ In contrast, MMPs degrade IL-1β, thus demonstrating anti-inflammatory activity.⁵⁵

CNS infections in which MMPs have specifically been shown to play a role include pneumococcal, gram-negative bacterial, and tuberculous meningitis, as well as viral infections such as HIV, human T-cell lymphotropic virus type I, and mumps and parasitic infections such as cerebral malaria and Angiostrongylus-associated meningoencephalitis.^56–61 We focus here on the MMPs and pneumococcal meningitis as an example of the complex roles played by MMPs in CNS infection.

Streptococcus pneumoniae is a gram-positive bacterium responsible for a number of infections in humans, including pneumonia, otitis media, sinusitis, sepsis, and meningitis. A rabbit model of experimental pneumococcal meningitis demonstrated a positive correlation between MMP-9 levels and CSF leukocyte counts and total protein levels.⁶² MMP-9 activity, as measured by gelatin zymography, localized predominantly to the intrathecal compartment and not to brain parenchyma, thus supporting the conclusion that the increase in MMP-9 activity during meningitis is derived from infiltrating leukocytes.⁶² MMP-9 may become activated during pneumococcal infection by ROS, as shown by the inhibition of MMP-9 activity in but not release from rat brain slices or neutrophils exposed to heat-inactivated pneumococci in the presence of ROS inhibitors.⁶³ Pneumococci can directly activate MMP-9 via production of ZmpC, a pneumococcal zinc metalloproteinase.⁶⁴ Interestingly, MMP-9 knockout mice (MMP-9^−/−) do not develop different CNS pathology than wild-type controls when S. pneumoniae is injected directly into the brain parenchyma, but these mice are less able to clear systemic bacteremia, which suggests that MMP-9 expression may play a more important role in clearing systemic pneumococcal infection.⁶⁵ Infant rat pups with experimental pneumococcal meningitis demonstrated a peak in brain parenchymal MMP-2 and MMP-9 levels 20 hours after infection and a peak in TIMP-1 levels 24 hours after infection.⁶⁶ The MMP-9/TIMP-1 ratio was significantly elevated during the first 20 hours of infection, thus supporting an imbalance in proteinase and inhibitor levels.⁶⁶ The elevation in MMP-9 was associated with proteolysis of collagen type IV in the meninges, perivascular spaces, and brain parenchyma, and parenchymal gelatinolytic activity correlated well with the degree of cortical damage.⁶⁶ In a rat model of pneumococcal meningitis, brain levels of MMP-9 increased in infected rats treated with ceftriaxone but not in saline-injected rats, whereas treatment with ceftriaxone plus dexamethasone reduced MMP-9 levels in comparison to untreated controls or animals treated with just ceftriaxone.⁶⁷ Glucocorticoids are known to inhibit MMP expression, and these findings are consistent with a potential protective effect of dexamethasone during pneumococcal meningitis. Tetracyclines inhibit the proteolytic activity of many MMPs, as well as the activity of TNF-α–converting enzyme (TACE), another proteolytic enzyme implicated in propagating CNS damage during meningitis.⁶⁸ Infant rats with pneumococcal meningitis that were treated with ceftriaxone plus doxycycline had lower mortality, less cortical damage, and less BBB disruption than did rats treated with ceftriaxone alone.⁶⁸ Both groups had sterile CSF by 40 hours after infection, with no differences in the time-kill curves between these groups during this time frame, thus suggesting that the effect of doxycycline was not due to enhanced bacterial clearance from CSF.⁶⁸ These animal models provide evidence that MMPs, particularly MMP-9, play a pathophysiologic role in bacterial meningitis and that targeting MMPs with inhibitors may reduce mortality and cortical damage without significantly altering sterilization of the CNS.

Data regarding MMPs in human disease also exist and are consistent with the observations described in animal models. In 27 children with bacterial meningitis, 91% and 97% had elevated CSF levels of MMP-8 and MMP-9, respectively, when compared with uninfected control children.⁶⁹ The majority of these children were infected with Haemophilus influenzae (n = 14) or N. meningitidis (n = 11), with only 2 infected with S. pneumoniae. However, elevated MMP-9 levels in CSF were associated (P < .05) with an increased risk for neurologic sequelae, including hearing loss and postinfection seizures.⁶⁹ In 19 adults with bacterial meningitis (n = 7 with S. pneumoniae), all had elevated MMP-9 activity, as measured by gelatin zymography, in comparison to uninfected controls or patients with Guillain-Barré syndrome.⁵³ Patients with bacterial meningitis also demonstrated increases in CSF TIMP-1 levels, and the MMP-9/TIMP-1 ratio was significantly elevated when compared with noninfected controls.⁵³ MMP-9 levels in the infected patients correlated with CSF protein concentrations but not CSF leukocyte counts, whereas TIMP-1 levels correlated with CSF leukocyte counts but not protein levels.⁵³ An investigation of 111 paired CSF and serum samples from patients with a range of neurological disorders, including both aseptic and bacterial meningitis, found that a CSF leukocyte count greater than 5/µL correlated well with elevated CSF MMP-9 activity by zymography (Spearman r = .755, P < .0001).⁷⁰ A more recent study examined the correlation between serum MMP-2 levels and the α₂-macroglobulin (α2M) index as a marker of increased BBB permeability in patients with infectious meningitis. The α2M index was defined as the ratio of α2M (CSF/serum) to albumin (CSF/serum). This study found that serum MMP-2 levels, as measured with an enzyme immunoassay, correlated well (r = .64, P < .0001) with the α2M index and were higher in patients with bacterial meningitis than in those with viral or fungal meningitis.⁷⁰

In summary, MMPs contribute to the pathogenesis of bacterial meningitis by degrading components of the BBB, a process that favors the formation of brain edema and propagation of the inflammatory response via leukocyte transmigration across the impaired BBB. MMPs also have more complex actions on other components of the immune response to infection, including modulation of cytokines and chemokines involved in balancing a successful immune response to infection in the CNS. These proteinases are potential targets for therapeutic interventions aimed at minimizing the collateral damage that occurs during the immunologic assault on CNS pathogens. A major component of this collateral damage involves brain edema, the development of ischemia, and neurotoxicity from elements of the immune response and from the invading pathogens themselves. The next section addresses mechanisms in the development of brain edema, increased ICP, and neurotoxicity during CNS infections.

Brain Edema and Central Nervous System Infections

Many factors contribute to the development of brain edema during the course of CNS infection, but most share a common final target, the BBB. The major cellular element of the BBB is the brain microvascular endothelium, a monolayer of overlapping endothelial cells connected by organized tight and adherens junctions and exhibiting little transcytosis, unlike the capillary endothelium in other tissues. BMECs behave much more like epithelial layers with respect to low paracellular permeability to macromolecules and cells, a feature that helps protect the brain parenchyma from unregulated fluid shifts based solely on Starling’s forces, such as oncotic or hydrostatic pressure. Alterations in BBB permeability occur during CNS infection when pathogen- or host-derived factors influence the complex machinery responsible for maintaining low BBB permeability. During infection of the CNS by viruses, bacteria, fungi, or parasites, the innate and adaptive immune responses induce influx of leukocytes into the CNS, a process associated with increased BBB permeability. This increase in BBB permeability is the sum of many individual, complex and interacting processes that involve leukocytic infiltration of the CNS, but changes in BBB permeability are not always dependent on leukocytes. Myriad proinflammatory and anti-inflammatory mediators are associated with this leukocyte infiltration and change in BBB permeability. Figure 39-3 provides a schematic view of the complex processes involved in the initiation and propagation of brain edema and neurotoxicity during bacterial meningitis; this scheme can also be applied to infection caused by nonbacterial meningitis. Some examples of specific pathogens and the mechanisms underlying alteration of the BBB during infection with these pathogens are discussed in the following sections.

Bacterial Infections of the Central Nervous System

Increased permeability of the BBB most often occurs after bacteria have crossed this barrier and are already present in the CNS. Accordingly, a reduction in the BBB is not a prerequisite for initiation of a CNS infection but is a common consequence of infection and probably facilitates further invasion of the CNS by the offending organism. Bacterial products and the immunologic responses to bacteria both contribute to an increase in BBB permeability. Perhaps equally as important as the increased permeability of the BBB is the decreased clearance of CSF that occurs during meningitis. Scheld and colleagues monitored cerebrospinal hydrodynamics in rabbits during experimental meningitis by using a pressure device in direct continuity with the supracortical subarachnoid space.⁸¹ Rabbits were infected with either S. pneumoniae or E. coli (K1) to produce experimental meningitis, and a subgroup of animals were treated with penicillin G or methylprednisolone to assess the impact of antimicrobial or steroid therapy on CSF hydrodynamics. CSF outflow resistance increased 26-fold, from a baseline of 0.26 ± 0.04 mm Hg/µL per minute to 6.77 ± 3.52 mm Hg/µL per minute, during experimental meningitis and remained elevated up to 15 days later despite penicillin treatment. Administration of methylprednisolone early during pneumococcal meningitis reduced CSF outflow resistance 11.5-fold to 0.59 mm Hg/µL per minute. These observations highlight the extremely impaired drainage of CSF during bacterial meningitis, a process that itself promotes brain edema.

LPS, a common structural element of gram-negative bacterial outer membranes, is recognized by TLR4 on innate immune cells, such as microglia in the CNS, and triggers a proinflammatory response. TLR4 knockout mice demonstrate impaired leukocyte recruitment to the CNS after intracerebral LPS injection.²³ Intracisternal injection of LPS from H. influenzae type b (Hib) into rats produces increased BBB permeability, as measured by the accumulation of radiolabeled albumin in CSF, in a dose-dependent fashion (2 to 20 ng), although at higher doses (500 to 1000 ng) there is attenuation of the inflammatory response.⁸² The LPS-induced increase in BBB permeability does not occur in leukopenic rats, thus supporting the hypothesis that leukocytes are responsible for LPS-induced alterations in BBB permeability.⁸² However, an in vitro model of the rat BBB in which isolated primary rat cerebral microvascular endothelial monolayers were used demonstrated an increase in radiolabeled albumin flux across monolayers exposed to Hib LPS.⁸³ The absence of leukocytes in this in vitro model indicates that LPS may increase BBB permeability independent of leukocytes through a direct effect on endothelial cells. Murine brain endothelial monolayers exposed to LPS secrete IL-1α, IL-6, IL-10, granulocyte-macrophage colony-stimulating factor, and TNF-α.⁸⁴ These cytokines are released in a polarized fashion that favors secretion on the luminal side of the endothelial monolayer, and IL-6 production is further enhanced when the monolayer is exposed to LPS from the abluminal surface (akin to parenchymal exposure).⁸⁴ LPS also triggers the expression of E-selectin and P-selectin on BMECs, as well as vascular cell adhesion molecule-1 (VCAM-1) and intercellular adhesion molecule (ICAM).⁸⁵ The selectins are critical to the initiation of leukocyte margination and rolling, which allows these cells to slow down and interact with adhesion molecules that facilitate leukocyte transmigration across the BBB, such as VCAM-1, ICAM, and integrins. When rats are injected intracisternally with encapsulated or unencapsulated isogenic strains of Hib, an increase in BBB permeability is noted in both groups, but differences are noted in clearance of Hib from CSF.⁸⁶ Rats injected with unencapsulated bacteria are able to clear these bacteria from CSF more rapidly than are animals injected with encapsulated bacteria, a finding anticipated because of the known antiphagocytic properties of the capsule. Interestingly, when encapsulated or unencapsulated Hib is injected into leukopenic rats, an increase in BBB permeability is still observed despite reduced CSF pleocytosis in comparison to normal rats, and the degree of BBB permeability correlates with the CSF bacterial load, not the degree of CSF pleocytosis.⁸⁶ These data suggest that leukocytes are not an absolute requirement for the increase in BBB permeability noted during Hib meningitis; additional host or bacterial factors are involved in reducing the barrier function of the BBB during Hib meningitis. Consistent with this hypothesis is the finding that Hib peptidoglycan (PGN) also induces meningeal inflammation and an increase in BBB permeability when injected intracisternally into rats, probably via TLR2 signaling and the associated inflammatory response.⁸⁷

N. meningitidis is a gram-negative coccus responsible for epidemic outbreaks of bacterial meningitis, most commonly among human populations in developed countries sharing close living conditions, such as college students and military recruits. A murine brain endothelial cell line (MB114En) exposed to meningococcal whole cell lysates rapidly produces a large amount of NO via expression of iNOS (or NOS-2), a series of events that actually results in the death of endothelial cells and is prevented by an inhibitor of NOS-2 (L-NNA) or an inhibitor of poly(ADP-ribose) polymerase (3-aminobenzamide).⁸⁸ This process is mediated by TLR2 and TLR4 and activation of the MAPK p38/NF-κB pathway.⁸⁹

Although these studies have identified elements of gram-negative bacteria as modulators of BBB permeability, alterations in BBB permeability have also been shown with components of gram-positive bacteria. An in vitro model of the BBB was developed by using bovine BMECs cocultured with rat primary glial cells and subsequently exposed to the gram-positive cell wall components LTA and muramyl dipeptide (MDP, a subunit of PGN).⁹⁰ Exposure of glial cells to purified LTA (10 µg/mL) from Staphylococcus aureus or S. pneumoniae led to an increase in barrier permeability to fluorescein isothiocyanate (FITC)-inulin after 72 hours as opposed to complete loss of the BBB after 24 hours of exposure to LPS (10 µg/mL).⁹⁰ The increase in BBB permeability noted with LTA was augmented by the addition of MDP to LTA.⁹⁰ The changes in BBB permeability associated with glial activation by LTA with or without MDP were associated with glial production of TNF-α, IL-1β, and NO. In this model, an increase in BBB permeability to FITC-albumin and a decrease in transendothelial electrical resistance could be produced by exposure of endothelial monolayers to TNF-α or IL-1β and blocked by the addition of antibodies against TNF-α and IL-1β to LTA-activated glial cells. Addition of an iNOS inhibitor to activated glial cells also partly reversed the reduction in transendothelial electrical resistance associated with LTA.⁹⁰ These data indicate that glial cells contribute significantly to the increase in BBB permeability associated with LTA by the production of proinflammatory cytokines and NO.

The role of NO in BBB permeability has been examined further in vivo in a rat model of Hib meningitis. The CSF nitrite level, an indirect measure of NO production, increases significantly during the course of experimental Hib meningitis from basal levels to peak levels over the first 18 hours after intracisternal inoculation with live bacteria, but with no significant change in serum nitrite levels.⁹¹ The CSF nitrite level correlates well with increased BBB permeability, as measured by accumulation of radiolabeled albumin in the CSF from serum (r = .84, P = .018), and administration of a systemic NOS inhibitor, N-nitro-L-arginine methyl ester, reduces CSF nitrite levels and the degree of CSF pleocytosis associated with intracisternal Hib lipooligosaccharide.⁹¹

NO can also interact with other mediators to produce alterations in BBB permeability. Intrathecal administration of LPS to induce experimental meningitis in rats leads to a rapid increase in NO and prostaglandin E₂ (PGE₂) levels in the CSF that peaks about 6 to 8 hours after LPS administration.⁹² These increases in NO and PGE₂ were temporally associated with an increase in BBB permeability, as measured by ¹⁴C-sucrose accumulation in brain parenchyma, thus supporting a role for these mediators in altering the BBB during meningitis.⁹²

NO, in addition to myriad physiologic functions, can also react chemically with the ROS species ·O₂⁻ (superoxide) to form peroxynitrite (·NO + ·O₂⁻ → OONO⁻), an extremely reactive oxidant capable of damaging a wide variety of biomolecules. Peroxynitrite produces an increase in BBB permeability in response to experimental rodent viral CNS infections (Borna disease virus, rabies virus) via a TNF-α–independent mechanism, thus providing an additional mechanism for NO synthesis to modulate the BBB and immune response in the CNS.^93,94

Alterations in BBB permeability as measured by leakage of macromolecules into the CSF during in vivo experimental meningitis correspond to morphologic changes in the cellular elements of the BBB. Experimental infection of rats with the encapsulated bacteria commonly responsible for human meningitis, S. pneumoniae, E. coli, or Hib, induces early changes in brain microvascular endothelium as examined by transmission electron microscopy.⁹⁵ An early (4 hours after infection) and sustained increase in pinocytotic vesicles (10 and 18 hours after infection) was noted with all these bacterial species (Fig. 39-4A).⁹⁵ The duration of infection is also associated with an increased observation of separation of intercellular junctions between BMECs (Fig. 39-4B).⁹⁵

FIGURE 39-4 Transmission electron micrographs of rat cerebral capillary endothelium during bacterial meningitis. A, Arrowheads highlight pinocytotic vesicles forming at the luminal membrane, whereas arrows demonstrate fully formed vesicles in the cytoplasm 18 hours after intracisternal inoculation of Escherichia coli K1. E, intraluminal erythrocyte; N, endothelial nucleus. Magnification ×60,000. B, The side arrows highlight the wide separation between two cerebral capillary endothelial cells that share an intact intercellular junction elsewhere (top arrow) 18 hours after intracisternal inoculation with Haemophilus influenzae. Arrowheads demonstrate an intact basement membrane. E, intraluminal erythrocyte. Magnification ×15,000.

(From Quagliarello VJ, Long WJ, Scheld WM. Morphologic alterations of the blood-brain barrier with experimental meningitis in the rat. J Clin Invest. 1986;77:1087-1088.)

Functional changes in BBB permeability have micromolecular (changes in junctional protein levels, activation states, altered intracellular signaling pathways, and so on) and macromolecular (toxic injury to endothelial cells, endothelial apoptosis, altered endothelial migration, and so on) mechanisms that can and have been studied extensively. For example, an in vitro model of the human BBB using HBMECs examined changes in intercellular junctional regulation after exposure to E. coli K1.⁹⁶ HBMEC monolayers exposed to E. coli K1 exhibit decreased transendothelial electrical resistance before traversal of these bacteria across the monolayer, but only when these bacteria are expressing OmpA. OmpA interacts with a 95-kD endothelial glycoprotein expressed only in brain microvascular endothelium and induces actin cytoskeletal rearrangements before bacterial internalization, a process dependent on protein kinase C-α (PKC-α).⁹⁶ Activated PKC-α phosphorylates a specific protein at the adherens junction, vascular endothelial cadherin, which leads to dissociation of β-catenin from the adherens junction, decreased adherens junction intercellular adhesion, and increased paracellular permeability at the BBB.⁹⁶ This entire process of PKC-α activation, β-catenin dissociation, and increased BBB permeability can be prevented by antibodies directed against E. coli K1 OmpA or its glycoprotein receptor.⁹⁶

In addition to functional changes in brain endothelium, bacteria often produce toxins capable of directly damaging endothelial cells and thereby destroying the BBB via cellular loss. S. pneumoniae, the most common bacterial cause of meningitis, produces a pore-forming toxin called pneumolysin. Pneumolysin is a member of the family of cholesterol-dependent cytolysins (CDCs), a group of gram-positive bacterial toxins that share a carboxy-terminal undecapeptide sequence that binds to cholesterol in the membranes of eukaryotic cells and leads to oligomerization and pore formation.⁹⁷ The CDC family includes proteins produced by many gram-positive human pathogens; some examples are streptolysin O (Streptococcus pyogenes), perfringolysin O (Clostridium perfringens), listeriolysin O (L. monocytogenes), and anthrolysin O (Bacillus anthracis). Pneumolysin lacks an amino-terminal signal sequence, which suggests that it is not secreted from pneumococci and requires autolysis for release.^98,99 Pneumolysin is highly cytotoxic to HBMECs, and pneumococci unable to produce pneumolysin do not alter the permeability of HBMEC monolayers in vitro.¹⁰⁰ Pneumolysin also plays a major role in neuronal damage during pneumococcal meningitis, as discussed in more detail later with respect to neurotoxicity.

Hemolytic-uremic syndrome, the leading cause of acute renal failure in children, is highly associated with infection by Shiga toxin (Stx1 or Stx2)-producing E. coli strains, and severe manifestations of this disease are complicated by CNS malfunction.¹⁰¹ Stx1 and Stx2 are AB-type protein toxins composed of five B subunits and one A subunit.¹⁰² The B subunits are responsible for binding to a cell surface glycolipid, globotriaosylceramide (Gb3), which results in endocytosis of Stx1/2 and retrograde transport of toxin to the cytosol of eukaryotic cells expressing Gb3.¹⁰² Once in the cytosol, Stx1/2 inactivates ribosomes by removal of the adenine at position 4342 of the 28S RNA of the 60S ribosomal subunit, a reaction that inactivates protein synthesis.¹⁰² Stx1 induces apoptosis of HBMECs, especially in the presence of TNF-α, and Stx2 induces apoptosis in HBMECs via upregulation of C/EBP homologue protein/growth arrest (CHOP) and caspase-3 activation.^103,104

Neurotoxicity

In addition to alterations in BBB permeability, inflammation, brain edema, and increased ICP, CNS infections are associated with damage to or death of neurons, a process termed neurotoxicity. Neurotoxicity can result from direct infection of neurons, from collateral damage secondary to the immune response, or from pathogen-derived factors that damage neurons during the infection. Two particular pathogens that have been extensively studied with regard to neurotoxicity are HIV and S. pneumoniae, and they will be used as examples to highlight the mechanisms underlying infection-associated neurotoxicity.

Brain Abscess—Pus in the Parenchyma

Brain abscesses are space-occupying purulent infections within the substance of the brain. The cause of brain abscess is most often related to infection outside the CNS and can be subdivided according to the following associations: (1) related to infection of the paranasal sinuses or otologic structures, (2) related to odontogenic infection, (3) related to thoracopulmonary infections (e.g., lung abscess, empyema), (4) hematogenous spread from other sites (e.g., endocarditis), (5) direct inoculation (e.g., trauma, neurosurgical procedures), and (6) cryptogenic (≈20%). The microbiology of brain abscess is predictably related to the primary source of the abscess; Table 39-3 lists microbes associated with specific primary sources. There is a predominance of streptococcal species and anaerobes associated with primary sources in the sinuses, mouth, and lung, whereas staphylococci and Enterobacteriaceae are commonly found in brain abscesses associated with direct inoculation from trauma or neurosurgical procedures.

TABLE 39-3 Specific Pathogens Associated with Anatomic Sources for Brain Abscesses

Adapted from Tunkel AR. Brain abscess. In: Mandell GL, Bennet JE, Dolin R, eds. Principles and Practice of Infectious Diseases, 6th ed. Philadelphia: Elsevier; 2005:1150-1163.

The time course of brain abscess development in humans has been characterized by computed tomography and in animal models via necropsy and histology. Brain abscess begins as an early cerebritis (days 1 to 3), with edema formation, tissue necrosis, and neutrophil infiltration. This early phase is followed by an intermediate to late cerebritis with infiltration of macrophages and lymphocytes, and this process culminates in the formation of a capsule infiltrated with plasma cells and myofibroblasts.^128,129 Encapsulation allows isolation of the infection and is thought to be responsible for the lack of significant clinical symptoms in patients with chronic brain abscess.

Experimental study of the pathogenesis of brain abscess was significantly advanced with the development of a rat model of brain abscess.¹³⁰ Early studies involved stereotactic injection of a mixture of bacteria, including aerobes and anaerobes, to simulate the known microbiology of brain abscesses in humans.^130–132 An important study by Costello and associates demonstrated that injecting microaerophilic or obligate anaerobic bacteria alone did not produce brain abscess, whereas the injection of facultative anaerobic organisms such as E. coli, S. aureus, or Streptococcus pyogenes produced abscesses, albeit with differences in potency (Table 39-4).¹³² The E. coli strains in this study were more virulent with respect to abscess formation than were the gram-positive organisms tested, and the K1-encapsulated strain was more virulent than the nonencapsulated E. coli strain. These important experiments demonstrate that microaerophilic and obligate anaerobes are probably not involved in the initiation of brain abscess despite the common isolation of these organisms from abscesses derived from different primary processes. The authors point out that this experimental model may not adequately reflect the initiation of brain abscesses in humans inasmuch as many of them are associated with mixed facultative aerobe/obligate anaerobe infections of the paranasal sinuses or dental structures, thereby leading to chronic exposure of the brain to mixed bacteria. Mixed facultative aerobe/obligate anaerobe infections are known to be synergistic in establishing infections at other sites in the body, and the clinical picture in human brain abscess probably reflects this synergistic advantage in establishing a brain abscess in normal human brain tissue.

TABLE 39-4 Relative Potency of Different Bacteria Producing Experimental Brain Abscesses after Intraparenchymal Injection

Data from Costello GT, Heppe R, Winn HR, et al. Susceptibility of brain to aerobic, anaerobic, and fungal organisms. Infect Immun. 1983;41:535-539.

Before development of the rat brain abscess model, rhesus macaques (Macaca mulatta) were used to investigate the development and characteristics of brain abscesses in primates.¹³¹ S. epidermidis was injected either intracerebrally (n = five macaques) or intra-arterially via the carotid artery (on silicone cylinders, n = eight macaques); brain abscesses developed in five of five macaques injected intracerebrally as opposed to six of eight macaques injected with contaminated cylinders intra-arterially. Interestingly, abscesses in the animals injected intracerebrally developed thick capsules with an exuberant inflammatory response, beneficial prognostic features, whereas abscesses caused by “septic emboli” developed thin capsules, features historically associated with poorer clinical outcomes. Figure 39-5 displays the coronal sections of abscesses from intracerebral injection and septic embolization.¹³¹ Only recently have experiments begun to address the immunopathogenesis of brain abscesses, although in rodent models such as mice or rats as opposed to primates.

FIGURE 39-5 Coronal brain sections from rhesus macaques with experimental brain abscesses. A, Abscesses from intracerebral injection of Staphylococcus epidermidis. Note the thick abscess walls and associated shift of the midline structures away from the abscesses. B, Abscess from septic embolization of S. epidermidis–coated silicone cylinders. Note the thin abscess wall with cavitation of infarcted tissue and the lack of a midline shift.

(From Wood JH, Lightfoote WE, Ommaya AK. Cerebral abscesses produced by bacterial implantation and septic embolisation in primates. J Neurol Neurosurg Psychiatry. 1979;42:65.)

Kielian and Hickey published the first study that examined the host cytokine response to rat brain abscesses induced by direct inoculation of S. aureus.¹³³ Using RNase protection assays and reverse transcriptase PCR, they found early induction of mRNA for IL-1α, IL-1β, IL-6, and TNF-α. Within 24 hours of S. aureus injection there was an increase in mRNA for KC, the murine homologue of IL-8, as well as increases in MCP-1 and MIP-1α. The increase in KC correlated histopathologically with the appearance of neutrophils at the injection site, and the increases in MCP-1 and MIP-1α heralded the influx of macrophages and lymphocytes. Additionally, the authors noted an increase in the expression of ICAM-1 and platelet endothelial adhesion molecule in microvessels at 24 and 48 hours after injection. These cytokine/chemokine/adhesion molecule changes agree well with the progression of brain abscess from an early, neutrophil-predominant cerebritis to an organizing lesion infiltrated with macrophages and lymphocytes.

The importance of chemokines and neutrophils in the early inflammatory response to brain abscess was highlighted in a study using an S. aureus mouse brain abscess model.¹³⁴ Depletion of neutrophils from mice before the implantation of S. aureus–laden beads in the cerebral cortex led to higher bacterial counts and more severe abscesses than seen in control infected mice, thus supporting a role for neutrophils in the rapid containment of bacteria in brain parenchyma. Several chemokines were upregulated within 6 hours of S. aureus injection, including CCL1 to CCL4, CXCL1, and CXCL2; CXC1 and CXC2 are major neutrophil chemoattractants and share a common receptor, CXCR2. When CXCR2 knockout mice were injected with S. aureus–laden beads, the animals demonstrated poor neutrophil extravasation at the site of abscess formation, as well as an increased bacterial burden in these abscesses. As in the rat brain abscess model, IL-1 and TNF-α have been shown to play an important “containment” role in regulating the inflammatory response to S. aureus in the murine brain abscess model.^135,136 TNF-α knockout mice injected intracerebrally with S. aureus have a higher mortality rate (54% versus 0%), higher bacterial loads, and higher inflammatory infiltrates, as well as delayed bacterial clearance rates, when compared with infected isogenic control mice.¹³⁶

The expression of proinflammatory cytokines and chemokines is commonly triggered by engagement of TLRs, and this has been demonstrated in the murine brain abscess model. TLR2 and TLR4 are both important for the innate immune response to S. aureus brain abscess, as shown by the significantly delayed clearance of bacteria and increased mortality in TLR2 or TLR4 knockout mice injected intracerebrally with this bacterium.¹³⁷ Interestingly, the TLR2 and TLR4 receptors share a common intracellular adaptor molecule, MyD88, and MyD88 knockout mice are also subject to more severe brain abscesses, with significantly increased tissue necrosis, than infected isogenic control mice are.¹³⁸ However, unlike TLR2 or TLR4 knockout mice, MyD88 knockout mice do not have significantly increased bacterial loads, thus suggesting that MyD88 plays an important role in the containment of brain abscesses but not in clearance of bacteria from the initial cerebritis.

In addition to recruitment and activation of immune cells from outside the CNS, injection of S. aureus into cerebral cortex leads to the activation of astrocytes and microglial cells.^130,133,139 The involvement of astrocytes in the pathogenesis of brain abscess is highlighted by the observation that mice lacking GFAP, an intermediary filament upregulated in activated astrocytes, demonstrate poor containment of primary S. aureus cerebritis and exhibit severe inflammation, increased bacterial loads, ventriculitis, vasculitis, and contralateral cerebral leukocyte infiltration.¹⁴⁰ A major downregulator of microglia and astrocyte activation is the peroxisome proliferator–activated receptor-γ (PPAR-γ), and agonists of this nuclear regulator may modulate the inflammatory evolution of S. aureus brain abscesses in the murine model.¹⁴¹ In vitro studies of PPAR-γ agonists (15-deoxy-Δ^12,14-prostaglandin J₂, ciglitazone) on primary microglia or astrocytes do indeed demonstrate downregulation of proinflammatory cytokines, chemokines, and membrane inflammatory markers in these CNS cell types, although these effects in astrocytes were independent of PPAR-γ.^142,143 Administration of ciglitazone to mice 3 days after injection of S. aureus intracerebrally was associated with reduced bacterial loads secondary to enhanced microglial phagocytosis; reduced expression of TNF-α, IL-1β, CXCL2, CCL3, and iNOS; and enhanced abscess encapsulation.¹⁴³ These remarkable effects support the concept that effective control of brain abscesses requires a balance between proinflammatory and anti-inflammatory processes; factors favoring inflammation lead to excessive tissue damage, whereas impairment of specific inflammatory processes (e.g., CXCR2 knockout) leads to uncontrolled infection.

More evidence to support this “balance” hypothesis comes from observations on the effects of minocycline on experimental murine brain abscess. Minocycline has been shown to exert a number of neuroprotective effects, including modulation of microglial activation, inhibition of CNS cell apoptosis, and inhibition of proinflammatory signal transduction cascades.¹⁴⁴ Kielian and colleagues injected mice intracerebrally with a minocycline-resistant strain of S. aureus to induce brain abscess and then administered minocycline to these mice at various time points.¹⁴⁵ Minocycline administration, even as long as 3 days after injection, was associated with a reduction in brain abscess size, reduced mortality in the first 24 hours, transient reductions in IL-1β and CCL2 expression, and reduced astrocyte and microglial activation. Thus, the clinical effects of drugs such as ciglitazone and minocycline on human brain abscesses deserve further investigation as adjuncts to surgical and chemotherapeutic interventions.

Cerebrospinal Fluid Shunt Infections—The Role of Biofilms

We finish this chapter with a discussion of a sticky problem for neurosurgeons, bacterial biofilms. As biotechnology progresses, more and more patients will have the privilege of experiencing “bionic” medicine in the form of implanted prosthetic devices. We will use CSF shunt devices as an example of a neurosurgical prosthetic device that can become infected, a complication that offers many diverse challenges for the neurosurgeon and patient alike.

CSF shunts come in many different varieties, each specifically designed to specialize in specific methods to drain excess CSF from the CNS. Despite their variety, CSF shunts become contaminated with bacteria through a limited number of mechanisms:

Regardless of the mechanism of contamination, a common process that prevents subsequent sterilization of the shunt and often prompts removal of the device is the formation of bacterial biofilm. Biofilm formation occurs with both infection and colonization of prosthetic medical devices, and many features of bacterial communities living in biofilm present major challenges for device sterilization: (1) most bacteria in biofilms are in a stationary phase of growth, thus making them less susceptible to antibiotics whose mechanism depends on active division; (2) biofilms are much less permeable to most antibiotics than the surrounding milieu; and (3) prosthetic devices lack a vascular supply, thus impairing delivery of immune effectors and antibiotics to the site of colonization.

Coagulase-negative staphylococci (CoNS) are common pathogens involved in the colonization and infection of CSF shunt devices, and these bacteria are notorious for their ability to form biofilm on prosthetic devices (Fig. 39-6). Several factors probably play a role in the initial adherence of CoNS to CSF shunt devices, including the surface materials of the shunt that interact with host tissues or fluids and bacterial virulence factors such as adhesins. Soon after placement, bioprosthetic materials are coated with host proteins, and these proteins may serve as receptors for adhesion molecules expressed by bacteria. These adhesion molecules are collectively known as “microbial surface components recognizing adhesive matrix molecules,” or MSCRAMMs.¹⁴⁶ CoNS and S. aureus appear to adhere more readily to bioprosthetic material than do S. pneumoniae and E. coli, and the staphylococci adhere better to Teflon than to silicone.¹⁴⁷ Live bacteria may not be necessary to initiate bacterial adherence to CSF shunt material; “dead” biofilm on sterile, unused silicone CSF shunt devices has been shown to facilitate the adherence of bacteria, with subsequent new biofilm formation.¹⁴⁸ Once adherence has occurred, CoNS initiate the process of biofilm formation almost immediately. Biofilms are complex surface structures consisting of sessile bacteria embedded in an extracellular matrix that is permeated by water channels.¹⁴⁹ Regulation of the transition state between planktonic (free-living) bacteria and sessile bacteria in biofilms is very often dependent on a phenomenon known as quorum sensing. Quorum sensing itself is a complex set of processes that allow bacteria in a community to regulate the gene expression and phenotypes of other bacteria in the same community. Biofilm formation by S. epidermidis is a process associated with phase variation and regulation of a DNA sequence known as the intercellular adhesion (ica) gene cluster.¹⁴⁶ The ica gene cluster is organized as a four-gene operon (icaADBC) and encodes for the machinery necessary to synthesize a polysaccharide intercellular adhesin (PIA) that forms the extracellular matrix of S. epidermidis biofilms. PIA is composed of polymeric N-acetylglucosamine and accounts for the bulk of the extracellular matrix in CoNS biofilms.¹⁴⁶ The icaADBC operon is under the transcriptional control of several systems in staphylococci, including repressors such as IcaR, TcaR, and the LuxS system, a quorum-sensing apparatus active in CoNS. Additionally, σ^B, a global stress regulator, acts as a positive regulator of biofilm formation in S. epidermidis, but not in S. aureus.¹⁴⁶ More recent studies have also identified PIA-independent mechanisms of biofilm formation in staphylococci; an excellent and extensive review of staphylococcal biofilm regulation can be found in the review by O’Gara.¹⁴⁶ Future advances in bioprosthetic device design, such as adhesion-resistant materials, antibiotic-impregnated devices, and devices designed to interfere with quorum sensing or transcriptional regulation of biofilm formation, may significantly reduce the infection rates associated with implanted bioprosthetic devices. Additionally, new chemotherapeutic interventions are needed; examples include bactericidal antimicrobials that penetrate biofilm and have a high threshold for the development of resistance, compounds that interfere with quorum sensing or transcriptional regulation of biofilm formation, or with both, and compounds that regulate phase variation among bacteria and drive sessile, biofilmed bacteria into the more susceptible planktonic form.

FIGURE 39-6 A and B, Staphylococcus epidermidis biofilm formation on a catheter surface. Note the deposition of extracellular matrix, manifested as a complex web of material around and between microcolonies. Magnification ×3600 (A); ×18,000 (B).

(From Yassien M, Khardori N. Interaction between biofilms formed by Staphylococcus epidermidis and quinolones. Diagn Microbiol Infect Dis. 2001;40:79-89.)