Altered intestinal microbiota are associated with sleep disturbances in patients with minimal hepatic encephalopathy caused by hepatitis B-related liver cirrhosis

Objective Minimal hepatic encephalopathy (MHE) caused by liver cirrhosis is quite prevalent, and approximately one-half of MHE patients have experience sleep disturbances. This study systematically evaluated the association between sleep disturbances and altered intestinal microbiota in patients with MHE caused by hepatitis B-related liver cirrhosis. Methods Ninety-eight and 45 MHE patients were respectively included in the exploration and validation cohorts. The Chinese version of the Pittsburgh Sleep Quality Index (PSQI) questionnaire was used to evaluate sleep disturbances. The intestinal microbiota of self-collected fecal samples was analyzed using the amplicon sequencing of bacterial 16S ribosomal RNA gene.


Introduction
Minimal hepatic encephalopathy (MHE) is a frequently occurring neurological comorbidity due to liver cirrhosis and/or portosystemic shunting, with a spectrum of mild neurocognitive impairments in working memory, psychomotor speed, and attention span 1,2 . MHE prevalence in cirrhotic patients varies between 30% and 70%, according to epidemiological surveys based on various diagnostic methods in different populations 3,4 . MHE affects daily functioning, impairs driving ability, and is associated with an increased risk of developing overt hepatic encephalopathy 5,6 .
Sleep disturbances is prevalent, severely in uences quality of life, and is related to poor prognosis in MHE patients 7,8 . Existing studies have suggested that altered intestinal microbiota is closely correlated with sleep disturbances in neuropsychiatric diseases such as Parkinson's disease and bipolar disorder 9,10 .
Through impairment of the gut-liver-brain axis in cirrhotic patients, altered intestinal microbiota also play signi cant roles in the pathophysiological mechanisms of MHE 11 . Thus, we hypothesized that altered intestinal microbiota might be associated with sleep disturbances in patients with MHE caused by liver cirrhosis.
In the presented study, we systematically and comprehensively analysed the intestinal microbiota in MHE patients with and without sleep disturbances, evaluating the association between sleep disturbances and altered intestinal microbiota in patients with MHE caused by hepatitis B-related liver cirrhosis.

Patients
All included MHE patients were recruited from the Department of Gastroenterology in People's Hospital of Ningxia Hui Autonomous Region between January 2019 and January 2021. Hepatitis B-related liver cirrhosis was the underlying etiology of MHE in all of the patients studied. The diagnosis of liver cirrhosis was according to reviews of medical history, radiographic or ultrasound imaging, liver biopsy, and endoscopic manifestations of portal hypertension. Child-Turcotte-Pugh grade was applied to assess the severity of liver cirrhosis.
MHE was diagnosed using the Chinese version of the psychometric hepatic encephalopathy score (PHES), which includes the number connection test-A, number connection test-B, line tracing test, digit symbol test, and serial dotting test, as recommended by the updated guideline from the Chinese Society of Hepatology 12,13 . When the score of PHES subtest exceeded mean ± 2 standard deviations (SD) of the age/education level/gender-matched normal range calculated from 146 healthy Chinese volunteers, the PHES subtest was considered as abnormal 13 . A value of -1 point was assigned to the abnormality in each subtest of PHES, and recruited patients with a total PHES value of < −4 points were diagnosed as having MHE 12,13 .
The following were used as exclusion criteria: clinical manifestations of overt hepatic encephalopathy; recent intake history of probiotics, lactulose, rifaximin in the last 12 weeks; recent intake history of psychotropic drugs in the last 12 weeks; history of recent alcoholism in the last 12 weeks; recent history of overt hepatic encephalopathy, acute gastrointestinal hemorrhage, infection, or electrolyte imbalance in the last 12 weeks; history of primary hepatocellular carcinoma or other malignant tumors; history of fatty liver disease; history of portosystemic shunt or transjugular intrahepatic portosystemic shunt; history of obstructive sleep apnea; obesity (body mass index [BMI] ≥ 30 kg/m 2 ); neurocognitive diseases including Alzheimer's disease, Parkinson's disease, or ischemic cerebral infarction; vital organ dysfunction including cardiac insu ciency, renal failure, or chronic obstructive pulmonary disease.
In total, 180 patients with MHE caused by hepatitis B-related liver cirrhosis were recruited in this study, and 120 of whom were randomly assigned to the exploration cohort and the remaining 60 patients were included in the validation cohort ( Figure 1). The protocol for this study was approved by the Ethics Committee of People's Hospital of Ningxia Hui Autonomous Region, and it adhered to the ethical principles from the Declaration of Helsinki (6th version, 2008). Before being included in this study, all individuals submitted their written informed consent to participate in it.

Assessment of sleep disturbances
The Chinese version of the Pittsburgh Sleep Quality Index (PSQI) questionnaire was used to assess sleep disturbances among the included MHE patients. The PSQI is divided into seven subcomponents: sleep quality, sleep latency, sleep duration, sleep e ciency, sleep disturbance, use of sleep medication, and daytime dysfunction 14 . Each subcomponent of the PSQI is scored on a scale of 0 to 3, with the sum of all subcomponent values resulting in a total score ranging from 0 to 21. In the Chinese population, a PSQI score of 5 has been demonstrated to be an effective threshold for identifying MHE patients with and without sleep disturbances 14 .

Laboratory tests
Venous blood samples were taken from all enrolled MHE patients after an overnight fast, and were transmitted for routine hematological, biochemical, and virological testings. Within 20 minutes, plasma The data concerning the sequencing of 16S rRNA gene were analyzed using the Quantitative Insights Into Microbial Ecology software (QIIME2 Version 2020.8) 15 . The sequencing reads were demultiplexed and ltered using the DADA2 software from QIIME2. Using the Vsearch software, the sequencing dataset was clustered to yield operational taxonomic units (OTUs) with criterion of 97% minimum identity. Chimeric sequences and OTUs with a sequence count of less than 0.005% of the overall sequence count were identi ed and removed 16 . An OTU count table with taxonomy was generated based on the detected OTU dataset and accompanying taxonomic assignments. The rare ed OTU table was used to estimate alphadiversity based on species richness. Variations in bacterial diversity across subjects in the different groups were evaluated using beta-diversity analyses based on unweighted UniFrac tests, which were displayed using the principal coordinate analysis (PCoA). The linear discriminant analysis effect size (LEfSe) was applied to determine genera that were signi cantly different in the relative abundances between two groups. A conditional stepwise logistic regression analysis was performed to identify the independent predictors of sleep disturbances from genera which were found to be signi cantly different in the LEfSe analysis. Simultaneously, odds ratios (OR) and 95% con dence intervals (CI) of the independent predictors were calculated. Furthermore, the analysis of multivariate association with linear models algorithm (MaAsLin) was used to assess the multivariable correlation between clinical covariates and relative abundances of microbial genera 17

Statistical analysis
Differences in continuous data and bacterial composition between two groups were examined by the Mann-Whitney rank-sum test. The χ 2 -test or Fisher's exact test were used to compare differences in categorical variables between two groups. The correlation analysis was performed using the partial Spearman's rank-based test. In order to identify the genera that best distinguished MHE patients with and without sleep disturbances, a multivariable stepwise logistic regression analysis was performed using the caret package. A receiver operating characteristic (ROC) curve was established to assess the discriminatory capacity of different genera, and the area under the ROC curve (AUC) was calculated using the pROC program. Signi cant differences were de ned as those with a two-sided p<0.05, and statistical analyses were conducted by the R software (Version 2.15.3).

Baseline characteristics of included MHE patients
Based on the inclusion and exclusion criteria, 22 MHE patients were excluded, and 98 MHE patients were ultimately included in the exploration cohort ( Figure 1). Fifty-seven percent of included MHE patients in the exploration cohort were male, and the median age was 58 years (range, 29-68 years). The median BMI was 22.5 kg/m 2 (range 17-27.5 kg/m 2 ), with a median educational background of 9 years (range, 5-17 years). Of these MHE patients, 82 (84%), 14 (14%), and 2 (2%) were graded as Child-Turcotte-Pugh grade A, B, and C, respectively. Moreover, 70 (71%) MHE patients had a medical record of previous hospitalization. In total, 89 (91%) MHE patients have received the treatment of nucleoside analogs including tenofovir and entecavir, and 73 (74%) of them tested positive for HBV-DNA. Table 1 showed the laboratory values for plasma ammonia, serum sodium, C-reactive protein, hemoglobin, and creatinine.
Meanwhile, 15 MHE patients were excluded, and 45 MHE patients were ultimately included in the validation cohort ( Figure 1). The demographic and clinical features of the included MHE patients in the validation cohort were presented in Table 1. Between the exploration and validation cohorts, no statistically signi cant difference in demographic and clinical features was identi ed (all p>0.05) ( Table  1).  Table S1).

Altered microbial composition in MHE patients with sleep disturbances
The richness of the intestinal microbiota, as evaluated by the count of detected OTUs, was lower in the 43 MHE patients with sleep disturbances than in the 55 MHE patients without sleep disturbances (p<0.001) (Figure 2A). After controlling for age, gender, and BMI, the lower species richness was con rmed. In MHE patients with sleep disturbances, the Shannon indices, which evaluated the richness and evenness of the intestinal microbiota, were signi cantly lower than in those without sleep disturbances (p<0.001) ( Figure  2B).  Figure 3A). Notably, in MHE patients with sleep disturbances, Salivarius and Veillonella presented a more than three-fold increase in comparison to those without sleep disturbances (p<0.001 and p=0.004, respectively) ( Figure 3B).

Salivarius distinguished MHE patients with sleep disturbances
In the exploration cohort, the logistic regression analysis presented that the relative abundances of Salivarius, Veillonella, Klebsiella, and Eubacterium were independent predictors of sleep disturbances in MHE patients (Table 2). A logistic regression model that was composed of relative abundances of these sleep disturbances-associated genera, including Salivarius, Veillonella, Klebsiella, and Eubacterium, could distinguish MHE patients with sleep disturbances from those without sleep disturbances, with an AUC of 0.81 (95% CI: 0.74-0.87). Moreover, using only Salivarius as a predictor generated an AUC of 0.75 (95% CI: 0.68-0.82), and an AUC of 0.64 (95% CI: 0.55-0.71) was yielded using only Veillonella as a predictor. The AUC was found to be insigni cantly different between the logistic regression model and Salivarius (p=0.625), while the AUC of both the logistic regression model and Salivarius was signi cantly higher than that of Veillonella (p=0.036 and p=0.015, respectively) ( Figure 4A).
In the validation cohort, the logistic regression model that was composed of relative abundances of sleep disturbances-associated genera had an AUC of 0.73 (95%CI: 0.65-0.80). Salivarius alone yielded an AUC of 0.73 (95%CI: 0.64-0.81), and Veillonella alone generated an AUC of 0.61 (95%CI: 0.53-0.69). In the AUC, there was no statistically signi cant difference between the logistic regression model and Salivarius (p=0.966), and the AUC of both the logistic regression model and Salivarius was statistically greater than that of Veillonella (p=0.029 and p=0.032, respectively) ( Figure 4B).

Correlation between altered intestinal microbiota and sleep disturbances
In the exploration cohort, after correcting for possible in uence factors including age, gender, and BMI using the MaAsLin analysis, the relative abundances of Salivarius were closely correlated with PSQI scores in MHE patients with sleep disturbances (rho=0.25, p<0.001). Furthermore, the relative abundances of Veillonella were also correlated with PSQI scores in these patients (rho=0.38, p=0.019). Moreover, a positive correlation was identi ed between plasma ammonia levels and relative abundances of Salivarius in MHE patients with sleep disturbances (rho=0.69, p<0.001) ( Figure 5A).
In the validation cohort, the PSQI scores were correlated with the relative abundances of Salivarius and Veillonella in MHE patients with sleep disturbances (rho=0.42, p<0.001 and rho=0.31, p=0.007, respectively). Furthermore, there was also a correlation between plasma ammonia levels and relative abundances of Salivarius in these patients (rho=0.56, p=0.033) ( Figure 5B).

Microbial functions altered in MHE patients with sleep disturbances
In the exploration cohort, the LEfSe analysis found 30 KEGG categories that had signi cantly different abundances between MHE patients with and without sleep disturbances (LDA score> 2.0, p<0.05). After adjusting for confounders in the MaAsLin analysis, seven functional modules were shown to be substantially related to sleep disturbances in MHE patients (all p<0.05) ( Figure 6). Notably, functional modules related to lipopolysaccharide biosynthesis, as well as protein digestion and absorption, were increased in the microbiome of MHE patients with sleep disturbances.

Discussion
Accumulating evidences have suggested that altered intestinal microbiota are associated with comorbidities of liver cirrhosis, including MHE and spontaneous bacterial peritonitis 19 . Using the amplicon sequencing of bacterial 16S rRNA gene, we delineated the composition of the intestinal microbiota in MHE patients with and without sleep disturbances. Our ndings revealed that the altered intestinal microbiota of MHE patients with sleep disturbances were characterized by lower bacterial diversities and distinct microbial composition in comparison to those without sleep disturbances. The multivariable logistic regression analysis presented that the relative abundances of Salivarius, Veillonella, Klebsiella, and Eubacterium were independent predictors of sleep disturbances in MHE patients. Based on the microbial pro les, a logistic regression model that was composed of these sleep disturbancesassociated genera was able to distinguish MHE patients with sleep disturbances from those without sleep disturbances. The most interesting nding was that higher relative abundances of Salivarius and Veillonella were respectively associated with sleep disturbances, and both may be novel diagnostic biomarkers and therapeutic targets for sleep disturbances in MHE patients.
Relevant investigations by Bajaj et al. have indicated that cirrhotic patients with MHE had higher relative abundances of Veillonella than those without MHE, indicating that the overgrowth of Veillonella may be correlated with MHE presence 20,21 . In this study, we have found that highly enriched Veillonella was associated with sleep disturbances in MHE patients. Meanwhile, functional modules related to lipopolysaccharide production were increased in the microbiome of MHE patients with sleep disturbances, according to the imputed metagenomic analysis by the PICRUSt. This may be explained by the reason that endotoxin produced by Veillonella may be involved in the pathophysiological mechanisms of sleep disturbances in MHE patients. In cirrhotic patients, the overgrowth of Veillonella is associated with increased production of endotoxin, the lipopolysaccharide in the outer membrane of Gram-negative bacteria 22 . Due to impaired intestinal barrier integrity, endotoxin translocates from the intestine to the circulatory system, in uences the permeability of the blood-brain barrier, and consequently induces microglial activation in the cerebrum 23,24 . A study using positron emission tomography has found that increased microglial activation in the substantia nigra is related to impaired dopaminergic function in the putamen, which may contribute to sleep disturbances during the idiopathic rapid-eye movement (REM) phase 25  Our study also revealed that highly enriched Salivarius was correlated with sleep disturbances in MHE patients. It has been suggested that the relative abundances of Salivarius were signi cantly higher in cirrhotic patients with MHE compared to those without MHE, and that the overgrowth of Salivarius was closely associated with MHE presence in patients with liver cirrhosis 27 . Similarly, our investigation indicated a correlation between plasma ammonia levels and relative abundances of Salivarius in MHE patients. Furthermore, we also found that functional modules associated with protein digestion and absorption were highly enriched in the microbiome of MHE patients with sleep disturbances, which has been attributed mainly to the involvement of Salivarius in the production of ammonia 27,28 . Salivarius has a large number of urea catabolism genes capable of activating urease activity, using urea-derived ammonium as its primary source of nitrogen in the intestine, which may contribute to ammonia accumulation and consequently to hyperammonemia 28 . Hyperammonemia induces astrocyte swelling and subsequent low-grade cerebral edema, and further interferes with the inhibition and activation of associated neural nuclei in the parietal cingulate and bilateral anterior, which play signi cant roles in the REM sleep regulation 29,30 . In the rat model of MHE, Llansola et al. reported that hyperammonemia led to signi cantly decreased REM sleep phases and increased sleep fragmentation 31 . Lactulose, which has been recommended for the treatment of MHE, has been reported to result in decreased levels of arterial ammonia and improvement of sleep quality in MHE patients 32 . These evidences suggest that lowering ammonia levels by inhibiting the increased abundances of Salivarius might be a possible alternative for the treatment of sleep disturbances in MHE patients.
This study had a few limitations that should be mentioned. First, although the MaAsLin analysis was used to adjust for age, gender, and BMI, the results of this study may have been in uenced by other possible confounders, including environmental factors and dietary habits. Second, the PICRUSt, which was used in the 16S rRNA gene sequencing and KEGG pathway analysis, was not able to accurately reveal the microbial composition and function at the species level, compared with the metagenomic sequencing 33 . Third, only patients with MHE caused by hepatitis B-related liver cirrhosis were included in our investigation; as such, ndings of our investigation may apply only to these patients and not to those with MHE caused by other aetiologies of liver cirrhosis. Finally, this was a single-center investigation. The ndings of this investigation should be con rmed and generalized in larger-scale, multi-center studies with MHE patients from different regions, in part because the composition and structure of the intestinal microbiota varies depending on where the host originates 34 .

Conclusion
Page 12/20 Our study comprehensively and systematically investigated the intestinal microbiota of MHE patients with and without sleep disturbances. We have found that the increased relative abundances of Salivarius and Veillonella were associated with sleep disturbances in MHE patients. Salivarius and Veillonella may be potential diagnostic biomarkers and therapeutic targets for sleep disturbances in patients with MHE caused by hepatitis B-related liver cirrhosis.

CONFLICTS OF INTEREST
The authors have no con icts of interest to declare.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.