Resting State Cortical Electroencephalographic Rhythms in Covert Hepatic Encephalopathy and Alzheimer's Disease

Keywords: Electroencephalography (EEG), Low resolution brain electromagnetic tomography (LORETA), Minimal hepatic encephalopathy (MHE), Covert hepatic encephalopathy (CHE) Mild cognitive impairment (MCI), Alzheimers' disease (AD), hepatic encephalopathy (HE), liver cirrhosis. 2 Abstract Patients suffering from prodromal (i.e. amnesic mild cognitive impairment, aMCI) and overt Alzheimer's disease (AD) show abnormal cortical sources of resting state electroencephalographic (EEG) rhythms. Here we tested the hypothesis that these sources show extensive abnormalities in liver cirrhosis (LC) patients with a cognitive impairment due to covert and diffuse hepatic cortical sources were estimated by LORETA. Widespread sources of theta (all but frontal), alpha 1 (all but occipital), and alpha 2 (parietal, temporal) rhythms were higher in amplitude in all LC patients than in the Nold subjects. In these LC patients, the activity of central, parietal, and temporal theta sources correlated negatively, and parietal and temporal alpha 2 sources correlated positively with an index of global cognitive status. Finally, widespread theta (all but frontal) and alpha 1 (all but occipital) sources showed higher activity in the subgroup of LC patients with CHE than in the patients with aMCI or AD. These results unveiled the larger spatial-frequency abnormalities of the resting state EEG sources in the CHE compared to the AD condition.


Introduction
Hepatic encephalopathy (HE) is a brain dysfunction caused by liver insufficiency and/or portal-systemic shunting.HE produces a wide spectrum of nonspecific neurological and psychiatric manifestations [1].In its subclinical phase (i.e.no clinically-overt cognitive dysfunction), HE causes a kind of minimal cognitive impairment concerning attention, working memory, psychomotor speed and visuo-spatial ability detectable by neuropsychological or psychophysical tests, as well as neurophysiological alterations [2].This condition is called minimal HE (MHE) [3,4,5].Approximately, 30-80% of patients with cirrhosis may show MHE [2,6].MHE is associated with poor quality of life and a high risk of traffic violations and accidents [7,8].It is also associated with an increased progression to overt HE (OHE; i.e., from confusion to coma).
The severity of hepatic encephalopathy is graded with the well known West Haven Criteria; this scale is based on the level of impairment of autonomy, changes in consciousness, intellectual function, behavior, and the dependence on therapy [6].The West Haven Criteria span from grade 1 (i.e. a poorly reproducible stage of HE in which an observer can detect some patient's negligible symptoms, such as trivial lack of awareness, euphoria or anxiety, shortened attention span and impaired performance of addition or subtraction) to grade 4 (i.e.coma as unresponsive to verbal or noxious stimuli) on the basis of a structured clinical evaluation and use of standardized items (e.g.Does the patient know which month it is?Does the patient know the day of the week?Can the patient count backward from 10 to 1 without making mistakes or stopping?Etc.).Recently, the term 'covert HE' (CHE) was coined, pooling MHE with West Haven grade I HE (Mullen and Prakash, 2012).The concept of CHE is more and more used in the literature (Kappus and Bajaj, 2012), and we will use it to name these patients in the following.
Several studies have shown that spectral analysis of electroencephalographic (EEG) rhythms in awake subjects at rest (eyes closed) is a low-cost, easy to perform, and widely available neurophysiological approach for the study of HE [8,9,10,11,12,13,14,15,16,17,18].Indeed, resting state EEG markers are virtually not affected by anxiety for performance, emotional variables, skilfulness, and subjects' social compliance.Furthermore, recording of the resting state EEG rhythms can be repeated countless times in the follow up of patients with liver cirrhosis (LC) to monitor for CHE without the learning effects or the need of cooperation which plague psychometric testing.Finally, resting state EEG rhythms seem to provide -at least at group level-useful markers/end points to evaluate disease progression in subjects with HE.
Previous studies have shown that spectral analysis of resting state EEG rhythms allows a classification of HE subjects in line with clinical severity of the disease [11,13].It also allows the detection of CHE [10,12,19] and some prediction of the progression to OHE and liver-related death, at least in patients with advanced liver disease [8].In particular, mean dominant EEG frequency was significantly "slower" in HE patients compared to healthy controls [16,18].In addition, theta (4-8 Hz) and alpha (8)(9)(10)(11)(12) rhythms showed different magnitude in individuals with even CHE compared to healthy controls [16,17,20].Such changes were associated to a remarkable topographical anteriorization of the maximum power density at theta and alpha rhythms from the occipital towards the parietal and central derivations, thus resembling some neural correlates of drowsiness preceding sleep onset [21,22,23,24].In CHE patients, the anteriorization of the dominant theta or alpha rhythms has been also quantified by analytical techniques such as spatiotemporal decomposition, which separated the main components of the EEG and defined their distribution across the scalp [25].Finally, EEG power density was found to be correlated to psychometric hepatic findings [20,26].These findings complement preceding evidence showing that, at group level, cortical sources of resting state eyes-closed cortical EEG rhythms present topographical and frequency differences in healthy normal elderly subjects (Nold) and patients with cognitive impairment [27,28,29,30].When compared to Nold subjects, subjects with amnesic mild cognitive impairment (aMCI; a sort of prodromal AD condition) showed a power density decrease of posterior alpha sources [27,28,29,30], whereas AD subjects were characterized by a power density increase of topographically widespread delta and theta sources and a power decrease of posterior alpha sources [27,31,32].In addition, posterior alpha sources had lower power density in AD than in cerebrovascular dementia (CVD) and in Parkinson disease with dementia (PDD), whereas topographically widespread theta rhythms were higher in power density in CVD and PDD than AD subjects [27,33].
In the present study, we tested the hypothesis of peculiar abnormalities in the spatial distribution and in the frequency bands of the cortical sources of resting state EEG rhythms in patients with CHE when compared to patients with possible prodromal or overt AD.As a novelty, the present study used a validated procedure for the estimation of these cortical sources (i.e.LORETA) to directly compare the results obtained in CHE and AD populations.It is expected that both CHE and AD pathologies reflect peculiar abnormalities in the abnormal fluctuation of cortical arousal, which are expected to be reflected by resting state EEG rhythms.The EEG features able to capture these peculiar abnormalities would be candidate markers to enlighten the neurophysiological mechanisms of action of the drugs for the two pathologies.This is a promising perspective for the development of new drugs and therapy monitoring in AD and CHE patients.Indeed, the present EEG methodological approach has a sufficient spatial and frequency resolution to unveil narrower EEG abnormalities in spatial and frequency domain in AD than in CHE patients.This prediction is based on the notion that at earlier stages of the clinical disease, MCI and AD patients show neurodegenerative processes circumscribed in few brain regions such as hippocampus, enthorinal paleocortex, and posterior neocortex (During et al., 2011).On the other hand, CHE patients are characterized by effects of blood neurotoxins on widespread brain regions spanning several sub-cortical and cortical structures (Jones and Mullen, 2012).
The main novelty of the present study is that for the first time the same EEG methodological approach was used in two matched populations with similar cognitive deficits but quite different neuropathologies such as AD and CHE.This is the first exploratory comparison of the spatialfrequency patterns of the cortical sources of resting state cortical EEG rhythms in these two pathological conditions affecting human cognition, beyond the same or similar outcome at mere behavioral/cognitive level.

Subjects
We enrolled 64 (28% of females) patients with liver cirrhosis (LC) and 64 (33% of females) cognitively intact age-matched elderly (Nold) subjects as a control group.The mean subjects' age was 58.2 (± 0.9 standard error, SE) years in the LC patients, and 59.5 (±0.8 SE) years in the Nold subjects.Despite age and gender were comparable (p>0.05),they were used as covariates in the subsequent statistical analysis, to adjust for the slight trend for an excess of women in Nold subjects.
Out of the 64 LC patients, 21 (33% of females; 61.1±2.1 years) were recognized to have CHE on the basis of abnormalities revealed by the psychometric evaluation and/or signs of grade 1 HE on a structured clinical examination performed by a clinician with international experience in HE studies (S.M.).This procedure reflected our daily clinical routine in the framework of an exploratory investigation.To study the peculiar cortical sources of resting state EEG rhythms of the CHE sub-group, we selected the following age-and gender-matched control populations: 21 (33% of females; 61.2±0.8 years) Nold subjects, 21 (33% of females; 63.4±1.2 years) a-MCI subjects and 21 (33% of females; 62.6±1.7 years) AD patients.Analysis of variance (ANOVA) using the factor Group (CHE, Nold, aMCI, AD) for age confirmed the match goodness (p > 0.5).
Local institutional ethics committees approved the study.All experiments were performed with the informed and overt consent of each participant or caregiver, in line with the Code of Ethics of the World Medical Association (Declaration of Helsinki) and the standards established by the authors' institutional review board.

Diagnostic criteria
The inclusion diagnostic criteria to enrol LC patients were the following: (i) the presence of hepatic stigmata on routine clinical examination, together with biochemical indices of decompensated liver disease (i.e.low serum albumin, high bilirubin, prolonged prothrombin time, low platelet count); (ii) endoscopic or ultrasound signs of portal hypertension, or a history of previous decompensation (ascites, jaundice, bleeding from esophageal varices); (iii) mental or motor disorders including subtle cognitive changes/confusion, deep altered levels of cognitive functioning/consciousness, irritability, tremor, difficulties with coordination.
Exclusion criteria included patients with past disorders that could have negative effects on cerebral functioning such as heart failure, endocrinological diseases, respiratory or renal insufficiency, neuropsychiatric diseases, cerebrovascular diseases, dementia, epilepsy or who was taking psychotropic drugs.To this regard, it should be remarked that brain imaging (i.e.magnetic resonance imaging, MRI) was not routinely performed, as it is not required for the diagnosis of HE.MRI was used in a minority of the cases when the CHE patients' history and medical documentation suggested that the cognitive decline might be due to not only LC but also to other co-morbidities.In the majority of the cases, the patients' history and medical documentation were directly used to exclude other causes of cognitive decline (note that all enrolled LC patients have received long-term follow ups to exclude major comorbidities).Furthermore, all LC patients underwent EEG recordings to exclude signs of local brain suffer and epilepsy.
The LC patients underwent psychometric evaluation by the porto-hepatic encephalopathy score (PHES), comprising the following procedures: the Trial Making Test A and B (TMT-A and TMT-B, respectively), the Digit Symbol test (DS), the Serial Dotting (SD), and the Line Trait Test (LTT) [34,35].For each test, age and education-adjusted Z score was calculated.The PHES is the sum of the integer scores of each test computed from the adjusted Z values as follows: <Z <1, score = 1 for Z >1.CHE was diagnosed as follows: 1) existence of hepatic cirrhosis, 2) orientation for time and space, 3) alteration of PHES (≤-4), standardized for the national population [36], 4) exclusion of concurrent disorders that could have negative effects on cerebral functioning (heart failure -NYHA≥2-, endocrinological diseases (in particular, hypothyroidism), respiratory or renal insufficiency, neuropsychiatric diseases, cerebrovascular diseases, dementia, epilepsy or consumption of psychotropic drugs.
In addition, the mean Z psychometric score (MZPS) of the above mentioned tests was evaluated as a synthetic psychometric index of global cognitive status both for simplicity and to reduce bias due to multiple comparison [36].Psychometric data for unimpaired LC and CHE patients are summarized in Table 1.

Insert here Table 1
Inclusion and exclusion criteria for a-MCI subjects were based on previous seminal reports [37,38].Summarizing, the inclusion criteria were as follows: (i) objective memory impairment on neuropsychological evaluation probing cognitive performance in the domains of memory, language, executive function/attention, etc; (ii) normal activities of daily living as documented by the history and evidence of independent living; and (iii) clinical dementia rating (CDR; [39]) score of 0.5.The exclusion criteria included: (i) mild dementia of the AD type, as diagnosed by standard protocols including NINCDS-ADRDA [40] and DSM-IV; (ii) evidence (including magnetic resonance imaging -MRI− procedures) of concomitant cerebral impairment such as frontotemporal degeneration, cerebrovascular disease as indicated by Hatchinski ischemic score [41], MRI vascular lesions, and reversible cognitive impairment (including depressive pseudodementia); (iii) marked fluctuations in cognitive performance compatible with Lewy body dementia and/or features of mixed cognitive impairment including cerebrovascular disease; (iv) evidence of concomitant extra-pyramidal symptoms; (v) clinical and indirect evidence of depression as revealed by the Geriatric Depression Scale (GDS; [42]) scores > 14; (vi) other psychiatric diseases, epilepsy, drug addiction, alcohol dependence (as revealed by a psychiatric interview) and use of psychoactive drugs including acetylcholinesterase inhibitors or other drugs enhancing brain cognitive functions; and (vii) current or previous uncontrolled or complicated systemic diseases (including diabetes mellitus) or traumatic brain injuries.
Probable AD was diagnosed according to NINCDS-ADRDA [40] and DSM IV criteria.
The recruited AD patients underwent general medical, neurological, neuropsychological, and psychiatric assessments.Patients were rated with a number of standardized diagnostic and severity instruments that included Mini Mental State Evaluation (MMSE, [43]), Clinical Dementia Rating Scale (CDR; [39]), GDS [42], Hachinski Ischemic Score (HIS, [41]), and Instrumental Activities of Daily Living scale (IADL, [44]).Neuroimaging diagnostic procedures (MRI) and complete laboratory analyses were carried out to exclude other causes of progressive or reversible dementias, in order to have a clinically homogenous mild AD group.Exclusion criteria included any evidence of (i) frontotemporal dementia, (ii) vascular dementia, diagnosed according to NINDS-AIREN criteria [45], (iii) extra-pyramidal syndromes, (iv) reversible dementias (including pseudodementia of depression); and (v) Lewy body dementia, according to the criteria by McKeith [46].
Finally, the Nold subjects were recruited mostly from non-consanguineous relatives of AD patients.All Nold subjects underwent physical and neurological examinations as well as cognitive screening (including MMSE).Among them, those affected by chronic systemic illnesses (i.e.diabetes mellitus or organ failure) were excluded, as were subjects receiving psychoactive drugs.The Nold subjects with history of present or previous neurological or psychiatric disease were also excluded.All Nold subjects had a GDS score lower than 14 (no depression).

EEG recordings
EEG data were recorded by specialized clinical units in the recruited subjects at resting state (eyes-closed).The EEG recordings were performed (0.3-70 Hz bandpass) from 19 electrodes positioned according to the International 10-20 System (i.e.Fp1, Fp2, F7, F3, Fz, F4, F8, T3, C3, Cz, C4, T4, T5, P3, Pz, P4, T6, O1, O2).A specific reference electrode was not imposed to the recording clinical units of this study, since preliminary data inspection and EEG source analysis were carried out after EEG data were re-referenced to a common average reference.To monitor eye movements, the horizontal and vertical electrooculogram (0.3-70 Hz bandpass) was also collected.All data were digitized in continuous recording mode (few minutes of EEG; 256 Hz sampling rate).The EEG recordings were performed in the late morning.In order to keep constant the level of vigilance, an experimenter controlled on-line the subject and the EEG traces to avoid sleep onset.

Preliminary EEG-EOG data analysis
The recorded EEG data were analyzed and segmented off-line in consecutive epochs of 2 s.The EEG epochs with ocular, muscular, and other types of artifact were preliminary identified by a computerized automatic procedure.The EEG epochs with sporadic blinking artifacts (less than 10% of the total) were corrected by an autoregressive method [47].Two independent experimenters (R.L. and F.V.) blind to the diagnosis manually confirmed the EEG segments accepted for further analysis.Of note, a special attention was devoted to avoid the inclusion of EEG segments and individual data sets with EEG signs of drowsiness or pre-sleep stages.
Furthermore, the experimenters were blind to the diagnosis of the subjects at the moment of the preliminary EEG data analysis.Finally, we re-referenced artifact free EEG data to common average for further analysis.
Choice of the fixed EEG bands did not account for individual alpha frequency (IAF) peak, defined as the frequency associated with the strongest EEG power at the extended alpha range [73].The mean IAF peak was 8.8 ±0.2 Hz for the LC patients and 9.2 ±0.2 Hz for the Nold subjects.Furthermore, the mean IAF peak was 7.9 ±0.3 Hz for the CHE patients, 9.7 ±0.3 Hz for the a-MCI subjects, and 8.9 ±0.4 Hz for the AD subjects.To control the effect of IAF on the EEG comparisons, the IAF peak was used as a covariate (together with age, and gender) for further statistics.

Cortical source analysis of the EEG rhythms by LORETA
Low resolution electromagnetic source tomography (LORETA) was used for the EEG source analysis as provided at http://www.unizh.ch/keyinst/NewLORETA/LORETA01.htm[76,77,78].LORETA is a functional imaging technique belonging to a family of linear inverse solution procedures modeling 3D distributions of EEG sources [78].With respect to the dipole modeling of cortical sources, LORETA estimation implies no a priori decision on the dipole position.Furthermore, LORETA is quite efficient when compared to other linear inverse algorithms like minimum norm solution, weighted minimum norm solution or weighted resolution optimization [77,79,80].Finally, LORETA has been successfully used in recent EEG studies on pathological brain aging [27,28,29,30,31,33,53 2394 voxels of the brain volume.After the normalization, the solutions lost the original physical dimension and were represented by an arbitrary unit scale.The general procedure fitted the LORETA solutions in a Gaussian distribution and reduced inter-subject variability [51].
Solutions of the EEG inverse problem are under-determined and ill conditioned when the number of spatial samples (electrodes) is lower than that of the unknown samples (current density at each voxel).To account for that, the cortical LORETA solutions predicting scalp EEG spectral power density were regularized to estimate distributed rather than punctual EEG source patterns [76,77,78].In line with the low spatial resolution of the adopted technique, home-made MATLAB software averaged the amplitude of LORETA solutions for all voxels belonging to each macroregion of interest such as frontal, central, parietal, occipital, temporal, and limbic.
Each of these macroregions of interest (ROIs) was constituted by all the voxels of the Brodmann areas listed in Table 3.The belonging of a LORETA voxel to a Brodmann area was defined by original LORETA package.
Finally, the main advantage of the regional analysis of LORETA solutions -using an explicit source model coregistered into Talairach space-was that our modeling could disentangle rhythms of contiguous cortical areas (namely those from the occipital source were disentangled with respect to those of the contiguous parietal and temporal sources etc).

Statistical analyses
Results are expressed as mean ± standard error (SE), unless otherwise indicated.To test the working hypothesis, the following three statistical sessions were performed.
The first session tested the hypothesis that cortical (LORETA) sources of EEG rhythms differed in amplitude between LC and Nold subjects.To this aim, the regional normalized LORETA solutions from LC (including subjects with and without cognitive deficits) and Nold subjects were used as an input for an ANOVA design using age, gender, and IAF as covariates.
Mauchley's test evaluated the sphericity assumption, and the correction of the degrees of freedom was made by Greenhouse-Geisser procedure.Duncan test was used for post-hoc test comparisons (p<0.05).The ANOVA used the factors Group (Nold, LC; independent variable), Band (delta, theta, alpha 1, alpha 2, beta 1, beta 2), and ROI (central, frontal, parietal, occipital, temporal, limbic).The working hypothesis would be confirmed by the following two statistical results: (i) a statistical ANOVA effect including the factor Group (p<0.05);(ii) a post-hoc test indicating statistically significant differences of the regional normalized LORETA solutions with the patterns LC ≠ Nold (Duncan test, p<0.05).
The second session tested the hypothesis that in LC patients, the activity of cortical (LORETA) sources of EEG rhythms was related to psychometric measures.To this aim, we performed a correlation analysis (Pearson test, p<0.01) between the amplitude of regional normalized LORETA solutions and mean Z psychometric score (MZPS) in LC patients.To reduce the amount of univariate comparisons, only the regional normalized LORETA solutions fitting the pattern LC ≠ Nold were considered for that correlation analysis.
The third session tested the hypothesis that cortical (LORETA) sources of EEG rhythms differed in amplitude activity in CHE patients with respect to Nold, a-MCI and AD subjects.To this aim, the regional normalized LORETA solutions from the CHE, Nold, a-MCI and AD subjects were used as an input for an ANOVA design using age, gender, and IAF as covariates.
The ANOVA used the factors Group (Nold, a-MCI, AD, CHE; independent variable), Band (delta, theta, alpha 1, alpha 2, beta 1, beta 2), and ROI (central, frontal, parietal, occipital, temporal, limbic).The working hypothesis would be confirmed by the following two statistical results: (i) a statistical ANOVA effect including the factor Group (p<0.05);(ii) a post-hoc test indicating statistically significant differences of the regional normalized LORETA solutions with the patterns CHE ≠ Nold, a-MCI, AD (Duncan test, p<0.05).

Topography of the EEG cortical sources estimated in the LC and Nold groups
For illustrative purposes, Fig. 1 maps the grand average of the LORETA solutions (i.e. relative current density at cortical voxels) modeling cortical sources of delta, theta, alpha 1, alpha 2, beta 1, and beta 2 rhythms in the Nold and LC groups.The Nold group presented alpha 1 sources with the maximal values of power distributed in parietal and occipital regions.Delta, theta, and alpha 2 sources had moderate activity amplitude values when compared to alpha 1 sources.Furthermore, beta 1 and beta 2 sources were characterized by lowest activity.Compared to the Nold group, the LC group showed an activity increase of widespread theta, alpha 1, and alpha 2 sources.
Insert here Figure 1 Statistical comparisons of the EEG cortical sources estimated in the LC and Nold groups The ANOVA for the evaluation of the first working hypothesis (i.e.amplitude difference of cortical LORETA sources of EEG rhythms between the LC and Nold groups) showed a statistically significant interaction (F(25,3150) = 4.48; p < 0.0001) among the factors Group (Nold, LC), Band (delta, theta, alpha 1, alpha 2, beta 1, and beta 2), and ROI (central, frontal, parietal, occipital, temporal, limbic).Fig. 2 shows the mean regional normalized LORETA solutions relative to this statistical ANOVA interaction.In the figure, the regional normalized LORETA solutions had the shape of EEG relative power spectra.Notably, profile and magnitude of these spectra in the Nold and LC groups differed across diverse cortical macro-regions, thus supporting the idea that scalp EEG rhythms are generated by a distributed pattern of cortical sources.The Duncan planned post-hoc testing (Table 3) showed that the source pattern LC ≠ Nold was fitted by theta (central, parietal, occipital, temporal, and limbic areas; p<0.00003), alpha 1 (central, frontal, parietal, temporal, and limbic areas; p<0.001), and alpha 2 (central, parietal, and temporal areas; p<0.002) sources, where the amplitude of LORETA sources was stronger in the LC compared to the Nold subjects.

Insert here Figure 2 and Table 3
To explore the relationship between the cortical EEG rhythms and subjects' global cognitive functions, the normalized regional LORETA sources were used as an input for the correlation with the MZPS in all LC subjects (Pearson test, p<0.01).The MZPS showed a statistically significant negative correlation with central (r=-0.32,p=0.008), parietal (r=-0.45,p=0.0002), and temporal (r=-0.45,p=0.0002) theta sources.Furthermore, the MZPS showed a statistically significant positive correlation with parietal (r=0.34,p=0.005) and temporal alpha 2 (r=0.33,p=0.009) sources.Fig. 3 shows the scatterplots of these statistically significant correlations.
Insert here Figure 3.
Insert here Figure 3 Topography of the EEG cortical sources estimated in the CHE compared to the control groups For illustrative purposes, Fig. 4 maps the grand average of the LORETA solutions (i.e. relative current density at cortical voxels) modeling cortical sources of delta, theta, alpha 1, alpha 2, beta 1, and beta 2 rhythms in the Nold, aMCI, AD and CHE groups.Compared to the Nold group, the AD group showed an increase of widespread delta sources, along with a dramatic reduction of parieto-occipital alpha 1 sources.With respect to the Nold and AD groups, the aMCI group showed intermediate magnitude of alpha 1 sources.Finally, the CHE group was characterized by an amplitude increase of widespread theta and centro-parietal alpha 1 sources compared to the Nold, MCI and AD groups.

Statistical comparisons of the EEG cortical sources estimated in the CHE compared to the control groups
An ANOVA compared the cortical sources of EEG rhythms in the Nold (specifically the sub-group of Nold subjects matched to pathological groups), aMCI, AD, and CHE groups.Results showed a statistically significant interaction (F(75,2) = 4.68; p < 0.0001) among the factors Group (Nold, aMCI, AD, CHE), Band (delta, theta, alpha 1, alpha 2, beta 1, and beta 2), and ROI (central, frontal, parietal, occipital, temporal, limbic).Fig. 5 shows the mean regional normalized LORETA solutions relative to this statistical ANOVA interaction.The Duncan planned post-hoc testing showed that the source pattern of interest CHE ≠ Nold, aMCI, and AD was fitted by theta (central, parietal, occipital, temporal, and limbic areas; p<0.002) and alpha 1 (central, frontal, parietal, temporal, and limbic; p<0.02) sources, where the amplitude of LORETA sources was stronger in the CHE group than in the Nold, aMCI, and AD groups.Noteworthy, the power of the theta sources did not differ among the Nold, aMCI, and AD groups (p>0.05).

Control analysis of EEG functional coupling
As previously mentioned, theta (central, parietal, occipital, temporal, and limbic areas), alpha 1 (central, frontal, parietal, temporal, and limbic areas), and alpha 2 (central, parietal, and temporal areas) LORETA sources were higher in power in the LC then Nold groups.These results raised the issue about the functional meaning of increased alpha rhythms in the LC patients in the light of the well known assumption that healthy cortical neural synchronization is reflected by low-amplitude theta rhythms and high-amplitude alpha rhythms in the resting state eyes-closed condition.A control analysis was performed to evaluate whether the high amplitude of alpha sources in the LC patients was associated to an abnormal functional coupling of resting state EEG rhythms, as a sign of an abnormal "networking" between paired cortical regions.To address this issue, we evaluated the EEG spectral coherence in the LC and Nold subjects.Indeed, EEG spectral coherence between electrode pairs is commonly interpreted as an index of functional coupling [86,87], mutual information exchange [88], functional co-ordination [89], and integrity of cortical neural pathways [90].Its basic theoretical assumption is that when the activity of two cortical areas is functionally coordinated, the EEG rhythms of these cortical areas show linear interrelatedness and high spectral coherence.Shortly, EEG coherence is a normalized measure of the coupling between two signals at any given frequency [88,91].From a physiological point of view, EEG coherence reflects functional cooperation among the brain areas under study.In the present study, the coherence values were calculated for each frequency bin by (eq. 1) (1) The above equation is the extension of the Pearson's correlation coefficient to complex number pairs.In this equation, f denotes the spectral estimate of two EEG signals x and y for a given frequency bin (λ).The numerator contains the crossspectrum for x and y (f xy ), while the denominator contains the respective autospectra for x (f xx ) and y (f yy ).For each frequency bin (λ), the coherence value (Coh xy ) is obtained by squaring the magnitude of the complex correlation coefficient R.This procedure returns a real number between 0 (no coherence) and 1 (maximal coherence).
In line with previous studies of the present research group [57,92,94,95,96,97,98], the spectral coherence between couples of electrodes was calculated by a home-made software developed under Matlab 6.5 (Mathworks Inc., Natrick, MA; an extension of Pearson's correlation coefficient to complex number pairs; 1-Hz frequency resolution).Specifically, the calculation of the coherence was performed from the artifact-free EEG data, on the basis of all scalp electrodes used during EEG recordings.The coherence was computed at theta (4-8 Hz), alpha 1 (8-10.5Hz), and alpha 2 (10.5-13Hz) bands.
The ANOVA showed a statistically significant interaction (F(6,756) = 2.2; p < 0.05) among the factors Group, Band, and ROI (see Fig. 6 A).Duncan post-hoc testing indicated that the bilateral frontal-central and frontal-occipital theta coherence was higher in the LC than Nold subjects (p<0.01).On the contrary, bilateral frontal-parietal and frontal-temporal alpha 1 and alpha 2 coherence was lower in the LC than Nold subjects (p<0.01).
On the whole, the results of this control analysis suggest that compared to the Nold subjects, the LC patients were characterized by higher functional coupling of the pathological theta rhythms and lower functional coupling of the dominant alpha rhythms.These control results suggest that in the LC patients, the abnormally high activity of alpha sources is associated with an abnormal functional coupling of the resting state EEG rhythms.

Insert here Figure 6 Discussion
In the present study, we tested the hypothesis that compared to control subjects, LC patients are characterized by widespread abnormalities of the cortical sources of resting state EEG rhythms as spatial distribution and frequency bands, as an effect of the generalized action of circulating neurotoxins (e.g.ammonia) on the whole brain.As a methodological novelty, we used a validated approach for the estimation of the cortical sources of resting state EEG rhythms that allowed a useful spatial analysis of the abnormal brain synchronization mechanisms generating these rhythms in elderly subjects with cognitive decline (27,28,29,30,33).
Results showed that widely distributed sources of theta (all but frontal), alpha 1 (all but occipital), and alpha 2 (parietal, temporal) rhythms were higher in amplitude in all LC patients than in the Nold subjects.Furthermore, the amplitude of central, parietal, and temporal theta sources correlated negatively, and parietal and temporal alpha 2 sources correlated positively with an index of global cognitive status across all LC patients.These results confirm and extend to the fine detail of the cortical source level previous evidence showing that EEG theta and alpha band power computed at scalp electrodes was related to psychometric performance in LC patients, and differed in amplitude between LC patients and matched healthy controls [16,17,20].Furthermore, these effects are fully in line with the so called "anteriorization" of theta and alpha rhythms previously described in HE patients by independent groups, which were ascribed to some processes of drowsiness preceding sleep onset [21,22,23,24,25].However, it is improbable that occasional sleep onset during the EEG recordings can explain the present results, as this aspect was carefully monitored by the experimenters and alpha rhythms typically drop down at the sleep onset.
Therefore, a possible slight reduction of the vigilance might have other reasons related to HE pathophysiology per se.
As a novelty of the present study, we tested the hypothesis of peculiar abnormalities in the spatial distribution and frequency bands of the cortical sources of resting state EEG rhythms in patients with CHE when compared to patients with possible prodromal (i.e.aMCI) or overt AD.
The CHE patients were expected to show widespread abnormalities of the cortical sources of EEG rhythms due to the diffuse action of the pathological agents affecting brain activity and cognitive processes.In contrast, the MCI and AD patients were expected to be characterized by more circumscribed spatial and frequency abnormalities of these EEG rhythms as the main target of the neurodegenerative processes is typically localized in hippocampus and posterior cortical regions.
The confirmation of the working hypothesis would confirm that the present methodological approach can shed light on the understanding of the peculiar abnormalities in the generation of the resting state EEG rhythms in people with cognitive decline due to different patho-physiological mechanisms such as CHE and MCI/AD.It should be remarked that at this early stage of the research, we followed the standard clinical guidelines for the clinical diagnosis of CHE, which is based on the pooling of patients with both MHE (i.e.diagnosed based on objective cognitive deficits as revealed by psychometric tests) and West Haven grade I HE (i.e.diagnosed based on a structured clinical evaluation, which is less reproducible than psychometric testing).This approach was implemented in the framework of our daily clinical routine.In light of the present encouraging findings, future confirmatory studies should systematically carry out psychometric testing in the enrolled LC patients to select only those having statistically significant sub-clinical cognitive deficits.
Results showed a different spatial-frequency pattern of the EEG sources between CHE and aMCI or AD patients, thus disclosing the diverse early effects of the HE and AD processes on the resting state cortical EEG rhythms.Specifically, widely distributed theta (all but frontal) and alpha 1 (all but occipital) sources showed higher activity in the sub-group of LC patients with CHE than in the patients with aMCI or AD.A plausible explanation of the results is that enhanced theta sources in the LC patients may be interpreted according to the general view that the high amplitude of the low-frequency EEG rhythms (about 1-7 Hz) reflects the abnormal interactions between thalamic and cortical neural populations, resulting in a functional isolation of cortical modules [52,73,99,100,101,102,103,104].The main known causes of this condition are cortical hypoperfusion, neurotrauma or neurodegeneration [27,52,105,104,106].In this theoretical line, the present results lead support to the notion that hepatic metabolic disorders may widely influence these mechanisms as well.With respect to the progressive AD neurodegenerative processes mainly affecting hippocampus and posterior cortical regions at early disease stages, the LC processes might exert acute and widely spread effects on the brain, thus inducing stronger pathological effects on the thalamo-cortical mechanisms generating the resting state theta rhythms.
Another interesting result of the present study is that the cortical sources of alpha rhythms were higher in power in the CHE patients than in the control subjects (i.e.Nold, aMCI, and AD).These results may be interpreted at the light of the general theory on the generation of the resting state alpha rhythms in humans.Low frequency (about 8-10.5 Hz) alpha rhythms would denote the synchronization of diffuse neural networks regulating the fluctuation of subject's global arousal and consciousness states, whereas high frequency (about 10.5-13 Hz) alpha rhythms would denote the synchronization of more selective neural networks specialized in the processing of modal specific or semantic information [72,74,75].Keeping in mind the above theoretical considerations, a plausible explanation of the results is that AD neurodegeneration and CHE differently impinge upon the neural mechanism at the basis of the generation of resting state alpha rhythms, which is one of main mechanisms regulating the cortical arousal and vigilance.Therefore, the present results suggest that cognitive deficits of the MCI/AD and CHE patients may be related to a diverse alteration of the cortical arousal.The MCI/AD subjects were characterized by a tonic de-synchronization of the resting state alpha rhythms as a reflection of a tonic cortical excitation.In contrast, the CHE patients were characterized by a tonic hyper-synchronization of the resting state alpha rhythms as a reflection of a tonic cortical inhibition.Both up-and down-regulation of cortical arousal would be associated to an abnormal brain function and predict cognitive deficits.Effective cognitive processing is expected to stem upon the selectivity and flexibility of the excitation and inhibition across brain neural networks during both resting state condition and task demands (see for example the concept of "neural efficiency"; Vernon, 1993, Haier et al., 1988, 2004, Rypma et al., 2006).The plausibility of the above explanation is related to the well know relationship between the fluctuation of the cortical arousal in the condition of the resting state and the power of the cortical alpha rhythms [72,74,75].
Less clear are the neurophysiological and neurobiological fine mechanisms linking the AD neurodegeneration, CHE neurotoxins, and the neurophysiological mechanisms generating the resting state cortical alpha rhythms.At this early stage of the research, we can just speculate that AD neurodegeneration and CHE neurotoxins exert their effects on the ascending neuro-modulatory systems at the basis of the resting state EEG rhythms and cortical arousal.However, the heuristic value of the present results should not be underestimated.They motivate future studies investigating the correlation among the present neurophysiological (EEG) markers, the cognitive deficits, and direct indexes of up-and down-regulation of the mentioned ascending neuromodulatory systems.
These future investigations should enlighten these interactions, unveiling the relationships between the alpha source pattern during the resting state condition and relevant indexes of the cholinergic, glutamatergic, and GABA modulation (note that GABA is the principal inhibitory neurotransmitter in the human brain and modulates cholinergic and glutamatergic activity in the neocortex and the hippocampus) [116].These future investigations should test the hypothesis that the increase of the low-frequency alpha sources (i.e.tonic cortical inhibition) in the LC/HCE patients is related to an enhancement of the inhibitory GABA-ergic tone provoked by neurosteroids [117,118,119,120,121,122] and/or to a depression of the glutamate-nitric oxide-cGMP excitatory pathway [123].
Furthermore, the decrease of the low-frequency alpha sources (i.e.tonic cortical excitation) in the MCI or AD patients might be related to a reduction of the inhibitory GABA-ergic tone [116,124], to a depression of the cholinergic inhibitory activity, and/or to an enhancement of the excitatory glutamatergic non-NMDA receptor activity [125,126].Finally, it should be tested the role of the dopaminergic and serotoninergic systems on the modulation of the cortical sources of the resting state alpha rhythms in these patients [127,128,129].
In conclusion, the present study tested the hypothesis that LC and AD conditions are characterized by peculiar abnormalities of the cortical sources of resting state EEG rhythms, reflecting the relative patho-physiological mechanisms.Compared to the control healthy subjects, the LC/CHE patients were characterized by widespread abnormalities of the cortical sources of the resting state theta and alpha rhythms.Furthermore, the cortical sources of the resting state alpha rhythms appeared to change differently under the influence of the LC/CHE and AD conditions.The patients with LC/CHE were characterized by widespread abnormalities as spatial distribution and frequency bands of the cortical sources of the theta, low-and high-frequency alpha rhythms.These findings are compatible with the widespread effects on brain activity provoked by the circulating byproducts of liver failure passing the brain blood barrier.On the other hand, the patients with aMCI and AD showed a spatially selective derangement of the posterior cortical sources of the delta and low-frequency alpha rhythms, as expected by the typical specific track of the neurodegenerative processes in the earlier disease stages.These EEG markers are interesting candidates to capture the neurophysiological mechanisms of action of pharmacological interventions for the two pathologies, with promising perspectives for the development of new drugs, a better understanding of their neurophysiological mechanisms, and therapy monitoring.Finally, the findings of this exploratory study encourages further investigations on the cortical sources of the resting state EEG rhythms in the LC/CHE patients and in the aMCI/AD patients as a function of the effects of the therapy, disease progression, and activity of the main ascending neuromodulatory systems.This ANOVA design used the regional normalized LORETA solutions as a dependent variable.Subjects' age, gender and individual alpha frequency peak were used as covariates.Regional normalized LORETA solutions modeled the EEG relative power spectra as revealed by a sort of "virtual" intracranial macro-electrodes located on the macrocortical regions of interest.The

Figure legend Figure 1 .
Figure legend Figure 1.Grand average of the LORETA solutions (i.e.normalized relative current density at the cortical voxels) modeling cortical sources of delta, theta, alpha 1, alpha 2, beta 1, and beta 2 rhythms in the Nold and LC subjects.The left side of the maps (top view) corresponds to the left hemisphere.Legend: LORETA, low-resolution brain electromagnetic tomography.Color scale:all power estimates were scaled based on the averaged maximum value (i.e.alpha 1 power value of occipital region in the LC and Nold subjects).The maximal value of the LORETA solution is reported under each column.
rectangles indicate the cortical regions and frequency bands in which LORETA solutions presented statistically significant LORETA patterns LC ≠ Nold (p < 0.05).Legend: Nold = normal elderly subjects; LC= liver cirrhosis; LORETA= low resolution brain electromagnetic source topography.

Figure 3 .
Figure 3. Scatterplots showing the correlation between regional normalized LORETA solutions and mean Z psychometric score (MZPS) in the LC subjects.The r-and p-values refer to Pearson correlation are reported within the diagram.

Figure 4 .
Figure 4. Grand average of the LORETA solutions (i.e.normalized relative current density at the cortical voxels) modeling cortical sources of delta, theta, alpha 1, alpha 2, beta 1, and beta 2 rhythms in the Nold, MCI, AD, and CHE subjects.The left side of the maps (top view) corresponds to the left hemisphere.Color scale: all power estimates were scaled based on the averaged maximum value (i.e.alpha 1 power value of occipital region in the CHE and Nold subjects).The maximal value of the LORETA solution is reported under each column.Legend: MCI = mild cognitive impairment; AD= Alzheimer's disease; CHE= covert hepatopatic hencephalopathy.

Figure 5 .
Figure 5. Regional normalized LORETA solutions (mean across subjects) relative to a statistical ANOVA interaction among the factors Group (Nold, MCI, AD, CHE), Band (delta, theta, alpha 1, alpha 2, beta 1, and beta 2), and Region of interest (central, frontal, parietal, occipital, temporal, limbic).This ANOVA design used the regional normalized LORETA solutions as a dependent variable.Subjects' age, gender and individual alpha frequency peak were used as covariates.Legend: The rectangles indicate the cortical regions and frequency bands in which LORETA solutions presented statistically significant LORETA patterns CHE ≠ Nold, MCI, AD (p < 0.05).

Table legend Table 1 .
Mean value (± standard error, SE) of psychometric data in patients with liver cirrhosis (LC) and LC patients with covert hepatic encephalopathy (CHE).The following psychometric tests were considered: Trial Making Test A -Test B (TMT-A, TMT-B), the Symbol Digit Test (SDT), the Serial Dotting Test (SD) and the Line Trait Test (LTT).The mean Z psychometric score (MZPS) of the above mentioned tests is also reported as a synthetic psychometric index.

Table 2 .
Brodmann areas included in the cortical regions of interest (ROIs) of the present study.LORETA solutions were collapsed in frontal, central, parietal, occipital, temporal, and limbic ROIs.

Table 3 .
Statistical ANOVA comparison distinguishing LC and Nold groups.