Longitudinal osmotic and neurometabolic changes in young rats with chronic cholestatic liver disease

Type C hepatic encephalopathy (type C HE) is increasingly suspected in children with chronic liver disease (CLD), and believed to underlie long-term neurocognitive difficulties. The molecular underpinnings of type C HE in both adults and children are incompletely understood. In the present study we combined the experimental advantages of in vivo high field 1H magnetic resonance spectroscopy with immunohistochemistry to follow longitudinally over 8 weeks the neurometabolic changes in the hippocampus of animals having undergone bile duct ligation as pups. Rats who develop CLD early in life displayed pronounced neurometabolic changes in the hippocampus characterized by a progressive increase in glutamine concentration which correlated with plasma ammonia levels and a rapid decrease in brain myo-inositol. Other neurometabolic findings included a decrease in other organic osmolytes (taurine, choline-containing compounds and creatine), ascorbate and glutamate. At the cellular level, we observed an increase in glial fibrillary acidic protein (GFAP) and aquaporin 4 (AQP4) expression in the hippocampus at 4 weeks post bile duct ligation (BDL), together with astrocytic morphological alterations. These findings differ from observations in the brain of adult rats following BDL, and are in keeping with the commonly accepted theory of age-dependent vulnerability.

www.nature.com/scientificreports www.nature.com/scientificreports/ CLD have only been performed on MR scanners up to 3 Tesla (T), allowing the measurement of a limited number of metabolites, thereby limiting our understanding of the molecular events in the CNS of patients with CLD. Studying type C HE in an animal model at higher resolution offers more granular insight into the longitudinal, molecular changes occurring during HE progression 18 . High resolution in 1 H-MRS allows for the identification of more metabolites, which is warranted to tease out the impact of CLD from developmental processes and to detect subtle, early changes.
What happens at the molecular and cellular levels to the brain of young subjects with CLD is understudied, as cognitive and motor delays have historically be ascribed to chronic illness, not focusing on the specific risks of CLD on brain development, something which this study aimed to address longitudinally in an animal model of CLD acquired at a young age. Therefore, the novelty of the present study was to use the experimental advantages of high field 1 H-MRS to analyse longitudinally the hippocampus of rats having undergone bile duct ligation (BDL) as pups and to compare these findings to those of a similar study from our lab in adult rats, the underlying hypothesis being that the pups would display more significant neurometabolic changes than adults.

Methods
Study design. All animal experiments were conducted according to federal and local ethical guidelines, and the protocols were approved by the local Committee on Animal Experimentation for the Canton de Vaud, Switzerland (VD2761).
Twelve (12) male Wistar pups underwent BDL and 7 animals were sham operated on postnatal day 21 (P21). Blood sampling and in vivo brain 1 H-MRS scans were performed at week 2, 4, 6 and 8 post-BDL. Behavioural tests were performed to measure motor activity at week 4, 6 and 8 post-BDL, before MRS scans.
To approximate rat-and human-brain development, we used a previously published neuroinformatics approach 19,20 . According to this study, overall brain development of the P21 rat (day of BDL surgery) is the equivalent of an 8 month old human, while P77 old rats (end of study: week 8 post-BDL or post coital age 98 days) approximate an 8.5 year old human (http://translatingtime.org/translate).

Validation of CLD and type C HE induced by BDL in P21 rats. Biochemical measurements.
Plasma samples were analysed using Reflotron ® System for glucose, Integra ® 400 Plus for ammonium and COBAS ® 8000 for total bilirubin as markers of biliary obstruction and liver function (Roche, Switzerland). 18 , to confirm the presence of motor deficits (characteristic of type C HE) 21,22 .

Behavioural tests. Locomotor activity was assessed in the open field (OF) test
In vivo 1 H-MRS. In vivo 1 H-MRS was performed as previously described 18 . Measurements were conducted on a horizontal 9.4 T MR-system using SPECIAL sequence (TE = 2.8 ms) in a volume of interest (2 × 2.8 × 2 mm 3 ) placed in dorsal hippocampus. The rationale for studying the hippocampus was two fold: 1) it is implicated in learning and memory, both of which are impaired in humans with type C HE 23 and 2) it is technically easily accessible and offers an excellent signal at 9.4 T. 18 brain metabolites were quantified using LCModel and water as internal reference.
Histological methods -brain. 3 BDL rats and 3 shams were sacrificed at week 4 post-BDL for histological analysis. These rats did not undergo 1 H-MRS scans or behavioural tests. Astrocytes were analysed by an anti-glial fibrillary acidic protein (GFAP) mouse monoclonal antibody (MAB360 Merck Millipore). Changes in water channel aquaporin 4 (AQP4) were followed using a rabbit polyclonal antibody (AB3594 Merck Millipore) as a marker of osmoregulation. Secondary antibodies were goat or rat anti-mouse or anti-rabbit IgG labeled with Alexa Fluor ® 555 (red, Life Technologies). Immunostaining was performed as previously described 18 (see supplementary materials).
Statistical methods. All results are presented as mean ± SD. One way ANOVA (Prism 5.03, Graphpad, La Jolla CA USA) with the Bonferroni's multi-comparisons post-test were used to assess significance (*p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001) in measured parameters. Pearson correlation analysis was performed on all longitudinally acquired data to test for correlations between measured parameters. Three types of correlations were performed: (1) to test overall correlation between changes during the whole progression of the disease (all longitudinally-acquired measurements between weeks 2 and 8); (2) to test correlations between changes only in the beginning of the disease (only values acquired at weeks 2 and 4 were used for the correlations); and (3) to test correlations between changes only at the end of the disease (values acquired at weeks 6 and 8 were used). Brain metabolites are expressed in absolute values (mmol/kg ww (wet weight)) and in % difference between BDL rats and shams at each time-point in order to account for ongoing development.
More details on Methods can be found in the Supplementary data file.

Results
Validation of CLD and type C HE induced by BDL in P21 rats. Biochemical measurements. Increased plasma bilirubin confirmed the presence of CLD in BDL rats: from undetectable in shams and before BDL (<0.5 mg/dl) compared to 9.5 ± 2.2 mg/dl at 8 weeks post-BDL (Data from different time points are summarized in Fig. 1A and Table S1). Plasma NH 4 + averaged 103.9 ± 26.9 μM in shams throughout the study. BDL rats displayed a significant increase in plasma NH 4 + at week 8 (237 ± 55 μM) ( Fig. 1B and Table S1). In addition, BDL rats had measurably lower plasma glucose levels from week 4 (Fig. S1). Finally, BDL rats displayed significantly slower weight gain than sham operated animals from week 4 (Fig. S1).
Behavioural tests. Starting at 4 weeks post-BDL surgery, BDL rats showed significant decrease in the distance moved during the OF test compared to shams (Fig. 1C,D). Performance of BDL rats decreased further, reaching only 45% of shams at week 8. This decrease of motor activity in BDLs compared to shams correlated with plasma bilirubin levels (r = −0.52, p = 0.03, Fig. 1E). However no correlation was observed between deterioration in motor activity and increase in plasma NH 4 + . There was no difference between BDL rats and shams in latency to enter central zone or time spent in that zone (data not shown) at any time point, suggesting no difference in anxiety during the test between BDLs and shams.

In vivo 1 H-MRS assessment of CNS metabolic alterations in a model of CLD acquired in pups.
Spectra were characterized by a visible and quantifiable increase of Gln signal and decrease of mIns in the BDL rat 8 weeks post-BDL compared to sham (Fig. 2).
Gln and other main organic osmolytes. One of the earliest observable metabolic changes in the hippocampus was an increase in Gln concentration 2 weeks post-BDL (+44%). It then increased progressively throughout the course of the study, reaching +368% at week 8 (Fig. 3A,B). Figure 3C illustrates the correlation between brain Gln and plasma NH 4 + levels over the course of disease (week 2-8). Although the correlation was very strong in the early stages of the disease between weeks 2-4 (r = 0.94, p < 0.0001), it became non-significant between weeks 6-8 (r = 0.53, p = 0.14) (Fig. 3D,E).
Gln increase was mirrored by a decrease in mIns, which was significant from week 2 (−13%) and reaching −57% at week 8 (Fig. 4A,B). Other major brain organic osmolytes displayed a comparative delay in their decrease, reaching significance 4 weeks post-BDL. Maximal changes in concentration reached −60% for tCho, −17% for Tau, −19% for total creatine (tCr) at 8 weeks post-BDL. Correlations between these metabolites (mIns, tCho, Tau or tCr) and brain Gln or plasma NH 4 + are presented in Fig. 4C,D.
Metabolic changes during brain development. It is well known that brain metabolite concentrations change during brain development until P28 24,25 . Given that 1 H-MRS measurements in this study were performed on rats between P35-P77, we evaluated possible metabolic changes in the hippocampus of sham BDL rats. The only metabolites that showed significant change in shams during this period were mIns and tCho (Fig. 4A). www.nature.com/scientificreports www.nature.com/scientificreports/ Astrocyte changes and dysregulation of AQP4 in BDL rat pups. Using GFAP staining, we demonstrated that BDL rat pups, compared to sham-operated controls, developed time-dependent changes in astrocyte morphology, already observable at 4 weeks post-BDL (Fig. 7). Astrocytes showed an increase of GFAP expression and the

Discussion
This first detailed description of the longitudinal neurometabolic changes induced by experimental biliary cirrhosis in a young rat model, shows that rats having developed cholestatic liver disease as pups display previously unknown molecular changes in addition to the characteristic findings of HE described in adult rats and human subjects [10][11][12][13][14] . In addition, the kinetics of some of the neurometabolic changes are different than those observed in rats having undergone BDL as adults. Taken together these findings suggest that the developing brain displays a unique vulnerability to chronic cholestatic liver disease, possibly offering insight into the long-term neurocognitive changes observed in patients having developed biliary cirrhosis in childhood.
The neurometabolic profile in the hippocampus of rats with experimental biliary cirrhosis induced by BDL at P21 was characterized by a progressive rise in Gln which correlated tightly with plasma NH 4 + during the first four weeks of an 8-week longitudinal observation period. Osmoregulatory decrease of mIns preceded that of Tau, tCho and tCr. Further, a progressive decrease in Asc and Glu correlated negatively with plasma NH 4 + www.nature.com/scientificreports www.nature.com/scientificreports/ concentration. There was a very progressive decrease in PCr throughout the study, which culminated at 8 weeks, concurrent with an increase in Lac and a decrease in GSH concentrations. These observations were probably related to the metabolic consequences of end-stage liver disease, but also suggested brain energy metabolism perturbation, an area still incompletely understood. neurometabolic response to plasma ammonium. Liver disease is characterized by an increase in peripheral NH 4 + load which reaches the CNS, where the only detoxifying pathway is the production of Gln by glutamine synthetase (GS) in astrocytes. Eight weeks post-BDL surgery, young BDL rats experienced a 4.5-times increase of Gln in hippocampus, compared to 2.4-times in adult BDL rats 18 (Figs. S2 and S3). Plasma NH 4 + and bilirubin levels also reached higher values in young vs adult BDL rats. 8 weeks post-BDL, plasma NH 4 + was 75% higher in pups than in adults and plasma bilirubin was 41% higher (Fig. S2), suggesting that rats ligated as pups may be sicker than their adult counterparts, and therefore that variations in neurometabolic profile between pups and adults may be secondary to the severity of liver disease. While disease severity might contribute to higher brain Gln, we suggest that the difference in brain Gln between young and adult BDL rat may be also due to difference in the effect of NH 4 + on maturing metabolic processes. Not only was the correlation between plasma NH 4 + and brain Gln stronger in the young animals (r = 0.81, p < 0.0001) than in adults (r = 0.61, p = 0.004), but the slope of the linear increase between plasma NH 4 + and hippocampal Gln was twice as strong in young animals than adults (Fig. S2). Whether this was due to GS kinetics in the developing brain or BBB maturation are pending questions.
The tight correlation between Gln and NH 4 + held true until week 4 post-BDL, but was lost thereafter (Fig. 3C-E). This was likely due to fact that brain Gln continued to rise between weeks 6 to 8 post-BDL, whereas plasma NH 4 + reached a plateau. This biological decoupling of two tissues is probably explained by different NH 4 + concentrations between blood and CSF/brain. Interestingly, plasma bilirubin and brain Gln were also correlated in this model. Whether there is a genuine link between these two parameters beyond disease severity remains to be determined. Similarly, the negative correlation of distance moved with plasma bilirubin may be interpreted in one of two ways: either this finding points to bilirubin as a neurotoxin or circulating bilirubin levels serve as a surrogate for severity of liver disease. NH 4 + impacts the CNS both through direct toxicity and indirect metabolic pathways 26,27 , something which we have previously shown in BDL rats 18 . This finding is further supported by studies showing that HE disappeared when GS was blocked experimentally during chronic hyperammonemia 28,29 . In the present study, several neurometabolic changes correlated very strongly with plasma NH 4 + . Whether this is due to a direct effect of NH 4 + on a young brain or to a rapid succession of molecular events in the CNS downstream of elevated plasma NH 4 + remains to be determined. www.nature.com/scientificreports www.nature.com/scientificreports/ Response to osmotic stress. As expected, we observed an increase in hippocampal Gln concentration and a decrease in mIns and tCho. Pups displayed the first significant alterations in Gln and mIns at week 2 post-ligation while in adult BDL rats the increase in Gln was significant only from week 4 post-BDL and the mIns decrease from week 6 (Fig. S3).
The early and very profound decrease of mIns (almost −60% at week 8 compared to only −30% in adult BDL rats 18 ) in BDL rats having undergone BDL at P21, may have far reaching implications considering the pleiotropic role of mIns in neurometabolism and development. mIns is central to phospholipid metabolism, membrane formation and intracellular signaling during brain growth 30 . The significance of the observed decrease in the present model is highlighted by the physiological increase of mIns in the hippocampus of sham rats during this developmental window.
The observed decrease in tCho may also be interpreted as an intracellular osmoregulatory response. Choline is required for membrane phospholipid synthesis and myelination 31 , and as a precursor of acetylcholine is essential for cognition and for normal memory development 32 . Therefore, in the context of brain development, the observed −60% decrease 8 weeks post-BDL in the hippocampus of young rats, compared to only −24% in hippocampus of adult BDL rats 18 may have far-reaching neurocognitive consequences (Fig. S3).
Likewise, Tau decreased sooner in rats with CLD acquired as pups than as adults (4 weeks post-BDL compared to 6 weeks in adults) and the absolute decrease at week 8 was −17% compared to −10% in adults 18 . In the CNS, Tau is present in both glial cells and neurons 33 . Therefore, it is not clear whether its decrease is linked to astrocytic osmoregulation or to neuronal alterations. In the present study, Tau was the only osmolyte correlating more significantly with plasma NH 4 + than with brain Gln. Further, its concentration decreased 2 weeks after the mIns decrease, known to occur in astrocytes, suggesting that in HE Tau may play a role beyond osmoregulation in astrocytes. Given the antioxidant properties of Tau, this relative Tau deficiency may further contribute to cellular stress 34 , adding to the list of metabolically important molecules that are significantly decreased in the hippocampus of rats having acquired CLD early in life.
Despite molecular evidence of ongoing osmoregulation, astrocytes displayed a decreased number and shortening of processes, and a thickening of proximal processes at 4 weeks post-BDL, akin to what was observed in adult rats 18 . This was paralleled by an increase of AQP4 expression, a transmembrane channel regulating water homeostasis 35 , normally expressed in the astrocytic feet and in microcapillary endothelial cells around the BBB, as previously shown in culture and experimental animals 36 . This suggests that BDL-induced osmotic stress and increased AQP4 may contribute to morphological changes in astrocytes, something recently shown in adult BDL rats 18 . The pathophysiology and functional implications of the observed changes in astrocyte morphology are still incompletely understood, given that these changes have also been shown in hyperammonemic rats without CLD and may differ according to brain region [37][38][39][40] . Although the neurometabolism of rats having undergone BDL at day 21 differs from that of adults, our data suggest that the morphological changes are rather similar 18 .
Interestingly we show that GFAP is increased in the hippocampus of pups at 4 weeks post BDL. This may appear in contradiction with previous work showing decreased GFAP in type C HE, particularly in adults and in late stages of the disease 41 . However, one might rather look at these results as a continuum. Indeed, we recently showed that there is an increase of GFAP signal at 4 weeks post BDL in adult rats, followed by a decrease at the late stages of the disease (8 weeks post BDL) 18 . Taken together, our findings suggest that both in the pup (this study) and in adult BDL models 18 , astrocytes may first be highly reactive as illustrated by increased GFAP, increased size of proximal processes, and decreased number and retraction of distal processes. We argue that this phase likely precedes the characteristic decrease in GFAP, suggestive of major cellular changes associated with advanced disease, but these kinetics warrant further exploration. energy metabolism. Cr, PCr and tCr were significantly decreased in experimental animals compared to sham-operated animals 4 weeks post-BDL. The observed decrease in tCr of −20% 8 weeks post-BDL was greater than that observed in adult BDL rats (−8%) 18 . The significance of this finding lies in the documented association between a 15% decrease in tCr and impairment of axonal growth in organotypic brain cell cultures in development exposed to NH4+ 6,42 . In our in vivo experiments, we observed a decrease of Cr starting early in the disease. While osmotic regulation may contribute to this decrease, an inhibition of brain Cr synthesis may also be involved, in particular at later stages of the disease, as arginine:glycine amidinotransferase (AGAT), the first of the two enzymes involved in Cr synthesis, is inhibited in organotypic brain cell cultures in development exposed to NH4+ 42 . The 20% deficit in Cr observed over 8 weeks may thus be reached through a combination of the natural degradation of Cr (nearing 2% per day), an inhibition of brain Cr synthesis, and osmotic regulation. Regardless of the underlying mechanism, this decrease in tCr at a time of axonal growth may seriously impact brain development and induce energy metabolism disturbances.
Lac concentration peaked at week 8, something not observed in adult BDL rats (Fig. S3). Although this may reflect ongoing energy dysfunction, its increase more likely reflects the intense metabolic, inflammatory and oxidative stress at the end of the disease course. The most likely interpretation is that this Lac peak is the signature of a terminal, neurometabolic storm rather than a novel molecular mechanism in HE. However, more detailed investigation of energy metabolism in this developing brain model of type C HE is warranted.
Response to oxidative stress. Oxidative stress is known to play a role in type C HE pathophysiology.
Asc and GSH are the most concentrated non-enzymatic antioxidants in the brain. They have slightly different but complementary roles. Asc is localized preferentially in neurons and GSH in glial cells 43 . In our model, Asc concentration decreased first, and was significantly reduced from week 4 post-BDL. The observed decrease correlated with plasma NH 4 + elevation, suggesting a direct toxicity via NH 4 + -induced reactive oxygen species production 44,45 . As liver disease progresses, decreased synthesis by the liver may contribute to declining Asc concentrations in the brain, but this remains to be studied in detail. The previously described decrease in Tau  www.nature.com/scientificreports www.nature.com/scientificreports/ tCr concentrations, two molecules also known for their antioxidant properties, correlated with Asc decrease, suggesting that they may also contribute to the response to oxidative stress in this model 34,46 .
The role of Asc in the brain extends beyond its antioxidant properties to include neuroprotective and neuromodulatory functions, for example against Glu 47,48 , all of which could be significantly impacted by a −30% decrease at week 8. Akin to all the other quantified metabolites, changes in Asc concentration also displayed a significantly stronger decrease in young vs adult BDL rats (−13% decrease at week 8 18 ). In contrast to Asc, GSH decreased at week 8 post-BDL in young rats, something not observed in adult rats 18 . GSH synthesis is dependent on precursor availability: cysteine and Glu 49 . Therefore, in advanced stages of HE, its synthesis could be impaired due to lack of available Glu. It is also known that GSH is consumed in mitochondrial dysfunction and metabolic stress such as might be observed in end-stage liver disease at week 8. Elevated Lac concentrations in the hippocampus of young rats and its correlation with GSH between weeks 6-8 suggests that this mechanism might be involved in our model. neurotransmitters. The experimental advantages of high field 1 H-MRS and high quality shimming afforded us the opportunity to separate Gln and Glu resonances in the spectrum and to measure their concentrations individually. The decrease in Glu appeared slightly earlier in young BDL rats than in adults (4 weeks post-BDL compared to 6 weeks in adults), and at 8 weeks post-BDL the overall change was −18% in young BDL compared to −13% in adults 18 (Fig. S3). Decreased Glu concentration in HE may occur for several reasons. One possibility is its increased use for astrocytic Gln synthesis from NH 4 + . Unlike in adult BDL rats, the correlation of plasma NH 4 + with brain Glu was stronger than that between brain Gln and Glu, suggesting a direct effect of plasma NH 4 + on Glu decrease. This may be secondary to an immature glutamatergic system in the developing brain, or changes in glutamate transporter function or number. Finally, the correlation of Glu decrease with Tau and tCr suggests its consumption may also be related to energy metabolism or neuroinflammation 50 . The most likely, however, is that the origin of the observed decrease in Glu hippocampal concentration is multifactorial. The decrease in hippocampal GABA concentration observed over the course of the study seems to be a specific feature of the developing brain under BDL conditions, as no such change was observed in adult BDL rats 18 . These findings need to be interpreted with caution for two reasons. First, this decrease was observed at the very late stages of the disease, possibly reflecting terminal changes. Second, this may be secondary to the significant decrease in Glu, the main GABA precursor, observed in pups. The aforementioned Tau decrease may also play a role in the regulation of GABAergic neurotransmission, given that it is a GABA A receptor agonist 51 . These in vivo findings are in keeping with changes in GABAergic neurotransmission changes reported in the adult brain with type C HE 52,53 . However, they contrast with reports in developing brain cell aggregates which displayed no change in GABA-ergic neurotransmission under NH 4 + treatment 6 . How to reconcile these findings deserves further in vivo studies in young animals.
To conclude, we demonstrated for the first time longitudinal changes in a model of CLD and type C HE in young rats during late rat brain development between post-natal days 21 to 77 (corresponding to human brain development from 8 months-8 years) and compared them to adult rats with the same disease. Using in vivo, longitudinal measurements of more than 15 brain metabolites, we observed a rapid increase in hippocampal Gln correlated with plasma NH 4 + . This was followed by a decrease in other osmolytes, with evidence of oxidative and metabolic stress, as well as perturbations in neurotransmitters and, importantly, a decrease in tCr. Neurometabolic changes were accompanied by changes in astrocyte morphology and increased AQP4 expression, observed already 4 weeks post-BDL and in keeping with findings in adult BDL animal.
Taken together, CLD-induced HE in the developing brain appears to be a multifactorial disease akin to what is accepted in adult models. Many of the changes are similar in animals having undergone BDL as pups or adults. There were however several striking differences: (1) all neurometabolic changes were more pronounced in the developing brain, (2) some changes appeared earlier (Gln, osmolytes mIns, tCho, neurotransmitter Glu) (3) significant changes in Asp, GABA, Lac, PCr, and GSH appeared to be unique to the young brain. Furthermore, most of the metabolic changes in the developing brain correlated better with rising plasma NH 4 + than with increasing brain Gln, suggesting a direct effect of NH 4 + on metabolism. Therefore, we conclude that the developing hippocampus does indeed display increased vulnerability to the metabolic insults of CLD compared to what is observed in adult animals. Whether this paradigm extends to other brain areas, how they impact short-term and long-term function of these respective areas, and whether the developing brain is equally vulnerable during other developmental windows remains to be elucidated.

Data availability
The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request.