Ammonia and inflammation in the pathogenesis of hepatic encephalopathy: Pandora's box?


  • Gavin Wright MRCP,

    1. Liver Failure Group, Institute of Hepatology, University College London, London, UK
    Search for more papers by this author
  • Rajiv Jalan FRCP, Ph.D.

    Corresponding author
    1. Liver Failure Group, Institute of Hepatology, University College London, London, UK
    • Liver Failure Group, Institute of Hepatology, University College London, 69-75 Chenies Mews, London WC1E 6HX, UK
    Search for more papers by this author
    • fax: +44-2073800405.

  • See Article on Page 514.

  • Potential conflict of interest: Nothing to report.

Hepatic encephalopathy (HE), the neuropsychiatric manifestation of liver disease, incorporates a spectrum of manifestations ranging from minimal derangements in neuropsychological function to confusion and coma. Over the past 10 years, studies have confirmed the strong association between hyperammonemia due to liver dysfunction and infection/inflammation in the pathogenesis of HE, in acute liver failure,1 cirrhosis,2 and more recently in acute-on-chronic liver failure.3 An improvement in HE following interventions which modulate inflammatory responses, such as hypothermia,4 and the use of indomethacin (cyclo-oxygenase pathway),5, 6 albumin and albumin dialysis (free radicals, free metals, nonspecific protein-bound substances)7 and, probiotics (endotoxin)8 indicates that reducing inflammation is also a valid modality for treating HE.

In this issue, Cauli et al.9 report in a rat model improved learning ability in animals with HE following the administration of the nonsteroidal anti-inflammatory drug (NSAID), ibuprofen. They showed that the chronic administration of ibuprofen (from day 10 up to 4 weeks after portacaval shunting [PCS]) at 5-6 times the therapeutic doses resulted in a “normalization” of cyclo-oxygenase (COX) and inducible nitric oxide synthase (iNOS) activity. Moreover, administration of ibuprofen was also associated with improvement in the glutamate–nitric oxide–cyclic guanosine monophosphate (Glu-NO-cGMP) pathway. In PCS rats, brain interleukin-6 was elevated but tumor necrosis factor alpha (TNFα) remained unchanged, and controversially, TNFα increased significantly in the ibuprofen-treated animals. It is important to note that this model is more akin to that observed in “minimal HE” as opposed to more severe forms of liver disease and therefore must be interpreted in this light. This study touches on the complex processes and multiple cell types involved in the pathogenesis of HE, highlighting the importance of inflammatory pathways and their modulation in the treatment of “minimal HE”. The questions that need to be addressed in the interpretation of the data presented and implications for pathogenic mechanisms and therapy are:

1. Is Ammonia Associated with Brain Inflammation?

Ammonia is known to be important in the pathogenesis of HE, and it is likely to act as the priming stimulus on the background of which inflammation of the brain may produce HE. Ammonia is detoxified by the astrocytes in the brain, which swell during hyperammonemia due to the osmotic effect of glutamine.10 Astrocytes form an integral component of the blood-brain barrier and regulate cerebral blood flow through an arachidonic acid–dependent pathway (related to the COX pathway).11 Therefore, astrocytes may be the critical cells interacting between ammonia and inflammation/infection. Thus, hyperammonemia may “activate” astrocytes by “unlocking” the blood-brain barrier, making them susceptible to endotoxemia through a COX-dependent mechanism.12


COX, cyclo-oxygenase; Glu-NO-cGMP; glutamate–nitric oxide–cyclic guanosine monophosphate; HE, hepatic encephalopathy; iNOS, inducible nitric oxide synthase; NSAID, nonsteroidal anti-inflammatory drug; PCS, portacaval shunt; TNFα, tumor necrosis factor alpha.

2. How Does the Systemic Inflammatory Response Impair Cerebral Function?

Due to the presence of the blood-brain barrier which remains anatomically intact until the very late stages of HE, it is difficult to explain how systemic inflammatory responses produce deleterious brain effects. Current hypotheses suggest that the brain parenchyma may interact with systemic factors via the blood-brain barrier endothelial cells, barrier-deficient areas, or by stimulation of peripheral nerves.13 The brain parenchyma is composed of multiple cell types such as astrocytes, microglia, neurons, and endothelial cells that have the capacity to respond to and/or generate an inflammatory response. Individually or collectively, it is likely that these cells participate in the modulation of inflammatory responses involving cytokines, COX, NOS, and the Glu-NO-cGMP pathways.9 (Fig. 1)

Figure 1.

Systemically derived ammonia (NH3) and inflammatory mediators along with cerebral inflammatory responses are likely to modulate cerebral blood flow (CBF). Systemic factors interact with the brain at barrier deficient areas, through paracellular pathways (red arrows), peripheral nerve stimulation, or via secondary messenger signaling from the microvessel endothelial cells (ECs) of the blood-brain barrier (BBB). In the brain, the parenchymal cells (astrocytes, microglia, and neurons) interact with each other, through alteration of the extracellular environment. Astrocytes are central to the pathogenesis of HE and swell due to the osmotic effect of glutamine, which is formed during detoxification of ammonia. Increased lactate (LAC) is produced in the astrocyte both during aerobic and anaerobic metabolism which may exacerbate swelling. Glutamate (GLU) is principally produced in the presynaptic neurons and is integral to glutamatergic neurotransmission. This activates the NMDA receptor signaling resulting in the production of cGMP. Increasing cGMP levels (for example, using a phosphodiesterase [PDE] inhibitor, Sildenefil, in portacaval-shunted rats) improves learning ability. All cells use similar inflammatory pathways (COX, NOS, and cytokine-related) to produce, for example, prostanoids, nitric oxide (NO), and reactive nitrogen and oxygen species (RNS/ROS) which may cause cellular dysfunction and injury. There is significant cross-talk between COX and NOS mechanisms; however, their responses may vary significantly (enlarged green box). NSAIDs were shown in the paper by Cauli et al.9 to be beneficial in minimal HE by modulating the COX-NOS pathways and improving NMDA-induced cGMP generation.

3. What are the factors that regulate brain inflammation?

The key pathophysiological processes involved in the regulation of inflammation are:

i. Cytokine/endotoxemia.

Endotoxin and cytokines are major mediators of the systemic inflammatory response. Neither readily cross the blood-brain barrier, acting primarily on brain parenchyma via endothelial cell receptor interactions with downstream NO and/or prostanoid production.14 Because the astrocytes form a part of the blood-brain barrier, it is likely that they oversee the interaction of systemic derived factors to regulate cerebral blood flow and protect neurons from injury.15 Microglia also exhibit an extensive cytokine repertoire, and though little evidence exists to support a role in HE, their importance in other chronic neurodegenerative diseases suggests otherwise.16 In the late stages of acute liver failure, proinflammatory cytokines are observed to efflux from the brain, indicating brain production,17 but brain cytokine responses in minimal HE are unknown. In the PCS model, which is clinically akin to “minimal HE”, Cauli et al.9 found an elevation of interleukin-6 with no changes compared to controls, but controversially, the TNFα was significantly higher in the animals treated with ibuprofen. It is difficult to reconcile the finding in this study of significantly elevated TNFα in animals treated with ibuprofen, with the observation of normal TNFα in control animals with minimal HE. Better understanding of this observation requires more detailed studies.

ii. NOS/NO responses and the Glu-NO-cGMP pathway.

NO is a key second messenger involved in multiple inflammatory and physiological functions, including vascular relaxation. Studies in experimental models have been inconclusive whether NO in the brain is increased18, 19 or decreased.20 It is counterintuitive that in PCS rats (a model of chronic liver disease and minimal HE), iNOS activity should be up-regulated (as shown by Cauli et al.9) because clinical studies suggest that advanced cirrhosis is a condition associated with reduced cerebral blood flow.21 In the absence of information about protein expression and endothelial NOS and neuronal NOS, interpretation of significance of the iNOS activity data is difficult.

Altered glutamatergic neurotransmission in the cerebellum and cortex are important to learning and memory.22 Activation of mainly neuronal N-methyl-D-aspartic acid (NMDA)-glutamate receptors activates NOS, with increased NO production; NO activates guanylate cyclase, resulting in increased concentration of cGMP. Modulation of any step in this pathway affects many important cerebral processes such as sleep-wake cycle, neurotransmitter release and intercellular signaling, ion channels, enzyme systems, and learning and memory, which may all be altered in HE.22 Increasing the extracellular cGMP level is seen as a potential target to improve learning ability in HE.4 For example, sildenafil, a phosphodiesterase inhibitor, improved learning ability in the PCS model.23 The dependence of this pathway on NOS, with its potential cross-talk with COX, provides a mechanism by which NSAIDs can modulate intellectual functioning in HE. The demonstration by Cauli et al.9 of increased cGMP in response to NMDA in the ibuprofen-treated animals with improved learning ability is intriguing but requires more explanation because it implies that ibuprofen is correcting a “low NO” state. However, the basal levels of glutamate, NO, and cGMP are normal in both the cerebellum and cortex of PCS rats, with the pathway abnormality only evident as a blunting of the response to further stimulation. It is not clear whether this represents a phenomenon in the particular model studied or a failure of compensation resulting in “minimal HE”. Additionally, there are directly opposing effects of Glu-NO-cGMP pathway activation in different areas of the brain. Studies by the Felipo group have shown that NO-induced activation of soluble guanylate cyclase to produce cGMP varies in different areas of the brain (increased in cerebral cortex and reduced in cerebellum) in different models.24 This naturally raises concerns that any manipulation of the Glu-NO-cGMP pathway may impart differential responses between functionally different sites in the brain and perhaps suggests it is unlikely that this is the final common pathway as is implied by Cauli et al.9

iii. Arachidonic acid/COX/prostanoid responses.

Of the 2 main COX isoforms, COX-1 (constitutive) predominates outside the brain whereas COX-2 (inducible) is expressed at relatively high levels in the neuronal cells.25 COX-1 predominates in astroglial cells and infiltrating monocytes.26 These cellular variations in COX-1/COX-2 indicate functionality with COX-2 chiefly regulating synaptic activity and COX-1 regulating inflammatory processes. However, the effect of simple inhibition of either isoform is not entirely predictable; COX-2 may be transiently up-regulated in response to inflammation26 or may have anti-inflammatory effects during resolution of inflammation.27 Under pathological conditions, NMDA receptor stimulation may induce neuronal death in certain disease states,28 secondary to superoxide generation and free radical–mediated lipid peroxidation.29 This may provide a rationale for the use of antioxidants such as desferrioxamine and N-acetylcysteine. The observation of Cauli et al.9 showing increased COX activity in PCS rats and its reduction with ibuprofen is interesting, but conclusions are difficult in terms of their role in the pathogenesis of minimal HE without information about the type of COX involved, the affected cell type, and the associated gene and protein data.

iv. Interaction between NOS, COX, and Glu-NO-cGMP.

Both NOS and COX have constitutive isoforms predominantly involved in the regulation of general “housekeeping” cellular processes, and the inducible isoforms predominate in modulating inflammatory responses. Although the exact links between NOS and COX are not fully defined, a cross-talk between NOS and COX is likely but the nature of the interaction is dependent on the cell type, enzyme isoform, concentration of mediator release, and/or the specificity of the inhibitors being used.30 How these relate to the changes in the Glu-NO-cGMP pathway remains to be determined.

4. Should NSAIDs Be Used in the Treatment of HE?

The use of NSAIDs in HE is not novel. In patients31 and PCS rats administered ammonia,6 the nonselective COX inhibitor indomethacin improved brain edema. Systemically, selective COX-2 inhibition appears desirable, limiting COX-1–related side effects, such as gastrointestinal, respiratory, and renal effects. Also, brain penetration of NSAIDs at therapeutic doses is low, which may necessitate potentially toxic doses.32 This may result in systemic side effects as was evidenced by a significant decline in renal function by about 35% in the PCS rats treated with ibuprofen.9 Additionally, administration of ibuprofen to patients with sepsis was associated with effects on cardiovascular hemodynamics, temperature, oxygen consumption, lactate, and whole body and cellular prostaglandin metabolism.33 These processes, both individually and collectively, are known to affect brain function. The lack of any systemic physiological data makes it difficult to exclude whether the observed effect of ibuprofen was due to alteration of physiologic processes outside the brain.

The study by Cauli et al.9 provides further evidence to indicate that inflammation plays an important synergistic role in HE but the mechanism of the processes involved, the neuroanatomical basis, or the cell types implicated remain unclear. From the pathophysiological perspective, it is becoming clear that the pathogenesis of HE is likely to involve several cell types in the brain and it will be important to carefully start to break down which mediators are expressed in which cell types and how do they interact with each other. Due to differences in regional metabolism and interactions between multicellular inflammatory processes, NSAIDs are unlikely to be a panacea for the treatment of HE. Modulation of hyperammonemia and the inflammatory processes without compromising global brain homeostasis or inducing systemic complications remains the key to defining new therapeutic strategies for HE.1

Table 1. A Proposed Relative Importance of Inflammatory Cellular Processes in the Brain
Brain cell typeNOcGMP (on intellect and motor function)COX-1COX-2Cytokine Repertoire
  1. Abbreviations: cGMP, cyclic guanosine monophosphate; COX, cyclooxygenase; NO, nitric oxide.

Endothelial cell✓✓x✓✓
Neuron✓✓✓✓✓✓x (✓)


Members of the Liver Failure Group, who helped consolidate the inflammation and liver failure hypotheses. This work was undertaken at University College London Hospital/University College London which received a proportion of funding from the UK Department of Health's National Institute for Health Research Biomedical Research Centres funding scheme. The authors acknowledge the contribution of Dr J. Bernuau, who was involved in several discussions about ammonia and inflammation interactions in hepatic encephalopathy. Dr. Gavin Wright was supported by the Sheila Sherlock Fellowship from the European Association for the Study of the Liver.