Cerebral edema and liver disease: classic perspectives and contemporary hypotheses on mechanism

Abstract

Liver disease is a growing public health concern. Hepatic encephalopathy, the syndrome of brain dysfunction secondary to liver disease, is a frequent complication of both acute and chronic liver disease and cerebral edema (CE) is a key feature. While altered ammonia metabolism is a key contributor to hepatic encephalopathy and CE in liver disease, there is a growing appreciation that additional mechanisms contribute to CE. In this review we will begin by presenting three classic perspectives that form a foundation for a discussion of CE in liver disease: 1) CE is unique to acute liver failure, 2) CE in liver disease is only cytotoxic, and 3) CE in liver disease is primarily an osmotically mediated consequence of ammonia and glutamine metabolism. We will present each classic perspective along with more recent observations that call in to question that classic perspective. After highlighting these areas of debate, we will explore the leading contemporary mechanisms hypothesized to contribute to CE during liver disease.

Introduction

Liver disease is a public health concern that represents a major contributor to mortality, disability, and healthcare resource utilization worldwide.¹ With the obesity epidemic contributing to an increased incidence of non-alcoholic fatty liver disease and the US opioid epidemic contributing to increased rates of hepatitis C infection, the public health burden of liver disease will likely continue to increase.²

Liver disease results either from acute liver failure (ALF) or chronic liver disease (CLD). ALF represents a rapid hepatocellular necrosis, in the absence of pre-existing liver disease, resulting in uncompensated liver dysfunction within 26 weeks of jaundice onset.³ Acetaminophen intoxication is the most frequent cause of ALF and often presents in a hyperacute fashion. CLD results from persistent, long-term insults to the liver that cause fibrosis and scarring, which eventually culminates in liver cirrhosis. Alcoholism, viral hepatitis, and non-alcoholic fatty liver disease represent leading contributors to cirrhosis. When patients with CLD experience an acute deterioration in liver function they may present with decompensated cirrhosis characterized by jaundice, ascites, gastrointestinal hemorrhage, renal dysfunction, or encephalopathy. In the most severe cases, decompensated cirrhosis manifests as acute-on-chronic liver failure (ACLF). ACLF is a relatively recently recognized syndrome of multisystem organ failure phenotypically similar to ALF but in patients with pre-existing cirrhosis.⁴ Hepatic encephalopathy (HE) represents a major complication of both ALF and CLD. HE is a neuropsychiatric syndrome caused by liver disease and manifests as a spectrum of symptoms ranging from subtle neurologic dysfunction to coma and death from brain herniation.⁵ HE has been classified according to underlying disease etiology (type A: ALF, type B: portosystemic shunt without intrinsic liver disease, and type C: cirrhosis), severity (minimal, covert, and overt), and time course (episodic, recurrent, or persistent).⁵ These classifications of HE have typically been treated as distinct entities, though the extent to which they share underlying pathophysiologic mechanisms is debated, and there are suggestions that HE should be viewed as a spectrum, especially for mechanistic investigations.⁶

Cerebral edema (CE), an increase in brain water content leading to volume expansion, is a key feature of HE but is not synonymous with HE, given the contributions of other pathophysiologic mechanisms.^{7, 8} The classic teaching is that CE is a clinically relevant feature in ALF but not CLD.⁹ However, CE has been observed during HE both in the context of ALF and CLD.^10–19 Despite multiple lines of evidence suggesting that CE is a feature of CLD, the incidence and relevance of CE in CLD is a topic of ongoing debate.^{4, 20, 21} Much of the current discussion about mechanisms of CE in liver disease is occurring in the context of similar debates between classic perspectives and observations from more recent data. As such, this review will begin by presenting three classic perspectives that form a foundation for a discussion of CE in liver disease: 1) CE is unique to ALF, 2) CE in liver disease is only cytotoxic, and 3) CE in liver disease is primarily an osmotically medicated consequence of ammonia and glutamine metabolism. We will present each classic perspective along with more recent observations that question that classic perspective. After highlighting these areas of debate, we will explore the leading contemporary mechanisms hypothesized to contribute to CE during liver disease. The majority of the literature on mechanisms of CE in liver disease comes from studies of cultured astrocytes exposed to ammonia, animal models of ALF, or clinical and translational studies of patients with ALF. Therefore, by necessity, the majority of the discussion will focus on ALF; however, mechanistically informative data from CLD models and patients is becoming more common and will be reviewed.

Classic Perspectives and Recent Challenges to Them

Is cerebral edema unique to ALF?

CE was first described as a frequent feature of massive hepatic necrosis by Ware and colleagues in 1971, and subsequently CE was established as a cardinal feature of ALF.^{22, 23} In historical series, CE was reported in approximately 80% of ALF patients and cerebral herniation was considered the most common cause of death.^{24, 25} Due to improved recognition of ALF and earlier medical interventions, recent ALF series report an incidence of CE of 20%, and in those with CE, mortality has improved from 95% in the 1970s to 55% in the 2000s.²⁶

CE was believed not to occur in CLD until Jalan described the first cases of CE and intracranial hypertension in cirrhotic patients in 1997.⁹ Currently there is debate regarding the incidence and significance of CE in CLD. In a study of cirrhotic patients with ACLF, CE defined by intracranial hypertension was reported to be rare, occurring in approximately 5% of patients.²⁷ This is in contrast to data from multiple magnetic resonance imaging studies demonstrating CE in patients with compensated CLD and suggesting that the severity of minimal, covert, and overt HE in CLD correlates with the degree of CE.^{10–15, 19, 28} Furthermore, using volumetric analysis of serially acquired computed tomography scans, our group has demonstrated that CE is common in both critically ill patients with ALF and ACLF; in both groups, changes in whole brain edema on the order of 10 to 20 mL correspond to changes in clinical encephalopathy severity measured by the Glasgow Coma Scale.^{17, 18}

The discrepancy between the classic view of CE as unique to ALF and more recent literature describing it as a feature of both ALF and CLD may be reconciled by considering how CE has been identified in studies. Early clinical studies in ALF identified CE either by an increase in intracranial pressure, qualitative radiographic signs of edema (e.g. brain compression, tissue herniation, and loss of gray-white matter distinction), or development of cerebral herniation syndromes. ^{27, 29–31} These approaches are insensitive to detecting CE before it progresses to severe levels because compensatory mechanisms buffer increases in intracranial pressure (Figure 1).

Figure 1.

Intracranial compliance curves demonstrating the relationship between intracranial volume and pressure changes and compensatory mechanisms in patients with normal baseline brain volume and patients with baseline atrophy due to advanced age or chronic illness. CSF is cerebrospinal fluid.

The ability to displace cerebrospinal fluid (CSF), and to a lesser extent blood, from the cranium results in an exponential relationship between intracranial volume and intracranial pressure such that CE may increase for some time before a meaningful increase in intracranial pressure occurs.³² In patients with greater degrees of cerebral atrophy, as might occur with age or chronic cirrhosis, there is a greater amount of CSF available for displacement and, therefore, a greater buffer against intracranial hypertension.³² Several authors have proposed differences in intracranial buffering capacity as an explanation for why CE formation in CLD has a less dramatic clinical presentation than in ALF.^{4, 33} However, it remains unknown whether CE that does not progress to intracranial hypertension (so called “low-grade” CE) is benign. Data from stroke and traumatic brain injury suggest that CE, independent of intracranial hypertension, may be a mechanism of secondary brain injury.^34–36 Furthermore, MRI studies in cirrhotic patients have demonstrated that the quantity of CE correlates with neuropsychometric test scores and that improvement in CE after liver transplantion is associated with improved cognition.^{15, 28, 37}

Is cerebral edema in liver disease only cytotoxic?

CE may result from either cytotoxic or vasogenic mechanisms. Cytotoxic edema results from derangements in cellular metabolism with resulting alterations in ionic gradients. Vasogenic edema results from a dysfunctional blood brain barrier (BBB); the physical and metabolic barrier between the brain and the systemic circulation that is formed by endothelial cells, the tight junctions between endothelial cells, astrocytes, and pericytes. BBB dysfunction results in extravasation of macromolecules from the plasma with resulting increase in osmotic pressure and movement of water in to the brain tissue. There is little doubt that cytotoxic edema is a key feature of HE during ALF. Astrocyte swelling, particularly astrocyte endfeet swelling, is the most prominent and consistent finding on autopsy studies of brains from animals and patients who died from ALF.^38–41 The consistent observation of astrocyte swelling in autopsy samples, combined with electron microscopy studies demonstrating normal appearing capillary endothelial cells and lack of BBB permeability to horseradish peroxidase, form the foundation of the classic viewpoint that CE in ALF is exclusively cytoxic in nature.^{7, 40} However, there are a number of contradicting studies that suggest vasogenic edema occurs in ALF. Multiple studies using the albumen bound dye Evans blue have demonstrated BBB disruption and interstitial albumen extravasation contributing to CE in experimental models of ALF.^42–45 Studies suggest that the necrotic liver releases the proteolytic enzyme matrix metalloproteinase-9 (MMP-9), which has been implicated in vasogenic edema in ischemic stroke and traumatic brain injury, and that MMP-9 contributes to Evans blue extravasation and edema in ALF models.^{42, 43} This MMP-9 related extravasation of albumen-bound Evans blue was observed despite light and electron microscopic examinations showing a structurally intact BBB.⁴² Chen et al built on this work by demonstrating MMP-9 mediated degradation of tight junction proteins occludin and claudin-5 and increased paracellular permeability of the BBB in a mouse model of ALF.^{46, 47} Similar findings of tight junction protein degradation with increased BBB permeability have been reported by others.^{48, 49} Further support for the potential of vasogenic edema to develop in ALF came from a study in which ALF mice were treated with lipopolysaccharide in order to mimic comorbid infection, which is a common complication of ALF. The lipopolysaccharide treated ALF mice demonstrated evidence of vasogenic edema with up-regulation of MMP-9, frank structural disruption of the BBB, and extravasation of immunoglobulin G in to the brain.⁵⁰ Cauli and colleagues reported a study on the regional and temporal progression of brain changes in ALF rats that may reconcile the discrepancy between the classic viewpoint on CE and the studies noted above.Cauli found that BBB permeability increases early in the evolution of ALF followed by development of vasogenic edema primarily in the basal ganglia, motor cortex, and cerebellum with only focal cytotoxic edema in the hypothalamus.⁴⁵ In the later stages of ALF, BBB permeability increases more diffusely in the brain and CE evolves to become increasingly cytotoxic and diffuse in nature.⁴⁵

Cytotoxic astrocyte changes are well known to develop in advanced cirrhosis in the form of Alzheimer type II astrocytosis.⁵¹ Electron microscopy studies of bile duct ligation cirrhotic rats suggested a cytotoxic mechanism of CE with a structurally intact appearing BBB, and studies using Evans Blue and sodium fluorescein in the same animal model also suggested an intact BBB despite evidence of brain edema.^{52, 53} These observations are in contrast to studies using magnetic resonance apparent diffusion coefficient and monoexponential diffusion tensor imaging that suggested the presence of vasogenic edema in cirrhotic patients that increases with the development of early grade HE.^{12, 15, 54, 55} Moreover, using a biexponential analysis of diffusion tensor imaging, Chavarria et al observed vasogenic edema in the parietal white matter and mixed cytotoxic and vasogenic edema in the corticospinal tract of cirrhotic patients without overt encephalopathy.⁵⁶ In patients with ACLF, Nath et al also described a mixed pattern of baseline predimonately vasogenic edema with superimposed cytotoxic edema in proportion to the severity of the acute systemic insult..⁵⁷ Both Chavarria and Nath suggested that CE in cirrhosis may represent a chronic vasogenic edema that develops a superimposed cytotoxic component in states of acute decompensation. Interestingly, this hypothesis is similar to the evolution of CE that Cauli observed in ALF, which raises the question if phenotypically similar liver disease (e.g. ALF and ACLF) might share similar mechanisms of CE.

Ammonia, Glutamine, and the Osmotic Hypothesis

The ability of ammonia alone to produce coma and CE has been recognized at least since the 1950s due to observations from patients and animals with portosystemic shunts or genetic urea cycle disorders.^{58, 59} Ammonia is produced primarily in the bowel by bacteria and the enzyme glutaminase converting glutamine in to ammonia and glutamate. Ammonia passes from the bowel through the portal venous system to the liver where it is detoxified to urea. In advanced liver diease, the loss of hepatocytes leads to insufficient detoxification through the urea cycle, and ammonia detoxification becomes progressively more dependent on the activity of glutamine synthetase, primarily in the muscle and kidney, to combine ammonia and glutamate to form glutamine.³³ As the loss of liver function progresses and the capacity of muscle and kidney gultamine sythetase is overwhelmed, serum ammonia levels increase.

As a result of numerous studies, there is no doubt that ammonia plays a central role in the development of HE and CE in liver disease. In particular, the effects of hyperammonemia appear to target astrocytes, rather than neurons, as the principle site of cellular dysfunction.^{7, 8, 60, 61} This is likely because astrocytes are the only cell type in the brain that possesses glutamine synthetase and can metabolize ammonia that diffuses in to brain tissue.⁶² Cultured astrocytes treated with ammonia and animal models of hyperammonemia consistently demonstrate the development of astrocyte swelling.^{40, 41, 63, 64} Furthermore, treatment of ALF rats with L-ornithine-L-aspartate, which stimulates the urea cycle to metabolize ammonia, demonstrates both a reduction in serum ammonia and improvement in CE.⁶⁵ Additionally, several clinical studies in ALF suggest an association between hyperammonemia and clinical encephalopathy severity, intracranial hypertension, and risk of cerebral herniation.^{29, 66}

Early in the study of HE, it was observed that hyperammonemia and ALF lead to accumulation of glutamine in astrocytes and that the irreversible glutamine sythetase inhibitor methionine sulfoximine reduced HE severity, astrocyte swelling, and CE.^{63, 67, 68} These observations, combined with glutamine’s function as an osmolyte, lead to the development of the “glutamine osmolyte” hypothesis. Under this hypothesis, increased brain ammonia levels lead to accumulation of glutamine in astrocytes, which generates an osmotic load to draw water in to the brain. The “glutamine osmolyte” hypothesis became the classic explanation for CE in hyperammonemia and liver disease and continues to be invoked in mechanistic discussions.²¹ However, numerous studies have challenged the osmotic component of the “glutamine osmolyte” hypothesis. Firstly, cerebral glutamine levels do not correlate well with the degree of astrocyte swelling or CE, and therapies that improve CE in ALF do not demonstrate a commensurate reduction in glutamine levels.^69–73 Furthermore, in cultured astrocytes exposed to ammonia, glutamine concentration and degree of cell swelling do not temporally correlate. Glutamine levels are maximal and cellular swelling is absent at 1–4 hours after ammonia exposure, and glutamine levels are normalized and cellular swelling is maximal at 1–3 days after ammonia exposure.⁷² Furthermore, treating ammonia exposed astrocytes with the glutaminase inhibitor diazo-5-oxo-L-norleucine acutely increases intracellular levels of glutamine but reduces the severity of subsequent cellular swelling.⁷²

Contemporary Proposed Mechanisms

While the literature continues to support a role for glutamine in the neurologic manifestations of liver disease, studies challenging the glutamine osmotic mechanism of CE have required other mechanisms to be proposed. In fact, more contemporaneous evidence has fueled debate challenging the classic view that derangements in ammonia metabolism are sufficient to explain the CE and encephalopathy seen in liver disease.^{5, 7, 8, 74} There is growing appreciation that multiple mechanisms may contribute to CE and that the contribution of individual mechanisms may vary depending on the phenotype of liver diease. For the remainder of this review, we will explore the leading proposed mechanisms that have built on the classic view in response to an expanding literature on CE in liver disease.

The Trojan Horse Hypothesis: Ammonia, Glutamine, and Mitochondria

An alternative hypothesis to the “glutamine osmolyte” hypothesis has been proposed to explain the association between ammonia, glutamine, and CE. In the “Trojan Horse” hypothesis, hyperammonia leads to the accumulation of astrocytic glutamine, similar to the “glutamine osmolyte” hypothesis. However, rather than producing an osmotic effect to explain CE, glutamine is transported in to the astrocyte mitochondria and undergoes hydrolysis by phosphate-activated glutaminase to produce ammonia (i.e. glutamine acts as the Trojan Horse to get ammonia in to the mitochondria).⁷⁵ Elevated mitochondrial ammonia triggers oxidative stress and progressive mitochondrial energetic dysfunction that culminates in opening the mitochondrial permeability transition (MPT) pore.^{76, 77} The MPT is a calcium-dependent process whereby formation of a non-selective pore in the inner mitochondrial membrane leads to membrane depolarization and energy failure through collapse of oxidative phosphorylation.⁷⁷ Dysfunction of mitochondrial oxidative phosphoralation likely leads to CE by failure of ATP dependent ion transporters involved in cellular volume regulation.

Evidence in favor of the “Trojan Horse” hypothesis includes studies demonstrating that glutamine induces free radical production in cultured astrocytes and the MPT in both free mitochondria and cultured astrocytes; however, direct ammonia exposure leads to the MPT only in cultured astrocytes and not free mitochondria.^78–82 Furthermore, blocking the steps in the “Trojan Horse” hypothesis has been shown to prevent the MPT and mitigate CE. Astrocyte swelling, CE, and the MPT may be mitigated by: 1) treatment with methionine sulfoximine to reduce astrocyte glutamine (incidentally, one of the observations that fostered the “glutamine osmolyte” hypothesis), 2) treatment with L-histidine to prevent glutamine uptake in to mitochondria, 3) treatment with diazo-5-oxo-L-norleucine to prevent conversion of glutamine to ammonia and glutamate by mitochondrial glutaminases, and 4) treatment with cyclosporin A to block the MPT pore.^{63, 67, 68, 72, 79, 83–85} The “Trojan Horse” hypothesis was initially criticized because of the historical stance that brain glutaminase was only present in neurons; however, after introduction of the “Trojan Horse” hypothesis, multiple studies confirmed ample amounts of phosphate-activated glutaminase in astrocyte mitochondria.^{86, 87} While the “Trojan Horse” hypothesis was derived with a focus on ALF, the observation that cerebral glutamine levels correleate with the severity of neuropsychiatric impairment in cirrhotic patients with induced hyperammonemia suggests that this mechanism might also contribute to CE in CLD.⁸⁸

Oxidative and Nitrosative Stress

Closely linked to the “Trojan Horse” hypothesis and the MPT is the hypothesis that oxidative and nitrosative stress during liver disease may contribute to CE. The ability of reactive oxygen species to produce cellular damage by affecting DNA, RNA, and protein expression and function is well accepted in numerous diseases. Studies in the early 1990s identified that ammonia treated astrocytes and hyperammonemic rats demonstrated evidence of lipid peroxidation, suggesting increased oxidative stress.^{89, 90} Subsequent studies in cultured astrocytes and rats suggested that hyperammonemia results in reduced activity of the antioxidant enzymes glutathione perioxidase, superoxide dismutase, and catalase along with the formation of reactive oxygen and nitrogen species (RONS) through a process associated with excessive N-methyl-D-aspartate (NMDA) receptor activation and intracellular glutamine accumulation.^91–95 Highlighting the role of nitrosative stress, experimental models of HE demonstrate increased neuronal nitric oxide synthase gene expression and activity in the brain and improved mortality when this activity is inhibited.^96–98 Furthermore, post-mortem brain tissue demonstrates evidence of oxidative and nitrosative stress, with RNA oxidation and protein tyrosine nitration, in cirrhotic patients with HE but not in cirrhotic patients without HE.⁹⁹ This oxidative stress appears to be concentration dependent as oxidative stress is apparent in rat models at ammonia concentrations of 500 μmole but not at concentrations of 100–200 μmole.^{100, 101}.

The connection between oxidative/nitrosative stress and CE formation is supported by cultured astrocyte and animal models in which treatment with antioxidants, including N-acetylcysteine and allopurinol, is associated with concurrent improvement in edema.^{53, 102–104} The exact mechanism by which ammonia-associated RONS results in CE is unknown. However, RONS might lead to CE through processes such as: activation of ionic transporters (namely oxidation and nitration of Na/K/Cl cotransporter-1), activation of intracellular signaling cascades by phosphorylation of mitogen-activated protein kinases (including ERK 1/2, p38, and JNK 1/2/3), BBB disruption by protein tyrosine kinase dependent activation of matrix metalloproteinases, and/or opening of the MPT pore.^104–108 It is particularly noteworthy that, while the MPT may lead to cell swelling through failure of oxidative phosphorylation, the MPT also results in reactive oxygen species generation that may perpetuate a cycle of oxidative injury. Regardless of the specific mechanism, it is likely that oxidative/nitrosative stress has a synergistic effect with hyperammonemia in the neurologic manifestations of liver disease. For example, cirrhotic patients with minimal HE demonstrate increased plasma markers of oxidative stress (3-nitrotyrosine) compared to cirrhotic patients without HE, despite similar ammonia levels.¹⁰⁹

The Synergism of Inflammation (and Infection)

Systemic infections are a frequent complication of acute liver disease and a common reason for presentation with decompensated cirrhosis or ACLF. In the last several years, there has been increasing recognition that systemic inflammation, as may occur from infectious complications of liver disease, likely plays a role in the development of HE and CE. From clinical studies, the proinflammatory cytokines IL-6 and IL-18 are know to be increased in cirrhotic patients with covert HE compared to cirrhotic patients without HE, and serum levels of tumor necrosis factor-alpha (TNF-α) correlate with the severity of overt HE in both patients with ALF and CLD.^{110, 111} Diffusion tensor imaging studies in patients with ALF demonstrate a correlation between IL-6 and TNF-α and severity of both cytotoxic and vasogenic edema.¹¹² Additionally, Shawcross et al observed that inducing hyperammonemia in cirrhotic patients during low grade infections resulted in worse neuropsychological test performance compared to either the infected pre-hyperammonemic state or the hyperammonemic post-infected state, suggesting a synergism between inflammation and hyperammonemia.¹¹³ In fact, some evidence from patients with ACLF suggests that the severity of inflammation is more strongly associated with the degree of encephalopathy than is the level of hyperammonemia.⁷⁴ Studies in experimental models of HE support these clinical findings. Compared to control rats treated with lipopolysaccharide (as a model of infection), cirrhotic rats treated with lipopolysaccharide demonstrate greater cytotoxic edema, more severe encephalopathy, and higher brain levels of TNF-α and IL-6.⁵² Furthermore, transgenic mice deficient in TNF-α and IL-1 type receptors demonstrate increased resistance to ALF associated CE.¹¹⁴ Mechanistically, TNF-α and transforming growth factor beta (TGF-β), produced in response to systemic inflammatory insults such as hepatocyte necrosis or infection, contribute to vasogenic edema through increased BBB permeability that results from disruption of tight junction proteins and upregulation of MMP-9.^{48, 50, 60, 115–117}

In addition to systemic inflammation, there appears to be a potential contribution of local inflammatory responses in the brain to the development of edema. Ammonia exposure activates brain microglia, which appears to contribute to oxidative stress and cognitive decline in models of HE.^{118, 119} Furthermore, astrocyte swelling can be induced by exposure to cell media from either microglia treated with ammonia or endothelial cells treated with ammonia, lipopolysacchride, or various inflammatory cytokines, which suggests a role for microglial and endothelial activation in CE formation.^{120, 121} In addition, treatment of astrocytes with the inflammatory cytokines TNF-α, IL-1β, IL-6, and IFN-γ leads to astrocyte swelling that is potentiated by pre-treatment with ammonia.¹²² Moreover, systemic and local brain inflammation appears to be linked because systemic inflammatory cytokines can activate brain leukocytes, astrocytes, and microglia to produce inflammatory cytokines locally within the brain.^{60, 116, 117} Both systemic and brain inflammatory mediators can predispose to vasogenic edema through similar effects on BBB permeability.^{116, 123}

The mechanisms by which systemic and local inflammation lead to CE are unresolved. Cultured astrocytes exposed to IL-1β, but not ammonia, demonstrate altered gene expression of the water channel aquaporin-4 (AQP-4), and astrocytes exposed to IL-1β or ammonia demonstrate decreased gene expression of glial fibrillary acidic protein (GFAP).¹²⁴ Both AQP-4 and GFAP are likely to be involved in astrocyte volume regulation. Ammonia treated astrocytes and models of ALF demonstrate activation of the transcription factor nuclear factor-kappaB (NF-κB) in a manner augmented by concurrent exposure to inflammatory cytokines (TNF-α, IL-1β, IL-6); among its numerous functions, activation of NF-κB enhances inflammation and leads to nitrosative and oxidative stress, which may contribute to CE as discussed above.^{122, 125, 126} In support of this mechanism, transgenic mice with inactivation of NF-kappaB are resistant to CE formation¹²⁵. Systemic inflammation has also been found to contribute to mitochondrial dysfunction and organ failure in multiple other disease processes, and as discussed above, mitochondrial dysfunction may contribute to CE formation.^{127, 128}

Ion Transporters and Water Channels

Aquaporin-4 (Aqp-4) is a water channel highly expressed in the endfoot membranes of astrocytes and has been implicated in the formation of CE in diseases including traumatic brain injury, brain tumors, and ischemic stroke.^129–131 Aqp-4 is co-localized with ion channels, particularly potassium channels, in the astrocyte membrane for the purpose of maintaining ionic equilibrium. The localization of Aqp-4 with potassium channels may be especially relevant for CE formation in liver disease because evidence suggests that ammonium ions may pass through potassium channels and transporters, including the inwardly rectifying potassium channel (Kir) and the Na-K-2Cl cotransporter (NKCC), due to the similar ionic properties of ammonium and potassium.^{132, 133} Increase in AQP-4 protein expression has been observed to proceed CE development in ammonia treated astrocytes, and AQP-4 plasma membrane expression is increased in rats with ALF and postmortum cerebral cortex from ALF patients.^134–137. This increased expression of AQP-4 appears to be due to enhanced stabilization of the protein in the plasma membrane, rather than an increase in total cellular AQP-4, and appears to be in response to glutamine uptake in to mitochondria (i.e. the “Trojan Horse” hypothesis).^136–138 Furthermore, AQP-4 deletion in mice attentuates CE and encephalopathy severity in an ALF model.¹³⁹

Multiple ion transporters, including the NKCC cotransporter and the sulfonylurea recepter-1 – transient receptor potential melastatin 4 (SUR1-TRPM4) channel (previously named NCCa-ATP), are involved in the regulation of cell volume and ion gradients. Intracellular movement of ions by these transporters is accompanied by obligatory water entry in to the cell such that excessive activity would be expected to contribute to cell swelling. The NKCC cotransporter has been shown to be activated by ammonia, apparently through ammonia-associated oxidation and nitration.^{105, 140} The SUR1-TRPM4 channel regulates transport of most inorganic cations in the brain, and the pore-forming region of SUR1-TRPM4 is regulated by SUR1.¹⁴¹ Ammonia increases SUR1 protein expression in cultured astrocytes and SUR1 mRNA in ALF rat models.¹⁴¹ Blocking SUR1 by the sulfonylurea agent glibenclamide reduced ammonia associated cell swelling in cultured astrocytes, and overexpression of SUR1 in ammonia treated astrocytes was diminished by the NF-κB inhibitor BAY 11–7082.¹⁴¹

While the role of water channels and ion transporters in CE requires further clarification, the mechanisms contributing to cytotoxic edema, such as oxidative and nitrosative stress, altered kinase signaling, NF-κB activation, and MPT pore expression, must converge to alter ion transporter and aquaporin function in order to yeild increased cellular water content.

Cerebral Lactate Accumulation and Energy Impairment

Astrocytes are known to be more glycolytic than neurons, and lactate generated through glycolysis in astrocytes can function as an energetic substrate for neurons, a process known as the astrocyte-neuron lactate shuttle.¹⁴² Lactate generated from pyruvate by lactate dehydrogenase-5 in astrocytes is expelled to the extracellular space by monocarboxylate transporters 1 and 4. This lactate can then be taken up by neurons and converted back to pyruvate for use in the tricarboxylic acid cycle.¹⁴² During hyperammonemia and liver failure, there is an increase in cerebral lactate with retained concentrations of high energy phosphates (e.g. ATP), suggesting dysfunctional energy metabolism without frank energetic failure, until the end stages of HE.^143–146 This dysfunctional energy metabolism likely results from the inhibitory effect of ammonia on alpha-ketoglutarate dehydrogenase, the rate limiting enzyme of the tricarboxylic acid cycle.^143–145 This impaired energy metabolism may be further exacerbated by thiamine deficiency since thiamine is a necessary co-factor in the function of both alpha-ketoglutarate dehydrogenase and pyruvate dehydrogenase, which is necessary for utilization of pyruvate (and indirectly lactate) in the tricarboxylic acid cycle.¹⁴⁷ Incidentally, thiamine is critical for the generation of NADPH through the pentose phosphate pathway such that thiamine deficiency could exacerbate the oxidative stress of liver disease through decreased NADPH production. Thiamine deficiency is known to be common in CLD and has recently been found to be prevalent in both ALF and ACLF.^{147, 148}

Increased brain lactate has been associated with CE in multiple models of liver disease, and reduction in lactate levels and CE can be accomplished by therapies such as hypothermia and the molecular adsorbent recirculating system (MARS).^{71, 149, 150} While lactate accumulation may represent energetic dysfunction and oxidative stress that is responsible for CE formation, some have speculated that lactate accumulation might also lead to an osmotic stress that contributes to CE.¹⁴⁹

Mechanisms Early in Development

Multiple other mechanisms might contribute to CE in liver disease and are in need of further investigation. The glymphatic system is a relatively recently described process whereby subarachnoid CSF travels along paravascular spaces and then circulates through the brain parenchyma to facilate the clearance of interstitial waste products. Impaired glymphatic function has recently been described in a rat model of chronic liver disease with covert HE.¹⁵¹ It is possible that glymphatic dysfunction in liver disease might lead to accumulation of interstitial toxins that contribute to cytotoxic edema. Alternatively, impaired fluid flow through the glymphatic system might result in increased interstitial fluid in a pattern resembling vasogenic edema.

The brain depends on a tightly regulated blood flow for normal function. Insufficient flow results in tissue ischemia and excessive flow results in hypertensive encephalopathy and vasogenic CE; irreversible brain injury can result from either extreme. The process by which cerebral blood flow is continuously regulated is known as dynamic cerebral autoregulation and is a process in which the brain’s astrocytes play a crucial role.^{152, 153} Evidence suggests that cerebral autoregulation dysfunction occurs in liver disease, although its relation to HE severity and patient outcomes needs to be further defined.^154–156 It is possible that CE could result from dysfunctional autoregulation either due to hypoperfusion with subsequent cellular metabolic dysfunction or hyperperfusion with increased vascular hydrostatic pressure and vasogenic edema.

As discussed above, inflammation and oxidative stress can lead to BBB dysfunction that contributes to vasogenic edema and increased exposure of the brain to circulating toxic substances. It is possible that toxic substances and systemic derangements in addition to ammonia and inflammatory cytokines might contribute to disruption of the BBB. For example, animal studies have demonstrated that artificial elevation of serum osmolality increases the permeability of the BBB and penetration of bilirubin and chemotherapeutic agents into the brain.^157–160 Our group has described that elevated serum osmolality is present in the majority of patients admitted to intensive care with hepatic encpehalopathy and liver failure; moreover, the magnitude of hyperosmolality is associated with the severity of encephalopathy and biomarkers of intracranial water-solute composition.^{18, 161, 162} Additionally, the accumulation of serum bile acids during liver disease may contribute to increased BBB permeability through the phosphorylation of occludin and the subsequent disruption of tight junctions.¹⁶³ Furthermore, brain and CSF bile acids may contribute to brain inflammation and encephalopathy severity during liver disease.^164–166

Potential treatment approaches

Treatment approaches for CE in liver disease have historically focused on reducing serum ammonia and, when edema progresses to a severe point in ALF, managing intracranial hypertension and frank brain compression (management approaches to intracranial hypertension are reviewed elsewhere).^{32, 167} There is little data to inform whether treating CE with more than standard prophylactic measures prior to the development of intracranial hypertension or brain compression improves patient outcomes, though observational data from other disease processes suggest it may be reasonable to consider.^34–36 That said, prophylactic hypothermia in patients with ALF did not prevent intracranial hypertension or improve survival compared to maintaining normothermia.¹⁶⁸

Standard drugs for hyperammonemia, including lactulose and rifaximin, are typically given for decompensated cirrhosis and may be appealing in ALF or ACLF, though they are untested in liver failure and gaseous distension could complicate liver transplantation.¹⁶⁷ Other agents targeting hyperammonemia, such as L-ornithine-L-aspartate or glycerol phenylbutyrate, have suggested benefit in small randomized trials of overt HE when combined with lactulose.¹⁶⁹ In ALF and possibly ACLF, continuous renal replacement therapy is often considered in patients with serum ammonia more than 200 μg/dL, steadily increasing ammonia, or concern for clinically significant CE.¹⁶⁷ Continous renal replacement therapy does in fact decrease serum ammonia levels.^{18, 170} Furthermore, our group demonstrated that, compared to the standard renal replacement therapy approach, renal replacement therapy with concurrent hypertonic saline infusion (titrated to avoid acute decline in serum osmolality) was associated with attenuation of CE and improved clinical encephalopathy severity in patients with liver failure and high grade HE.¹⁸

There is considerably less experience in using agents to target specific mechanisms in liver disease besides hyperammonemia. Anti-inflammatory or antioxidant agents have suggested some potential in experimental models, as noted above. Incidentally, N-acetylcysteine has become the standard of care for ALF, and its antioxidant and anti-inflammatory effects could conceivably benefit select mechanisms of CE.¹⁶⁷ Recent studies in sepsis have suggested that thiamine may aide metabolic resuscitation; as discussed above, thiamine deficiency could exacerbate mechanisms that contribute to CE in liver disease.¹⁷¹ Given the multiple mechanisms that may contribute to CE during liver disease, one might expect that therapeutic approaches that target several pathways simultaneously could represent added promise, especially if initiated before the disease process reaches a critical stage. There is data suggesting that plasmapharesis or continuous renal replacement therapy may represent such therapies, but drug cocktails might serve a similar purpose.^{18, 172, 173} Unfortunately, albumen dialysis, as a therapy that might target multiple pathways, did not demonstrate improvement in ammonia levels or HE in a randomized controlled trial, though an effect on CE was not measured.¹⁷⁴

Summary and Conclusions

The classic perspective that CE results from ammonia metabolism leading to astrocyte glutamine accumulation and osmotic swelling is overly simplistic and likely does not explain the temporal evolution of CE. In addition, CE in liver disease is not just cytotoxic in nature and is not unique to the ALF phenotype. CE can progress to severe levels before intracranial pressure becomes elevated, and in cases with increased baseline compensatory mechanisms (e.g. atrophy in ACLF patients), intracranial pressure may not become elevated until CE is extreme. More senstive methods to appreciate CE are needed for both clinical and research purposes. Existing literature suggests that CE in ALF, ACLF,and CLD share underlying mechanisms along a spectrum of severity and that CE results from a complex web of mechanisms. Lastly, CE is not the only factor that contributes to HE. Substances such as endogenous benzodiazepines, neurosteroids, or false neurotransmitters might contribute to HE without necessarily inducing edema, and some mechanisms that produce CE might also have effects on encephaloapthy that are independent of edema.^{7, 8, 67, 117}

Highlights.

• Cerebral edema may occur during both acute and chronic liver disease.

• Intracranial pressure is not a sensitive measure of cerebral edema.

• Cerebral edema in liver disease is a mix of vasogenic and cytotoxic mechanisms.

• Ammonia causes brain edema by mechanisms more complex than cellular osmotic burden.

• Impaired energy metabolism, oxidative stress, and inflammation contribute to cerebral edema.