Pathogenic mechanisms and therapeutic targets of inflammation in acquired hydrocephalus

Abstract

Hydrocephalus is the most common neurosurgical disorder worldwide and is characterized by enlargement of the cerebrospinal fluid (CSF)-filled brain ventricles from failed CSF homeostasis. Since the 1840’s, physicians have observed inflammation in the brain and the CSF spaces in both post-haemorrhagic (PHH) and post-infectious hydrocephalus (PIH). Reparative inflammation is an important protective response that eliminates foreign organisms, damaged cells, and physical irritants; however, inappropriately triggered or sustained inflammation can initiate or propagate disease, respectively. Recent data have begun to uncover the molecular mechanisms by which inflammation — driven by toll-like receptor 4 (TLR4)-regulated cytokines, immune cells, and signalling pathways — contributes to the pathogenesis of hydrocephalus. We propose that therapeutic approaches that target inflammatory mediators in both PHH and PIH could address multiple drivers of disease, including choroid plexus CSF hypersecretion, ependymal denudation, and tissue damage and scarring of intraventricular and parenchymal (glia-lymphatic) CSF pathways. Here, we review the evidence for a prominent role of inflammation in the pathogenic mechanism of PHH and PIH, and highlight promising targets for therapeutic intervention. Focusing research efforts on inflammation could shift our view of hydrocephalus from that of a life-long neurosurgical disorder to that of a preventable neuro-inflammatory condition.

Introduction

Historically, hydrocephalus has been defined as the progressive distension of the brain ventricular system that results from inadequate passage of cerebrospinal fluid (CSF) from its main site of production at the choroid plexus epithelium (CPe) to its site(s) of reabsorption into the systemic circulation (for example, the arachnoid granulations)_¹. This view of hydrocephalus is based on the bulk flow model of CSF circulation and is being modified by the development of hydrodynamic models that account for additional factors such as cardiac pulsatility_^2,3. Furthermore, emerging data suggest alternative sources of both CSF production and reabsorption, for example, the glia-lymphatic (or “glymphatic”) pathways. In addition, recent genetic analyses indicate that many forms of congenital hydrocephalus, both inherited and spontaneous, are a result of altered regulation of neural stem cell fate_⁴. Regardless of aetiology, hydrocephalus is often characterized by increased intracranial pressure, ventricular enlargement from CSF build-up, and structural brain damage that, if left untreated, can progress to neurological decline, coma and death._⁵

Historically, hydrocephalus has been classified as either primary (congenital, developmental, and/or genetic) or secondary to CNS insults such as haemorrhage, infection, trauma or tumor growth (Fig. 1)_^5–8. Post-haemorrhagic hydrocephalus (PHH) and post-infectious hydrocephalus (PIH) are two of the most common forms of hydrocephalus worldwide (Supplementary Table 1, 2)_^5,8, and are traditionally thought to be caused by obstructive mechanisms that prevent CSF reabsorption_^9,10. However, many patients with PHH or PIH have no discernible physical impediment to CSF flow in the ventricular and subarachnoid spaces. Nevertheless, surgical CSF diversion via implantation of permanent ventricular shunts or endoscopic ventriculostomy remains the mainstay of care for many patients with PHH or PIH. Although these procedures are life-saving, they frequently fail or result in complications_¹¹. Additionally, in resource-poor countries, these neurosurgical treatments are often unavailable owing to limited resources and a lack of specialized personnel such as neuro-intensivists and neurosurgeons_⁵.

Figure 1 |. Classification and treatment of hydrocephalus.

Hydrocephalus can be dividied into primary and acquired forms. Hemorrhage and infection arc two of the most common causes of hydrocephalus worldwide. Both primary and acquired forms of hydrocephalus can involve intraventricular obstruction of CSF flow, which can be treated with a ventriculo-peritoneal shunt or endoscopic third ventriculostomy (ETV). ETV can be performed with or without choroid plexus coagulation (CPC). To date, all treatments for hydrocephalus are surgical, and have a high morbidity and failure rate.

One fundamental obstacle to the development of more effective treatments for hydrocephalus, including non-surgical therapies, is our limited knowledge of the molecular physiology of the disease. In this Review, we summarize the existing literature on the epidemiology, aetiology and treatment of PHH and PIH. We highlight the similarities between PHH and PIH, and synthesize recent findings on the contribution of inflammatory mediators, including toll-like receptor-4 (TLR4)-regulated cytokines and immune cells, to the pathogenesis of hydrocephalus. We suggest that in PHH and PIH, two critical functions of the CPe — immune function and CSF secretion —maladaptively engage in an epithelial response to injury that leads to acute inflammation-dependent hypersecretion of CSF. We also speculate that sustained inflammation propagated by ongoing injury to the CPe, ependymal cells and brain tissue is likely to affect CSF resorption. Impaired CSF homeostasis at the chronic stage of PHH or PIH might be a result of intraventricular obstruction via ependymal scar formation, or extraventricular obstruction via arachnoid scar impairment of glymphatic pathways. This model of acute and chronic inflammation might describe better than previous models the pathological changes occuring in PHH and PIH across time (that is, acute versus chronic changes) and space (for example, changes to the CPe, ependyma, aqueduct or glymphatic system). We emphasize that throughout this complex process, acute and chronic inflammation is mediated by specific molecular signals that could provide therapeutic targets. Thus, improved understanding of the shared pathophysiology of PHH and PIH could catalyze the discovery of therpauetic agents for both forms of hydrocephalus.

Global epidemiology of PIH and PHH

The prevalence of hydrocephalus described in the current literature varies four-fold among different reports_¹², This lack of consistency has prevented reliable estimation of the international prevalence and incidence of the conditions_⁸. However, a recent meta-analysis indicated a global prevalence of hydrocephalus of 88 cases per 100,000 individuals under 18 years of age, 11 cases per 100,000 individuals between 19 and 64 years of age, and 175 cases per 100,000 individuals over 64 years of age_¹². This meta-analysis controlled for study quality, publication bias, and population heterogeneity. In the same study, the prevalence of hydrocephalus in individuals under 18 years of age was nearly twofold higher in Africa than in North America, indicating a difference in hydrocephalus burden between resource-rich and resource-poor countries. An analysis of global incidence rates of hydrocephalus by Dewan et al,_⁸ also suggested that the epidemiology of PHH and PIH is driven by socio-economic status. According to this analysis, PIH is the predominant form of acquired hydrocephalus in resource-poor countries and PHH is the most common cause of acquired paediatric hydrocephalus in resource-rich, countries_^8,13.

PHH occurs primarily in preterm neonates with a very low birth weight (< 1500 grams), in whom the condition is secondary to germinal matrix haemorrhage_^14,15. In these neonates, PHH is often fatal unless adequate prenatal, neonatal intensive, and neurosurgical care is provided_¹⁶. Accordingly, PHH in infants is underrepresented in countries that lack the resources to provide this care_^8,13. For example, in East Africa, with sparse neonatal intensive care resources and only 1 neurosurgeon per ∼10,000,000 individuals_¹⁷, most neonates with a very low birth weight do not survive_¹³. PHH is also a common cause of hydrocephalus in adults_^14,18,19, in whom the condition is often associated with intraventricular haemorrhage resulting from hypertension, aneurysm rupture, or traumatic brain injury._²⁰

PIH is the most common cause of pediatric hydrocephalus worldwide_⁵ and is most prevalent in Africa, Latin America and Southeast Asia_⁸. As mentioned above, in resource-poor countries, PIH is more common that PHH. This predominace of PIH is likely to result from the increased occurance of peripartum infections in these countries, caused by hygienically challenging neonatal environments and the lack of advanced obstetric care_^17,21. Within the region known as the African meningitis belt, seasonal increases in meningitis have been linked with PIH_²². In areas where tuberculosis is endemic, such as South Africa,_²³ India,_^24,25 China,_²⁶ and Philippines,_²⁷ post-tuberculosis hydrocephalus constitutes a considerable disease burden. Interestingly, congenital Zika virus has been shown to cause severe hydrocephalus in Brazi_²⁸. In resource-rich countries, PIH associated with prenatal infection is often caused by Toxoplasma gondii and cytomegalovirus, whereas typical neonatal aetiologies include bacterial sepsis from Escherichia coli, Streptococcus agalactiae, and Listeria monocytogenes._^5,29,30 Among adults, the most common causes of PIH include the bacteria Neisseria meningitidis and Streptococcus pneumoniae,_³¹ although viral, fungal, and protozoan infections have been implicated in the development of PIH in patients who are immunocompromised._^32,33 Identifying the bacterium responsible for PIH in patients in resource-poor countries has been difficult owing to limited access to advanced clinical microbiological diagnostics_³⁰. Additionally, factors such as proximity to farm animals_³⁰, access to prenatal care_^8,30, and seasonal changes in rainfall,_³⁴ among other differences in living conditions, can result in a wide variety of bacterial infections, some of which cannot be detected with standard methods.

Current treatments

The current approach to treating hydrocephalus involves either ventriculo-peritoneal shunting of CSF or endoscopic third ventriculostomy (ETV). ETV is often combined with choroid plexus cauterization (ETV/CPC). Shunting is the most common treatment for PHH and PIH across all age groups_{^5,29,35–38}. The treatment involves the subcutaneous tunneling of silicone elastomer tubing from the cerebral ventricles to the peritoneal cavity, thus draining excess CSF. An interposed valve is used to prevent retrograde fluid flow or excessive loss of CSF5. However, mechanical obstructions and/or malfunctions, tubing complications, and infections frequently occur in patients with shunts, which substantially decreases their quality of lifes,_^41–44. In the U.S., >50% of shunts fail within 2 years of insertion and 70% fail within 10 years, making shunt failure the most common medical device failure in the country_^5,39. The high likelihood of shunt failure means that these patients need life-long access to immediate neurosurgical care_⁴⁰.

ETV/CPC is the alternative to shunting and has been increasingly used worldwide to treat both PHH and PIH in infantS_^41,42. ETV/CPC involves endoscopic fenestration of the floor of the third ventricle to provide an alternate pathway for CSF reabsorption, coupled with electro-thermal destruction of the CPe which might reduce CSF production. In infants aged <6 months, ETV/CPC results in cognitive development outcomes and brain growth similar to those seen in shunting, although the newer technique reduces ventricular size more slowly and to a smaller degree_^29,43,5,38. ETV/CPC also has lower long-term failure rates than shunting,_⁴⁰ and is not affected by hardware complications_⁴⁴. However, ETV/CPC requires more advanced technical expertise_^45,46 and has a higher short-term failure rate than shunting_^38,40. Furthermore, the long-term effects of ETV/CPC are unknown, and the procedure can negatively affect other functions of the CPe, including immune function, nutrient reabsorption and multiple aspects of neurodevelopment_^47–49. Recent data indicate that ETV/CPC is the preferred treatment option in resource-poor countires where access to urgent neurosurgical care is limited_²⁹.

Effective pharmacological treatments for PHH and PIH have yet to be developed. Clinical trials for furosemide (a loop diuretic) and acetazolamide (carbonic anhydrase inhibitor) aimed to decrease CSF production by inhibiting ion flux across the basolateral and apical membranes of the CPe that provides the the osmotic gradient for water transport50. These drugs were not effective in treating PHH caused by neonatal germinal matrix hemorrhages_^51–53 and were associated with higher rates of shunt placement and worse neurological outcomes than placebo_⁵¹. This lack of efficacy is likely to reflect systemic administration and the poor blood-brain barrier permeability of acetazolamide and loop diuretics_^54–57. The International Posthaemorrhagic Ventricular Dilatation Drug Trial Group, which carried out one of the trials, concluded that these drugs could not be recommended as therapeutics_⁵¹.

The classical model of pathogenesis

That PHH and PIH result from intraventricular CSF accumulation owing to failed CSF homeostasis is widely accepted. According to the classical model of CSF dynamics, PHH and PIH result from obstruction of intraventricular CSF flow and/or dysfunction of extraventricular arachnoid granulations, which causes a decrease in CSF reabsorption^10,16. This model is supported by some PHH case series that reported occlusion of the fourth ventricular outflow tracts by fibrous thickening of the leptomeninges, known as “tetra-ventricular” PHH_^58,59. Evidence from other PHH case series suggests that blood and its breakdown products acutely obstruct narrow CSF passages such as the cerebral aqueduct_^60,61. Some authors have implicated the arachnoid granulations in post- intraventricular haemorrhage communicating hydrocephalus, suggesting that microthrombi and debris from intraventricular haemorrhage can plug arachnoid villi and impair CSF reabsorption, and that inflammation and scarring of the arachnoid at the posterior fossa might hinder CSF flow_⁶². Indeed, obstruction of CSF flow by an intraventricular blood clot, or a scarred-over aqueduct are apparent and almost certainly causative in some cases of PIH and PHH. Combinations of all of the above mechanisms are likely to contribute to development of hydrocephalus, especially in the chronic stage.

Despite these reports, the classical model is supported by sparse experimental evidence,_^7,14 neglects the potential role of increased CSF secretion6,_^7,63, and overlooks clinical and pre-clinical evidence of CPe inflammation in PHH and PIH_^7,64–67. In addition, this model fails to acknowledge that development of arachnoid granulations, which are believed to reabsorb much of adult CSF, is gradual_^68,69. This gradual development means that arachnoid granulations are not yet present in human infants or in most animal models of PHH, suggesting that arachnoid scarring does not have a central role in the pathophysiology of PHH. The classical model does not account for other sites of CSF reabsorption, such as the ventricular ependyma, perineural space, leptomeninges, glymphatics, and nasal mucosa,_{^{62,68,70–72}} or the role of ependymal ciliary beating on CSF bulk flow_⁷³. Moreover, CSF hypersecretion caused by CPe villous hyperplasia or choroid plexus tumors is sufficient to cause non-obstructive hydrocephalus⁶. Finally, intracerebroventricular injection of blood metabolites causes CPe inflammation_^7,64–66 and hydrocephalus_⁷⁴ in animal models. Collectively, these observations suggest that the classical model is unable to adequately explain PHH and PIH, and that alternative or complementary pathogenic mechanism(s) need to be considered.

The pathogenic role of inflammation

Evidence from humans

The possibility of a role for inflammation in the pathogenesis of human PHH and PIH was discussed in the scientific literature as early as 1840_^75–79 and is supported by several lines of clinical evidence. For example, in infants and adults who have had an infection or brain haemorrhage, levels of IL-6, IL-4, TNF-α, TGF-β1, and other inflammatory markers in the CSF and the peripheral blood correlate with the likelihood of subsequently developing hydrocephalus, and with the severity of the condition_^80–83. In addition, neuropathological examination of brain tissue from fetuses and infants with PHH or PIH shows signs of neuroinflammation, such as microglial activation and reactive gliosis_^58,84,85. Although these laboratory and neuropatho logical studies have identified correlations between inflammation and ventriculomegaly, the strongest evidence for the pathogenic role of inflammation in hydrocephalus comes from a randomized, double-blind, placebo-controlled trial of the corticosteroid dexamethasone in 545 adults with tuberculous meningitis_⁸⁶. Intriguingly, dexamethasone administration decreased the frequency of hydrocephalus in these individuals. These findings are consistent with an earlier retrospective study, in which higher dexamethasone doses were associated with decreased frequency of hydrocephalus following aneurysmal subarachnoid haemorrhage_⁸⁷. Altogether, the evidence discussed in this section suggests that reducing inflammation could be a powerful approach for treatingPHH and PIH.

Evidence from animal models

Evidence from pre-clinical animal models provides further support for the pathogenic role of inflammation in PHH and PIH. As in patients, haemorrhage or infection in rabbit and rodent models results in upregulated expression of inflammatory markers as well as activation of TLR4-dependent inflammatory signalling pathways_^64–66,88. In particular, inflammation of the choroid plexus and ependymal layer of the lateral ventricles was observed in an animal model of PHH_⁶⁶. Histological studies of brain tissue showed leukocytic infiltration, microglial activation and reactive gliosis in rodent models of PIH and PHH_^89–91.

Experimental manipulation of inflammatory pathways provides strong evidence that acute inflammation is both necessary and sufficient to induce PHH and PIH. In animal models, pharmacological inhibition or genetic ablation of TLR4 signalling attenuates molecular and histological correlates of inflammation as well as ventricular dilation_^7,92–94. Furthermore, pharmacological or genetic hyperactivation of inflammation causes ventricular dilation_^95–97. Interestingly, in rabbit and dog studies, steroid administration decreased CPe-mediated CSF production_^98–100, which is reflected in the benefits of steroid administration observed in patients with PHH and PIH_^86,87.

These preclinical findings show that inflammation is a shared pathogenic event of PHH and PIH, thus supporting use of anti-inflammatory agents such as glucocorticoids to treat patients with these conditions.

The immune function of the CPe

The CPe is located within the cerebral ventricles and consists of a single cell layer of polarized cuboidal epithelial cells surrounding fenestrated capillaries. These epithelial cells actively secrete sodium (Na₊), potassium (K₊), chloride (C1₋) and bicarbonate (HCO₃₋) ions, amongst others, resulting in an osmotic gradient that drives transport of water from the blood to the ventricular space. The CPe is responsible for ∼80% of CSF production in rodentS^_3,7,50,101 and probably contributes a similar percentage in humans. The remainder of the CSF is likely to be derived from bulk flow of brain interstitial fluid3. The CPe is the most actively secreting epithelium in the human body, producing CSF at a rate of ∼400–500 mL per day^₅₀. As such, the CPe receives more blood flow per gram of tissue than any other tissue in the body^_50,102 and metabolizes more ATP than any other epithelium^₁₀₃. CSF secretion by the CPe is subject to strict regulation and can be modified by multiple neuro-humoral mechanisms ^₅₀.

Epithelial barrier cells are constantly exposed to microbes and other environmental insults that can compromise tissue function, either by excessive activation of inflammation or by direct cell damage. The major challenge facing the immune system is to neutralize foreign invaders and resolve injury without inflicting the collateral damage that perpetuates a chronic inflammatory cycle. Maintaining immune homeostasis is particularly challenging at barrier sites where constant exposure to immunogenic agents can induce destructive inflammation. Although the function of the innate immune system at barrier epithelia in the intestine, skin, and respiratory tract has been well studied, the immune functions of the blood-CSF barrier (CPe) and the brainCSF barrier (ependyma) have received less attention.

The CPe functions as a tightly regulated gate that separates the blood and CSF, but allows circulating immune cells to enter the brain for defense and repair_¹⁰⁴. Like other epithelial cells, the cells of the CPe express toll-like receptors (TLRs) on their surface. Pathogen-associated molecular patterns (PAMPs) in the CSF bind to these TLRs, resulting in activation of nonspecific innate immune responses_^105,106. Several different TLRs_^64,107, including TLR4_⁷, are highly expressed in the CPe and are regulated in specific ways by different pro-inflammatory stimuli_^7,64,65. The gram-negative bacterial cell wall component lipopolysaccharide, which is common in patients with PIH in Western countries, is a classic PAMP and canonical TLR4 ligand that activates NF-κB-dependent cytokine production and immune cell recruitment_^106,108 (figure 1).

TLRs also recognize damage-associated molecular patterns (DAMPs), or “alarmins”, which are released from tissue in response to injury and are interpreted by adjacent cells as foreign danger signals_¹⁰⁹. DAMPs that bind to TLRs include heat shock proteins,_¹¹⁰ matrix degradation products_^111–113, the S100A8-S100A9 protein complex_¹¹⁴, lysophosphatidic acid_¹¹⁵, and intraventricular haemorrhage-derived blood-breakdown products such as methemoglobin (metHgb) and iron_{^{74,110,111,116}} (figure 1)

TLR4-dependent CPe hypersecretion in PHH

The CPe is one of the first brain structures to encounter extravasated blood after intraventricular haemorrhage _{^{48,64,117,118}}. Recent studies in animal models have shown that intracerebroventricular injection of autologous blood into the lateral ventricles is sufficient to cause ventriculomegaly, as well as NF-κB activation and cytokine production in CPe cells_^7,66. In rabbit pups and human infants with intraventricular haemorrhage, the level of metHgb in CSF strongly correlates with that of the TLR4-dependent cytokine TNF-α_^64,65. Furthermore, at physiological concentrations, intracerebroventricular delivery of metHgb activates TLR4 homodimers or TLR4/2 heterodimers,_^119,120 promoting nuclear translocation of NF-κB and TNF-α and IL-1β secretion,_^116,119,120 and is sufficient to cause ventriculomegaly_^64,65.

Interestingly, many secretory epithelia respond to pro-inflammatory stimuli by increasing fluid secretion_¹²¹, which helps maintain homeostasis by clearing pathogens or debris from the epithelial surface_^122,123. However, inappropriately initiated or sustained inflammation of secretory epithelia can lead to disease _^124,125. For example, dysregulated epithelial inflammation and associated fluid hypersecretion can be observed in several conditions, including chemical, autoimmune or infectious forms of pleuritis, colitis and pancreatitis_¹³,. In addition, chronic inflammation can cause tissue damage, which propagates and amplifies the initial inflammatory response via the release of host-derived DAMPs.

Until recently, the impact of inflammation on the secretory capacity of the CPe has been difficult to study, reflecting a paucity of techniques that can adequately measure and manipulate rates of CSF secretion in vivo. However, a recently developed microneurosurgical technique that enables the real-time measurement of CSF secretion rate in live rats _¹²⁶ has facilitated several novel observations of CSF dynamics in experimental PHH7. Infusion of autologous blood into the right lateral ventricle provoked a TLR4-NF-κB-dependent CPe inflammatory response that was associated with a >3-fold increase in CSF secretion. This increase in secretion was detected from 24h to at least 7 days after experimental intraventricular haemorrhage and could be inhibited by administration of the NKCC1 transporter inhibitor bumetanide. At the 7 day timepoint ventriculomegaly was also osberved_⁷. CPe inflammation was characterized by greatly up-regulated phosphorylation of NF-κB, production of TNF-α and IL-1β, as well as infiltration of activated ED-1₊ microglia and macrophages (Figure 2)_⁷. The same study also found that this intraventricular haemorrhage-induced CSF hypersecretion resulted from TLR4-dependent activation of the NF-κB-regulated STE20/SPS1-related, proline-alanine-rich kinase (SPAK).

Figure 2. Proposed mechanism of CSF hypersecretion in PHH and PIH.

Host-derived danger-associated molecular patterns (DAMPs) such as methemoglobin (metHgB) enter the CSF during intraventricular haemorrhage and pathogen-associated molecular patterns (PAMPs), such as lipopolysaccharide (LPS) enter the CSF during bacterial meningitis. These DAMPs and PAMPs are thought to bind toll-like receptor 4 (TLR4) on the surface of the choroid plexus epithelium (CPe). This binding stimulates a TLR-4-MyD88 signalling cascade leading to nuclear translocation of nuclear factorkB (NF-κB). Nuclear NF-κB stimulates production of pro-inflammatory cytokines, for example, tumour necrosis factor-α (TNF-α) and interleukin 1β (IL-1β), which increase activity of Ste20-type stress kinase (SPAK). SPAK phosphorylates its canonical substrate, the Na₊/K₊/2Cl ion co-transporter- (NKCC1), and probably also phosphorylates other ion transporter targets. NKCC1 phosphorylation increases activity of the transporter, which results in a net increase in cerebrospinal fluid (CSF) production by the CPe. DAMPs and PAMPs in the CSF also bind to TLR-4 on the surface of microglial cells that are resident on the choroid plexus. This binding results in the production of pro-inflammatory cytokines by the microglia. These cytokines can bind receptors on the CPe and likely progogate CPe inflammation and CSF hypersecretion. AQP1, AE2 and NCBE are some additional transporter proteins that facilitate the passage of water (AQP1) and ions (AE2 and NCBE) across the plasma membrane. Cl₋

SPAK integrates and transduces environmental stress signals, including NF-κB-regulated inflammatory cytokines _^127–130 such as TNF-α, IL-1β, and IFN-γ. In addition, NF-κB is itself a transcriptional regulator of SPAK_¹³⁰. Interestingly, in models of colitis,_¹³¹ IgA nephropathy, _¹³² and hypoxic lung injury_¹³³, TNF-α_¹³⁰ and IFN-γ_¹³⁴ have been shown to stimulate SPAK in an NF-κB-dependent manner, indicating that positive feedback occurs. SPAK directly interacts with the TNFα receptor RELT to activate downstream stress response pathways mediated by p38 and JNK1/2 signalling_¹³⁵. When activated, SPAK binds, phosphorylates, and stimulates the cotransporter NKCC1 at the CPe apical membrane_^7,136,137. NKCC1 accounts for >50% of the total CSF production by the CPe, and SPAK is the most important regulator of NKCCl_^7,101. However, SPAK also binds and/or regulates multiple other ion transporters_¹³⁸, for example, the CPe basolateral membrane transporters Na₊-dependent Cl₋/HCO₃₋ exchanger (NCBE) and (in the presence of scaffolding protein spinophilin) Na₊-independent CI-/HCO₃₋ exchanger (AE2)_¹³⁹. Interestingly, stimulatory phosphorylation of NCBE_¹⁴⁰ was recently implicated in CSF hypersecretion in neonatal PHH, suggesting that SPAK might be responsible for inflammation-induced up-regulation of NCBE_¹⁴⁰. Therefore, SPAK seems to be a crucial link between TLR4-dependent CPe inflammation and CSF hypersecretion. In an animal model of PHH, genetic inhibition of TLR4 or SPAK returned CSF secretion to healthy levels and prevented hydrocephalus by decreasing intraventricular haemorrhage-induced phosphorylation of NKCC1, as did treatment with inhibitors of TLR4-NF-κB or SPAK-NKCC1 before intraventricular haemorrhage_⁷. These data suggest that the ability of CPe CSF production capacity to be dynamically regulated in response to inflammation, known as ‘immuno-secretory plasticity’, is important in the pathogenesis of acute PHH, and that pharmacological targeting of TLR4 or SPAK could be a promising treatment approach.

TLR activation in PIH

Bacterial CNS infection is probably the most common cause of PIH_^5,8. Bacteria can cross the blood-brain barrier (BBB) and CPe to gain entry to the CNS_¹⁴¹. Once in the CNS, bacteria replicate within subarachnoid and ventricular CSF spaces, and cause intense CPe, ependymal, and CSF inflammation by releasing cell wall fragments that are highly immunogenic_¹⁴². The robust acute inflammatory response associated with PIH has been assumed to cause noncommunicating (obstructive) hydrocephalus via blockage of the aqueduct, the 4th ventricle outlets, or the basal subarachnoid spaces around the 4th ventricle. However, the acute onset of PIH, often within 12 hours of infection, precedes the expected onset of post-inflammatory scarring and aqueductal obstructions.

PAMPs contained in PIH-causing organisms promote local inflammation through recognition by antigen-presenting cells, for example, microglia and CPe cells, that express cell surface patternrecognition receptors, for example, TLRs_¹⁴⁴. TLR2, TLR4 and TLR5 are all expressed by the CPe_^64,107, and exhibit exhibit ligand-specific regulation in response to pro-inflammatory stimuli from PIH-associated microorganisms. TLR4 recognises lipopolysaccharide found on gram-negative bacterial cell walls_^7,30,64,65 and S. pneumoniae-derived pneumolys_¹⁴⁵, and TLR4 activation by these stimuli causes the CPe to produce inflammatory cytokines such as TNF-α, IL-1β and IL-6_^146–148, which results in recruitment of additional immune cells into the CNS from across the BBB and CPe_¹⁴⁹. TLR2 recognizes lipoteichoic acids from S. pneumoniae,_¹⁵⁰ L. monocytogenes,_^151–153 and S. agalactiae,_¹⁵⁴ whereas TLR5 recognizes flagellin of flagellated bacteria_¹⁵⁵. There is some evidence that activation of TLR2_¹⁵⁶ in the CPe leads to chemotaxis and leukocyte infiltration; however, additional work is needed to elucidate the function of the TLRs in the CPe.

The role of the ependymal epithelium

The pathogenesis of hydrocephalus after hemorrhage or infection is certainly not limited to CSF hypersecretion from the CPe. The ependymal epithelium, ventricular zone and subventricular zone also contribute to the development of PHH in neonates; these structures are likely involved in PIH,_^157–159 however, more research directly investigating this is required. The ventricular zone is a single layer of mono-ciliated neural stem cells that lines the embryonic ventricular system. In utero, these cells develop into the multi-ciliated ependymal ventricular wall that separates the CSF-filled venrticles from underlying brain parenchyma_^160,161. The subventricular zone lies adjacent to the ventricular zone and is a region of densely populated neural progenitor cells _^160,161. Together, the ventricular zone and subventricular zone are critical regions for the birth of newborn neurons and glia during perinatal development_¹⁶¹. Postmortem histological analysis of frontal and subcortical brain regions showed that, compared with controls, infants with intraventricular haemorrhage had a loss of neural stem cells in the subventricular zone, reduced numbers of multi-ciliated ependymal cells, cytoplasmic relocation of N-cadherin (the adhesion protein connecting the cells of the ventricular zone), periventricular heterotopia, and abnormal invasion of astrocytes into areas affected by hemorrhage_¹⁶². These findings were associated with abnormal brain development, altered CSF dynamics, and the development of ventriculomegaly_^162,163. The histological findings were supported by findings from in vitro models of mouse intraventricular haemorrhage_¹⁶³. Together, the clinical and preclinical data suggest that intraventricular haemorrhage leads to loss of developing and mature ciliated epithelial cells, resulting in disrupted ciliary beating and abnormal CSF flow. This disruption, in combination with developmental abnormalities such as periventricular heterotopias and glial activation, could contribute to the pathogenesis of PHH and associated neurodevelopmental sequelae.

PHH and PIH: shared therapies?

Given that TLRs recognize both DAMPs and PAMPs, PIH and PHH might share common pathogenic mechanisms driven by PAMP and DAMP-triggered innate immune responses, raising the possibility that anti-inflammatory treatments could modulate development of hydrocephalus in both conditions. Before permanent CSF shunting, many patients with PIH and PHH require urgent placement of temporary CSF diversion devices such as external ventricular drains, implanted access reservoirs, or lumbar drains. In patients with these devices, intraventricular administration of medications targeting TLR4-dependent inflammation seems particularly attractive. Such medications could include anti-inflammatory agents that target the TLR4-NF-κB pathway, for example, the TLR4 inhibitor Tak242, pyrrolidine dithiocarbamate_¹⁶⁴ or melatonin_^165,166. Data from an animal model suggest that systemic administration of Tak242 is effective in treating PHH_⁷. Additionally, systemic administration of Tak242 was tested in a clinical trial in patients with sepsis_¹⁶⁷. Although Tak242 administration did not significantly alter mortality rate or suppress cytokine levels, the authors noted that gram-status of bacteria was not an inclusion criteria. Reterospective analysis showed that only 40% of patients in the study had gram-negative infection, which would be more likely to respond to TLR-4 inhibition_¹⁶⁷ than gram-positive or fungal infection. Agents that sequester DAMPs, PAMPs or cytokines, for example, neutralizing antibodies or decoy receptor "sponges" _^168,169, might also be effective treatments for PIH and PHH, and could be more specific than corticosteroids, a more general anti-inflammatory agent. Obviously, each of these potential agents will require experimental validation in models of PIH and PHH.

Given the importance of NKCC1 in CSF secretion_¹⁰¹ and intraventricular haemorrhage-induced CSF hypersecretion_⁷, the NKCC1 inhibitor, bumetanide could reduce the CSF secretion induced by acute inflammation in PIH and PHH. Systemic bumetanide was found to reduce the symptoms of autism in children_^170–172; however the drug and its derivatives show a low level of CNS penetration that indicates that systemic administration of these agents to treat neurological disorders might not always be effective_^7,55. Therefore, intracerebroventricular delivery of bumetanide via the CSF diversion devices discussed above might be a more suitable method of delivery. However, bumetanide has been associated with hearing loss when added to phenobarbital for treatment of seizures in neonates_¹⁷³, suggesting that bumetanide should not be administered to infants at this age.

SPAK might be preferable to NKCC1 as a therapeutic target for PIH and PHH, as SPAK is more highly expressed in CPe than in any other epithelial tissue, is an amplifier of the TLR4-dependent inflammatory reaction and cytokine production? and is a master regulator of multiple ion transporters_¹³⁶. Regardless of the potential of targeting SPAK, we propose that targeting inflammation as opposed to CPe ion transport is the most promising therapeutic approach for PIH and PHH, because, in addition to driving the initial CSF hypersecretory response, inflammation is likely to contribute to the ensuing tissue damage and release of DAMPs that ultimately leads to sustained hydrocepahlus. Additionally, it might be beneficial to preserve the acute CSF hypersecretion response as it could clear debris from the CPe and ependymal surface.

Reparative vs. damaging inflammation

Although recent studies have begun to identify inflammatory mediators of PHH,_{^{7,64–66,116,174}} numerous gaps in our understanding remain. For example, we do not know the identities of the intraventricular haemorrhage-induced metabolite(s) that bind to TLR4, although metHgb has already been identified as a TLR4 ligand_¹¹⁶. In addition, further work is needed to identify the components of the TLR4 signalling cascade induced by acute or chronic intraventricular haemorrhage, the dynamic spectra and profiles of TLR4-dependent cytokines and immune cells, and the contribution of additional inflammation-dependent mechanisms (for example, those resulting from accompanying tissue injury) to PHH. Establishing the duration of inflammation-induced CPe hypersecretion and whether TLR4 inhibition after intraventricular haemorrhage can prevent PHH will also be important.

It seems likely that a CSF hypersecretory response from an inflamed CPe contributes to development of acute hydrocephalus, which occurs before chronic inflammation can lead to scarring. However, additional TLR-dependent or innate immune mechanisms, for example, activation of microglial NOD-like receptors _^175,176, can be triggered by inflammation-induced tissue damage of CNS barrier epithelia (the CPe and ependyma) and associated DAMP release. Activation of these additional mechanisms could propagate and sustain the neuro-inflammatory reaction in the CPe, and affect other CSF homeostatic pathways such as the recently characterized and still controversial glymphatic system (Figure 2). Neuroimaging studies of patients with idiopathic normal pressure hydrocephalus revealed significant suppression of glymphatic clearance in these individuals_^177,178. This observation is significant, because the glymphatic system is a possible alternative pathway for CSF efflux.

In the glymphatic model of CSF circulation, arterial pulsatility drives CSF influx into the periarterial space. A combination of arterial pulsatility and high levels of the water channel aquaporin-4 in the vascular endfeet of astrocytes then facilitates movement of CSF from the periarterial space to the brain parenchyman, where the CSF mixes with interstitial fluid (ISF). This mixture of CSF and ISF is drained via the perivascular spaces surrounding the deep veins, as well as by cranial and spinal nerves, and is collected by meningeal and cervical lymphatic vessels_¹⁷⁹. The recent identification of the glymphatic system raises the question of whether inflammation-dependent impairment of glymphatic fluid efflux contributes to PHH or PIH. In a recent animal study, germinal matrix haemorrhage resulted in impaired glymphatic transport. In this study, inhibition of reactive astrogliosis after haemorrhage improved glymphatic function and attenuated progression of PHH_¹⁸⁰. Impaired glymphatic fluid transport was also noted in mouse models of subarachnoid bleeding, traumatic brain injury, and inflammation_¹⁸¹. However, some experimental evidence does not support a major role for the glymphatic pathway in CSF homeostasis_^182–187.

After injury, cytokines derived from microglia and epithelia induce the production of additional cytokines and growth factors by underlying connective tissue fibroblasts, promoting epithelial proliferation and repair_¹⁸⁸. As in the intestine and respiratory tract_¹⁸⁹, chronic tissue damage or inflammation in the CPe and ependyma, such as that associated with extensive intraventricular haemorrhage (for example, grade IV germinal matrix haemorrhage) and with partially treated ventriculitis, might drive conversion of activated connective tissue fibroblasts to extracellular matrix-producing myofibroblasts. This conversion could lead to fibrosis, recruitment of inflammatory cells, and excessive production of inflammatory mediators, thus driving pathological inflammation that exacerbates tissue damage. In support of this theory, damage to the CPe, intraventricular fibrosis and septation, and friable ependyma have all been observed in individuals with chronic PIH or PHH_^10,190,191. In addition, ependymal denudation and ventricular zone disruption were identified in patients with chronic PHH and in animal models of the condition_¹⁶².

Some studies have shown reduced CSF secretion in chronic hydrocephalus_^192–194. Silverberg et al.,_¹⁹³ suggest that this reduced CSF production results from prolonged elevation in intracranial pressure. Kosteljanetz et al.,_¹⁹⁴ noted great variation (within and among patients) in the rate of CSF production after subarachnoid haemorrhage, though measurements from different individuals were taken at different times after the initial bleeding event. In our proposed model of the contribution of inflammation to hydrocephalus (Figure 3), CSF hypersecretion occurs acutely (1–7 days) after infection or haemorrhage and is then followed by scarring, inflammation (fibrosis), and CSF malabsorption. The latter phase would be accompanied by normal or even decreased levels CSF production as the CPe becomes chronically scarred and fibrotic.

Figure 3. Glymphatic CSF transport.

The glymphatic system is a perivascular cerebrospinal fliud (CSF) and interstitial fluid (ISF) exchange network that mediates waste clearance and CSF efflux from the brain to outlets such as the cervical or meningeal lymphatic system and the major draining venous sinuses. In the glymphatic system, as arteries on the surface of the cortex penetrate the brain, CSF enters the parenchyma alongside the vessels, ensheathed by astroglial endfeet. Driven by cardiac-driven arteriole pulsations, and facilitated by the high expression of aquaporin 4 (AQP4) in the astroglial endfeet, CSF exits the perivascular space and mixes with the brain’s ISF. Either by bulk flow or diffusion, the mixture of CSF and ISF flows through the parenchyma into either perivenous or perineural spaces (perineural spaces not illustrated). The fluid then travels along the perivenous or perineural spaces until it drains into to the dural sinuses or lymphatic vessels on the way to the general circulation for clearance.

At present, our ability to investigate CSF circulation in individuals with hydrocephalus is limited by a paucity of appropriate non-invasive molecular imaging tools. CSF circulation is thought to occur either according to the bulk flow model or the hydrodynamic model. In the bulk flow model, CSF is secreted by the CPe and enters the venous system via arachnoid granulations. According to this model, hydrocephalus results from the obstruction of CSF as it flows from the ventricles to the arachnoid granulations⁵. In the hydrodynamic model, arterial systolic pressure waves entering the brain are transmitted to the subarachnoid spaces, venous capacitance vessels, and interact with intraventricular pulsations transmitted by the choroid plexus_^195,196. The intraventricular pulsations facilitate CSF egress through the ventricular outlet foraminas. In this model, hydrocephalus is caused by decreased elasticity (secondary to increased ICP or hypertension) of these pulsation absorbing structures, contributing to abnormally high pulsation amplitudes that result in ventricular expansion. Elements of both models are likely to be operative in both normal CSF homeostasis and hydrocephalus, with possible contributions from glymphatic and lymphatic pathways.

Imaging modalities such as contrast magnetic resonance imaging have been used to assess CSF flow in the Sylvian aqueduct between the third and fourth ventricles. In some cases of communicating hydrocephalus, these imaging measurements have revealed up to 6-fold increased retrograde fluid flow through the aqueduct._¹⁹⁷ Other non-invasive imaging modalities such as time-spatial labelling inversion pulse imaging_^198,199 enable measurement of CSF movement in real-time and could be applied to the study of hydrocephalus. Innovative imaging reagents such as bis-5-hydroxy-tryptamide-diethylenetriaminepentaacetate gadolinium with cross-linked iron oxide nanoparticles_²⁰⁰ or europium-doped very small superparamagnetic iron oxide particles_²⁰¹ seem to be particularly sensitive detectors of CPe neuroinflammation. However, these reagents have not yet been applied to the study of inflammatory hydrocephalus.

Conclusions

Emerging data have identified inflammatory pathways involving TLR4-regulated CSF cytokines and immune cells, that are likely to be important for the pathogenesis of hydrocephalus, suggesting that pharmacological prevention of PIH and PHH is feasible. Nonetheless, much additional work is needed before any of the potential treatment strategies discussed in this Review can be tested in clinical trials. This work will involve the continued identification of specific inflammatory mechanisms that contribute to the pathogenesis of PHH and PIH; development of pharmacological agents that modulate these targets; and pre-clinical trials of these agents in relevant experimental models. A therapeutic approach that addresses neuroinflammation might not only prevent shunt-dependence, but might also ameliorate the neurodevelopmental sequelae of PIH and PHH that are not addressed by surgical CSF diversion, for example, inflammation-induced tissue damage and resultant cerebral palsy. Such an approach would reduce the lifelong morbidity and economic burden associated with hydrocephalus surgery, and could be life-saving in regions with limited access to neurosurgical care.

Figure 4: Proposed inflammatory contributors to PHH and PIH.

Illustration showing the relative contribution of inflammation to hydrocephalus in the days week and months following haemorrhage or bacterial infection. CNS exposure to foreign pathogen-derived damage-associated molecular patterns (PAMPs), for example bacterial cell wall fragments, or host-derived damage-associated molecular patterns (DAMPs), for example blood-breakdown products, leads to an acute inflammatory response (red) in the CSF that takes place in the days to weeks after haemorrhage or infection. This response is characterized by recruitment of immune cells (e.g. microglia) and TLR4-dependent CSF hypersecretion by the CPe. Tissue damage, including friability and denudation at CSFbrain (ependyma) and CSF-blood (CPe) barrier sites, is likely to propagate and sustain the initial infectious or traumatic insult via release of other DAMPs, resulting in a transition from acute reparative inflammation (red) to chronic pathological inflammation (blue). This chronic inflammation is likely to result in scarring and obstruction of CSF drainage pathways (e.g., brain parenchymal glymphatics and meningeal lymphatics; green), which would impair CSF reabsorption. Early modulation of TLR4 activity in post-infectious and post-haemorrhagic hydrocephalus could reduce the acute CPe hypersecretory response, and prevent chronic inflammation-induced scarring. In addition, anti-inflammatory therapies offer the potential advantage of preventing the need for surgical CSF diversion, and alleviating inflammation-induced brain damage that contributes to poor long-term neurodevelopmental outcomes.

Key points.

• Hydrocephalus (enlarged brain ventricles associated with failed cerebrospinal fluid [CSF] homeostasis) is the most common neurosurgical disorder and is treated mainly by neurosurgical CSF diversion procedures with high rates of morbidity and failure.

• Post-hemorrhagic hydrocephalus (PHH) and post-infectious hydrocephalus (PIH), the most common causes of hydrocephalus, are both characterized by inflammation in the brain tissue and CSF space.

• Recent data have begun to uncover the molecular mechanisms by which inflammation, driven by activation of to 11-like receptor-4 (TLR4), contributes to the pathogenesis of hydrocephalus.

• Pharmacotherapeutic approaches that target inflammation have the potential to address multiple drivers of PHH and PIH, for example, acute hypersecretion of CSF by the choroid plexus epithelium and scarring of CSF drainage pathways.

Glossary

Time-spatial labelling inversion pulse imaging

A non-contrast magnetic resonance imaging (MRI) technique using cerebrospinal fluid (CSF) as a tracer to measure CSF flow.

Bulk flow model

Movement of cerebrospinal fluid (CSF) from the choroid plexus through the cerebro-ventricles and cisterns to the subarachnoid space, where reabsorbtion through the arachnoid granulations occurs.

Neuro-humoral mechanisms

Sympathetic and hormonal regulation of CSF production. Periventricular heterotopia: Bilateral nodules of grey matter lining the lateral ventricles consisting of neurons that failed to migrate during fetal development.