Brain metabolic and microstructural alterations associated with hepatitis C virus infection, autoimmune hepatitis and primary biliary cholangitis

Neuropsychiatric symptoms in hepatitis C (HCV) patients resemble those of patients with autoimmune hepatitis (AIH) or primary biliary cholangitis (PBC), whilst the mechanisms behind them are unknown. Here we looked for cerebral metabolic and/or microstructural alterations in patients with HCV, AIH or PBC as possible causes behind these symptoms.


| INTRODUC TI ON
Hepatitis C virus (HCV) infection is known not only to induce chronic progressive liver disease but also neuropsychiatric symptoms, such as chronic fatigue, mood alterations and cognitive dysfunction, which occur in about 50% of HCV-infected patients. [1][2][3] These symptoms are independent of the grade of liver disease, PCR-status and virus replication rate. 4 In concordance with these clinical observations different research groups detected negativestrand HCV-RNA in human brain samples and in cerebrospinal fluid of HCV-infected patients, which indicated that HCV is able to surpass the blood-brain barrier and to replicate in human brain cells. [5][6][7][8][9] HCV replication has been especially shown in microglia and astrocytes. 9 However, the mechanisms behind brain dysfunction and neuropsychiatric symptoms in HCV-afflicted patients have not yet been clarified. Of note, similar neuropsychiatric symptoms were also observed in patients with other liver diseases. Two frequent autoimmune liver diseases-autoimmune hepatitis (AIH) and primary biliary cholangitis (PBC)-have often been linked with fatigue, and patients with PBC also showed significantly impaired cognitive function compared to controls. 10,11 In contrast to HCV infection, the hepatitis B virus (HBV) infection was not linked with fatigue, suggesting that a liver disease per se does not necessarily provoke fatigue. 12 Previous brain magnetic resonance imaging (MRI) studies in HCV patients with only mild if any liver dysfunction compared to healthy controls showed an alteration of brain metabolism, neurotransmission and connectivity in HCV patients. 13,14 Quantitative magnetic resonance measurements reportedly can detect pathological or physiological microstructural alterations that remain undetected in conventional MRI. 15,16 Therefore, aiming to achieve further insight into the pathophysiology of brain dysfunction in these liver diseases, we conducted this single-centre observational study to investigate brain functional, metabolic and microstructural alterations in non-cirrhotic patients with HCV, AIH and PBC as well as a sample of healthy subjects as a reference. Therefore, combined laboratory tests on blood cytokine levels and immune cell phenotypes in patients, neuropsychometric assessment and quantitative MRI (qMRI) method T2 relaxometry and MR spectroscopy (MRS) were performed. 17 Here we report on the results derived with qMRI and MRS as well as possible associations between abnormal neuropsychological findings and brain alterations in these patients, whilst the detailed results of laboratory tests and neuropsychometric assessments have been reported elsewhere. 17

| Subjects
The subjects included in this study were a subpopulation of patients recruited as previously described in detail. 5 In brief, all patients registered in the database of the local hepatitis outpatient clinic were informed about the study by letter and were invited to take part. Furthermore, patients who attended the hepatitis outpatient clinic were informed by a leaflet and were asked if they were interested to take part in the study. The inclusion criteria were diagnosis of hepatitis C, hepatitis B, AIH or PBC or overlap of AIH and PBC.
The exclusion criteria were as follows: HBV/HCV co-infection, the combination of AIH or PBC and virus hepatitis, other causes of liver disease besides HCV, HBV, AIH or PBC, liver cirrhosis, accompanying neurological or psychiatric diseases or diseases that might affect brain function such as HIV-co-infection, renal dysfunction, alcohol or drug abuse and medication affecting the central nervous system. All patients underwent neurological and neuropsychological assessment and blood draw for the assessment of blood cytokine levels and immune cell phenotypes. Amongst the 88 patients taking part in the whole neurological and neuropsychological study, 70 volunteered to take part in the MRI/MRS study as well. Finally, 47 patients were picked according to their availability at an assigned time slot for the MR examinations: 17 patients with HCV infection (group HCV), 5 with HBV infection (group HBV), 14 with AIH (group AIH, and 11 with PBC (group PBC). Due to the low number of subjects, trend in the HCV group in pWM. Correlation analysis did not reveal significant associations between MRI/MRS alterations and neuropsychological dysfunction.

Conclusion:
The findings suggest similar pathophysiological mechanisms behind neuropsychiatric symptoms associated with HCV infection, AIH and PBC.

K E Y W O R D S
autoimmune hepatitis, hepatitis C virus, magnetic resonance imaging, magnetic resonance spectroscopy, primary biliary cholangitis

Lay summary
Magnetic resonance imaging (MRI) and spectroscopy (MRS) revealed similar brain metabolic and microstructural alterations in patients with HCV infection, AIH and PBC, suggesting autoimmune neuroinflammation as a common pathophysiological mechanism behind the neuropsychiatric symptoms.
the HBV group was excluded from further analysis. Eighteen healthy controls adjusted for age were recruited from the local environment and were studied as a reference group for MRI/ MRS. All subjects gave written informed consent.

| Clinical assessment
As described previously, 5 all patients underwent a clinical assessment consisting of a neurological examination and documentation of the current medication, co-diseases and actual level of liver enzymes. All patients were asked to complete the 36-item Short-Form Health Survey (SF-36), 18

| MRI and data processing
All subjects underwent MRI at a 1.5 T MR system (Avanto, Siemens, Erlangen, Germany) with a standard quadrature head coil. The MRI protocol included -amongst others -a T2-weighted spin-echo se- for estimation of brain metabolites. The SVS was carried out at two voxels, one located in frontal white matter (fWM) and the other in parietal white matter (pWM). Corresponding MR spectra without water suppression were also acquired. An aqueous phantom was also scanned with the Triple TE sequence. Proton density and T2weighted images were inspected by 2 experienced neuroradiologists to exclude subjects with morphological abnormalities.
Brain maps of proton density as well as T2 relaxation time were obtained first on the MR console with an extended image reconstruction provided by the manufacturer, which used monoexponential fitting to the signal-intensity decay curves of the data acquired with Triple echo sequence. The proton density of the phantom (PD H2O ) was also determined in the same manner for normalization of brain proton density measured in subjects. Using region of interest (ROI) analysis with free software ImageJ (National Institutes of Health, Bethesda, Maryland) 26 numeric values of brain proton density and tissue relaxation time T2 were determined from corresponding brain maps by using mean values over 17 ROIs within each hemisphere located in the following brain structures: the frontal and parietal white matter (fWM and pWM), centrum semiovale (CS), splenium of the corpus callosum (sCC), genu of the corpus callosum the lower pole (CblPol) of the cerebellum. Each ROI was drawn manually as a circle with an area of about 28 mm 2 on the corresponding brain maps. All ROIs were carefully chosen according to anatomical landmarks to avoid partial volume effects. For normalization brain proton density measured from each ROI of subjects was presented as a ratio to that of phantom PD H2O (PD, in %). Considering the limitations of the sample size related to the handedness, the left and right PD or T2 values for each ROI were averaged for further analysis. The coefficient of variation (COV = standard deviation/mean value) was used as data quality criteria for measured PD and T2 values, that is the data with a COV of >20% was not sampled for further analysis.
Magnetic resonance spectroscopy data were analyzed offline with the software LCModel to estimate the concentrations of brain metabolites N-acetylaspartate (NAA), choline-containing compounds (Cho), creatine (Cr), glutamine/ glutamate (Glx) and myoinositol (mI). 27 The water-unsuppressed data were used for eddy current correction and for obtaining internal water signal that was used as a reference to estimate metabolite concentration by spectral analysis, that is the concentrations of the metabolites were determined in ratio to the obtained internal water. Cramer-Rao lower bound (CRLB) derived from the spectral analysis was used as data quality criteria for measured metabolite concentrations, that is those with a CRLB of >20% for NAA, Cho and Cr and a CRLB of >30% for Glx and mI were not considered for further analysis.
The study was a priori approved by the local ethics committee and performed according to the World Medical Association Declaration of Helsinki (revised in 2008).

| Statistical analysis
Demographical and clinical data (age, liver enzyme levels, fibrosis-4score [FIB4-score]) and the results of self-report questionnaires as well as memory and attention tests were found to be not normally distributed and thus are presented as median and 25th/75th percentile. The global null hypothesis was tested using the Kruskal-Wallis test. In case of significance, the Mann-Whitney U test was used to look in in-between group comparisons for statistical differences between the groups.
The values of brain PD and T2 relaxation time derived from qMRI measurements and brain metabolite concentrations measured with MRS were compared between the patient groups and the healthy controls to estimate possible brain metabolic and microstructural alterations associated with each disease. Since the independent T2 tests did not reveal significant gender differences for measured PD, T2 and metabolite values in each brain area of the healthy controls (P > .05), the values measured from females and males within each group were combined used in further analyses. After the normal distribution of the data was checked with Shapiro-Wilk test and quantile-quantile plots, the global null hypothesis was tested using one-way analysis of variance (ANOVA). In case of significant effects (global P < .05), a post hoc Dunnett t test was used to look for statistical differences between each patient group and the healthy controls. Possible correlations between altered neuropsychological findings and abnormal MRI/MRS observations were estimated with the Spearman ρ test. All statistical analyses were performed using SPSS version 24.

| Baseline characteristics of patients and controls
Patients and controls did not significantly differ in age. The liver enzyme levels and FIB-4 scores did also not differ significantly between the patient groups. The basic characteristics of all groups are summarized in Table 1. The results of the self-report questionnaires and psychometric test battery of the patient groups and controls are presented in Table   S1A-C. All patient groups showed worse results than the controls regarding fatigue, anxiety and depression scores, attention and memory function as well as health-related quality of life scores. The MRS study cohort achieved similar results as the main study cohort and thus can be considered representative for the whole study cohort (Table S1A-C). 5

| Results of quantitative MR measurements and MRS
Example localizations of the 17 ROIs in the right brain hemisphere selected for qMRI measurements are shown as white filled circles on the proton density maps displayed in Figure 1A and example MR spectra from fWM and pWM, which were obtained from a 49-year-old healthy male volunteer, in Figure 1B.
The group means and the corresponding standard deviations of brain PD measured from each ROI together with the results of the ANOVA analysis and the post hoc Dunnett tests are shown in Table 2. The analysis with ANOVA revealed significant group differences of brain PD values in all selected brain areas but the cerebellar white matter, that is in 16 of 17 measured ROIs (ANOVA P (=0.000-0.014) <.05). Post hoc tests proved that in comparison to healthy controls, the patients showed a significantly decreased PD in a majority of measured brain areas: PBC patients showed decreased PD in all the 16 ROIs, AIH patients in 14/16 ROIs and the HCV group in 7/16 ROIs (Dunnett t test: all P < .05, Table 2). For brain T2 relaxation time, the analysis with ANOVA revealed significant group differences only in 2 of 17 ROIs -in occipital grey matter and in ventral pons (ANOVA: P = .035 and .023, respectively), as shown in Table 3, where the ROIs that did not reveal significant

| Correlation analysis
To look for an association between brain abnormalities and cognition or mood alterations in the patients, correlation analyses were performed both for the 3 patient groups separately as well as for all patients (HCV, AIH, PBC) grouped regarding abnormal psychometric findings and MRI/MRS results for those parameters that had differed between patients and controls.
After correction for multiple comparisons, the analysis showed no significant results.

| D ISCUSS I ON
In this work, we determined changes in brain microstructure and metabolism in non-cirrhotic patients with HCV infection, autoimmune hepatitis or primary biliary cholangitis, compared to a sample of healthy subjects using quantitative MRI (qMRI) for the assessment of relative proton density and T2 relaxation time and proton MR spectroscopy. With qMRI measurements, we found that in comparison to healthy controls, all patients showed significant decreases of brain PD with varying numbers of respective brain regions in the groups. Most brain regions showing a decreased PD were observed in the PBC group (16 ROIs), followed by the AIH group (14 ROIs) and the HCV group (7 ROIs). The HCV patients revealed in addition significantly decreased T2 relaxation times in two ROIs. In previous reports, the changes of brain PD or T2 relaxation times in patients or in healthy ageing humans were interpreted as microstructural changes caused by pathological processes, such as demyelination (increased T2), or increased cellpacking density in morphologically normal-appearing brain areas (increased PD). 15,16 Accordingly, our findings of altered PD and T2 prove brain microstructural alterations in the patients with HCV, AIH and PBC. The observation of similarly altered brain PD though with varied numbers of brain regions affected may indicate similar brain microstructural changes associated with HCV infection, AIH and PBC but with the different extent depending on the diseases.
The altered brain T2 relaxation times in HCV patients may suggest some additional brain microstructural changes related only to HCV infection, but this finding is less significant than the PD results. An increasing number of studies give evidence of altered brain structure and function in HCV patients. 13 Glx, glutamine/glutamate; mI, myo-inositol; NAA, N-acetylaspartate; tCr, total creatine TA B L E 2 Results of ANOVA analysis including post hoc Dunnett t test for brain regional proton densities measured from subject groups with quantitative MRI

TA B L E 2 (Continued)
the rsFC was increased in the patients, which was interpreted as a compensatory response. 36 These results are expanded with the present findings of qMRI measurements, indicating the coexistence of brain alterations and neuropsychiatric symptoms in PBC patients. Of note, PBC patients with visible brain lesions were excluded in our study, and it must be emphasized that the significant brain PD alterations were found in multiple brain regions, including brain white matter, cortical and deep grey matter regions, suggesting diffuse brain microstructural changes in the PBC patients.
To our knowledge, this study for the first time assessed the brain microstructure of AIH patients. Of interest, a decrease of proton density similar to the findings in PBC was also found in patients with AIH in various subcortical and cortical regions, though to a slightly less amount. This is in line with the findings of our previous work, where we were able to show that both non-cirrhotic AIH and PBC patients do worse than healthy controls especially in verbal learning and word memory as well as fatigue and depression scores. 5

TA B L E 3
Results of ANOVA analysis including post hoc Dunnett t test for brain T2 relaxation times measured from subject groups with quantitative MRI. Only results with significant overall P are shown Moreover, we observed that the HCV patients revealed significantly decreased PD in 7 ROIs, where significantly decreased PD was also found in patients with AIH and with PBC, indicating that the brain microstructural changes associated with HCV in part resemble those associated with AIH and PBC. Four of these 7 ROIs were located in the deep grey matter -a region that is affected in patients with HCV encephalopathy before. [28][29][30][31] Using proton MRS, we additionally observed that in comparison to healthy controls in fWM, the concentration of NAA -a marker of number and function of neurons -was significantly increased in patients with HCV, AIH and PBC. This might result from a compensatory mechanism to improve brain function as interpreted by Bokemeyer et al in a previous study. 28 The concentration of choline was significantly increased in PBC patients in both fWM and pWM, in AIH patients in pWM and with a weak trend in patients with HCV.
Former studies in HCV patients had observed an increase in choline in the white matter and some also in the basal ganglia. 28  and indicates cerebral microglia activation, was investigated in HCV patients via PET. 43 Increased binding potential of PK11195 was observed in the caudate, thalamus and putamen, which was positively correlated with HCV viral load. 30,44 Furthermore, gliosis might be one of the reasons for the altered brain microstructure.
Immunohistochemical analysis of frontal cortex samples of patients who died with HIV encephalopathy, HCV infection and controls showed neuronal damage and in addition significant gliosis in both -patients with HIV encephalitis and patients with HCV infection. 45 However, such data are sparse as is the information upon the clinical presentation of these cases. Thus, the interpretation must be done with care.
In this study, we did not find correlations between MR measurements and the neuropsychological assessments. The reason could be that presently used assessment techniques may not be adequately sensitive or specific. Further studies with more specific methods are needed to verify the results.
Some limitations apply to our study: It is limited by the fact that the sample size of each patient group was rather small, which has the impact upon the statistical power. Moreover, the sizes and locations of selected ROIs by qMRI on 17 brain regions and MRS measurements on only two regions were different. Thus, it was not clear, whether the microstructural alterations and the metabolic changes occurred simultaneously in the same brain structures. The inclusion of a group of hepatitis B patients would have been desirable but was not realizable at our site.
In conclusion, using qMRI and MRS, we demonstrated similar brain microstructural and metabolic alterations in non-cirrhotic HCV, AIH and PBC patients, suggesting autoimmune neuroinflammation as a common pathophysiological mechanism underlying neuropsychiatric symptoms in the different diseases. Furthermore, our results support the hypothesis of HCV-induced neuroinflammation suggested by previous studies. However, further studies must be performed in larger groups of patients to prove this hypothesis.

ACK N OWLED G EM ENT
Open access funding enabled and organized by Projekt DEAL.

CO N FLI C T O F I NTE R E S T
None of the authors has to declare any conflict of interest.