Post-Hemorrhagic Hydrocephalus Development after Germinal Matrix Hemorrhage: Established Mechanisms and Proposed Pathways

Abstract

In addition to being the leading cause of morbidity and mortality in premature infants, germinal matrix hemorrhage (GMH) is also the leading cause of acquired infantile hydrocephalus. The pathophysiology of post-hemorrhagic hydrocephalus development after GMH is complex and vaguely understood, although evidence suggests fibrosis and gliosis in the periventricular and subarachnoid spaces disrupts normal cerebrospinal fluid dynamics. Theories explaining general hydrocephalus etiology have substantially evolved from the original bulk flow theory developed by Dr. Dandy over a century ago. Current clinical and experimental evidence supports a new hydrodynamic theory for hydrocephalus development involving redistribution of vascular pulsations and disruption of Starling forces in the brain microcirculation. In this review, we discuss cerebrospinal fluid flow dynamics, history and development of theoretical hydrocephalus pathophysiology, and GMH epidemiology and etiology as it relates to post-hemorrhagic hydrocephalus development. We highlight known mechanisms and propose new avenues that will further elucidate GMH pathophysiology, specifically related to hydrocephalus.

Introduction

Germinal matrix hemorrhage (GMH) occurs in approximately 3 live births per 1,000, has a 20–30% mortality rate, and accounts for 1.7% of all neonatal deaths in the United States (Ballabh 2014; Osterman et al. 2015). Premature infants have a much higher rate of occurrence; for infants born before 32 weeks of gestation up to 20%, (about 12,000 infants) develop GMH each year in the US (Kochanek et al. 2012). Fortunately, the premature birthrate and percentage of low birthweight (<2500g) infants have steadily declined between 2006 and 2013, although remaining higher than in the 1980s and 1990s. In 2006, the preterm birthrate was 12.8% and, in 2013, the preterm birthrate declined to 11.39% while the percentage of low birthweight infants was relatively unchanged at 8.02%. The percentage of very low birthweight (<1500g) was 1.41% in 2013 (Osterman et al. 2015). A study investigating premature infants dating back to 1914 determined median postnatal survival increased from 2 to 26 days, and median gestational age decreased from 33 to 27 weeks. Interestingly, GMH incidence was 4.7% before 1960, but it increased to 50.0% between 1975 and 1980, and then decreased to 12.5% after 2005 (Hefti et al. 2015). The introduction of positive pressure ventilation in preterm clinical management after the 1960s increased survival while simultaneously increasing GMH incidence, which may be attributed to cardiorespiratory and hemodynamic instability associated with mechanical ventilation, and the decline in GMH incidence after the 1980s may be attributed to improvements in mechanical ventilation methodology as well as the use of antenatal steroids and surfactant. Despite improving trends in premature birth incidences and outcomes, GMH remains the leading cause of morbidity and mortality in premature and/or very low birthweight infants, and its incidence has remained steady in the past decade.

Premature and very low birthweight infants are prone to hemodynamic and cardiorespiratory instability, leading to abrupt fluctuations in cerebral blood flow (Ballabh 2014). The fetal brain is hypothesized to lack vascular autoregulatory mechanisms to adequately prevent cerebral blood flow fluctuations, although clinical research involving cerebral blood flow monitoring in preterm infants has produced ambiguous results (Alderliesten et al. 2013; Caicedo et al. 2011; du Plessis 2008; Soul et al. 2007; Tsuji et al. 2000; Wong et al. 2008). The germinal matrix layer, which is present in the fetus and matures by term, contains many neuronal and glial precursor cells and is a site of rapid angiogenesis relative to other parts of the brain (Ballabh et al. 2004; Ballabh et al. 2007). The germinal matrix neurovascular unit, consisting of neurons, astrocytes, pericytes, vascular smooth muscle cells, and vascular endothelial cells, is deficient in fibronectin at the endothelial basal lamina, glial fibrillary acidic protein at astrocyte end-feet, and pericyte coverage (Ballabh 2010; Ballabh 2014). Thus, the germinal matrix vasculature is inherently weak and vulnerable to hemorrhage under abnormal conditions, regardless if the premature infant brain has autoregulatory mechanisms to adequately prevent cerebral blood flow fluctuations.

GMH severity is graded on an I-IV scale based on the extent and localization of bleeding. Incidence of higher GMH grades (III-IV) increases as gestational age and/or birthweight decreases (Robinson 2012). Between 50–75% of GMH survivors develop long-term neurocognitive sequelae, including cerebral palsy, learning disabilities, psychiatric disorders, and post-hemorrhagic hydrocephalus (PHH), and higher-grade GMH survivors are most vulnerable to worse long-term outcomes (Ballabh 2010; Ballabh 2014). The mortality rate for severe grade (III-IV) GMH is approximately 44%, with 60% of survivors developing PHH and 25% requiring surgical installation of shunts (Vassilyadi et al. 2009). Another study estimates 10% of GMH patients (any grade) and 20% of severe GMH patients (III-IV) will require surgical insertion of permanent shunts (Robinson 2012). Shunt dependency is not desirable, given the large, costly, detrimental complications that occur due to shunt infection, occlusion, and displacement. Additional approaches to manage or prevent PHH include serial lumbar punctures, ventricular taps, external ventricular drainage, ventricular access device, ventricular-subgaleal shunt, endoscopic third ventriculostomy, and endoscopic coagulation of the choroid plexus (Tully and Dobyns 2014). A non-invasive, therapeutic approach towards ameliorating PHH would significantly improve long-term quality of life for GMH patients.

PHH pathophysiology after GMH remains vague and complex, and minimal advancements have been made in its clinical management. In this review, we discuss CSF flow dynamics, particularly focusing on its importance in GMH. We highlight advancements made in hydrocephalus research after Dr. Dandy first proposed the bulk flow theory over a century ago. We discuss the current hydrodynamic theory for hydrocephalus pathophysiology and how it applies to PHH development after GMH. Special attention is given to CSF dynamics, CSF production at the choroid plexus, and CSF circulation through the glymphatic system. We identify gaps in current research and propose avenues for further exploration.

Cerebrospinal Fluid Flow Dynamics

Cerebrospinal fluid (CSF) is an isotonic solution that primarily acts as a mechanical cushion for the brain, although it serves many other physiologically vital functions as well (Chakravarthi 2012). CSF has a lower specific gravity than brain tissue, creating a buoyant force that reduces the effective mass of the brain. CSF [H⁺] concentration is detected by central chemoreceptors located at the ventrolateral medullary surface, which help regulate pulmonary ventilation and cerebral blood flow to ensure the brain receives ample oxygen and nutrients. CSF also maintains a stable external environment for growth and development of neurons and glia (Chakravarthi 2012). Importantly, CSF removes brain metabolic waste and transporting neuropeptides, glucose, and lipids (Iliff et al. 2012; Xie et al. 2013).

Production

In adults, between 400–600 mL of CSF is produced per day and the brain renews its CSF between 3–4 times within a 24 hour period (Cutler et al. 1968; Pierce et al. 1962; Sahar 1972; Sato et al. 1975). CSF is primarily produced by the choroid plexus epithelial lining and, to a minimal extent, the cerebral ventricular ependymal lining, which compose the blood-CSF barrier. Choroid plexus epithelial cells are interconnected by tight junctions that are leakier than endothelial cells of the blood-brain barrier. Over two thirds of produced CSF originates from the choroid plexus (Pollay 1975; Segal and Pollay 1977). The choroid plexus lines the lateral ventricles from the inferior horns to the interventricular foramen, where it becomes continuous into the third ventricle and continues into the fourth ventricle. The choroid epithelium protrudes into the ventricles through invaginations of the pia matter containing choroidal capillaries, called tela choroidea, which significantly increase the surface area of the choroidal epithelium (Davson and Segal 1970; Johanson et al. 2011; Keep and Jones 1990; Speake and Brown 2004). The choroid plexus vasculature is also fenestrated to better facilitate CSF production. Non-choroidal ependymal cells, brain interstitial fluid, and capillaries may be other CSF sources as well, which is secreted by transependymal seepage into the brain ventricles or transpinal seepage into the subarachnoid space (Davis and Milhorat 1975; Milhorat et al. 1975; Pollay and Curl 1967; Saunders et al. 1999).

The posterior choroidal, anterior choroidal, inferior cerebellar and superior cerebellar arteries supply the choroid plexus of the lateral ventricles, third ventricle, fourth ventricle, and temporal horns, respectively (Chakravarthi 2012; Milhorat 1978; Sakka et al. 2011). In adults, blood flow to the choroidal epithelium is estimated at 4 – 6 mL / minute / gram tissue, which is significantly greater than blood flow to other brain tissue estimated at 0.9 – 1.8 mL / minute / gram tissue (Maktabi et al. 1991). The choroidal interstitial compartment is the region between choroidal capillaries and choroidal ependymal cells. Choroidal capillaries lack tight junction proteins in their endothelial cells, making them more permeable, and blood plasma filtrate passively crosses into the choroidal interstitial compartment from the choroidal capillaries primarily by Starling forces (Welch 1975; Wright 1972). Starling forces are hydrostatic, and oncotic forces that govern the movement of fluid across capillary membranes. Hydrostatic forces refer to the difference in fluid pressure between the capillary and interstitium, where higher capillary fluid pressure will drive water into the interstitium. Oncotic forces refer to the difference in solute concentration between the capillary and interstitium, where higher interstitial solute and macromolecule concentration will drive fluid from the capillary into the interstitium. Net fluid movement is the net combined hydrostatic and oncotic forces. Thus the main source for produced CSF is technically choroidal capillaries, not the choroid plexus itself (Bulat and Klarica 2011; Oreskovic and Klarica 2010; Oreskovic and Klarica 2011), although this assertion is contentious.

[Na⁺] and [Cl⁻] from choroidal interstitium are actively exchanged for [H⁺] and [HCO₃⁻], generated by cytosolic carbonic anhydrase on choroidal ependymal cells, using carrier proteins in the choroidal ependymal basolateral membrane. Pumps on the choroidal ependymal apical membrane then expel [Na⁺], [Cl⁻], [K⁺], and [HCO₃⁻] into the ventricle lumen, which generates an osmotic pressure (Keep and Jones 1990; Pollay 1975; Spector and Johanson 1989). Water flows down the created osmotic gradient with the help of aquaporin 1 on the choroidal ependymal apical membrane (Reiber 2003). The CSF contains higher concentrations of [Na⁺], [Mg²⁺], and [Cl⁻] than blood plasma but less [Ca²⁺], [K⁺], [HCO₃⁻], [PO₄⁺], protein (contains 0.3% plasma proteins), amino acids, and glucose (Felgenhauer 1974). Several holes exist in the assertion that CSF is produced by cerebral capillaries via filtration, which need to be kept in mind. The difference in small solutes between plasma and interstitium is small and the concentration of plasma protein is significantly less in brain tissue interstitium, which significantly diminishes the oncotic pressure gradient. Additionally, flux between cerebral capillaries and the interstitium is not unidirectional in reality, as there is almost as much back flux as forward flux (Hladky and Barrand 2014; Hladky and Barrand 2016).

The choroidal epithelium can alter CSF secretion in response to multiple factors and mechanisms. Most regulatory mechanisms target membrane transporters, carbonic anhydrase, and aquaporins (Faraci et al. 1990; Sakka et al. 2011). The NaK2Cl cotransporter, located on the choroidal ependymal apical membrane, helps regulate CSF composition and secretion by its bidirectional transport ability. Arginine vasopressin, atrial natriuretic peptide, serotonin, melatonin, and dopamine receptors are located on choroidal epithelium. Arginine vasopressin and atrial natriuretic peptide decrease CSF secretion. CSF secretion can also be increased by sympathetic innervation and decreased by cholinergic innervation (Chakravarthi 2012). Pharmaceutical drugs that inhibit carbonic anhydrase or sodium transporters, such as diuretics, reduce CSF production, while drugs that augment cerebral blood flow tend to increase CSF production. Increased intracranial pressure also tends to decrease CSF production, although evidence suggests CSF production tends to remain constant despite large increases in hydrostatic pressure (Sakka et al. 2011).

Circulation

CSF flows from the sites of secretion at the choroidal epithelium to the sites of absorption in the subarachnoid space. The mean CSF volume within the adult brain is 150 mL, with 25 mL in the ventricles and 125 mL in the subarachnoid space (Sakka et al. 2011). Generally, CSF flows from the lateral ventricles, passes through the interventricular foramen of Monro into the third ventricle, and finally passes into the cerebral aqueduct of Sylvius into the fourth ventricle. From the fourth ventricle, CSF enters through three openings, the lateral apertures of Lushka and median aperture of Magendie, into the subarachnoid space where it is absorbed. A portion of the CSF exits the cranium through arachnoid villi and cranial nerves while the remainder enters along the spinal cord and exit through spinal nerve roots (Chakravarthi 2012; Dichiro 1964; Milhorat 1976).

CSF circulates through the brain’s ventricular system and spinal cord in a pulsatile manner. Cerebral arterial pulse waves are the primary drivers of CSF circulation, although jugular venous pressure, respiratory waves, and even physical activity play minor roles as well (Post et al. 1974; Williams 1976). CSF flow, however, is very slow and sometimes occurs bi-directionally through ventricle compartments with each cardiac and/or respiratory cycle, but net CSF flow occurs from the lateral ventricles to the subarachnoid space. Additionally, ventricular ependymal cells have cilia that mix CSF while it is circulating. CSF pressure gradients, which are generated by continuous CSF secretion and arterial pulsations, are also important for maintaining CSF flow. This pressure gradient is particularly important in driving CSF flow through the subarachnoid spaces and venous sinuses. In adults, CSF flow across the subarachnoid epithelium is driven by a 6 cm H₂O pressure difference between subarachnoid CSF pressure (approximately 15 cm H₂O) and superior sagittal sinus pressure (approximately 9 cm H₂O), and the pressure continues to drop into the jugular vein and systemic venous system (Bradley 1970; Milhorat 1975; Sahar et al. 1970; Shulman et al. 1964).

Reabsorption

Conventionally CSF is reabsorbed in the subarachnoid space and enters through the dural venous sinuses, where it returns to the internal jugular system (Chakravarthi 2012). Subarachnoid villi, called pacchonian bodies, were originally thought to be the main reabsorption sites (Brierley and Field 1948; Welch 1975; Welch and Friedman 1960), but evidence suggests other potential CSF outflow and reabsorption routes through either cerebral lymphatic channels or the venous system [(Bradbury et al. 1981; Bulat and Klarica 2011; Oreskovic and Klarica 2010; Oreskovic and Klarica 2011; Zakharov et al. 2003). Evidence for the venous system being the main reabsorption site suggests the vast majority of CSF outflow occurs at the superior sagittal sinus with the remainder occurring at dural sinusoids in dorsal root nerves. CSF outflow is driven by pressure gradients between the subarachnoid space and venous sinuses (Cutler et al. 1968; Pollay 2010; Saunders et al. 1999; Zlokovic et al. 1990). Increased intracranial pressure tends to increase CSF outflow, but very high intracranial pressure that persists for a long period of time tends to actually decrease CSF outflow, mostly because venous pressure tends to increase with intracranial pressure while the overall pressure gradient diminishes. Evidence for cerebral lymphatic channels being the main reabsorption site suggests CSF flows along cranial nerves and spinal nerve roots and is reabsorbed in lymphatic channels (Bradbury et al. 1981; Zakharov et al. 2003). Indeed, CSF outflow in the nasal submucosal lymphatic channels through the cribriform plate, which feed into the cervical lymph nodes, is relatively important (Courtice and Simmonds 1951; Cserr et al. 1992; Erlich et al. 1986; Kida et al. 1993; Mollanji et al. 2002; Silver et al. 1999). Lymphatic vessels have also been recently characterized surrounding the dural sinuses, which are also connected to cervical lymph nodes, further suggesting the lymphatic system plays an important role in CSF outflow (Aspelund et al. 2015; Bradbury et al. 1981; Iliff et al. 2015; Louveau et al. 2015; Zakharov et al. 2003; Zervas et al. 1982). Lymphatic-mediated CSF reabsorption is thought to play a greater role in neonates, since subarachnoid granulations are more sparsely distributed.

While the central nervous system lacks a conventional lymphatic system, evidence suggests the presence of a functional waste clearance pathway involving exchange between CSF and interstitial fluid, occurring mostly within perivascular Virchow-Robin spaces in the brain parenchyma (Iliff and Nedergaard 2013; Iliff et al. 2012; Jessen et al. 2015). This exchange system is called the glymphatic system for its lymphatic-like function and dependence upon glial cells (Figure 1). Cerebral arteries at the cortical surface extend into pial arteries running through the subarachnoid space and subpial space, which turn into arterioles surrounded by astrocyte end-feet as they run deeper into the brain parenchyma. The Virchow-Robin space is the CSF containing perivascular space between the astrocyte end-feet and arteriole, with both walls lined by a leptomeningeal cell layer (Kulik et al. 2008; Prince and Ahn 2013; Zhang et al. 1990; Zlokovic 2011). Virchow-Robin spaces along veins lack this leptomeningeal cell layer. Arteriole Virchow-Robin spaces become continuous with the basal lamina, which has minimal resistance to CSF flow due to its loosely structured extracellular matrix (ECM). CSF flows along arteriole Virchow-Robin space, through basal lamina surrounding capillaries, and exits through the venous Virchow-Robin space. Arterial pulsation is the main force driving perivascular fluid bulk movement from the subarachnoid space into the Virchow-Robin spaces; although respiration, slow vasomotion, and CSF pressure gradients play minor roles too (Iliff and Nedergaard 2013; Iliff et al. 2012; Jessen et al. 2015). Astrocyte end-feet have high expression of aquaporin 4 and are important for CSF exchange with interstitial fluid, since astrocyte end-feet surround perivascular spaces and facilitates water movement across cell membranes down osmotic pressure gradients from periarterial to perivenous spaces. Interstitial fluid then drains into cervical lymph channels from perivenous spaces (Johnston et al. 2004; Murtha et al. 2014).

Figure 1:

Overview of the glymphatic system. Cerebrospinal fluid enters within para-arterial Virchow-Robin spaces in the brain parenchyma and an astroglia-mediated mechanism exchanges cerebrospinal fluid with interstitial fluid and flushes wastes out within para-venous Virchow-Robin spaces (A). Astrogliosis (B) from brain injury possibly disrupts this astroglia-dependent mechanism.

The glymphatic system is particularly important for removing soluble proteins and metabolites from the brain (Rangroo Thrane et al. 2013). Glymphatic-mediated exchange is greatest during sleep, which is thought to be important for removing metabolic waste during the resting state (Xie et al. 2013). In rodent models of Alzheimer’s disease, glymphatic-mediated exchange was reduced by 65% in aquaporin 4 knockout mice, resulting in increased accumulation of β-amyloid plaques (Iliff et al. 2012). In a mouse repeated traumatic brain injury model, glymphatic exchange was reduced at 24 hours after the last injury and persisted for up to 4 weeks, which was attributed to gliosis (Plog et al. 2015). Furthermore, the glymphatic system was significantly impaired after subarachnoid hemorrhage, due to blood clots occluding perivascular spaces, and during ischemic stroke, due to reduced arterial pulsations (Gaberel et al. 2014). The glymphatic hypothesis, however, has been challenged by a few groups who identified a few shortcomings with the model, which need to be taken into consideration (Abbott et al. 2018; Smith and Verkman 2018). For instance, it is not entirely clear the role aquaporin-4 plays in interstitial fluid flow and no evidence has been provided for its ability to transport solutes. In addition, the brain extracellular matrix significantly hinders fluid movement. The brain extracellular space also allows for the diffusion of small and large molecules naturally (Abbott et al. 2018; Smith and Verkman 2018). Regardless, a system in which CSF enters perivascular arterioles, diffuses with the brain extracellular space, and is cleared along with interstitial fluid and waste products through perivascular venules does have a presence and warrants further investigation. More research is further elucidating the pathophysiological role the glymphatic system plays in multiple neurodegenerative diseases and injuries, and this system may be particularly important in neonatal GMH and consequent PHH pathophysiology due to the role it plays in CSF dynamics.

Hydrocephalus

The International Hydrocephalus Imaging Working Group defines hydrocephalus as “an active distension of the ventricular system resulting from inadequate passage of cerebrospinal fluid from its point of production within the cerebral ventricles to its point of absorption into the systemic circulation” (Rekate 2008). Clinical consequences can include increased intracranial pressure, seizures, mental deterioration, tunnel vision, gait disturbance, headaches, mental impairment, urinary incontinence, dementia, vomiting, and nausea. Most treatments involve surgical implantation of shunts that divert CSF from the brain or surgery, if possible, to repair any malformations that contribute towards hydrocephalus development (Kahle et al. 2015). Dr. Dandy and Dr. Blackfan classified hydrocephalus into communicating and non-communicating hydrocephalus in 1914 after inducing hydrocephalus in dogs by obstructing the foramen of Monro (Dandy 1914). They proposed the bulk flow theory, which states that CSF flows in bulk from the sites of production in the ventricles to the sites of reabsorption in the subarachnoid space. Hydrocephalus, according to bulk flow theory, had to result from an imbalance in CSF production and absorption. Using the same conceptual framework, Dr. Russell proposed a more specific classification of hydrocephalus in 1949 into non-obstructive and obstructive hydrocephalus, which corresponds to communicating and non-communicating hydrocephalus, respectively (Russell 1949). The original terms proposed by Dr. Dandy, however, remain the most pervasively utilized. Advancements involving CSF tracers and imaging technology, however, has produced evidence challenging the bulk flow theory (Symss and Oi 2013). New hydrocephalus classifications have been proposed based on more recent experimental and clinical evidence, which will be discussed.

Non-communicating Hydrocephalus

Hydrocephalus resulting from an obstruction of CSF flow through ventricular and subarachnoid spaces is called non-communicating hydrocephalus, also known as obstructive hydrocephalus (Kahle et al. 2015; McAllister 2012). Non-communicating hydrocephalus is typically caused by congenital cerebral malformations. Arnold-Chiari malformation, which is the displacement of the cerebellar tonsils through the foramen magnum, often obstructs the fourth ventricle, leading to dilation of the lateral ventricles and cerebral aqueduct (Gardner 1965). Dandy-Walker malformations, characterized by the absence of the cerebellar vermis, often obstruct the foramina of Luschka and foramen of Magendie, resulting in prominent dilation of the fourth ventricle (Hirsch et al. 1984). Colloid cysts may obstruct the Foramen of Monro, resulting in lateral ventricular dilation (Camacho et al. 1989). Other lesions may cause abhorrent narrowing of the aqueduct of Sylvius, called aqueductal stenosis, resulting in third and lateral ventricular dilation.

Communicating Hydrocephalus

Communicating hydrocephalus is impaired CSF reabsorption in the absence of any obstruction to CSF flow through the ventricles and subarachnoid spaces (Kahle et al. 2015; McAllister 2012). Communicating hydrocephalus was believed to primarily result from impaired arachnoid granulations, resulting in reduced reabsorption of CSF. Indeed, cerebral malformations resulting in the absence of arachnoid villi has resulted in hydrocephalus development (Gutierrez et al. 1975). Subarachnoid hemorrhage and intraventricular hemorrhage, which induce inflammation and glial scarring in the subarachnoid space, can cause communicating hydrocephalus as well (Korobkin 1975; Vassilouthis and Richardson 1979). Accumulating evidence, however, challenges the presumption that CSF is mostly absorbed by subarachnoid villi (Greitz 2004; Oreskovic and Klarica 2011). Normal pressure hydrocephalus is a form of communicating hydrocephalus that results in ventriculomegaly without increased CSF pressure. CSF pressure readings are within normal range because ventricular dilation compensates for accumulated CSF in the ventricles, thus increased CSF pressure is compensated by increased ventricular volume in this pressure-volume compensatory relationship (Black and Ingraham 2008). In general, the elderly population is most vulnerable to normal pressure hydrocephalus, and causes are either idiopathic or related to other central nervous system diseases and injuries, particularly subarachnoid hemorrhaging. Hydrocephalus ex vacuo is different from normal pressure hydrocephalus because ventricular dilation results from brain tissue atrophy, usually due to a neurodegenerative disorder, and not as a compensatory mechanism for increased CSF pressure (Rekate 2009). Normal pressure hydrocephalus, however, seems to contradict bulk flow theory, because the ventricles should not dilate without increased mean CSF pressure, although differing pressure waveforms may lead to hydrocephalus despite average pressures remaining the same.

Current Hydrodynamic Theory

Although bulk flow theory is congruent with non-communicating hydrocephalus development, when a ventricular obstruction creates back pressure that dilates the ventricles preceding the obstruction, it is incongruent with communicating hydrocephalus, because the apparent obstruction is within the subarachnoid space, which does not dilate or increase in volume (Greitz 2004; Oreskovic and Klarica 2011). In 1914, Dr. Weed injected Prussian blue into the ventricles of dog and cat brains and found the dye accumulated near pacchonian bodies (Weed 1914). Prussian blue, however, was also found in other brain parenchymal areas, and further research concluded Prussian blue cannot cross pacchonian bodies under normal conditions (Symss and Oi 2013). Even Dr. Dandy recognized reduced bulk flow across pacchonian bodies should result in subarachnoid CSF pressure being greater than ventricular CSF pressure and the subarachnoid space should expand before the ventricles, neither of which is observed. Dr. Dandy concluded CSF is primarily reabsorbed in the subarachnoid space and quickly enters the circulatory system, based on intrathecal dye injections that rapidly entered the blood and urine (Dandy 1929). The idea that CSF is mostly reabsorbed at pacchonian bodies, however, remained pervasive. In 1960, Dr. Welch reported pacchionian bodies could act as mechanical valves, although future anatomical studies found no mechanical valve presence (Welch and Friedman 1960). Dr. Di Chiro started experimenting with radionuclide cisternography and, in 1966, suggested CSF was reabsorbed at pacchionian bodies because radionuclide accumulated there after 24 hours (Di Chiro 1966). Future studies, however, challenged this conclusion since other radionuclides enter the circulatory system within minutes and most are reabsorbed in the spinal canal (Greitz 1993; Greitz et al. 1997; Greitz and Hannerz 1996). Furthermore, sites where radionuclides accumulate after a long period of time may indicate sites where CSF reabsorption is actually very limited. A radionuclide cisternography study in patients with venous vasculitis and high intracranial pressure, performed by Dr. Greitz and Dr. Hannerz in 1996, found no tracer in vessel outlets near capillary beds of pacchonnian bodies, providing evidence for an alternative site of CSF reabsorption (Greitz and Hannerz 1996). Another major issue is pacchonian bodies are absent in infants and young children, suggesting CSF must be reabsorbed by a different mechanism (Papaiconomou et al. 2002).

Some scientists investigated if abnormal vascular and CSF pulsations may be the root cause for communicating hydrocephalus. In 1943, after observing normal pressure hydrocephalus patients and noting inconsistencies with the bulk flow theory, Dr. O’Connell proposed communicating hydrocephalus may result from increased ventricular pulse pressure (O’Connell 1943). Dr. Bering provided experimental evidence in 1962 that choroid plexus pulsations deliver the means for ventricular enlargement instead of increased mean CSF pressure (Bering 1962). Dr. Bering used a kaolin-induced hydrocephalic dog model and excised the choroid plexus from one lateral ventricle, which resulted in asymmetric ventricular dilation. Increased mean CSF pressure, therefore, could not account for asymmetric ventricular dilation. Dr. Di Rocci provided additional experimental evidence in 1978 in which extreme ventricular pulsation, caused by inflating and deflating a microballon inserted into the lateral ventricles, can produce hydrocephalic ventricular dilation in sheep (Di Rocco et al. 1978). Concurrently, Dr. Guinane in 1977 produced olfactory ventricular dilation, which lacks a choroid plexus, in rabbits by obstructing surrounding subarachnoid spaces with silicone rubber (Guinane 1977). The silicone rubber obstruction decreased subarachnoid arterial and venous compliance as well as increased capillary pulsations. Increased capillary pulsations, therefore, had to generate the force necessary for the observed ventricular dilation. Using magnetic resonance imaging and radionuclide cisternography, Dr. Greitz reported in the early to mid-1990s arterial pulsation and expansion provides the force necessary for CSF pulsatile circulation in both the brain and spinal cord, and arterial compliance is important for keeping capillary and venous pulsation low (Greitz 1993; Greitz 2004; Greitz et al. 1997; Greitz and Hannerz 1996).

According to bulk flow theory in which CSF malabsorption is a causative factor for communicating hydrocephalus, the subarachnoid CSF-venous pressure gradient would increase, the subarachnoid space would expand, and the ventricles would dilate after subarachnoid space compliance is at maximum. In actuality, the subarachnoid space is smaller, and the subarachnoid CSF-venous pressure gradient is diminished, although both the subarachnoid CSF pressure and venous pressure increase. In 2002, Dr. Egnor developed a mathematical model of communicating hydrocephalus caused by a redistribution of CSF pulsations in the brain (Egnor et al. 2002). Decreased intracranial compliance causes abnormal distribution of vascular pulsations, such that arterial pulsations are weaker while capillary and venous pulsations are stronger, and stronger pulsations reach the ventricles while weaker pulsations reach the subarachnoid space. Thus, this vascular pulsation redistribution causes the ventricles to expand at the expense of the subarachnoid space and decreases the subarachnoid CSF-venous pressure gradient. Dr. Edgor’s model, based on alternating current electric circuitry, accounted for experimentally and clinically observed CSF malabsorption, increased resistive index, ventricular dilation, intracranial pressure waves, reduced cerebral blood flow, and diminished CSF-venous pressure gradient. Dr. Greitz elaborated on this concept in 2004 in his discussion of hydrodynamic theory of chronic hydrocephalus development. Reduced intracranial compliance causes decreased arterial pulsations and increased compensatory capillary pulsations, generating transmantle pulsatile stress responsible for hydrocephalus. CSF malabsorption, therefore, is not a causative factor of communicating hydrocephalus but an effect from vascular pulsatile redistribution (Greitz 2004). Dr. Oreskovic further suggests that disruption of Starling forces in the brain parenchymal microvasculature lead to an imbalance in interstitial fluid and CSF exchange, contributing to hydrocephalus development (Oreskovic and Klarica 2011).

In light of our increased understanding of hydrocephalus pathophysiology, Dr. Oi and Dr. Di Rocco proposed a new classification based on the involved pathway: major pathway hydrocephalus and minor pathway hydrocephalus (Oi and Di Rocco 2006). Major pathway hydrocephalus accounts for CSF circulation disruption from the ventricles to the subarachnoid spaces. Major pathway hydrocephalus encompasses most obstructive / non-communicating hydrocephalus cases. Minor pathway hydrocephalus accounts for disruptions in CSF circulation within the subarachnoid space and brain parenchyma. Evidence suggests this pathway is very important for CSF reabsorption in the embryo, fetus, and infants, making it critical for infantile hydrocephalus development (Papaiconomou et al. 2002). Minor pathway hydrocephalus is disruption of CSF flow and reabsorption in newly elucidated channels in the brain parenchyma, which involve deep vascular structures and lymphatic channels (Figure 2). Dr. Nedergaard further characterized this pathway in rodents using in vivo two photon imaging and coined the term “glymphatic system”, since this functional waste clearance pathway involves astroglia and lymphatic-like paravascular channels (Iliff et al. 2012). CSF enters from the subarachnoid space into paravascular artery channels and exchanges with interstitial fluid, which is cleared through paravascular veins. Additional lymphatic channels lining the dural sinuses and meningeal arteries were characterized by Dr. Louveaue and Dr. Aspelund (Aspelund et al. 2015; Louveau et al. 2015). However, it is unclear if these meningeal lymphatic vessels are anatomically connected with the glymphatic system. As the cerebral glymphatic / lymphatic systems are further characterized, more research is warranted on their potential pathophysiological roles played in hydrocephalus development.

Figure 2:

Axial non-contrasted adult brain MRI demonstrating perivascular spaces (arrows) that appear hypointense to brain tissue and isointense to CSF in T1-weighted (A) and T2-weighted (B) sequences.

Post-hemorrhagic Hydrocephalus Pathophysiology and Potential Mechanisms

PHH is a common debilitating consequence of severe grade GMH, and the mechanisms contributing to PHH development remain to be elucidated. Cerebroventricular expansion leads to mechanical compression of surrounding brain tissue, causing injury and consequent neurological deficits in patients surviving the initial bleed (Robinson 2012). PHH was commonly theorized to be caused by blood clots obstructing the cerebral aqueduct or foramina of Luschka and Magendie or by microthrombi obstructing small CSF outflow passages in the subarachnoid space. Much evidence suggests a variety of inter-related pathophysiological mechanisms that potentially alter normal CSF dynamics play significant roles in PHH development as well (Strahle et al. 2012; Tang et al. 2016; Whitelaw and Aquilina 2012). Applying concepts in current hydrocephalus theory towards PHH development after GMH may better illuminate potential mechanisms for therapeutic intervention (Figure 3).

Figure 3:

Know pathways and potential mechanisms disrupting cerebrospinal fluid dynamics and contributing to post-hemorrhagic hydrocephalus development after germinal matrix hemorrhage.

Blood Clots, Hemoglobin, and Iron

Non-communicating / obstructive PHH may result from cerebroventricular blood clots and microthrombi directly impairing CSF circulation and absorption by obstructing the cerebral aqueduct, foramina of Luschka and Magendia, and subarachnoid CSF outflow passages. Subsequently, it was hypothesized intraventricular fibrinolytic therapy would remove cerebroventricular blood clots and reduce PHH incidence (Whitelaw and Aquilina 2012). In an adult intraventricular hemorrhage dog model, in which intraventricular blood injection resulted in 80% of dogs developing PHH, intraventricular urokinase injection reduced PHH incidence to 10% (Pang et al. 1986). Clinical investigations of intraventricular streptokinase, urokinase, or tissue plasminogen activator injections after GMH, however, concluded fibrinolytic therapy did not improve long-term dependence on ventriculo-peritoneal shunts (Whitelaw 1993). Thus, cerebroventricular obstruction from thrombi may play only a minor role in long-term PHH development.

Although intraventricular fibrinolytic therapy failed to improve clinical PHH outcomes, evidence suggests hemoglobin and iron may play an important role in PHH development (Strahle et al. 2014). Erythrocyte lysis after hemorrhage, typically from complement activation and consequent membrane attack complex formation, releases hemoglobin and iron into surrounding brain tissue. Experimental adult cerebral hemorrhage models conclude hemoglobin metabolites and iron contributes towards brain edema (Chen et al. 2011). Hemoglobin metabolites were also found in the CSF of rabbit pups with intraventricular hemorrhage, and iron was elevated in the CSF of preterm infants with PHH (Lee et al. 2010; Savman et al. 2001). Intraventricular injection of hemoglobin or iron into neonatal rat pups also resulted in significant acute ventricular dilation (Strahle et al. 2014). Additionally, acute and delayed iron chelation by Deferoxamine reduced long-term PHH development in neonatal rats after GMH (Klebe et al. 2014). Iron, thus, is a quintessential player in PHH formation, although the exact mechanisms remain unclear.

Gene deletion studies determined iron transport and iron-dependent metabolic proteins are highly expressed in the ependymal lining compared to other brain tissue (Keep and Smith 2011). Thus, the ependymal lining may be adversely affected from iron overload due to GMH. Indeed, ependymal cells are theorized to prevent iron diffusion into the brain parenchyma by up-taking it from the CSF (Moos 2002). Additionally, iron overload has been associated with increased expression of aquaporin 4 in adult rats with cerebral hemorrhage, and Deferoxamine treatment reduced aquaporin 4 expression (Qing et al. 2009). Iron, thus, may regulate expression of ependymal ion and water channels, such as aquaporin 4, and contribute towards PHH by altering CSF production dynamics at the ependymal layer. It should be noted, however, that combinatorial furosemide and acetazolamide diuretic treatments targeting choroid plexus epithelial transport were evaluated in clinical trials of preterm GMH patients and determined to have no clinical benefit, although other diuretics have been recommended for further investigation (Whitelaw et al. 2001). More research is needed to further elucidate iron’s pathophysiological role in development of hydrocephalus.

Inflammation, Fibrosis, and Gliosis

Inflammation has been associated with subependymal gliosis, fibrosing arachnoiditis, and meningeal fibrosis after GMH (Cherian et al. 2004b; Oi and Di Rocco 2006). GMH patients also have increased expression levels of inflammatory markers in their CSF, including TNF-α (Savman et al. 2002). Vessel rupture results in blood and serum components entering the brain parenchyma. Resident immune cells, namely microglia, are activated by stimulating toll-like receptors and nod-like receptors with damage-associated molecular patterns, molecules that induce a non-infectious inflammatory response (Klebe et al. 2015). Activated microglia secrete pro-inflammatory cytokines, extracellular proteases, and oxidative species, which damage surrounding tissue and recruit leukocytes that exacerbate inflammation (Chen et al. 2015; Yang et al. 2015). In neonatal rat pups with GMH, microglia proliferation was observed in the perihematoma region, microglia activation was associated with phosphorylated ERK, and modulating microglia activation with minocycline or cannabinoid receptor 2 agonist ameliorate inflammation and improved outcomes (Tang et al. 2015a; Tang et al. 2015b). Additionally, in an IVH adult rat model, IVH caused a TLR4 and NF-κB-dependent inflammatory response in the choroid epithelium, causing an up-to 3-fold increase in CSF production and consequent PHH (Karimy et al. 2017). In the same study, genetic depletion of TLR4 or SPAK as well as pharmacological inhibition of TLR4-NF-κB or SPAK-NKCC1 signaling ameliorated excess CSF production in the choroid epithelium and attenuated PHH. Interestingly, microglia may play an important role in hematoma resolution, since stimulating PPARγ improved short-term hematoma resolution, which was dependent upon CD36 scavenger receptor and was associated with inducing the alternatively activated M2 microglia/macrophage phenotype (Flores et al. 2016).

Fibrosis is the forming of excess connective tissue as a consequence of a reparative process after inflammation (Birbrair et al. 2014). Excess fibrous tissue formation may disrupt the normal functioning of surrounding tissue. Multiple factors trigger fibrosis after GMH. Thrombin, which is significantly active up to 10 days after GMH in neonatal rats, cleaves fibrinogen into fibrin to form fibrin clots, activates the complement pathway to augment inflammation, and stimulates protease-activated receptors (PARs), a family of G protein-coupled receptors (Babu et al. 2012; Lekic et al. 2015; Luo et al. 2007). PAR stimulation has been associated with fibrosis in several tissues, including liver, renal, pulmonary, and cardiac tissues. PAR stimulation upregulates mammalian target of rapamycin (mTOR), which is associated with ECM protein proliferation. Additionally, PAR stimulation exacerbates inflammation by upregulating cyclooxygenase 1 and 2 activity (Kataoka et al. 2003; Luo et al. 2007; Steinhoff et al. 2005). Phosphorylated mTOR and cyclo-oxygenase 2 levels were increased by 72 hours after GMH in rats, which were both reduced by combinatorial PAR-1,4 inhibitor administration (Lekic et al. 2015). ECM proteins are theorized to deposit within the cerebroventricular system, similar to blood clots and microthrombi (Bowen et al. 2013; Strahle et al. 2012; Tang et al. 2016). ECM protein overproduction, therefore, may obstruct normal CSF flow pathways. Indeed, fibronectin and vitronectin expression levels are significantly increased in GMH rats with long-term PHH (Klebe et al. 2014; Manaenko et al. 2014). Inhibiting mTOR with rapamycin and inhibiting cyclo-oxygenase 2 activity ameliorated long-term PHH and neurocognitive deficits in GMH rats, although expression levels of ECM proteins was not determined in this study (Lekic et al. 2015).

TGF-β stimulates mesenchymal stem cells and fibroblasts, which produce ECM matrix proteins and deposit connective tissue (Bowen et al. 2013). TGF-β can be secreted from activated microglia, and TGF-β secretion can be induced by thrombin (Schuliga 2015). ECM production induced by TGF-β stimulation may deposit in the cerebroventricular system, disrupting CSF dynamics (Tada et al. 1994). A rabbit pup GMH model indicated TGF-β, fibronectin, and laminin expression levels were significantly increased in the ependymal and subependyma tissue after GMH (Cherian et al. 2004a). Mice with transgenic TGF-β overexpression developed hydrocephalus with higher expression of ECM proteins in the brain than wild-types (Wyss-Coray et al. 1995). In a clinical study, increased TGF-β1 and ECM protein expression in the CSF were associated with PHH development in preterm infants (Aquilina et al. 2012; Douglas-Escobar and Weiss 2012). The TGF-β1 isoform is most associated with PHH after IVH in neonates and adults (Gomes et al. 2005). Intrathecal TGF-β1 injection in mice resulted in hydrocephalus development, and TGF-β1 expression was significantly increased in brains of neonatal rats with PHVD after intraventricular blood injection (Cherian et al. 2004a; Tada et al. 1994). Indeed, TGF-β1 was elevated in both animal models and premature infants with PHH, although some studies dispute this (Heep et al. 2004). In a rat GMH model, TGF-β1 was elevated within hours after GMH, but normalized by 24 hours post-ictus (Tang et al. 2015a). Additionally, inhibiting TGF-β1 ameliorated long-term PHH and neurocognitive deficits as well as reduced vitronectin and GFAP expression in rats (Manaenko et al. 2014). Although the mechanism of TGF-β signaling after GMH and its association with PHH development has been established, studies are lacking that discern the changes to CSF dynamics as a consequence of TGF-β signaling and fibrosis.

Gliosis results from damage to the central nervous system and is characterized by the nonspecific reactive proliferation of astrocytes, microglia, and oligodendrocytes (Sofroniew 2009). Hydrocephalus development is also associated with neuroinflammation and reactive gliosis (Del Bigio et al. 2003; Deren et al. 2009). Gliosis was observed in cerebral cortical biopsies from hydrocephalic children with shunts (Glees and Hasan 1990). Increased expression of Iba-1 and GFAP, markers for microglia and astrocytes respectively, were also observed in the brains of neonatal rats with hydrocephalus (Deren et al. 2010). Reactive gliosis in the subarachnoid space was associated with hydrocephalus development after subarachnoid hemorrhage in rats. In an IVH rat model, long-term GFAP expression is markedly increased, and injection of umbilical cord blood-derived mesenchymal stem cells reduced GFAP expression as well as long-term PHH development (Ahn et al. 2013). Aquaporin 4 knockout mice more rapidly developed hydrocephalus after kaolin injection, although increased aquaporin 4 expression is observed in hydrocephalus too (Bloch et al. 2006; Mao et al. 2006). Aquaporin 1 is expressed on the choroid plexus apical membrane and aquaporin 1 knockout mice had decreased CSF production (Oshio et al. 2005). Given the important role astrocytes play in the blood-brain barrier function as well as in glymphatic mediated CSF-interstitial fluid exchange, gliosis may have a profound effect on CSF dynamics and PHH development, which warrants further investigation.

Conclusions

Our understanding of hydrocephalus has changed significantly since Dr. Dandy’s first experiments in the early 20^th century and the bulk flow theory was proposal. CSF is produced by the choroid plexus epithelial lining and, to a minimal extent, the cerebral ventricular ependymal lining, and, following glymphatic mediated CSF-interstitial fluid exchange, CSF outflow occurs through perivascular channels, meningeal lymph vessels, spinal nerve roots, and the cribriform plate. Our purpose is to reconcile our knowledge of GMH and PHH development with the current hydrodynamic theory of hydrocephalus. PHH after GMH may be obstructive, non-communicating hydrocephalus during the acute phase due to the hematoma, but generally develops as chronic communicating hydrocephalus into adolescence and adulthood. Our proposed mechanisms explore the latter. Indeed, many GMH/IVH studies suggest PHH is a consequence of obstructions within the cerebroventricular system and subarachnoid drainage pathways due to thrombi, gliosis, and fibrosis. In line with current hydrocephalus school of thought, we suggest thrombi, gliosis, and fibrosis after GMH are not merely obstructing CSF passages but are altering barrier dynamics in the microvasculature and ependymal lining, altering CSF dynamics and CSF-interstitial fluid exchange to cause PHH development. Future research should elucidate these potential mechanisms.

Significance Statement.

Germinal matrix hemorrhage (GMH) is a leading cause of morbidity and mortality in preterm and very low birthweight infants. A common long-term consequence of GMH is hydrocephalus. This comprehensive review discusses cerebrospinal fluid dynamics, our current knowledge of hydrocephalus development and GMH pathophysiology, and proposes new mechanisms that warrant further investigation.