Header
Header
Article

A mouse model of hepatic encephalopathy: bile duct ligation induces brain ammonia overload, glial cell activation and neuroinflammation


Currently, the best-characterized animal model in hepatic encephalopathy (HE) research is the bile duct ligation (BDL) rat model. Despite availability of mouse models for advanced chronic liver disease (CLD)7, central nervous system (CNS) changes in these models are not well characterized. Here, we are the first to systematically describe the development and temporal evolution of key characteristics of HE in chronic BDL mice. Additionally, we characterize the neurometabolic and gliovascular changes in BDL mouse brains which provides a framework for outcome measures in future studies, but also identifies possible contributing factors in HE development.

First, we show that the criteria for an HE model according to the International Society of Hepatic Encephalopathy and Nitrogen Metabolism (ISHEN) guidelines have been fulfilled5. We establish that mouse BDL induces behavioural changes and liver function failure with hyperammonemia. Importantly, hyperammonemia and motor dysfunction is absent in Swiss mice, a strain with known cirrhosis development 6 weeks after BDL8,9. Genetic differences likely account for these striking differences in phenotype. Interestingly, one group has reported that C57Bl/6j mice exhibit striking levels of gut barrier failure, gut dysbiosis and bacterial translocation compared to A/J mice. As ammonia is a gut-derived toxin and bacterial translocation is known to provoke HE, this might provide a clue as to the origin of this differential phenotype46. Additionally, another group has reported alterations in hepatic ammonia metabolization after BDL in C57Bl/6j mice, while no such observation has been reported in Swiss mice11. Careful consideration with regards to strain selection is therefore imperative in future research. Notably, other groups have independently reported hyperammonemia in BDL mice11,14, suggesting good reproducibility. Kinetics however differ, with one group describing hyperammonemia as early as 7 days post-induction11. Processing differences might explain variable results, as ammonia measurement is highly susceptible to preanalytical differences. Therefore, strict adherence to processing advice recently postulated by ISHEN is essential, as was done in our study5. In line with other studies10,11,12, we report that BDL mice reproduce the phenotype of decompensated cirrhosis already at 14 days after induction. Later timepoints are useful for the investigation of long-term effects of chronic HE.

Next, we show that BDL, only in C57Bl/6j mice, induces clear motor function deficits. It should be noted that the motor behavioural changes occur before hyperammonemia and are thus likely affected by non-ammonia related factors. Early systemic inflammation might contribute to these behaviours55. However, levels of circulating IL-6 are however only very mildly elevated in C57Bl/6j mice at early timepoints. Moreover, levels of circulating cytokines are comparable between Swiss and C57Bl/6j mice after BDL, while Swiss mice do not develop motor dysfunction. As suggested previously13,30,31, influx of inflammatory monocytes into the brain could induce this sickness behaviour in inflammatory liver injury in C57Bl/6j mice. Because of this marked motor dysfunction, reliable cognitive function assessment was not possible. In the future, cognitive tests less affected by motor phenotype, like the passive avoidance test14, might be considered to distinguish behavioural deficits related to hyperammonemia and inflammatory liver injury.

With the use of cerebrospinal fluid (CSF) metabolomics, we investigate for the first time the metabolome in CSF of BDL mice. In line with patient data18 and BDL rats19,28, glutamine levels reproducibly rise secondary to plasma ammonia in BDL mice, consuming glutamate. With regards to other osmolytes, transient decreases in creatine and taurine are detected as possible compensatory osmoregulatory mechanisms. Remarkably, creatine and taurine levels return to baseline in later timepoints, while glutamine levels remain stably elevated. This might suggest that taurine and creatine levels decrease in response to the rapid relative changes in glutamine concentrations, while in later timepoints, a new steady state develops. Alternatively, increased transport into the brain might account for the replenishing of creatine and taurine amounts in the brain56.

Energy metabolism dysregulation, through the direct inhibitory effect of ammonia on tricarboxylic acid (TCA) cycle enzymes and/or the preferential allocation of energy towards the activated immune system, has been put forward as an important pathogenic mechanism in HE4. Lactate accumulation secondary to energy failure could additionally play a role in brain edema in BDL rats23. To assess energy status, we measured levels of ATP and its derivatives ADP and AMP. In our study, AMP depletion in BDL mice could suggest that energy production in the brain is altered. However, since we were unable to detect ATP and ADP, these results should be interpreted with caution. Importantly, ADP and ATP levels have also been found to be decreased in BDL rat brains, suggesting other mechanisms like defective nucleotide synthesis as a possible cause for decreased purine levels20. In line with patient data15, we could not detect differences in TCA cycle intermediates. Additionally, we could not detect differences in CSF lactate levels, suggesting these mechanisms are not directly involved in mouse HE.

The role of cerebral, ammonia-induced, oxidative stress in HE is controversial49,50,51. In line with previous data19,20, we observe no alterations in glutathione-related metabolites in the CSF of BDL mice, suggesting cerebral oxidative stress in BDL mice is limited.

Another finding of interest for future investigation is the accumulation of (tauro-)conjugated bile acids, which has been described in both HE patients15 and experimental animals24. Interestingly, bile acids accumulate before the development of liver function failure. In a recent study, CNS accumulation of conjugated bile acids before the development of hyperammonemia in a model of non-alcoholic steatohepatitis with hepatocellular carcinoma is associated with astrocyte and neuronal loss, suggesting an independent role for bile acids in the brain47. In acute HE models, bile acids have been found to influence neuroinflammatory signaling48. In agreement with our data, taurocholic acid has been shown to induce neuronal MCP-1 signaling through sphingosine-1-phosphate receptor 257.

In decompensated cirrhosis, tryptophan metabolization through the tryptophan–kynurenine pathway is amplified25. Surprisingly, but in line with patient data15, we find increased levels of tryptophan in CSF in BDL mice, already 7 days after induction. Possible explanations include a higher free tryptophan concentration, which can pass the blood–brain barrier (BBB), due to hypoalbuminemia25. Also, active transport of tryptophan over the BBB is in competition with branched chain amino acids, which are depleted in cirrhosis58. Interestingly, we also detect kynurenine, a marker of neuroinflammation26, from 14 days onward, the timepoint where we detected microglial morphological changes consistent with neuroinflammation. As targeting the kynurenine pathway has proven beneficial in preclinical models of depression, this could prove to be an interesting therapeutical target for future investigation59.

In short, mouse BDL in C57/Bl6j mice presents with a CSF metabolic profile consistent with HE. Our data do not support a significant contribution of CNS energy metabolism dysfunction or oxidative stress, while cerebral bile acid signaling and the role of the tryptophan–kynurenine pathway in mouse HE deserve further investigation, with special focus on their role in neuroinflammation. As the CSF is an approximator of brain metabolic status, future studies should investigate whether these metabolic changes are present in brain tissue. As brain regions have been found to react differently in HE, investigation of other brain regions like the cerebellum might uncover similar regional variability in this model.

Finally, we are the first to perform extensive morphological and longitudinal evaluations of glial cells in BDL mice. With regards to astrocytes, alterations in glutamine already suggest their involvement in HE mice. For the first time, we report time-dependent morphological and molecular alterations in astrocytes in BDL mice. Astrocytes are instrumental in BBB homeostasis60. We have established that BBB permeability is altered, suggesting that astrocyte morphological alterations go together with pathological BBB perturbation52. The fact that clear changes in gene expression and protein subcellular redistributions of tight junction components seem to be absent suggests there is no major structural breakdown of the BBB in BDL mice. As suggested in BDL rats, functional BBB breakdown through decreased astrocytic vessel coverage might be the operant mechanism24.

We also observe activated microglial morphology as soon as 14 days after BDL induction. This is likely to be caused, at least in part, by systemic inflammation61,62 rather than ammonia alone63. It should be noted that accompanying soma enlargement and thickening of processes seen in classically activated microglia32 is not obvious in microglia after BDL, possibly indicating a transitional state of these microglia. Alternatively, senescent microglia can present as deramified cells with retracted processes64 and might account for the specific morphological phenotype observed in these animals. Interestingly, ammonia has been found to induce senescence in astrocytes and senescence markers have been detected in brain tissue of HE 65,66. However, whether ammonia with or without systemic inflammation similarly affects microglia in HE remains unknown.

Strikingly, only limited changes in CNS cytokines are detectable despite obvious morphological changes in microglial cells. However, this is in line with previous data on the effects of chronic low-grade inflammation on brain tissue cytokine levels62. Most likely, cyto/chemokine changes are relatively subtle and therefore lost due to bulk assessment of all neuronal cell types, as has been reported before in aged microglia67. Alternatively, microglia may present with a chronic effector phenotype, characterized by low cytokine release68. More specific investigation of enriched microglial populations in our model might provide the answer in the future.

Surprisingly, microglial activation is associated with a downregulation of Aif1, the gene encoding for allograft inflammatory factor 1 (IBA1), transcription in BDL mouse brains. Similar results were reported in sorted microglia after LPS injection69 as well as in disease associated microglia in Alzheimer’s disease70, suggesting the often reported increased Aif1 expression in neuroinflammation is rather due to microglial proliferation than activation. Accordingly, microglia do not proliferate in BDL mice, confirming earlier data13. Interestingly, as senescent cells are unable to divide due to cell cycle arrest71, a senescent phenotype would possibly account for this absence in proliferation.

From the presented data, it is evident that glial cell function is altered in HE. As this study is descriptive in nature, whether these changes are causal or rather secondary to the observed phenotype remains unclear. Future research should focus on determining the role of glial cell dysfunction in behavioural HE phenotype, as well as on the molecular mechanisms underpinning astrocyte reactivity and microglial activation and how these cell types interact in BDL mice. Based on our data, the contribution of hyperammonemia, (systemic) inflammation and bile acid accumulation in the development of glial cell dysfunction are obvious points of future research. Additionally, it has become increasingly clear that astrocytes72 and microglia73 are very heterogeneous, and efforts should be undertaken to assess the role of different types of astrocytes in HE.

In conclusion, we show that mouse BDL in C57Bl/6j mice can be used as a model for HE in CLD. This model has the benefit of possible genetic manipulation, which will allow research into CNS specific changes in HE, a drawback of the current most widely used model, the rat BDL model5. It however also has disadvantages compared to the rat BDL model, most prominently the confounding sickness behaviour that develops before the development of hyperammonemia. In future studies, depending on the research question, it is advisable to consider advantages and disadvantages of each rodent species. While BDL rats remain ideally suited for e.g. the study of the effects of ammonia-lowering compounds5,74, BDL mice provide a valuable tool for fundamental research into the brain-specific pathophysiology of HE. Additionally, we identify several mechanisms possibly involved in disease development in this model, including bile acid accumulation, tryptophan–kynurenine pathway activation and (neuro-)inflammation. This warrants further research into their contribution to the HE phenotype, as well as the therapeutic potential of targeting these pathways in clinical HE.



Source link

Back to top button