Probiotics combined with rifaximin influence the neurometabolic changes in a rat model of type C HE

Type C hepatic encephalopathy (HE) is a neuropsychiatric disease caused by chronic liver disease. Management of type C HE remains an important challenge because treatment options are limited. Both the antibiotic rifaximin and probiotics have been reported to reduce the symptoms of HE, but longitudinal studies assessing their effects on brain metabolism are lacking and the molecular mechanisms underpinning their effects are not fully understood. Therefore, we evaluated in detail the effects of these different treatments on the neurometabolic changes associated with type C HE using a multimodal approach including ultra-high field in vivo 1H MRS. We analyzed longitudinally the effect of rifaximin alone or in combination with the probiotic Vivomixx on the brain metabolic profile in the hippocampus and cerebellum of bile duct ligated (BDL) rats, an established model of type C HE. Overall, while rifaximin alone appeared to induce no significant effect on the neurometabolic profile of BDL rats, its association with the probiotic resulted in more attenuated neurometabolic alterations in BDL rats followed longitudinally (i.e. a smaller increase in Gln and milder decrease in Glu and Cr levels). Given that both rifaximin and some probiotics are used in the treatment of HE, the implications of these findings may be clinically relevant.

www.nature.com/scientificreports/ inflammatory signals in one of three ways: either by promoting anti-inflammatory cytokines such as IL-10 11 , by reducing pro-inflammatory mediators (IL-6, TNF-α) or by reducing systemic endotoxin levels 12 , none of these clearly contributing to lowering ammonia absorption from the gastrointestinal tract. Nonetheless, rifaximin has been shown to improve cognition in patients with mHE and to prevent hospitalizations and the recurrence of overt HE episodes if combined with lactulose 9,13 . However, the combined administration of rifaximin with lactulose in many of these studies is an important limitation to assess the efficacy of the rifaximin alone 14 .
Animal models of type C HE are necessary to improve the understanding of the effects of these treatments. Very few studies evaluating the effect of rifaximin and/or probiotics have been performed in animal models of type C HE 15,16 . One recent study suggested that rifaximin (50 mg/kg/day) did not reduce inflammation (TNF-α levels) and fibrosis in bile duct ligated rats (BDL) 17 , but there is a lack of knowledge about how rifaximin administration impacts changes in brain metabolites, which are frequently present in HE. Therefore, further studies are needed to characterize the potential role and mechanisms of action of rifaximin in the setting of type C HE 18 .
Probiotics modulate the gut bacterial flora by altering the microbial population, thereby decreasing the production of gut derived bacterial toxins 19 and promoting non-ammoniogenic bacteria. A probiotic formulation containing a mixture of eight strains (Vivomixx in EU, Visbiome in USA) has been recently shown to improve cognitive function and inflammatory response in patients with cirrhosis 20 . This probiotic formulation has also been associated with significant improvement in minimal HE symptoms in humans, decreasing hospitalization rates and preventing encephalopathy episodes in patients with cirrhosis 21 . While the effect of probiotics was often similar to lactulose, they were better tolerated clinically 22,23 . Further, it is important to emphasize that there is significant variability among published studies in the strains of probiotics, daily dose or the duration of the administration of the probiotics, all of which potentially influencing treatment efficacy.
To date, there are very few reports in animal models of type C HE using probiotics. D'Mello et al. found that probiotics (formulation sold as VSL#3 until 2016, but now exclusively available under the brands Vivomixx and Visbiome) improved inflammation-associated sickness behavior in a mouse model of liver inflammation 24 , while Lactobacillus rhamnosus GG reduced liver fibrosis and hepatic gene expression of IL-6 in BDL rats 25 . Our group has recently shown that the administration of the probiotic Vivomixx had a beneficial effect on the development of HE in BDL rats. Both the hippocampal neurometabolic profile and the performance in behavioral tests were improved in BDL + probiotic rats compared to the BDL group at week 8 post-BDL 16 .
The effects of both the antibiotic rifaximin and probiotics have been evaluated through neurological testing in humans. They are currently used to reduce the symptoms of HE in some clinical settings, but longitudinal studies assessing their effects on brain metabolism are lacking, and the molecular mechanisms underpinning their effects are not fully understood 26 . Therefore, we aimed to use the advantages of highly resolved 1 H MRS to analyze longitudinally, in an established model of type C HE, the effect of the antibiotic rifaximin alone or in combination with the probiotic Vivomixx on the neurometabolic profile of BDL rats. We focused on the hippocampus and cerebellum, as they are key regions involved in HE 27,28 . We hypothesized that a lesser increase in brain glutamine would be observed in the treated animals compared to the non-treated ones, followed by a milder decrease in osmolytes and neurotransmitters. We also expected that probiotics and rifaximin would act synergistically to induce better outcomes than rifaximin alone. The present study evaluates for the first time the effect the antibiotic rifaximin and its use associated with a specific probiotic formulation on the neurometabolic changes both in vivo and longitudinally in a rat model of type C HE.

Material and methods
Study design. Adult Wistar male rats (150-175 g, Charles River Laboratories, L' Arbresle, France) underwent bile duct ligation (BDL), a model of chronic liver disease-induced type C HE 29 . Two groups of BDL rats were used. The first group was treated with rifaximin at human pharmacological doses ('rifaximin' , n = 12, 15.7 mg/ kg/day). It was administered orally twice daily in the morning (8 am) and in the evening (6 pm) starting 2 weeks after BDL surgery ('week 2'). The second group was treated with the same dose of rifaximin combined with the multi-strain probiotics mixture Vivomixx (n = 9, 'rifaximin + probiotics' , Vivomixx in EU, Visbiome in USA, 60 billion bacteria/kg of rat). Vivomixx contains 8 lyophilized, highly viable bacterial strains: 4 lactobacilli (Lactobacillus acidophilus DSM24735, L. plantarum DSM24730, L. paracasei DSM24733, L. bulgaricus DSM24734), 3 bifidobacteria (Bifidobacterium infantis DSM24737, B. longum DSM24736, B. breve DSM24732) and Streptococcus thermophilus DSM24731. Probiotics administration started 2 weeks before BDL surgery until the end of the study. In this group, rifaximin was administrated at 8 am and 3 pm, and Vivomixx daily at 7 pm.
Starting from week 0 (before BDL surgery) and every 2 weeks thereafter, all rats underwent 1 H MRS scan, blood sampling (sublingually), and feces collection for microbiology studies. Behavioral tests (open field test) were performed to evaluate motor activity as previously described 4 . During the MR experiments and blood sampling (in the sublingual vein), the animals were kept under 1.5-2% isoflurane anesthesia in a mixture of 50% air and 50% oxygen, with respiratory rate maintained at 60-70 breaths/min and body temperature at 37.5-38.5 °C. The study design is summarized in Table 1.
Each animal served as its own control for all measurements. The treated groups were compared to one of our previously published groups of BDL non-treated rats (n = 29) and sham-operated rats (n = 18) 4 for comparison of open field test and 1 H MRS measurements in the hippocampus region. In addition, another group of our BDL non-treated rats (n = 15) 30 was used for comparison of 1 H MRS measurements in the cerebellum region. 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). The present study is reported in accordance with ARRIVE guidelines.  28 and in cerebellum in a VOI of 2.5 × 2.5 × 2.5 mm 3 , for its important role in motor control 27 which is affected in type C HE 32 , using the SPECIAL 4,31 sequence (TE = 2.8 ms, TR = 4000 ms, 160 averages). Metabolite concentrations were estimated using LCModel combined with an in vitro metabolite basis set of metabolites and the spectrum of macromolecules measured in vivo 33 . Unsuppressed water signal from the same VOI was used as internal reference for absolute metabolites quantification. The ultra-high magnetic field and ultra-short TE allow to detect and quantify in vivo 18 metabolites involved in osmoregulation, neurotransmission, energy or antioxidant metabolism: alanine (Ala), ascorbate (Asc), aspartate (Asp), glycerophosphocholine (GPC), phosphocholine (PCho), creatine (Cr), phosphocreatine (PCr), γ-aminobutyric acid (GABA), glucose (Glc), glutamine (Gln), glutamate (Glu), glutathione (GSH), myo-inositol (Ins), lactate (Lac), N-acetylaspartate (NAA), N-acetylaspartylglutamate (NAAG), phosphoethanolamine (PE), and taurine (Tau). Cramer-Rao lower bounds (CRLB) calculated from the LCModel were used as a reliability measure for the metabolite concentration estimate. Only metabolites with CRLB lower than 30% were included in the results.

Bifidobacteria measurements.
Feces were collected at week 0, 4 and 8 to measure bifidobacteria concentration. Measurements were performed as described previously by Mastromarino et al. 34 .

Statistical analysis.
All results are presented as mean ± SD and % of increase/decrease compared to week 0.
One-way ANOVA (Prism 5.03, Graphpad, La Jolla CA USA) with respect to each metabolite in the neurochemical profile followed by Bonferroni's multi-comparisons post-test was used within a single group. Two-way ANOVA (Prism 5.03, Graphpad, La Jolla CA USA) followed by the Bonferroni's multi comparisons post-test was used to assess significance (p < 0.05) in bifidobacteria, distance moved, brain and plasma metabolite's changes between the groups ('treatment' factor). As variations in metabolite concentrations between groups may appear as early as week 0, our analysis is based on the results of the two-way ANOVA on metabolites changes both in absolute value and relative to week 0. All tests were 2-tailed. In addition, for metabolites with a significant difference between groups ('treatment' factor) both in absolute concentration and relative to week 0, additional two-way ANOVA tests were performed by comparing the 'non-treated' group and each of these two groups individually. This was done to eliminate the potential influence of the differences between the 'rifaximin' group and the 'rifaximin + probiotics' group.
Pearson correlation analysis was performed on longitudinally acquired data to test for correlations between brain metabolites and plasma values.

Results
First, the biological and clinical features of the BDL rats were characterized using plasma bilirubin levels. Bilirubin levels were increased in the three groups of rats (Fig. 1A), without significant difference between groups. All rats displayed a similar ammonium increase regardless of the group (Fig. 1B). During the open field test, the distance traveled decreased from week 4 to week 8 in all groups of BDL rats and was significantly shorter than the sham-operated animals. No significant differences were observed between the three BDL groups throughout of the study (Fig. 1D). Table 1. Longitudinal study design. The number of rats (n) measured in every group are indicated for each week and each type of measures: 'H' stands for MRS scan in the hippocampus and 'C' for MRS scan in the cerebellum, 'B' for the rats whose bifidobacteria in the feces were analysed, and 'O-F' for the rats who undergone open field test. Of note, for blood sampling we used the same number of rats as for the MRS scan (data not included in the table for readability).
Week-2 www.nature.com/scientificreports/ There were significant differences in the longitudinal evolution of the amount of bifidobacteria in feces between the three groups (two-way ANOVA, 'treatment' factor, p = 0.03, Fig. 1C). The difference was particularly pronounced at week 8 where the amount of bifidobacteria in the 'rifaximin + probiotics' group was on average twelve times higher compared to the 'non-treated' group (one-way ANOVA, p = 0.0106).
We further noted that the general condition of the rats in the 'rifaximin + probiotics' group was overall better than in the 'rifaximin' group, as reflected by the fact that only one rat reached the humane endpoint (euthanasia) in this group before the 8 weeks of the study compared to four rats in the 'rifaximin' group (Fig. 1E). Of note, the 'non-treated' BDL rats come from a previously published study 4 where no such assessment was performed since some of the rats were sacrificed before reaching the humane endpoint for additional measures.
To assess the neurometabolic consequences of BDL we performed in vivo 1 H magnetic resonance spectroscopy ( 1 H MRS) at 9.4 Tesla. The high quality of the spectra throughout the study allowed the separation of Gln from Glu and thus quantification of 16 brain metabolites (Fig. 2). All the groups exhibited a significant increase of brain Gln due to ammonia detoxification, and a subsequent reduction of other brain osmolytes (Ins, Tau, tCho), as well as antioxidants and neurotransmitters.
Brain Gln significantly increased in all three groups both in the hippocampus and the cerebellum (Fig. 3). However, 1 H MRS revealed significant differences in longitudinal Gln changes between the 'rifaximin + probiotics' , the 'rifaximin' and the 'non-treated' rats: Gln increase was significantly lower in the 'rifaximin + probiotics' group at week 2, 4, 6, and 8 in the hippocampus compared to the 'non-treated' group, both in absolute value and relative to week 0 (+ 99% in the 'rifaximin + probiotics' group vs + 136% in the 'non treated' group at week 8, Fig. 3A  Percentage of BDL rats at each week in the 'rifaximin + probiotics' group and in the 'rifaximin' group. All results are presented as mean ± SD. One-way ANOVA was used to assess the significance intra-group (A, B) and twoway ANOVA to assess significance between-groups ('treatment' factor; C, D). Statistical significance: *p < 0.05; **p < 0.01; ***p < 0.001.  Table 2). Similarly, in the cerebellum region, the 'rifaximin + probiotics' rats showed a lower rise in brain Gln at early stage of HE compared to 'non-treated' rats (+ 26% in the 'rifaximin + probiotics' group vs + 66% in the 'non treated' group at week 4, Fig. 3B) with the differences observed between groups being significant both in absolute value and relative to week 0 (Table 2). However, no significant differences were observed between 'rifaximin' rats and the 'non-treated' rats, even though the % increase in Gln was higher in the 'rifaximin' group.  Statistical significance intra-group reported: one-way ANOVA (*p < 0.05; **p < 0.01; ***p < 0.001). The relative increases shown are relative to week 0. (C) Correlation between brain Gln in the hippocampus and plasma ammonium in rats treated with the combination of rifaximin and probiotics. www.nature.com/scientificreports/ Furthermore, brain Gln in the hippocampus correlated positively with blood ammonium in the 'rifaximin + probiotics' group (r = 0.85, p < 10 −3 , Fig. 3C), while the correlation did not reach significance in the 'rifaximin' group (r = 0.27, p = 0.056). The correlations between cerebellar Gln and plasma ammonium were similar to those observed in the hippocampus (r = 0.83, p < 10 −3 in the 'rifaximin + probiotics' group; r = 0.26, p = 0.081 in the 'rifaximin' group). Interestingly, in both treated groups the distance moved in the open field test correlated negatively with Gln in the cerebellum, an important region implicated in motor control (r = − 0.51, p = 0.009 in the 'rifaximin + probiotics' group; r = − 0.65, p = 0.0001 in the 'rifaximin' group).

Scientific
The decrease of the osmolyte Ins was significant in all groups in both hippocampus and in cerebellum. It was less marked in the 'rifaximin + probiotics' group compared to the other groups at week 6 and at week 8, both in absolute value and relative to week 0 (− 10% in the 'rifaximin + probiotics' group vs − 15% in the 'non treated' group at week 6, Fig. 4).
Among the other brain organic osmolytes, both Tau and tCho concentrations decreased significantly in all three groups in both the hippocampus and cerebellum, with no significant difference between groups (Figs. 4 and 5).
Cr, a metabolite known for its role in energy metabolism and reported to participate in osmoregulation and neuroprotection 35 showed a decrease in the hippocampus in the three groups without any significant difference between groups. However, in the cerebellum, Cr concentration was stable over time in the 'probiotics + rifaximin' group (− 4% at week 8, Fig. 5), while its decrease was stronger in the two other groups (− 11% in the 'rifaximin' group vs − 15% in the 'non-treated' group at week 8, Fig. 5). These differences between groups were significant both in absolute and relative values ( Fig. 5 and Table 2). Although not reaching significance, there was a trend toward an attenuated reduction of tCr in the 'rifaximin + probiotics' group compared to the 'non-treated' group both in the hippocampus (− 2% vs − 6% at week 6) and in cerebellum (− 4% vs − 10% at week 6).
Among the neurotransmitter-related metabolites, Glu decrease was less marked in the 'rifaximin + probiotics' group and in the 'rifaximin' groups compared to the 'non-treated' group in both hippocampus (− 4% in the 'rifaximin + probiotics' group vs − 8% in the 'non-treated' group at week 6, Fig. 4) and cerebellum (− 2% in the 'rifaximin + probiotics' group vs − 15% in the 'non-treated' group at week 6, Fig. 5). Interestingly, these differences between groups were significant only in the cerebellum ( Table 2). Asp decreased similarly between groups without reaching significance (Fig. 4).
When considering antioxidants measurable by 1 H MRS, the Asc decrease was slightly less marked in the two treated groups while no significant changes were observed for GSH.
Lac exhibited a slight tendency to increase its concentration in the three groups in the hippocampus, but none of these changes reached significance. In the cerebellum, however, the Lac increase was more substantial and significant at week 8 in all groups (+ 80%, + 109%, + 92% at week 8 in the 'non-treated' , 'rifaximin + probiotics' , 'rifaximin' groups respectively) but did not significantly vary between the groups (Figs. 4, 5 and Table 2). GABA, PE, NAA, NAAG nor tNAA did not change significantly during the course of the study in any of the groups or brain region.
We also measured the correlations between Gln, for its central role in the development of HE, and other neurometabolites to assess inter-metabolite dependencies in treated BDL rats. Overall, with each of the two treatments, the hippocampal Gln increase was strongly correlated with the decrease in Ins (p < 0.001 for both groups) and in tCho (p < 0.001 for both groups). We observed a negative correlation between Gln and Glu, which was more pronounced in the 'rifaximin + probiotics' group (r = − 0.36, p = 0.017) compared to the 'rifaximin' group (r = − 0.29, p = 0.027). In the cerebellum region, Gln also correlated negatively with Ins in both treated groups (p < 0.001 in both groups) and with tCho (p < 0.01 in both groups). Gln correlated negatively with Glu in the 'probiotics + rifaximin' group only (p < 0.01). Unlike the hippocampal region, a strong correlation was measured with Tau (p < 0.001 in both groups). Table 2. Summary of statistical significance of the two-way ANOVA for 'treatment' factor (i.e. 'rifaximin' , 'rifaximin + probiotics' , 'non-treated') for each metabolites whose concentrations significantly varied over time, both in absolute concentrations and relative to week 0 in hippocampus and cerebellum. Metabolites which significantly differ between the groups both in absolute and relative values are highlighted in red. Of note, the interaction was not significant for any of the metabolites. The relative increase reported are relative to week 0. Statistical significance: *p < 0.05; **p < 0.01; ***p < 0.001. www.nature.com/scientificreports/ Furthermore, for metabolites with a significant difference between groups ('treatment' factor) both in absolute concentration and relative to week 0 (Gln, Cr, Glu, Table 2) as well as for bifidobacteria (Fig. 1C), we performed an additional two-way ANOVA test to compare only the 'non-treated' group and the 'rifaximin + probiotics' group. The differences between the groups were all significant: Gln hippocampus, p < 0.001; Gln cerebellum, p < 0.05; Glu cerebellum, p < 0.05; Cr cerebellum, p < 0.01; bifidobacteria, p < 0.05.

Discussion
The present study reports for the first time the beneficial effect of combined rifaximin and the probiotic Vivomixx on the longitudinal neurometabolic changes in vivo in a rat model of type C HE. The characteristic longitudinal increase of Gln and decrease of Glu, known to characterize HE, and the recently reported decreased in Cr, were significantly less pronounced in rats treated with rifaximin and probiotics compared to the two other groups.
In each group, we observed the characteristic features of type C HE previously described using 1 H MRS 4 (Fig. 2). In all three groups, there was a gradual increase of brain Gln followed by a gradual decrease in Ins and other osmolytes, such as Tau and Cr. These observations are in agreement with previously published studies in BDL rats and CLD patients 3,4,36 and confirm the already well-known partial compensatory effect to maintain osmotic balance in the presence of Gln accumulation.
The measured plasma bilirubin values as well as the lack of significant changes in treated animals are in agreement with recently published studies on BDL rats treated with probiotics 16,25 . We observed a significant increase . Longitudinal 1 H MRS data of brain metabolites in the hippocampus region of the 'rifaximin' treated group, the 'rifaximin + probiotics' treated group and the group without treatment. For better readability, relative changes indicated in the graph are shown only at week 6 relative to week 0 and are indicated for metabolites whose concentrations significantly varied over time. Statistical significance: One-way ANOVA (*p < 0.05; **p < 0.01; ***p < 0.001). www.nature.com/scientificreports/ in plasma ammonium in all groups to a level similar to that reported in other studies using BDL rats 4,37 , with a high intra-group variability. No significant differences were observed between groups, despite the administration of rifaximin combined with probiotics or rifaximin alone, whose purported role is to reduce the production of gut ammonium. Of note, the lack of significance in plasma NH 4 + measurements between groups is in agreement with a recently published study on BDL rats 16 . However, the effect of rifaximin on plasma ammonium levels is still disputed, and some studies reported that its use had no significant effect on plasma ammonium levels either in BDL rats 15 or in humans 11 . In contrast, some studies reported a reduction in blood ammonium after treatment 38,39 , but it is now generally considered that despite the irrefutable and central role of ammonium in the pathogenesis of HE, its plasma levels display inter-individual variability and are not always useful to predict the severity of HE or to assess the response to therapy, especially in type C HE 40,41 . Given these aforementioned elements and the challenges of measuring ammonium, additional methods must be combined in a multi-modal approach to accurately assess the effect of different therapeutic options in type C HE, as put forth in the present study.
Gln increase was significantly less pronounced in the 'rifaximin + probiotics' group at weeks 4, 6 and 8, compared to the two other groups, in agreement with a recent study using the probiotic Vivomixx alone on the same animal model 16 . Therefore, the significantly lower decrease of brain Glu in the 'rifaximin + probiotics' group suggests that the reduction in brain Glu is mainly the result of ammonium detoxification through glutamine synthetase, which catalyzes the formation of Gln from ammonium and glutamate in astrocytes 42 . This might also explain the positive correlation between brain Gln in hippocampus and plasma ammonium in the 'rifaximin + probiotics' group. The significant correlation between Gln and Glu (in all of the three groups) is consistent with this hypothesis. In addition, the correlations observed between hippocampal Gln, Ins, Tau and plasma NH 4 + are in agreement with the previously published results in non-treated BDL rats 4 .
While Glu showed a significantly less pronounced decrease in the 'rifaximin + probiotics' group, Asp, a selective NMDA receptor agonist, showed a similar trend of decreasing concentration in all groups, although not significant (Fig. 4). Even if the role of Asp as an excitatory neurotransmitter is still debated, its decrease might be linked to alteration in the neurotransmission system, which is known to play an essential role in the pathophysiology of HE even though the precise mechanisms are not yet fully identified 43 . In addition, no significant changes in the inhibitory neurotransmitter GABA concentrations were observed, consistent with our previous studies in non-treated BDL rats 4 . Figure 5. Longitudinal 1 H MRS data of the brain metabolites in the cerebellum region of the 'rifaximin' treated group, the 'rifaximin + probiotics' treated group and the group without treatment. Relative changes indicated in the graph at week 6 are calculated relative to week 0 and are indicated for metabolites whose concentrations significantly varied over time. Statistical significance: One-way ANOVA (*p < 0.05; **p < 0.01; ***p < 0.001). Note that in the cerebellum many concentrations estimated with LCModel of Asp and Ala were excluded due to their unreliability (CRLB > 30%). As this made difficult to interpret the variations of these metabolites, their evolutions are not shown. GSH at week 8 is not shown for the same reason. www.nature.com/scientificreports/ An attenuated decrease of Cr, tCr and Ins was measured in the 'rifaximin + probiotics' group, consistent with previous reports 16 . These smaller variations of Cr and tCr, metabolites also involved in energy metabolism, suggest a lower osmotic imbalance and possibly improvement of the energy metabolism in the 'rifaximin + probiotics' group.
Regarding microbiota, rifaximin alone did not significantly alter bifidobacteria concentration in feces. This is in keeping with two previously reported observations: first, that no significant change in microbial composition after rifaximin administration in humans was observed 44 , and second, that rifaximin may exert a beneficial modulation of the gut microbiota metabolic profile rather than a significant rearrangement of the intestinal microbial community 45 . The probiotics used in our study resulted in a significantly higher bifidobacteria concentration at week 4 and week 8 in the intestines of rats treated with rifaximin combined with Vivomixx compared to the two other groups. These modifications may have decreased the production and absorption of ammonium, as reported in various studies in humans 23,46 and might explain the lower brain Gln measured in this group. These results are consistent with a previous preliminary study showing an increase of bifidobacteria in gut microbiota in BDL rats treated with the probiotic Vivomixx only 16 .
The present study has several limitations. First, stool bifidobacteria concentrations were assessed at only 3 time points (weeks 0, 4 and 8) using a limited number of samples. Moreover, alterations in gut microbiota composition and function in CLD contribute to the systemic inflammation and accumulation of gut-derived toxins 47 , something for which we could not control in the present study. As both rifaximin and probiotics influence systemic inflammation 18 , the longitudinal assessment of inflammatory markers in the blood (TNF-α, CRP, WBC counts) and markers of neuroinflammation remains to be performed.
It is important to emphasize that different factors influenced our study design: first, various studies suggested that probiotics improve antibiotic therapy as they possess immunomodulatory properties, promote the recovery of commensal microbiota and increase treatment tolerability 48 . To date, the evidence is unclear as to the best time to administer probiotics. For now, patients with type C HE are advised to take probiotics before, at the same time as, or even after taking antibiotics. Second, one purpose of the present study was to compare two similar groups. Therefore, modulation of the gut flora started 2 weeks before BDL surgery in the 'rifaximin + probiotics' group and rifaximin was administered at the same dosage per day in both treated groups ('rifaximin' and 'rifaximin + probiotics'). In the 'rifaximin + probiotics' group, drug administration was timed carefully (alternating times for rifaximin and probiotics) in order to minimize any direct interaction in the gastrointestinal tract between the two treatments. These strategies used in our experimental design may have affected the magnitude of the results.
Overall, while rifaximin alone at this specific dose appeared to induce no significant effect on the neurometabolic profile of treated-compared to non-treated BDL rats, its association with probiotics resulted in more attenuated neurometabolic alterations in BDL rats followed longitudinally. Our results on rifaximin alone are consistent with a recent study that explored the effects of rifaximin in BDL rats, where the authors concluded that rifaximin at this dose was not efficacious in the treatment of HE 49 . We thus hypothesize that the positive effects observed are mainly due to indirect effects of this specific probiotic formulation, namely modulation of the intestinal barrier, anti-inflammatory properties, and decreased production of bacterial toxins 19 . Our results are in keeping with our recent study showing a beneficial effect of the probiotic Vivomixx alone in BDL rats 16 .

Conclusion
We conclude that in BDL rats rifaximin treatment combined with the probiotic Vivomixx can be associated with positive effects on brain Gln, Glu and Cr levels in a model of type C HE. Given that both rifaximin and some probiotics are used in the treatment of HE, the clinical implications of these findings may be far-reaching.