Acta Neuropathologica

, Volume 118, Issue 2, pp 197–217

Pathology and new players in the pathogenesis of brain edema

  • Sukriti Nag
  • Janet L. Manias
  • Duncan J. Stewart

DOI: 10.1007/s00401-009-0541-0

Cite this article as:
Nag, S., Manias, J.L. & Stewart, D.J. Acta Neuropathol (2009) 118: 197. doi:10.1007/s00401-009-0541-0


Brain edema continues to be a major cause of mortality after diverse types of brain pathologies such as major cerebral infarcts, hemorrhages, trauma, infections and tumors. The classification of edema into vasogenic, cytotoxic, hydrocephalic and osmotic has stood the test of time although it is recognized that in most clinical situations there is a combination of different types of edema during the course of the disease. Basic information about the types of edema is provided for better understanding of the expression pattern of some of the newer molecules implicated in the pathogenesis of brain edema. These molecules include the aquaporins, matrix metalloproteinases and growth factors such as vascular endothelial growth factors A and B and the angiopoietins. The potential of these agents in the treatment of edema is discussed. Since many molecules are involved in the pathogenesis of brain edema, effective treatment cannot be achieved by a single agent but will require the administration of a “magic bullet” containing a variety of agents released at different times during the course of edema in order to be successful.


Angiopoietins Aquaporins Blood–brain barrier Brain edema Caveolin-1 Claudin-5 Cold injury Cytotoxic edema JAM-A Matrix metalloproteinases Occludin VEGF-A VEGF-B Vasogenic edema 


Brain edema is defined as an increase in brain volume resulting from a localized or diffuse abnormal accumulation of fluid within the brain parenchyma. This definition excludes volumetric enlargement due to cerebral engorgement which results from an increase in blood volume on the basis of either vasodilatation due to hypercapnia or impairment of venous flow secondary to obstruction of the cerebral veins and venous sinuses. Initially, the changes in brain volume are compensated by a decrease in cerebrospinal fluid (CSF) and blood volume. In large hemispheric lesions, progressive swelling exceeds these compensatory mechanisms and an increase in the intracranial pressure (ICP) results in herniations of cerebral tissue leading to death. Hence the significance of brain edema, which continues to be a major cause of mortality after diverse types of brain pathologies such as major cerebral infarcts, hemorrhages, trauma, infections and tumors. The lack of effective treatment for brain edema remains a stimulus for continued interest and research into the pathogenesis of this condition.

The realization that brain edema is associated with either extra- or intra-cellular accumulation of abnormal fluid led to its classification into vasogenic and cytotoxic edema [81, 83]. Vasogenic edema is associated with dysfunction of the blood–brain barrier (BBB) which allows increased passage of plasma proteins and water into the extracellular compartment, while cytotoxic edema results from abnormal water uptake by injured brain cells. Other types of edema described later include hydrocephalic or interstitial edema [52, 83] and osmotic or hypostatic edema. Despite this classification of the distinct forms of edema, in most clinical situations there is a combination of different types of edema depending on the time course of the disease. For example, early cerebral ischemia is associated with cellular swelling and cytotoxic edema; however, once the capillary endothelium is damaged there is BBB breakdown and vasogenic edema results. In meningitis, cytotoxic, vasogenic and interstitial edema may coexist, while in traumatic brain injury both vasogenic and cytotoxic edema coexist [106].

This review will provide basic information on the types of brain edema for those new to this field and to aid in the understanding of the expression pattern of some of the newer molecules implicated in the pathogenesis of edema.

Vasogenic edema

Brain diseases such as hemorrhage, infections, seizures, trauma, tumors, radiation injury and hypertensive encephalopathy are associated with BBB breakdown to plasma proteins leading to vasogenic edema. Vasogenic edema also occurs in the later stages of brain infarction as discussed below.

Vasogenic edema may be localized or diffuse depending on the underlying pathology (Fig. 1a). The overlying gyri become flattened and the sulci are narrowed (Fig. 1b). When diffuse edema is present the ventricles are slit-like. Breakdown of the BBB to plasma proteins can be demonstrated by immunohistochemistry using antibodies to whole serum proteins, albumin, fibrinogen or fibronectin in human autopsy brain tissue (Fig. 1c) or brains of experimental animals. The white matter is more edema-prone since it has unattached parallel bands of fibers with an intervening loose extracellular space (ECS). The grey matter has a higher cell density with many inter-cellular connections which reduce the number of direct linear pathways making the grey matter ECS much less subject to swelling. Light microscopy in acute edema shows vacuolation and pallor of the white matter (Fig. 1d, e). In long standing cases of edema there is fragmentation of the myelin sheaths which are phagocytosed by macrophages resulting in myelin pallor. An astrocytic response is present in the areas of edema (Fig. 1e, f). An increase in glial fibrillary acidic protein (GFAP) mRNA occurs as early as 6 h after a cortical cold injury [22]. mRNA levels are maximal on days 4–5 and they remain elevated up to day 14 post-injury. Spatial mRNA expression in this study follows the pattern of post-injury edema being present in the subpial regions, the cortex adjacent to the lesion, and the ipsilateral and contralateral callosal radiations.
Fig. 1

Computed tomographic (CT) (a, g, h), gross (b, i), microscopic (cf) and magnetic resonance (MR) (j) images demonstrating edema in human brains. a An axial CT scan post-gadolinium from a case diagnosed with glioblastoma multiforme showing a mass in the right hemisphere with midline shift. A serpiginous area of enhancement is present in the center of the mass indicating breakdown of the BBB. Around this, there is decreased density due to tumor and vasogenic edema. b The brain of the same patient shown in a shows cerebral edema which is marked in the right hemisphere which shows flattening of gyri and indistinct sulci. c Microscopy of a glioblastoma shows massive exudation of immunoglobulin G (IgG) into the extracellular space. IgG was demonstrated by immunostaining with antibody against human IgG. d Light microscopic appearance of normal white matter stained with hematoxylin–eosin and Luxol fast blue. e White matter from an area of edema adjacent to a meningioma (not shown) shows myelin pallor and an increased number of astrocytes (arrowheads). f Astrocytes and their processes in edematous white matter are demonstrated by immunostaining for glial fibrillary acidic protein. Axial CT scans of a 52-year-old patient taken on days 1 (g) and 3 (h) post-presentation with a left-sided facial droop, left-sided hemiplegia and decreased sensation on the left side. g An area of decreased density and loss of grey/white differentiation is present in the right insular region which represents an infarct. h On day 3, a large area of decreased density involving almost the whole right hemisphere is present due to infarction associated with vasogenic edema. i Horizontal section of the brain corresponding to the CT scan (h) shows a large infarct in the territory of the right internal carotid artery with shift across the midline and right subfalcial herniation. Also present is a hemorrhagic infarct in the right posterior cerebral artery territory related to right uncal herniation. j An axial MR image of a 4 year old with hydrocephalus involving the lateral and third ventricles due to a posterior fossa tumor (not shown). The flair sequence highlights the transependymal edema. c–f ×300. c, g, h reproduced with permission [120]

The blood–brain barrier (BBB)

It is well known that cerebral vessels differ from non-neural vessels and have a structural, biochemical and physiological barrier, which limits the passage of various substances including plasma proteins from blood into brain [66, 120, 126, 236, 250]. Cellular components of the BBB include endothelium, pericytes and the perivascular astrocytic processes, which together with their associated neurons form the “neurovascular unit”. The best studied cell type is cerebral endothelium which has two distinctive structural features that limit their permeability to plasma proteins. These cells have fewer caveolae or plasmalemmal vesicles than non-neural vessels and circumferential tight junctions are present along the interendothelial spaces [125, 170].

Breakdown of the BBB is assessed by tracers. Gadolinium DPTA is the most commonly used tracer in human studies (Fig. 1a), while 125Iodine-labeled serum albumin, Evans blue, horseradish peroxidase (HRP) and dextrans, having molecular weights of 60,000–70,000 Da, are used in experimental animals [124]. The diameter of the HRP molecule is 600 nm which is very close to the diameter of albumin which is 750 nm, making HRP a good tracer for protein permeability studies [233]. Tracers having molecular weights less than 3,000 Da such as lanthanum, small molecular weight dextrans, sodium fluorescein or 14C sucrose are indicators of BBB dysfunction to ions. Although small amounts of water may also enter brain, the magnitude is not sufficient to produce edema. Therefore, studies using these tracers have no relevance to the BBB breakdown to plasma proteins which is a key feature of vasogenic brain edema. Permeability properties of cerebral endothelium are not uniform in all brain vessels. In rodents, aside from regions outside the BBB such as the median eminence and area postrema, a significant number of normal cerebral vessels in the entorhinal sulcus and entorhinal cortex are permeable to HRP [130]. Thus, the demonstration of increased permeability in these areas cannot be ascribed to pathology. Also, freeze fracture studies show that there is variation in the number of interconnected strands that make up tight junctions in the different types of brain vessels, with cortical vessels having junctions of the highest complexity, while junctions of the postcapillary venules are least complex [133]. The latter would explain why increased permeability of the postcapillary venules occurs in inflammation.

The cold injury model

This model was developed by Klatzo [85] to study the pathophysiology of vasogenic edema and has been used extensively in the literature and in our studies [122, 123, 127, 129]. A unilateral focal cortical freeze lesion is produced by placing the tip of a cold probe cooled with liquid nitrogen on the dura for 45 s. There are variations in the method of producing the cold lesion which makes it difficult to compare the results obtained from different laboratories. The ensuing edema was initially studied using exogenous tracers such as Evans blue and HRP. In our study, BBB breakdown to HRP was present at 12 h, which was the earliest time point studied and the BBB was restored on day 6 post-injury [122]. Similar results were obtained using immunohistochemistry to demonstrate endogenous serum protein extravasation using an antibody to serum proteins, fibrinogen or fibronectin. Two peaks of active BBB breakdown occur in the cold injury model [127, 129]. An initial phase which extends from 6 h to day 2, affects mainly arterioles and large venules at the margin of the lesion and leads to extravasation of plasma proteins at the lesion site (Fig. 2a). There is spread of edema fluid through the ECS into the underlying white matter of the ipsilateral and contralateral side (Fig. 2b). The second phase of BBB breakdown accompanies angiogenesis and is maximal on day 4 (Fig. 2c). Arterioles, veins and neovessels at the lesion site show extravasation of plasma proteins which remain confined to the lesion site. There is florid angiogenesis in this model which will not be discussed in this review.
Fig. 2

Confocal images obtained using single channels (a, c, d), merged confocal images (ei), and a coronal paraffin section of rat brain (b) depict the findings in the rat cortical cold injury model. a The cold injury site on day 0.5 shows several vessels with BBB breakdown to fibronectin (arrowheads). b On day 1, immunostaining with an antibody to serum proteins demonstrates extravasation of serum proteins from the lesion site into the underlying white matter of the ipsilateral hemisphere with spread via the corpus callosum to the white matter of the opposite hemisphere. c On day 4, there is spread of fibronectin from permeable vessels into the extracellular spaces of the surrounding neuropil. d On day 0.5, increased Cav-1 immunoreactivity is present in the endothelium of an arteriole which shows BBB breakdown to fibronectin which is demonstrated in the merged confocal image (e). f A vein with BBB breakdown to fibronectin shows endothelial phosphorylated Cav-1 (PY14Cav-1). Note the increased Cav-1 (d, e) and PY14Cav-1 (f) immunoreactivity in neutrophils. g On day 4, a vein with BBB breakdown shows focal endothelial immunostaining with VEGF-A antibody (yellow, arrowhead). The surrounding endothelial cells show cytoplasmic VEGF-A. h On day 1, there is loss of endothelial Ang1 in two vessels showing BBB breakdown to fibronectin (arrowheads). Neutrophils show cytoplasmic Ang1 immunoreactivity. i On day 1, a vein shows mural active caspase-3 indicating the presence of apoptosis. This vessel shows focal endothelial Ang2 immunoreactivity (arrowhead). Neutrophils show co-localization of Ang2 and active caspase-3. aScale bar 50 μm, c–iscale bar 20 μm

BBB breakdown in vasogenic edema

Ultrastructural studies demonstrate an increase in the number of endothelial caveolae only in the vessels with BBB breakdown to HRP within minutes after the onset of pathological states such as hypertension, spinal cord injury, seizures, experimental autoimmune encephalomyelitis, excitotoxic brain damage, brain trauma, and BBB breakdown-induced by bradykinin, histamine, and leukotriene C4 [100, 120, 126]. These findings suggest that enhanced caveolae are the major route by which early passage of plasma proteins occurs in brain diseases associated with vasogenic edema. Caveolae allow protein passage across endothelium via fluid-phase transcytosis and transendothelial channels. Enhanced caveolae represent the response of viable endothelial cells to injury since both caveolar changes and BBB breakdown are reversed 10 min after the onset of acute hypertension induced by a single bolus of a pressor agent [123]. No alterations in tight junctions were noted in the studies mentioned above. Convincing demonstration of tight junction breakdown has only been reported following the intracarotid administration of hyperosmotic agents using the tracer lanthanum, which is a marker of ionic permeability [21]. Thus, junctional breakdown to proteins occurs late in the course of brain injury probably during end-stage disease and precedes endothelial cell breakdown.

Research in the last decade has led to the isolation of novel proteins in both caveolae and tight junctions and studies are underway to define their role in brain injury.

Caveolin-1 (Cav-1)

The specific marker and major component of caveolae is Cav-1, an integral membrane protein (20–22 kDa), which belongs to a multigene family of caveolin-related proteins that show similarities in structure but differ in properties and distribution. Of the two major isoforms of Cav-1 only the α-isoform is predominant in the brain [73]. Cav-2 has a similar distribution as Cav-1 and non-neural endothelial cells express both Cav-1 and -2 [117, 183, 190]. Cav-1 has been localized in human [221] and murine cerebral endothelial cells which also show co-localization of Cav-1 and Cav-2 [132]. In addition to their role in protein passage across endothelium, caveolae appear to serve as signaling platforms by compartmentalizing a multitude of signaling molecules including the molecular machinery that promote vesicle formation, fission, docking and fusion with the target membrane [120]. The properties of Cav-1 are the subject of many reviews [31, 115, 156, 234].

Brain injury is associated with increased expression of Cav-1. Time course studies in the rat cortical cold injury model demonstrate a threefold increase in Cav-1α expression at the lesion site on day 0.5 post-injury [132]. At the cellular level, a marked increase in endothelial Cav-1 protein is present in vessels showing BBB breakdown to fibronectin (Figs. 2d, e, 3). Further studies demonstrate that the endothelial Cav-1 in vessels with BBB breakdown is phosphorylated [121] (Fig. 2f). It is well established that dilated vascular segments show enhanced permeability and leak protein [110, 213]. Within dilated vascular segments, endothelial cells are subjected to hemodynamic shear stress which is known to increase plasma membrane levels of Cav-1 due to redistribution of Cav-1 from the Golgi complex to the plasma membrane and this is associated with an increase in the surface density of caveolae [19, 175]. Shear and cellular stress can also result in phosphorylation of Cav-1 [175, 222]. Phosphorylation of Cav-1 is known to be an essential step for fission of caveolae and the internalization of bound and fluid-phase macromolecules within caveolae, a step which precedes transcytosis of caveolae [116]. Thus, phosphorylation of Cav-1 is essential for transcytosis of proteins across cerebral endothelium leading to BBB breakdown and brain edema following brain injury.
Fig. 3

Temporal expression of caveolins and tight junction proteins during BBB breakdown in the cold injury model are shown. Using both immunofluorescence and western blotting, increased expression of both caveolin-1 and phosphorylated caveolin-1 (PY14 Caveolin-1) was observed during the period of BBB breakdown. Decreased expression of junctional adhesion molecule-A was observed on day 0.5 only and of claudin-5 on day 2 only, while decreased expression of occludin was present on day 2 and persisted throughout the period of observation

In summary, caveolae and Cav-1 have a significant role in early BBB breakdown; hence, they could be potential therapeutic targets in the control of early brain edema.

Tight junction proteins

Tight junctions are localized at cholesterol-enriched regions along the plasma membrane associated with Cav-1 [138]. Tight junctions are formed of three integral transmembrane proteins: occludin, and the claudin, and junctional adhesion molecule (JAM) families of proteins. The extracellular loops of these proteins originate from neighboring cells to form the paracellular barrier of the tight junction, which selectively excludes most blood-borne substances from entering brain. Several accessory cytoplasmic proteins have also been isolated which are necessary for structural support at the tight junctions. They include zonula occludens (ZO)-1 to -3 [62, 65, 201], all-1 fused gene on chromosome 6 [240], 7H6 [249] and cingulin [29]. Occludin, the first tight junction protein to be identified, is an approximately 60-kDa tetraspan membrane protein with two extracellular loops, a short intracellular turn, and N- and C-terminal cytoplasmic domains [57]. High expression of occludin in brain endothelial cells as compared to nonneural endothelia provides an explanation for the different properties of both these endothelia [68]. Claudins are 18- to 27-kDa tetraspan proteins with a short cytoplasmic N-terminus, two extracellular loops and a C-terminal cytoplasmic domain, and they do not show any sequence similarity to occludin [56, 118, 210]. The claudin family consists of 24 members in mice and humans and exhibits distinct expression patterns in tissue [210, 212, 217]. Claudins may be the major transmembrane proteins of tight junctions as occludin knockout mice are still capable of forming interendothelial tight junctions having normal morphology and barrier function in intestinal epithelial cells [189] while claudin knockout mice are nonviable [61]. The JAMs belong to the immunoglobulin superfamily and to the cortical thymocyte marker for xenopus family of molecules that lie at the crossroads between antigen-specific receptors and adhesion molecules. JAM-A, the first member of the family to be isolated has been implicated in a variety of physiologic and pathologic processes involving cellular adhesion including tight junction assembly [109] and leukocyte transmigration [109, 144]. Details of the properties of the tight junction proteins are the subject of many reviews [28, 53, 58, 236, 250].

Occludin, claudins-3, -5 and -12, JAM-A and ZO-1 proteins have been localized in normal cerebral endothelium [86, 120, 136, 223, 242]. Occludin and claudin-5 strands run parallel to the long axis of brain vessels in murine brain [120]. Human studies show decreased occludin immunoreactivity in lesion vessels with BBB breakdown in acquired immunodeficiency virus infection [36], and in multiple sclerosis plaques [162]. Decreased expression of the tight junction proteins in vessels with BBB breakdown in the cold injury model follows a specific sequence with transient decreases in expression of JAM-A on day 0.5 only [242], of claudin-5 on day 2 only while occludin expression is attenuated from day 2 onwards and persists up to day 6 [132] (Fig. 3). These studies support our previous observations that caveolae and Cav-1 changes precede significant tight junction changes during early BBB breakdown. In our studies, lesion vessels without BBB breakdown to fibronectin also show a decrease in endothelial occludin immunoreactivity. Therefore, decreased occludin expression cannot be interpreted as evidence of BBB breakdown to proteins and does not have relevance to brain edema unless the permeability status of the vessel is also investigated. Alterations in expression of occludin and claudin-5 in cerebral ischemia is discussed later in this review.

Reduced Cav-1, occludin and ZO-1 expression has been reported in cultured brain microvascular cells exposed to the chemokine CCL2 [199]. In this model, targeted knockdown of Cav-1 by the adenoviral-mediated small interfering RNA approach is associated with reduced expression of these tight junction proteins and increased paracellular permeability of the endothelial cells to fluorescein dextran [198]. These results are diametrically opposed to the in vivo studies described above [121, 132]. These divergent results may be due to heterogeneity in endothelial cells present in the different type of cerebral vessels since the in vivo studies have reported BBB breakdown of predominantly arterioles and large veins [132], while the in vitro study observed increased permeability of microvascular endothelial cells [198]. However, there are sufficient differences between the in vivo and in vitro systems that preclude comparisons of the findings of both systems. Cultured endothelial cells show alterations in both caveolae and tight junctions both of which are important determinants of BBB permeability in vivo. Endothelial cells removed from their natural flow environment show loss of caveolae [175], while the tight junctions of cultured brain endothelial cells lack the structural complexity or permeability properties of endothelial cells in vivo that have an estimated electrical resistance that is significantly higher than present in vitro [155, 197]. Finally, hemodynamic factors such as blood pressure and vasodilatation, which have an important role in BBB breakdown to proteins in brain injury in vivo, are absent in vitro.

Resolution of edema

Much of our information about the resolution of vasogenic edema is derived from the earlier studies of the cortical cold injury model. During the period of BBB breakdown to plasma proteins there is progressive increase in 125I-labeled albumin, paralleled by an increase in water content [88]. Disappearance of serum proteins from the ECS coincides with the return of water content to normal values. Resolution of edema occurs immediately after closure of the BBB to proteins and continues on, although the BBB may remain open to small molecular substances such as sodium fluorescein for at least 24 h. Progression of edema is associated with hydrostatic pressure gradients that cause the edema fluid to move preferentially through the white matter into the ventricles via bulk flow [173, 174, 211]. This is further supported by the observation that reduction of CSF pressure accelerates the clearance of edema fluid into the ventricle [172]. Recent evidence suggests that aquaporin 4 channels located in the ependyma and astrocytic foot processes have an important role in the clearance of the interstitial water as discussed later.

Other mechanisms for clearance of edema fluid include digestion of serum proteins in the extracellular spaces by astrocytes [84] and passage of extravasated proteins via the abluminal plasma membrane of endothelial cells back into blood [224]. Edema fluid can also pass across the glia limitans externa into the CSF in the subarachnoid space and enter the arachnoid granulations for clearance into the superior sagittal venous sinus. A component of the edema fluid reaches the lymphatics and through this route reaches lymph nodes in the head and neck. Quantitative studies of the relative involvement of the various routes indicate that the clearance of edema by bulk flow into the CSF is restricted to the early phase of edema. Clearance by brain vasculature is small compared to that of CSF, while lymphatic clearance of edema appears to be negligible during the early phase [107].

For further reading on specific types of edema, such as brain tumor edema [148, 203], and edema associated with intracerebral hemorrhage [206, 238] the reader is referred to reviews on these subjects.

Cytotoxic edema

The most commonly encountered cytotoxic edema occurs in cerebral ischemia [12, 89, 97], which may be focal due to vascular occlusion, or global due to transient or permanent reduction in brain blood flow. Other causes include traumatic brain injury [106, 215], infections, and metabolic disorders including kidney and liver failure [219]. Intoxications such as exposure to methionine sulfoxime, cuprizone, isoniazid are associated with cytotoxic edema and swelling of astrocytes [112]. Triethyl tin and hexachlorophene intoxications cause accumulation of water in intramyelinic clefts and produce striking white matter edema, while axonal swelling is a hallmark of exposure to hydrogen cyanide. Since toxins are not involved in all cases of cytotoxic edema some authors prefer the term “cellular edema” rather than cytotoxic edema.

Experimental models used to study cytotoxic edema include the focal and global ischemia models and the water intoxication model [229]. In cytotoxic edema astrocytes, neurons and dendrites undergo swelling with a concomitant reduction of the brain ECS [15, 82]. This cellular swelling does not constitute edema which implies a volumetric increase of brain tissue. Astrocytes are more prone to pathological swelling than neurons because they are involved in clearance of K+ and glutamate, which cause osmotic overload that in turn promotes water inflow. Astrocytes outnumber neurons 20:1 in humans and astrocytes can swell up to five times their normal size, therefore glial swelling is the main finding in this type of edema [80].

Cytotoxic edema is best studied in focal ischemia models where an interruption of energy supply due to decrease in blood flow below a threshold of 10 ml/100 g leads to failure of the ATP-dependent Na+ pumps. This results in intracellular Na+ accumulation, with shift of water from the extracellular to the intracellular compartment to maintain osmotic equilibrium. This can occur within seconds. The Na+ is accompanied by influx of Cl, H and HCO3 ions. These changes are reversible. However, ischemia of less than 6 min results in irreversible brain damage forming the “ischemic core”. This infarcted tissue is surrounded by a region referred to as the “penumbra” where the blood flow is greater than 20 ml/100 g per min. Neurons and astrocytes in the penumbra undergo cytotoxic edema. If hypoxic conditions persist, death of these neurons and glia results in release of water into the ECS. Elimination routes for excess water may be the same as those in vasogenic edema.

Temporary vascular occlusion followed by reperfusion is associated with an interruption of energy supply leading to cessation of ionic pump function with intracellular accumulation of Na+ and water as described above. Damage to endothelium leads to vasogenic edema which can be demonstrated within 4 h after an ischemic/hypoxic insult in rodent brains [193] and is apparent by computed tomography in human brain by 24–48 h after the onset of ischemic stroke (Fig. 1g–i). The vasogenic component of ischemic brain edema is biphasic [90, 181]. The first opening of the BBB is hemodynamic in nature and occurs 3–4 h after the onset of ischemia. There is marked reactive hyperemia which develops in the previously ischemic area due to a rush of blood into vessels that are dilated by acidosis and devoid of autoregulation. This opening may be brief but it allows the entry of blood substances into the tissue which may influence post-ischemic pathophysiology. The second opening of the BBB follows the release of ischemic occlusion and affects cerebral vessels at the margins of infarcts and may be associated with a progressive increase in the infarct size [83]. Exudation of protein into the infarct area combined with an increase in osmolarity due to breakdown of cell membranes results in an increase in local tissue pressure. This leads to depression of regional blood flow below the critical thresholds for viability in penumbral regions and to further extension of the territory which undergoes irreversible tissue damage. Glutamate [8, 42] and the reactive oxygen radical [98], which have a pivotal role in neuronal damage in ischemia, are also known to promote BBB breakdown in ischemia [79, 214].

Clinically, increased brain impedance is used to monitor the decline in the extracellular space in cytotoxic edema, and hence, is an indirect means of monitoring cell swelling. The apparent diffusion coefficient is reduced in cytotoxic edema due to expansion of cell volume with encroachment upon and increased tortuosity of the ECS, shrinkage of neuronal cell bodies and alterations in the chemical constituents of the ECS [182].

Hydrocephalic or interstitial edema

This is best characterized in noncommunicating hydrocephalus where there is obstruction to flow of CSF within the ventricular system or communicating hydrocephalus where the obstruction is distal to the ventricles and results in decreased absorption of CSF into the subarachnoid space. The etiology of hydrocephalus is beyond the scope of this review. In hydrocephalus, a rise in the intraventricular pressure causes CSF to migrate through the ependyma into the periventricular white matter, thus, increasing the extracellular fluid volume [113, 114] (Fig. 1j). The edema fluid consists of Na+ and water and has the same composition as CSF. The white matter in the periventricular regions is spongy and on microscopy there is widespread separation of glial cells and axons. Astrocytic swelling is present followed by gradual atrophy and loss of astrocytes [114]. In chronic hydrocephalus, increase in the hydrostatic pressure within the white matter results in destruction of myelin and axons and this is associated with a microglial response. The end result is thinning of the corpus callosum and compression of the periventricular white matter. Other changes reported are destruction of the ependyma which may be focal or widespread, distortion of cerebral vessels in the periventricular region with collapse of capillaries and occasionally there is injury of neurons in the adjacent cortex [39].

In normal pressure hydrocephalus where normal intraventricular pressure is recorded, ependymal damage with backflow of CSF is postulated to produce edema. Functional manifestations in these cases are minor unless changes are advanced when dementia and gait disorder become prominent.

Osmotic edema

In this type of edema an osmotic gradient is present between plasma and the extracellular fluid and the BBB is intact, otherwise an osmotic gradient could not be maintained. Edema may occur with a number of hypo-osmolar conditions including improper administration of intravenous fluids leading to acute dilutional hyponatremia, inappropriate antidiuretic hormone secretion, excessive hemodialysis of uremic patients and diabetic ketoacidosis [44, 154]. There is a decrease of serum osmolality due to reduction of serum Na+ and when serum Na+ is less than 120 mmol/L, water enters the brain and distributes evenly within the ECSs of the grey and white matter. Astrocytic swelling may be present. The spread of edema occurs by bulk flow along the normal interstitial fluid pathways. Following a 10% or greater reduction of plasma osmolarity, there is a pronounced increase in interstitial fluid volume flow, and extracellular markers are cleared into the CSF at an increased rate [226]. The formation of osmotic edema can lead to a significant increase in the rate of CSF formation [40] without any contribution of the choroid plexuses. Since osmotic edema is vented rapidly, the increase in brain volume tends to be modest [229]. Experimentally, this type of edema is induced following intraperitoneal infusion of distilled water [229, 230]. The BBB is not affected and cytotoxic mechanisms are not involved.

Osmotic brain edema can also occur when the plasma osmolarity is normal but tissue osmolarity is high in the core of the lesion as occurs following brain hemorrhage, infarcts or contusions [78, 112].

Mediators of brain edema

It is well established that vaso-active agents can increase BBB permeability and promote vasogenic brain edema (Table 1). The properties of these agents are the subject of several reviews [2, 79, 191, 225] and will not be discussed. Of interest are the newer molecules such as the aquaporins, the matrix metalloproteinases, growth factors such as the vascular endothelial growth factors: A (VEGF-A) and B (VEGF-B) and the angiopoietins, along with their role in the pathogenesis of brain edema. All these agents have many properties other than their effects on BBB permeability and brain edema which will not be reviewed.
Table 1

Vasoactive agents that increase blood–brain barrier permeability

Arachidonic acid [24, 25, 47, 152, 232]

Bradykinin [169, 192]

Complement-derived polypeptide C3a-desArg [2]

Glutamate [79]

Histamine [76]

Interleukins: IL-1α, IL-1β, IL-2 [2, 38, 231]

Leukotrienes [16]

Macrophage inflammatory proteins MIP-1, MIP-2 [2]

Nitric oxide [146]

Oxygen-derived free radicals [214, 232]

Phospholipase A2, platelet activating factor, prostaglandins [2]

Purine nucleotides: ATP, ADP, AMP [139]

Thrombin [92]

Serotonin [195]

Aquaporins and brain edema

Aquaporins (AQP) are a growing family of molecular water-channel proteins that assemble in membranes as tetramers [4, 220]. Each monomer is ~30 kDa and has six membrane-spanning domains surrounding a water pore that allows bidirectional passage of water. At least 13 AQPs have been found in mammals and more than 300 in lower organisms [205]. Expression of AQP 1, AQP3, AQP4, AQP5, AQP8 and AQP9 has been reported in rodent brain [13, 49, 134, 239]. Only AQP1 and AQP4 are reported to have a role in brain edema and will be discussed.

Aquaporin1 (AQP1)

Localization of AQP1 in the apical membrane of the choroid plexus epithelium [64] suggests that it may have a role in CSF secretion. This is supported by the finding that AQP1 is upregulated in choroid plexus tumors [99], which are associated with increased CSF production. AQP1 is also expressed in tumor cells and peritumoral astrocytes in high grade gliomas [187]. Although AQP1 is present in endothelia of non-neural vessels, it is not observed in normal brain capillary endothelial cells [45, 135]. Brain capillary endothelial cells cultured in the absence of astrocytes [45] and those in brain tumors that are not surrounded by astrocytic end-feet [187] do express AQP1, suggesting that astrocytic end-feet may signal adjacent endothelial cells to switch off AQP1 expression.

AQP1-null mice show a 25% reduction in the rate of CSF secretion, reduced osmotic permeability of the choroid plexus epithelium and decreased ICP [143]. These findings support the role of AQP1 in facilitating CSF secretion into the cerebral ventricles by the choroid plexuses and suggest that AQP1 inhibitors may be useful in the treatment of hydrocephalus and benign intracranial hypertension, both of which are associated with increased CSF formation or accumulation.

Aquaporin4 (AQP4)

AQP4, the principal AQP in mammalian brain, is expressed in glia at the borders between major water compartments and the brain parenchyma [134, 168] (Fig. 4). AQP4 is expressed in the basolateral membrane of the ependymal cells lining the cerebral ventricles and subependymal astrocytes which are located at the ventricular CSF fluid–brain interface. Expression of AQP4 in astrocytic foot processes brings it in close proximity to intracerebral vessels, and thus, the blood–brain interface. Water molecules moving from the blood pass through the luminal and abluminal endothelial membranes by diffusion and across the astrocytic foot processes through the AQP4 channels. AQP4 is also expressed in the dense astrocytic processes that form the glia limitans which is at the subarachnoid–CSF fluid interface. Two AQP4 splice variants are expressed in brain, termed M1 and M23, which can form homo- and hetero-tetramers, respectively, which in turn assemble into square lattices of tetramers within astroglial cell plasma membranes [55, 167, 196].
Fig. 4

Pathways for water entry into and exit from brain are shown. The AQP4-dependent water movement across the blood–brain barrier, through ependymal and arachnoid barriers is shown. Reproduced with permission [220]

The location of AQP 4 at the brain–fluid interfaces suggests that it is important for brain water balance and may play a key role in brain edema. AQP4 overexpression in human astrocytomas correlates with the presence of brain edema on magnetic resonance imaging [188, 228]. However, decrease in AQP4 protein expression is associated with early stages of edema in rodents subjected to permanent focal brain ischemia [27] and hypoxia-ischemia [111]. In traumatic brain injury AQP4 mRNA is decreased in the area of edema adjacent to a cortical contusion [204].

AQP4-null mice provide strong evidence for AQP4 involvement in cerebral water balance in the various types of edema [18, 150, 151, 205, 220, 245].

Vasogenic edema

Data derived from AQP4-null mice suggest that AQP4 is involved in the clearance of extracellular fluid from the brain parenchyma in vasogenic edema. A number of models in which vasogenic edema is the predominant form of edema, including the cortical cold injury, tumor implantation and brain abscess models, demonstrate that the AQP4-null mice have a significantly greater increase in brain water content and ICP than the wild-type mice suggesting that brain water elimination is defective after AQP4 deletion [147, 150]. Melanoma cells implanted into the striatum of wild-type and AQP4-null mice produce peritumoral edema and comparable sized-tumors in both groups after a week. However, the AQP4-null mice have a higher ICP and water content [147]. These authors suggest that in vasogenic edema, excess water enters the brain ECS independently of AQP4, but exits the brain primarily through AQP4 channels into the CSF and via astrocytic foot processes into blood.

Cytotoxic edema

Swelling of astrocytic foot processes is a major finding in cytotoxic edema and since AQP4 channels are located in the astrocytic foot processes, it was hypothesized that they may have a role in formation of cell swelling. This was found to be the case since water intoxicated AQP4-null mice show a significant reduction in astrocytic foot process swelling, a decrease in brain water content and a profound improvement in survival [105]. Since water intoxication is of limited clinical significance, AQP4-null mice were subjected to ischemic stroke [105] and bacterial meningitis [149]. In both models AQP4-null mice showed decreased cerebral edema and improved outcome. AQP4 protein in plasma membranes is thought to be bound on an aggregate of intracellular proteins including α-syntrophin [5]. Reduced brain swelling after cerebral ischemia and water intoxication is also observed in α-syntrophin-null mice which have reduced AQP4 expression in astrocyte foot processes [6]. These studies imply that AQP4 has a significant role in water transport and development of cellular edema following cerebral ischemia.

Hydrocephalic edema

Obstructive hydrocephalus produced by injecting kaolin in the cistern magna of AQP4-null mice show accelerated ventricular enlargement compared with wild-type mice [17]. Reduced water permeability of the ependymal layer, subependymal astrocytes, astrocytic foot processes and glia limitans produced by AQP4 deletion reduces the elimination rate of CSF across these routes. Thus, AQP4 induction could be evaluated as a nonsurgical treatment for hydrocephalus.

In summary, AQP4 has opposing roles in the pathogenesis of vasogenic and hydrocephalic edema when compared to cytotoxic edema. Therefore, AQP4 activators or upregulators have the potential to facilitate the clearance of vasogenic and hydrocephalic edema, while AQP4 inhibitors have the potential to protect the brain in cytotoxic edema. This is an area of ongoing research since none of the AQP4 activators or inhibitors investigated thus far are suitable for development for clinical use [151].

Matrix metalloproteinases (MMPs) in vasogenic edema

The MMPs are zinc- and calcium-dependent endopeptidases which are known to cleave most components of the extracellular matrix including fibronectin, laminin, proteoglycans and type IV collagen [178, 200]. Activation of MMPs involves cleavage of the secreted proenzyme, while inhibition involves a group of four endogenous tissue inhibitors of metalloproteinases (TIMPs) [35] and α2-macroglobulin. The balance between production, activation, and inhibition prevents excessive proteolysis or inhibition.

Type IV collagenases are members of the larger MMP gene family of proteolytic enzymes that have the ability of destroying the basal lamina of vessels and thereby play a role in the development of many pathological processes including vasogenic edema in multiple sclerosis and bacterial meningitis [94] and ischemic stroke. MMPs are found in all of the elements of the neurovascular unit, but different MMPs have a predilection for certain cell types. Endothelial cells express mainly MMP-9, pericytes express MMP-3 and -9, while astrocytic end-feet express MMP-2 and its activator, membrane-type MMP (MT1-MMP) [23]. MMP-2 is constitutively expressed, has a molecular weight of 72 kDa and is normally present in a latent form tethered to the cell surface by MT1-MMP and requires the presence of TIMP-2 in order to undergo activation. This restricts the proteolytic action of MMP-2 to the immediate vicinity of the protease. MMP-9 has a molecular weight of 92 kDa. Normally it is expressed at low levels but is markedly up regulated in many brain diseases.

In human ischemic stroke, active MMP-2 is increased on days 2–5 compared with active MMP-9 which is elevated up to months after the ischemic episode [30]. Molecular studies in experimental permanent and temporary ischemia have shown that MMPs contribute to disruption of the BBB leading to vasogenic cerebral edema and hemorrhage [23, 119, 181]. Permanent middle cerebral artery occlusion in spontaneously hypertensive rats (SHR) resulted in the production of MMP-9 by 24 h and a marked increase in MMP-2 by day 5 [180]. Middle cerebral artery occlusion for 90 min with reperfusion in SHR causes biphasic opening of the BBB in the piriform cortex with a transient, reversible opening at 3 h which correlates with a transient increase in expression of MMP-2, and its activator MT1-MMP [26, 241]. This is associated with a decrease in claudin-5 and occludin expression in cerebral vessels after 3 h of reperfusion. By 24 h the tight junction proteins are no longer observed in lesion vessels, an alteration that is reversed by treatment with the MMP inhibitor, BB-1101 [241]. The later BBB opening between 24 and 48 h is associated with a marked increase of MMP-9 which is released in the extracellular matrix where it degrades multiple proteins, and produces more extensive blood vessel damage. The role of MMPs in BBB breakdown is further supported by the observation that treatment with MMP inhibitors or MMP neutralizing antibodies decreases infarct size and prevents BBB breakdown after focal ischemic stroke [9, 10, 179].

The MMP inhibitors used so far restore early integrity of the BBB in rodent ischemia models but are ineffective in the later opening at 48 h. Since these inhibitors block MMPs involved in angiogenesis and neurogenesis as well, they slow recovery. Therefore, the challenge is to identify agents that will protect the BBB and block vasogenic edema without interfering with recovery.

Growth factors and brain edema

Vascular endothelial growth factor-A (VEGF-A)

VEGF, the first member of the six member VEGF family to be discovered is now designated as VEGF-A. Initial reports described the potent hyperpermeability effect of VEGF-A on the microvasculature of tumors hence its designation ‘vascular permeability factor’ [33, 95, 194]. Alternative exon splicing was initially shown to generate four isoforms (VEGF121, VEGF165, VEGF189, and VEGF206), with VEGF165 being the major secreted isoform and the most abundant in human tissues [72, 209]. VEGF-A has a significant role in vascular permeability and angiogenesis during embryonic vasculogenesis and in physiological and pathological angiogenesis [3, 51, 145, 166, 185, 243]. There is agreement that vascular endothelial growth factor receptor-2 (VEGFR-2), which is present on endothelial cells, is the major mediator of the mitogenic, angiogenic and permeability-enhancing effects of VEGF-A [51].

The permeability inducing properties of VEGF-A have also been demonstrated in the brain. Intracortical injections of VEGF-A produces BBB breakdown at the injection site [43, 128, 163]. Normal adult cortex shows basal expression of VEGF-A mRNA and protein, while high expression of VEGF-A mRNA and protein is present in normal choroid plexus epithelial cells and ependymal cells [20, 127, 131]. Although several studies reported VEGF-A gene up regulation in cerebral ischemia models, increased expression was related to angiogenesis and not to BBB breakdown [67, 87, 108, 158].

The mechanism by which VEGF-A induces vascular hyperpermeability has been the subject of several studies. In non-neural vessels, VEGF-A is reported to cause vascular hyperpermeability by opening of interendothelial junctions and induction of fenestrae in endothelium [176, 177] and vesiculo-vacuolar organelles [50]. A single ultrastructural study reported interendothelial gaps and segmental fenestrae-like narrowings in brain vessels permeable to endogenous albumin following a single intracortical injection of VEGF-A [43]. In contrast, VEGF-A-induced hyperpermeability of the blood–retinal barrier endothelium is associated predominantly with enhanced numbers of endothelial caveolae, while tight junction alterations or fenestrae in endothelium were not observed [69]. These divergent results may be due to the different time points that permeability was assessed in these studies. Members of the Src family have been implicated in VEGF-dependent vascular hyperpermeability [48, 157] since Src−/− mice show reduced brain damage after induction of cortical ischemia, and a Src inhibitor has protective effects in wild-type mice in a similar brain injury model. VEGF-A can also increase permeability by inducing changes in expression of tight junction proteins. Reduced occludin expression occurs in retinal [7] and brain [227] endothelial cells exposed to VEGF-A. In addition, disruption of ZO-1 and occludin organization occurs leading to tight junction disassembly [227].


This member of the VEGF family displays strong homology to VEGF-A [141, 142]. VEGF-B has two known isoforms, which bind to VEGFR-1 and neuropilin-1 [104, 140]. Mice embryos (day 14) and adults show high expression of VEGF-B mRNA in most organs with very high levels in the heart and the nervous system [1, 91]. Moderate down regulation of VEGF-B occurs prior to birth (E17) and VEGF-B is the only member of the VEGF family that is expressed at detectable levels in the adult CNS. Constitutive expression of VEGF-B protein is present in the endothelium of all cerebral vessels including those of the choroid plexuses [127]. Thus, VEGF-B may have a role in maintenance of the BBB in steady states and VEGF-B may be protective against BBB breakdown and edema formation.

Angiopoietin (Ang) family

Four members of this family have been isolated thus far and designated Ang1–4 [37, 75, 216]. Ang1 and 2 are best characterized, approximately 75-kDa secreted proteins with considerable sequence homology; each containing a coiled-coil domain in the amino terminal region and a fibrinogen-like domain in the carboxyl-terminal region [37, 103]. Both Ang1 and Ang2 function as ligands for the Tie-2/Tek receptor with similar affinity; [46, 103, 186] and do not bind to Tie-1 [165]. Tie-2 (tyrosine kinase with immunoglobulin and epidermal growth factor homology domain) is ubiquitously expressed in endothelial cells of all type of vessels in all organs, including brain [137, 237]. Ang1 induces autophosphorylation of Tie-2 and has a remarkable chemotactic effect on endothelial cells, whereas Ang2 competitively inhibits this effect, suggesting that it may be a naturally occurring inhibitor of Ang1/Tie-2 activity [103, 235]. Endothelial Ang1 is expressed widely in normal adult tissues, consistent with it playing a constitutive stabilization role by maintaining normal endothelial cell to cell and cell to matrix interactions [103]. Our studies of the rodent brain show constitutive expression of Ang1 protein in endothelium of all cerebral cortical vessels and only weak expression of Ang2, while the Tie-2 receptor is present in the endothelium of all intracerebral vessels [128, 137].

Functional studies indicate that Ang1 and Ang2 have reciprocal effects in many systems. Ang1 has an anti-apoptotic effect on endothelial cells through the Akt/survivin/PI-3 kinase pathway [54, 153], while Ang2 is reported to promote endothelial apoptosis [32, 128, 246]. Ang1 has a potent anti-leakage property that was established using transgenic mice [208]. Ang1 blocks Ca2+ entry, Rho activation and elevated monolayer permeability induced by VEGF, platelet-activating factor, histamine, bradykinin and thrombin [59, 74, 96, 159]. Presence of Ang1 is also associated with fewer and smaller gaps in the endothelium of postcapillary venules during inflammation [14]. Ang1 is reported to stabilize interendothelial junctions by increased expression of platelet endothelial cell adhesion molecule-1 (PECAM-1) and decreased phosphorylation of PECAM-1 and vascular endothelial cadherin [207] in non-neural endothelial cells and by upregulating ZO-2 in brain endothelial cells [93]. These studies demonstrate that Ang1 is a potent anti-leakage factor.

Time course of growth factor expression post-injury

The cold injury model was used to study the temporal and spatial alterations in expression of growth factors and their relation to BBB breakdown (Fig. 5). In the early phase post-injury up to day 2, there is increased expression of VEGF-A protein, VEGFR-2 protein and a sevenfold increase in Ang2 mRNA [128, 137]. During this period, vessels with BBB breakdown show endothelial immunoreactivity for VEGF-A (Fig. 2g) and Ang2 but not for VEGF-B or Ang1 [127, 128, 137] (Fig. 2h). Dual labeling shows co-localization of Ang2 only with active caspase-3 in endothelium of lesion vessels at day 1 [128] (Fig. 2i) implicating Ang2 as an effector of endothelial apoptosis and BBB breakdown in damaged vessels. This is supported by the finding that intracortical injections of Ang2 can produce BBB breakdown [128] and by the finding that Ang2 produces vascular leakage in non-neural vessels [184]. On days 4 and 6 post-injury, there is progressive increase in Ang1 and VEGF-B mRNA and protein and decrease in Ang2 and VEGF-A mRNA coinciding with maturation of neovessels and restoration of the BBB (Fig. 5). No changes were observed in the Tie-2 receptor either at the mRNA or at the protein level.
Fig. 5

Temporal expression of growth factor proteins and their receptors is shown during the period of BBB breakdown in the cold injury model. Protein expression was determined by immunohistochemistry and/or immunofluorescence. Increased expression of VEGF-A, Ang2 and VEGFR-2 is present on days 2 and 4, while there is decreased expression of VEGF-B and Ang1 on days 0.5 and 2. Tie-2 receptor expression remains unchanged throughout the period of observation

The temporal expression of VEGF-A, Ang1 and Ang2 in focal cerebral ischemia models has been summarized previously [34, 63] and the expression pattern is very similar to that observed in the cold injury model. There is early up regulation of VEGF-A and Ang2 mRNA and protein coinciding with the first opening of the BBB post-ischemia. At 48 h post-ischemia, there is increased expression of VEGF-A, Ang1 and Ang2 coinciding with the period of the second BBB opening post-ischemia and the onset of angiogenesis.

Increased expression of growth factors has been reported in gliomas. VEGF-A is overexpressed up to 50-fold in the peri-necrotic tumor cells in glioblastomas suggesting hypoxia-mediated transcriptional activation of the VEGF-A gene [102, 160, 161]. This is associated with up regulation of VEGF receptors in tumor endothelium suggesting a paracrine mechanism of VEGF-driven tumor angiogenesis [160, 161]. Increased expression of the angiopoietins has also been reported in glioblastomas. High expression of Ang1 has been reported in areas of high vascular density in all stages of glioblastoma progression [11, 70, 71, 171], while high expression of Ang2 has been reported in endothelial cells in glioblastomas [41, 171, 202, 244, 247]. In these studies a strong association is made between these growth factors and tumor angiogenesis. Since time course studies are difficult in human biopsy tissue, one assumes that these angiogenic factors play the same role in the genesis of vasogenic edema in brain tumors as observed in the cold injury and the focal brain ischemia models. This is supported by the studies that provide a strong correlation between VEGF-A expression and vascular permeability in gliomas [101], and between VEGF-A expression and the amount of peritumoral edema in meningiomas [60, 77, 164].

There is the potential of using growth factors to treat early and massive edema associated with large hemispheric lesions which are lethal due to the effects of early edema. Potential candidates include inhibitors of VEGF-A or administration of Ang1 or VEGF-B. Inhibitors of VEGF-A or recombinant Ang1 have been tried in rodent models of ischemia. Pretreatment of rodents with VEGF-A receptor chimeric protein [Flt-(1-3)-IgG], which inactivates endogenous VEGF-A [218] or recombinant Ang1 [248] attenuates BBB breakdown and edema associated with cerebral infarcts. The long-term effects of administering these agents on angiogenesis and repair were not studied in these models. This must be assessed before these agents can be used for the treatment of brain edema.

Research in the last decade has led to an appreciation of the complexity of brain edema pathogenesis and to the awareness that many molecules are involved acting simultaneously or at different stages during the edema process. This suggests that effective treatment of brain edema cannot be achieved by a single agent, but will require the administration of a “magic bullet” containing a variety of agents released at different times during the course of edema in order to be successful.


This review is dedicated to Dr. Igor Klatzo, an experimental neuropathologist par excellence, an innovative thinker and friend who died on 5th May 2007. Work in the author’s laboratory is supported by the Heart and Stroke Foundation of Ontario, Grant 6003.

Copyright information

© Springer-Verlag 2009

Authors and Affiliations

  • Sukriti Nag
    • 1
  • Janet L. Manias
    • 1
  • Duncan J. Stewart
    • 2
  1. 1.Department of Laboratory Medicine and Pathobiology, Banting InstituteUniversity of TorontoTorontoCanada
  2. 2.Ottawa Health Research InstituteUniversity of OttawaOttawaCanada

Personalised recommendations