Potential conflict of interest: Nothing to report.
Supported by the German Research Foundation through Sonderforschungs bereich 974 “Kommunikation und Systemrelevanz bei Leberschädigung und Regeneration” (Düsseldorf, Germany). We thank the Australian Brain Donor Program's NSW Tissue Resource Centre, which is supported by the University of Sydney, National Health and Medical Research Council of Australia, Schizophrenia Research Institute, National Institute of Alcohol Abuse and Alcoholism, and NSW Department of Health for tissue support.
Hepatic encephalopathy (HE) is a frequent complication of liver cirrhosis and is seen as the clinical manifestation of a low-grade cerebral edema associated with oxidative-nitrosative stress. However, comprehensive data on HE-associated molecular derangements in the human brain are lacking. In the present study, we used a whole human genome microarray approach for gene expression profiling in post mortem brain samples from patients with cirrhosis with or without HE and controls without cirrhosis. Altered expression levels were found for a total of 1,012 genes in liver cirrhosis patients without and with HE, and HE-characteristic gene expression changes were identified. Genes with altered expression pattern in HE were related to oxidative stress, microglia activation, receptor signaling, inflammatory pathways, cell proliferation, and apoptosis. Despite an up-regulation of genes associated with microglia activation, pro-inflammatory cytokine messenger RNA profiles remained unchanged in the brains of patients with liver cirrhosis and HE compared with controls. Interestingly, many genes counteracting pro-inflammatory signaling and inflammatory cytokine expression were up-regulated in the cerebral cortex of patients with liver cirrhosis and HE. Conclusion: Pathogenetic mechanisms of HE deduced from cell culture and animal experiments, such as oxidative stress, altered Zn2+ homeostasis and microglia activation also apply to human brain from patients with liver cirrhosis and HE. The study also revealed a not-yet recognized increased expression of genes antagonizing proinflammatory signaling and inflammatory cytokine expression. (HEPATOLOGY 2013;57:2436–2447)
Hepatic encephalopathy (HE) represents a neuropsychiatric syndrome, which frequently accompanies acute or chronic liver failure.1 Studies on the human brain in vivo suggest that HE reflects the clinical manifestation of a low-grade cerebral edema that develops after exhaustion of the volume-regulatory capacity of the astrocytes in response to ammonia and other HE-precipitating factors.2-4 As a consequence, astrocyte swelling triggers an oxidative-nitrosative stress response through activation of nicotinamide adenine dinucleotide phosphate oxidase and nitric oxide synthase (NOS).4-6 Because oxidative stress in turn induces astrocyte swelling, a self-amplifying signaling loop between osmotic- and oxidative/nitrosative stress stress was proposed.1, 4 Consequences of this oxidative/nitrosative stress response are protein-tyrosine nitration, oxidation of RNA, and activation of Zn2+-dependent transcription.1, 4-8 More recently, evidence for the presence of oxidative stress, protein tyrosine nitration, and RNA oxidation was found in post mortem brain biopsies from patients with liver cirrhosis and HE when compared with those without HE, indicating that cerebral oxidative stress is a hallmark of HE in patients with liver cirrhosis.9
In addition to osmotic and oxidative stress, microglia activation may contribute to the pathogenesis of HE. Microglia activation was found not only in rodents with acute or chronic liver failure or after ammonia intoxication, but also in human cerebral cortex from patients with cirrhosis with HE, but not in those without HE.10-13 Whereas microglia activation in animal models of acute liver failure was accompanied by an increased synthesis of pro-inflammatory cytokines, microglia activation in patients who had cirrhosis with HE was not associated with an increased formation of pro-inflammatory cytokines, indicating that under these conditions, microglia is activated but not reactive.10, 13
Additional mechanisms have been implicated in the pathobiology of HE, such as loss of blood brain barrier integrity or neuronal apoptosis, but direct evidence for an involvement of such mechanisms in the pathobiology of HE in humans is still missing.14, 15
Techniques for analyzing gene expression changes have advanced considerably in recent years, and profiling of the entire human transcriptome by microarray analysis in post mortem brain tissue from patients with neurodegenerative diseases (e.g., Alzheimer disease, Parkinson disease) has uncovered new pathophysiologically relevant signaling pathways.16, 17 Such a broad screening approach has not been applied to human post mortem brain samples from patients with liver cirrhosis with or without HE. In the present study, we used Agilent whole human genome micro-array technique to analyse transcriptomic changes in post mortem brain biopsies from the cerebral cortex of patients with liver cirrhosis with or without HE. This approach allowed us to identify a specific set of genes whose expression is altered in the brains of patients with liver cirrhosis and HE when compared with patients who had cirrhosis without HE or controls without cirrhosis. These genes were related to oxidative stress, microglia activation, receptor signaling, inflammatory pathways, cellular proliferation, and apoptosis.
Post mortem human brain tissue was obtained from autopsies of eight control subjects and eight patients with liver cirrhosis and accompanying HE who died while in hepatic coma (HE grade IV according to West Haven Criteria18). Patient characteristics and histories are described in detail in the Supporting Information and Supporting Tables 1-3). Controls were free from hepatic or neurological disorders. Written consent for tissue analysis had been given either by the patients or their relatives or was included in the body donor program of the Department of Anatomy at the University of Düsseldorf. Tissue from three patients with liver cirrhosis without HE was obtained from the Australian Brain Donor Programs NSW Tissue Resource Centre. All brain samples analyzed in this study were taken from the intersection parietal to occipital cortex area.
RNA Isolation and Quality Control.
A detailed description of RNA isolation and quality control can be found in the Supporting Materials and Methods.
A detailed description of the microarray analysis can be found in the Supporting Materials and Methods.
Data Preprocessing and Functional Grouping Analysis.
A detailed description of data preprocessing and functional grouping analysis can be found in the Supporting Materials and Methods.
A detailed description of array data analysis can be found in the Supporting Materials and Methods.
Differential Gene Expression in Cerebral Cortex From Patients Who Have Cirrhosis With and Without HE.
By comparing global gene expression profiles using a whole genome gene array approach, we identified a total of 1,012 genes in post mortem brain biopsies from patients with liver cirrhosis with and without HE with an altered expression level of more than 1.5-fold at a statistical significance level of P ≤ 0.01.
In patients with liver cirrhosis and HE, gene expression levels in the cerebral cortex were changed in about 1.5% out of 42,405 transcripts analyzed when compared with controls without cirrhosis (Fig. 1A). Expression levels of 434 genes were increased, whereas those from 204 genes were reduced. Altered gene expression levels were also found in almost 1% of all genes in patients with liver cirrhosis without HE. Here, messenger RNA (mRNA) expression levels of 145 species were up-regulated and 251 were reduced (Fig. 1B). Likewise, comparison of individual gene expression levels between patients with cirrhosis with HE and patients with cirrhosis without HE showed different gene expression levels in about 3% of all transcripts analyzed (Fig. 1C).
When genes with altered expression levels were illustrated in heat maps, it became evident that the vast majority of genes with altered expression levels in patients with cirrhosis with HE is not affected in patients with cirrhosis without HE (Fig. 2A). This was found for both, up-regulated and down-regulated genes (Fig. 2A). Only a small fraction comprising 22 genes was affected in both patient groups with liver cirrhosis (Fig. 2B). From those genes, the expression levels of 16 were altered in the opposite direction (i.e., they were up-regulated in patients with cirrhosis with HE and down-regulated in patients with cirrhosis without HE, whereas only six (∼0.01% of all transcripts analysed) were altered in parallel in both groups (Supporting Tables 4 and 5).
These data suggest specific alterations of gene expression in the human cerebral cortex triggered by liver cirrhosis per se on the one hand and by HE per se on the other.
Functional Grouping of Regulated Genes in Human Cerebral Cortex in Patients With Cirrhosis With and without HE.
By using annotations from the Gene Ontology database providing information about biological processes and molecular functions, genes with altered expression levels were assigned to different groups. Gene expression changes were found in patients with cirrhosis with and without HE with varying frequency in all different functional classes listed in Fig. 3A. Depending on the presence or absence of HE in patients with cirrhosis, a clear tendency for either up-regulation or down-regulation of genes was observed in every functional class (Fig. 3A). Figure 3B,C further resolves functional classes into groups of specific biological function and provides an overview about biological categories that were found most frequently among regulated genes. Because the size of the bars does not necessarily represent a particular biological significance, but may result from a higher number of members within a specific group, additional statistical testing was performed (Supporting Table 6). The results show that genes up-regulated in the cerebral cortex of patients with cirrhosis and HE are significantly associated with biological processes that have been formerly suggested to play a role in the pathobiology of HE, such as receptor signaling, innate immunity, inflammation, responses to toxins or reactive oxygen species, and nitric oxide–mediated signaling, among others. In contrast, down-regulated genes could not be assigned to any particular functional category (data not shown).
In patients with cirrhosis without HE, down-regulated genes were clustered in biological categories such as immune response pathways, proliferation, or cell differentiation (Supporting Table 6), whereas up-regulated genes could not be assigned to any particular functional category (data not shown). Due to space limitations, this manuscript can only focus on a subset of individual genes with altered expression level.
Expression Levels of Individual Genes Associated With Oxidative Stress Responses in Human Cerebral Cortex From Patients with Cirrhosis with and without HE.
Findings derived from cultured brain cells, vital brain slices, animal models for HE, and post mortem brain biopsy material from patients with liver cirrhosis and HE consistently point to a major role of cerebral oxidative stress in the pathophysiology of HE.4-9 As illustrated in Fig. 4A, expression levels of genes involved in oxidative stress defense, such as heme oxygenase-1 (HO-1), selenoprotein-V, peroxiredoxin-4, and peroxisome proliferator-activated receptor α (PPARα) were elevated in patients with cirrhosis with HE but not in patients with cirrhosis without HE when compared with controls. No increased gene expression was found for members of pro-oxidative signaling pathways such as NOS isoforms, which were recently suggested to be involved in the pathobiology of HE (Supporting Fig. 2).4, 7 Findings derived from cultured astrocytes treated with HE-precipitating factors (e.g., hyponatremia, ammonia, tumor necrosis factor α [TNF-α], and diazepam) and ammonium acetate-treated rats in vivo suggest that oxidative/nitrosative stress liberates Zn2+ from protein complexes, thereby increasing the intracellular concentration of “free” zinc and activating Zn2+-dependent transcription of genes such as Zn2+ chelating metallothioneins (MTs).6, 19, 20 As shown in Fig. 4A, a total of six different genes coding for MTs are up-regulated in the cerebral cortex of patients with cirrhosis and HE. In contrast, a significant down-regulation of the gene expression levels for MT1C and MT2A was found in patients with cirrhosis without HE.
Expression of Genes Associated with Microglia Activation and Inflammation in Cerebral Cortex From Patients with Cirrhosis with and without HE.
Recent findings derived from cultured microglia, animal models for HE, and human post mortem brain biopsies from the cerebral cortex of patients with cirrhosis with HE but not those without HE point to a role of microglia activation in the pathogenesis of HE.10-13 As shown in Fig. 4B, increased gene expression levels of the microglia activation marker perilipin-II are present in the cerebral cortex of patients with liver cirrhosis with HE but not without HE.21 Likewise, CD14 and CD163, which are specifically expressed on activated microglia in brain cells,22 were significantly up-regulated in the cerebral cortex of patients with cirrhosis with HE but not without HE. Moreover, significant up-regulation of genes associated with microglia activation–such as major histocompatibility complex type II (MHC-II/human leukocyte antigen) and MHC-II transactivator, which activates transcription of MHC-II genes—was found in patients with liver cirrhosis and HE but not without HE (Fig. 4B). These data suggest increased expression levels of genes that are involved in microglia activation in patients with cirrhosis and HE, but not in patients with cirrhosis without HE. In our recent study,13 we observed up-regulation of the microglia activation marker ionized calcium-binding adaptor protein 1 (Iba-1) at the protein expression level, whereas no increase was found at the mRNA level as analyzed by real-time polymerase chain reaction (data not shown) pointing to an involvement of posttranslational mechanisms in Iba-1 up-regulation in HE. This was also confirmed in the present study via microarray analysis (Supporting Fig. 2). Analysis of the surrogate marker expression characteristic for the pro-inflammatory M1 or anti-inflammatory M2 microglia phenotype suggests the presence of both microglia subpopulations, with one surrogate marker being up-regulated for M1 and three being up-regulated for M2 (Fig. 4B,C, Fig. 5A, and Supporting Fig. 3B).
Recently, HE in acute and chronic liver failure was proposed to be a neuroinflammatory disorder associated with increased cerebral synthesis of pro-inflammatory cytokines.10-12 As shown in the present study, no changes in pro-inflammatory cytokine or chemokine gene expression were observed in the cerebral cortex from patients with cirrhosis with or without HE (Supporting Fig. 1A-D). These array findings confirm our recently published polymerase chain reaction data on interleukin (IL)-1α, IL-β, IL-6, and TNF-α.13
However, a significant up-regulation of mRNA coding for chemokine (C-X-C motif) ligand (CXCL) 8 and CXCL2 and for the lymphotoxin-β receptor (whose activation triggers CXCL8 synthesis) as well as for receptors of the anti-inflammatory cytokines IL-4 and IL-10β was found in post mortem brain biopsies from patients with HE (Fig. 4C).23 In the absence of detectable changes of IL-1 gene expression (Supporting Fig. 1A), a significant up-regulation of IL-1 receptor type 1 mRNA was noted in the cerebral cortex of patients with HE (Fig. 4C).
Bacterial infections are sensed by the innate immune system through Toll-like receptors (TLRs), thereby inducing the expression of pro-inflammatory cytokines through activation of the nuclear factor kappa B (NF-κB) pathway.1, 4, 24 As shown in Fig. 5A, gene expression levels of TLR-1, which in the central nervous system is exclusively expressed in microglia, and its dimer-partner TLR-2, are significantly elevated in the cerebral cortex from patients with cirrhosis with HE but not from those without HE.24 Concurrently, a significant up-regulation of Lyn, which represses lipopolysaccharide-induced TLR activation and synthesis of pro-inflammatory cytokines was observed in patients with HE.25 Likewise, genes that inhibit or counteract with TLR-dependent or TLR-independent NF-κB signaling such as myeloid differentiation primary response gene 88 (MYD88) splice variant II,26 IL-1 receptor-associated kinase 3 (IRAK3)27, TNF-α–induced protein (TNFAIP3; A20),28 and suppressor of cytokine signaling 3 (SOCS3)29 were significantly up-regulated in patients with cirrhosis and HE but not without HE.
These results indicate not only an up-regulation of genes involved in microglia activation, but also of anti-inflammatory signaling pathways in the cerebral cortex of patients with cirrhosis and HE, but not in those without HE.
Expression of Genes Involved in Cell Differentiation, Proliferation, or Apoptosis in Cerebral Cortex From Patients with Cirrhosis with and without HE.
Findings derived from cell culture and animal models for HE suggest that ammonia may affect proliferation and differentiation of astrocytes, thereby inducing the characteristic Alzheimer type II phenotype.30 Figure 5B,C illustrates the relative expression levels of genes involved in the regulation of cell division, differentiation, or apoptosis in the cerebral cortex of patients with cirrhosis with and without HE.
Among the pro-apoptotic genes, the death receptor TNF receptor superfamily member 10A and caspases 1, 6, and 8 were significantly up-regulated, whereas apoptosis-inducing factor mitochondrion-associated 3 was significantly down-regulated in patients with cirrhosis and HE. Among antiapoptotic genes, TNF receptor superfamily members 1, 6, and 11B—the latter of which is known to act as a decoy receptor for the death receptor ligands Fas or TNF-related apoptosis-inducing ligand (TRAIL)—were significantly up-regulated in patients with cirrhosis and HE.
Among genes involved in proliferation, transcription factors or their regulators—such as STAT5A, tumor suppressor protein 53 (TP53), γ-interferon inducible protein 16, and calmyrin—were elevated in the brains of patients with liver cirrhosis and HE (Fig. 5C). Other genes with elevated expression levels affecting proliferation are involved in nuclear organization (lamin A/C) or cell cycle control, including cell division protein kinase 2, cyclin-dependent kinases regulatory subunit 1, minichromosome maintenance complex component 5, and kinesin family member 22.
Expression levels of none of these genes were affected in the cerebral cortex of patients with cirrhosis without HE (Fig. 5B,C).
Relation to In Vivo Data.
Gene expression levels of MTs and genes involved in inflammatory signaling were correlated with in vivo laboratory data that were available from six patients with HE (Supporting Table 3) within a 24-hour period prior to death. Highly significant correlations were found between plasma ammonia concentration and gene expression levels of MTs 1B, H, L, and X (Fig. 6A) and between C-reactive protein (CRP) blood levels and cerebral mRNA levels for IRAK3, CD163, Lyn, and caspase 8, which are involved in inflammatory signaling (Fig. 6B). The data suggest a role of ammonia for MT expression in the brains of patients with cirrhosis with HE in line with the reported MT induction by ammonia in cultured astrocytes.20 The data further suggest that peripheral inflammation can trigger cerebral gene expression.
This is the first analysis of expression profiles of the whole transcriptome in post mortem biopsies from the cerebral cortex of patients with cirrhosis without or with HE in comparison with brain samples from patients without liver or neurological disease. The study shows a total of 638 or 396 genes with altered expression levels in the brains of patients with cirrhosis with or without HE, respectively, when compared with controls without cirrhosis. However, few genes are affected in both patient groups with liver cirrhosis, and here most of the genes are altered in the opposite directions (16 out of 22), which may point to adaptive processes relating to specific metabolic conditions. These findings indicate that gene expression profiles can discriminate patients with cirrhosis with HE from those without HE and points to specific pathophysiological mechanisms underlying the neurological phenotype of HE (Fig. 1). Overall, gene expression changes were consistent in seven out of eight patients with cirrhosis with HE. However, patient 5 (Fig. 2A) frequently had a different gene expression pattern. The reason for this is unclear, but it should be noted that this patient had the highest CRP by far but had normal bilirubin levels in peripheral blood (Supporting Table 3).
Functional grouping of genes with altered expression levels in patients with cirrhosis and HE revealed an unequivocal association with biological processes that were formerly implicated in the pathobiology of HE, such as oxidative stress (Fig. 3A and Supporting Table 6).1, 4, 9 In line with this, the present study shows increased mRNA expression levels of HO-1 in patients with cirrhosis with HE but not in those without HE (Fig. 4A), which serves as a surrogate marker for the presence of various stressors such as oxidative stress.31 This is in line with recent reports showing up-regulation of HO-1 in cultured rat astrocytes in vitro and in the brains of acutely ammonia acetate challenged rats in vivo at the mRNA and protein level32 or after portocaval anastomosis in rat brain in vivo employing gene chip array analysis.33
Apart from HO-1, other mRNA species involved in anti-oxidative defense were up-regulated in patients with HE, including peroxiredoxin-4 (which has recently gained much attention in different neuropathologies such as Morbus Alzheimer34), selenoprotein-V, and PPARα35 (which strongly counteracts oxidative stress and signaling pathways toward the synthesis of pro-inflammatory cytokines).
Nitrosative stress is induced by ammonia and other HE-precipitating factors in cultured rat astrocytes and activates Zn2+-dependent gene transcription, thereby increasing MT mRNA levels.6, 20 In line with this, an up-regulation of mRNAs of several MT isoforms in the cerebral cortex of patients with cirrhosis with HE was found that paralleled the peripheral blood ammonia concentration (Figs. 4A and 6A). Cirrhosis and HE had no significant effect on NOS mRNA levels, a finding that is in line with previous data on NOS 1 (neuronal NOS) and NOS 2 (inducible NOS) protein expression in human brain.9
In the brain, pro-inflammatory cytokines are largely produced by activated microglia. However, despite strong evidence for microglia activation, no increased mRNA or protein levels of pro-inflammatory cytokines were found in the cerebral cortex of patients with cirrhosis and HE when compared with those without HE or control patients13 (Fig. 4B,C and Supporting Fig. 1A-C). With the exception of an increased CXCL2 and CXCL8 mRNA expression (Fig. 4C), mRNA expression levels of the vast majority of chemokine mRNAs were unaffected in the brains of patients with cirrhosis and HE (Supporting Fig. 1D). CXCL2 and CXCL8 both target chemokine receptor CXCR2, thereby triggering chemotaxis, but they also exert neuroprotective effects, and moderately elevated CXCL8 levels have been suggested to protect against severe brain damage in septic encephalopathy.36-39
In addition to up-regulation of mRNA species coding for anti-inflammatory IL-4 and IL-10 receptors, elevated gene expression levels of the pro-inflammatory IL-1 cytokine receptor (IL-1R) were observed in patients with HE. Interestingly, within the cerebral cortex IL-1 receptor is involved in sleep regulation, and deranged IL-1 receptor signaling is known to promote hypersomnolence.40 However, it is unclear to what extent IL-1 receptor up-regulation may contribute to the sleep abnormalities found in patients with HE.1
An activation of microglia in the cerebral cortex of patients with HE is also suggested by significantly increased mRNA expression levels of toll-like receptor (TLR) 1 and its dimerization partner TLR 2 (Fig. 5A), the former being exclusively expressed on microglia in the brain.24 As part of the innate immune system, TLRs 1 and 2 sense bacterial lipopolysaccharide (LPS) and promote pro-inflammatory signaling through activation of the NF-κB pathway.24 Because LPS exposure not only induces phenotypic transformation of microglia into a pro-inflammatory state, but also up-regulates TLR expression, TLR up-regulation may be attributed to the presence of systemic infections, which are frequent complications of cirrhosis and are known to trigger episodes of HE.1, 4, 26 This might also be reflected by elevated gene expression levels of the LPS receptor CD14 in brains from patients with cirrhosis with HE, but not those without HE (Fig. 4B).
In contrast to this, pro-inflammatory signaling through TLRs is not reflected at the level of cytokine expression13 (Supporting Fig. 1A-D). This discrepancy can be explained by a significant up-regulation of mRNAs coding for proteins that strongly counteract inflammatory signaling through TLRs and NF-κB activation such as Lyn, PPARα, MYD88 splice variant 2, IRAK3, SOCS3, selenoprotein-S, or TNFAIP3. Interestingly, a strong correlation between CRP blood levels and cerebral expression levels of genes involved in pro- and anti-inflammatory signaling such as IRAK3, Lyn, and CD163 was observed. This suggests that peripheral inflammation can affect cerebral gene expression in patients with cirrhosis with HE.
Thus, one might speculate that pro-inflammatory processes in the cerebral cortex of HE patients are counteracted by an increased transcription of anti-inflammatory signaling. In line with this, microglia activation in the brains of HE patients is associated with increased expression levels of several biomarkers characteristic of the anti-inflammatory M2 macrophage phenotype.
Oxidative stress and inflammation are also known to impair cell proliferation and can promote apoptosis.41 In view of an increasing number of reports suggesting that HE symptoms may not be fully reversible after liver transplantation, it was speculated that neuronal apoptosis in HE might have been overlooked in the past.15 The present investigation revealed an up-regulation of genes potentially involved in apoptosis, such as TRAIL and caspases 1, 6, and 8, but also of anti-apoptotic genes, such as decoy receptors that inactivate FasL and TRAIL (Fig. 5B). Thus, the present gene expression profiling data may not necessarily indicate increased apoptosis in HE, and the up-regulation of caspases may serve biological functions other than apoptosis induction.
More clearly, the data point to gene expression changes that indicate altered cellular proliferation, a finding that has already been discussed with regard to Alzheimer type II astrocytes.1, 30 Interestingly, the transcription factor TP53, which substantially affects cell proliferation and is involved in DNA repair, is up-regulated in patients with cirrhosis and HE, and it was suggested recently that TP53 participates in ammonia-induced astrocyte swelling and glutamate uptake inhibition.42
Taken together, the findings in the present study identified specific changes within the transcriptome in human brain in liver cirrhosis that depend on the presence of HE and therefore may reflect molecular changes associated with the development of neurological symptoms in liver cirrhosis. However, gene expression changes do not necessarily reflect changes in protein expression, and the results of the present study cannot be extrapolated to other brain areas, which differ greatly in terms of their cellular architecture and transcriptome.43 Although special attention must be paid to the low sample numbers available for the present gene array analysis, the present and former studies9, 13 suggest that current views on the pathobiology of HE,1-4 which were obtained in in vitro systems and animal models, may also apply to the human brain. Apart from the low-grade cerebral edema,2, 3 these pathomechanisms include not only oxidative stress and microglia activation but also offer new perspectives with regard to an involvement of anti-inflammatory signaling pathways, which warrants further research.
We thank Jutta Kollet, Lisa Zatrieb, and Jan Schäferkord at Miltenyi-Biotech for technical support, bioinformatic analysis, and helpful discussions. Expert technical assistance was provided by Torsten Janssen.