Dysbiosis of small intestinal microbiota in liver cirrhosis and its association with etiology

Cirrhosis-associated duodenal dysbiosis is not yet clearly defined. In this research, duodenal mucosal microbiota was analyzed in 30 cirrhotic patients and 28 healthy controls using 16S rRNA gene pyrosequencing methods. The principal coordinate analysis revealed that cirrhotic patients were colonized by remarkable different duodenal mucosal microbiota in comparison with controls. At the genus level, Veillonella, Megasphaera, Dialister, Atopobium, and Prevotella were found overrepresented in cirrhotic duodenum. And the duodenal microbiota of healthy controls was enriched with Neisseria, Haemophilus, and SR1 genera incertae sedis. On the other hand, based on predicted metagenomes analyzed, gene pathways related to nutrient absorption (e.g. sugar and amino acid metabolism) were highly abundant in cirrhosis duodenal microbiota, and functional modules involved in bacterial proliferation and colonization (e.g. bacterial motility proteins and secretion system) were overrepresented in controls. When considering the etiology of cirrhosis, two operational taxonomic units (OTUs), OTU-23 (Neisseria) and OTU-36 (Gemella), were found discriminative between hepatitis-B-virus related cirrhosis and primary biliary cirrhosis. The results suggest that the structure of duodenal mucosa microbiota in cirrhotic patients is dramatically different from healthy controls. The duodenum dysbiosis might be related to alterations of oral microbiota and changes in duodenal micro-environment.

Gut microbiota and bacterial translocation (BT) play an important role in the pathogenesis of complications of cirrhosis 1 . Although a lot of work has been done to characterize the gut microbiota in cirrhosis, fecal samples were used in most of these studies. The small intestine has been suggested as the preferred site for BT in cirrhosis 2 . Small intestinal bacterial overgrowth (SIBO) has the greatest potential for promoting BT. This is supported by the study showing bacterial translocate predominantly from the small intestines when inoculating equal concentrations of E.coli into small or large intestines 3 . Moreover, reduction in BT rates is associated with lower jejunal, but not cecal, bacterial counts in cisapride-treated cirrhotic rats with an increased intestinal motility 4 . SIBO is traditionally defined if bacterial culture more than 10 5 CFU/ml in upper jejuna aspirate. However, the majority of intestine microbiota could not be cultured, so the culture-based method cannot reveal the real changes of microbiota in the small intestines 6 .
Recently, the duodenal microbiota has been analyzed in a few cohort studies using 16S rRNA gene pyrosequencing method. Duodenal samples demonstrated greater biological diversity and possessed a unique microbial signature compared with the rectum 7 . The duodenal microbiota of obese individuals was found to have a higher proportion of anaerobic genera and a lesser proportion of aerobic genera in comparison with normal-weight individuals 8 . Altered duodenal microbiota composition was also demonstrated in celiac disease patients suffering from persistent symptoms on a long-term gluten-free diet 9 . The treated patients with persistent symptoms had reduced microbial richness with higher relative abundance of Proteobacteria and lower abundance of Bacteroidetes and Firmicutes. In 2011, Steed et al. analyzed the microbial composition of duodenal biopsies samples in cirrhosis by real-time PCR, but found no significant difference between controls and cirrhosis 10 . The results might be hampered by the small sample size, as only 6 distal duodenal biopsies were evaluated in their study.
Compositional and functional analysis of duodenal microbiota in cirrhosis. Principal coordinate analysis based on weighted UniFrac matrices revealed remarkable differentiation of bacterial communities between cirrhosis and controls (Fig. 1a). When comparing the within-group variance, it was observed that the duodenal microbiota in cirrhosis have significantly higher variance than in controls (p < 0.01), which suggests that duodenal microbiota across cirrhosis patients are less similar than across healthy individuals (Fig. 1b).
To identify the distinguishing taxa in the duodenal microbiota of cirrhosis and controls, linear discriminant analysis effect size (LEfSe) were performed based on the RDP taxonomy data (Fig. 1c). At the phylum level, Firmicutes was significantly enriched in cirrhosis group. And Proteobacteria and SR1 were more abundant in control group than in cirrhosis group. At the genus level, Veillonella, Megasphaera, Dialister, Atopobium, and Prevotella were found overrepresented in cirrhosis duodenum. And the duodenal microbiota of healthy controls was enriched with Neisseria, Haemophilus, and SR1 genera incertae sedis. To investigate the functional potential of duodenal microbiota, phylogenetic investigation of communities by reconstruction of unobserved states (PICRUSt) analysis was applied to generate the prediction profiles of kyoto encyclopedia of genes and genomes modules. Functional modules were compared between cirrhosis and controls with LEfSe (Fig. 1d). The functional modules involved in bacterial motility proteins and secretion system were highly abundant in healthy control group. Additionally, modules for arginine and proline metabolism, valine, leucine and isoleucine biosynthesis were also enriched in control group. Cirrhosis was associated with increased abundance of modules involved in transporters and amino sugar and nucleotide sugar metabolism. Functional modules involved in peptidases and fructose and mannose metabolism were also overrepresented in cirrhosis group.
Key OTUs in the duodenal microbiota of cirrhosis. Using partial least squares-discriminant analysis (PLS-DA), there were 12 OTUs identified as the key lineages contributing to the differentiation between cirrhosis and control duodenal microbiota. A clear separation between cirrhosis group and control group can be obtained by non-metric multidimensional scaling method based on these 12 key OTUs (Fig. 2a). Seven of these OTUs were found prevalent in cirrhosis group, along with other 5 OTUs declined in cirrhosis group (Fig. 2b). The OTU with the largest contrition to differentiation (OTU-2 with the variable importance in the projection value 2.3) belonged to Neisseria at the genus level.
OTUs associated with etiology of cirrhosis. HBV-related cirrhosis and PBC were two main etiologies of cirrhosis in this study. When considering the etiology of cirrhosis, no significant difference was observed between HBV-related cirrhosis and PBC at the genus level or above. However, when LEfSe was applied on the OTU profile, there were two OTUs, OTU-23 (Neisseria) and OTU-36 (Gemella), discriminative between the two types of cirrhosis (Fig. 3a). Both of these two OTUs showed higher relative abundance in PBC than in HBV-related cirrhosis patients (Fig. 3b). Microbiota alterations and other clinical factors. All the cirrhotic patients were detected with esophageal and gastric varices. Among of them, 12 cirrhotic patients had received endoscopic treatment for esophageal or gastric varices. The microbiota composition was compared between patients with endoscopic treatment (n = 12) versus those without (n = 18). Generally, the duodenal microbiota was similar at the family level or above. Endoscopic treatment was not associated with alterations of predicted functional genes. At the genus level, SR1 genera incertae sedis was found significantly higher in patients with endoscopic treatment than those without (median 0.23% in treated group versus 0.03% in untreated group, p = 0.04) (Fig. 4a). There was a moderate decrease of genus Staphylococcus in treated group than in untreated group (median 0.04% in treated group versus 0.20% in untreated group, p = 0.059) (Fig. 4b).
Microbiota composition between patients on proton pump inhibitors (PPIs) (n = 12) and those without (n = 18) was also compared. No significant difference was observed at the family level or above. PPIs did not change the profile of predicted functional genes. At the genus level, Cloacibacterium was significantly reduced in patients on PPIs (median 0.15% in patients on PPIs versus 0.03% in those without, p = 0.03) (Fig. 4c). Patients on PPIs were found to have moderate higher relative abundance of Dialister than those without (median 0.20% in patients on PPIs versus 0.06% in those without, p = 0.054) (Fig. 4d). The OTUs in the red oval (7 OTUs) showed higher abundance in cirrhosis group. The OTUs in the green oval (5 OTUs) showed higher abundance in control group. (b) Heat map of relative abundance for the 12 key OTUs identified by PLS-DA. The relative abundance of each OTU in each subject of this study was used to plot the heat map. Each column represents one subject. The group information was showed under the plot: cirrhosis patients on the left with red line, controls on the right with black line. Each row represents one OTU. The genus affiliation of the OTU was showed on the right. The OTUs in red were found enriched in cirrhosis group, and in green enriched in controls.

Discussion
Studies of the duodenal microbiota have been focused on SIBO, which relies on traditional culture-dependent method or breath testing. Cirrhosis-associated duodenal dysbiosis is not yet clearly defined. In the attempt to shed light on this issue, we used 16S rRNA metagenomics to determine whether the duodenum microbiota differed between cirrhotic patients and healthy controls. Our results suggest that the structure of duodenal mucosa microbiota in cirrhotic patients is dramatically different from normal controls.
As can be observed in this study, at the genus level, Veillonella, Prevotella, Neisseria, and Haemophilus, were the most discriminative taxa between cirrhosis and controls. All these taxa are commonly presented in the oral cavity 11 , which suggests oral microbiota has a great impact upon duodenal microbiota. The oral cavity is the entry point of bacteria into the body. The human oral microbiota not only play a role in disease of the oral cavity, but also interact with microbiomes from other parts of the human body 12 . Our earlier study has found that microbes of oral origin could be present in stool of cirrhosis patients 13 . Recently, a direct evaluation of the salivary microbiome in controls and patients with cirrhosis was performed by Bajaj et al. 14 . In salivary microbiome of cirrhotic patients with previous hepatic encephalopathy, relative abundance of autochthonous taxa (Lachnospiraceae and Ruminococcaceae) decreased whereas potentially pathogenic taxa (Prevotella and Fusobacteriaceae) increased. Partly in consistent with their results, our results also found the relative abundance of duodenal Prevotella and Fusobacterium in cirrhosis was significantly higher than in controls. It has been reported in a previous study that distinct bacterial populations in the oral microbiota are involved in production of high levels of H 2 S and CH 3 SH in the oral cavity. The H 2 S group showed higher proportions of the genera Neisseria, Porphyromonas and SR1, whereas the CH 3 SH group had higher proportions of the genera Prevotella, Veillonella, Atopobium, Megasphaera, and Selenomonas 15 . It is interesting that the duodenal bacterial groups enriched in cirrhosis and healthy controls are highly consistent with the oral microbiota in CH 3 SH group and H 2 S group, respectively. Blood levels of CH 3 SH have been suggested as important factors in the pathogenesis of hepatic encephalopathy 16 . The shift of To investigate if, in addition to compositional impact, there is an impact on bacterial function, PICRUSt was applied to the 16S rRNA amplicon sequencing data to infer bacterial metabolic functions. Interestingly, we found higher abundance of bacterial motility proteins and bacterial secretion system in healthy duodenal microbiota. The motility proteins play an important role in bacterial attachment on epithelial cells and travel to or away from stimulus 18 . Bacterial secretion system, which can be classified into type I-IV, operates generally on the principal of active transportation of protein from cytoplasm to bacterial surface, which play crucial roles in gut colonization through invasion on mucosal surface 19 . Both bacterial motility and secretion system are heavily involved in host adhesion and colonization. The enrichment of genes related to bacterial motility and secretion system might indicate a harsher and more competitive environment in healthy duodenum than in cirrhotic duodenum. In normal conditions, intestinal peristalsis, gastric acid, and bile secretion act to control control bacterial colonization, attachment, and infiltration into the host 20 . Abnormalities in one or more of these host defenses result in bacterial overgrowth of the small intestine 2 . In cirrhosis, marked decreases in intestinal intraluminal concentrations of bile acids have been ascribed to decreased secretion and increased deconjugation 21 . Abnormalities in small intestinal motility are related to the degree of chronic liver failure 22 . Decompensated cirrhotics were found to have slower intestinal transit times as compared with compensated cirrhotics and healthy controls 23 . Small intestine motility dysfunction is more severe in those with history of spontaneous bacterial peritonitis 24 . In our research, cirrhotic patients were observed with significantly higher inter-individual variations than healthy controls, which also supports the conclusion that cirrhosis abolish colonization control to certain exogenous bacteria and weaken the normal control of endogenous bacterial community.
The enriched pathways in cirrhosis were related to transporters, amino sugar and nucleotide sugar metabolism, which likely reflect the basic requirements of microbial life in the duodenum of cirrhosis. The bacterial transport systems enable bacteria to accumulate needed nutrients, extrude unwanted by products and maintain cytoplasmic content of protons and salts conducive to growth and development 25 . Some ABC transporter can be involved in resistance to antimicrobial peptides 26 . The drug transporters genes are overrepresented in the infant/ elderly gut microbiome, perhaps due to the frequent antibiotic treatment of infants and the elderly compared with adults 27 . Therefore, we speculate that the enrichment of transporter gene might be a selective result of occasionally antibiotics use in cirrhotic patients, who are more susceptible to infections. Although, at the genus level, Neisseria were found overrepresented in healthy controls. Two OTUs representing Gemella and Neisseria, respectively, were the most discriminative OTUs between two types of cirrhosis. The results here are in line with our recently published data, which showed alterations and correlations of the gut microbiome and immnunity in PBC patients 28 . The fecal microbiota of PBC patients were depleted of some potentially beneficial bacteria, such as Lachnobacterium and Acidobacteria, but were enriched in some bacterial taxa, such as Neisseriaceae and Klebsiella. Several altered gut bacterial taxa exhibited interactions with altered immunity and urine metabolism, such as Klebsiella with IL-2A and Neisseriaceae with urinary indoleacrylate. A close association between celiac disease and PBC has been extensively reported in literature 29 . In duodenum of adult celiac patients, members of Neisseria genus were significantly more abundant in active celiac disease patients than in controls 30 . Gemella has been found to be involved in pulmonary exacerbations of cystic fibrosis patients. The relative abundance of Gemella in airway microbiota increased in 83% of the patients with cystic fibrosis at exacerbation and was found to be the most discriminative genus between baseline and exacerbation samples 31 . It is hypothesized that specific species of Neisseria and Gemella may be involved in orchestrating inflammatory disease by promoting inflammation and remodeling normally benign microbiota into dysbiotic community. Recently, Sabino et al. found that primary sclerosing cholangitis is associated with alterations in intestinal microbiota independently of comorbidity with inflammatory bowel disease, which might indicate a potential effect of cholestatic and bile flow on gut microbiota 32 . Further studies are needed to confirm and assess these links and causality.
Our results showed that the duodenal microbiota is primary determined by cirrhosis itself. Only slight association can be observed between duodenal microbial alterations and endoscopy varices treatment or PPIs. An increase in the relative abundance of genus Dialister was observed in cirrhotic patients on PPIs. Certain Dialister species (D. pnermosintes, D. invisus) have been identified as pathogens, mainly in orthodontic infections 33 . It is hypothesized that PPIs could enhance gastrointestinal proliferation of these potential oral pathogen species by reduction in gastric acid. This hypothesis is supported by recent findings showing long-term PPIs use is associated with an increase of Holdemania filiformis 34 , which is also potential oral pathogens usually isolated from advanced periodontitis 35 . In the present study, genus SR1 genera incertae sedis and Staphylococcus were found increased and decreased in patients with endoscopic treatment, respectively. Endoscopic sclerotherapy and band ligation of esophageal varices have been showed to cause gastric hemodynamic changes, and thus increase the incidence and the severity of portal hypertensive gastropathy 36 . The duodenal microbiota alterations observed here might be related to mucosa changes in advanced portal hypertension. However, the present study was not designed to compare microbiota alterations before and after endoscopy treatment or PPIs use. Further large scale studies should be performed to confirm these links.
In conclusion, the study reported herein demonstrates that marked dysbiosis is associated with the duodenal mucosa of cirrhotic patients. The deviation of duodenum microbiota might be related to alterations of oral microbiota and duodenal microenvironment. Also, when considering the etiology of cirrhosis, PBC and HBV have slight difference of gut microbiota, indicating the effect mediated by immunity or bile acids between host and microbiota. We note that our study is only able to describe correlations between cirrhosis and duodenal microbiota, and causality cannot be inferred. Further studies, perhaps using experimental animals, are expected to shed light on the causal factors underlying this relationship.

Methods
Patients. The patients and controls recruitment and sampling were carried out at the first affiliated hospital of Zhejiang University. A total of 30 cirrhotic patients and 28 healthy subjects were recruited prior to attending for a pre-arranged upper gastrointestinal tract endoscopy in the hospital. All individuals underwent complete evaluation including biochemical tests of liver functions and ultrasongraphy. Cirrhosis was diagnosed on the basis of clinical and laboratory data supported by liver biopsy or ultrasongraphy. Patients were excluded if they were on or had received antibiotics in the last 2 months. Medication history in the last 2 months was recorded for each cirrhotic patient. There were 12 patients on PPIs. No other antacid medicine, rifaximin, or lactulose, was used in the cirrhotic patients. Child-Pugh score was used to assess the prognosis of cirrhosis 37 . Of the 30 patients included, there were 27 Child A, 2 Child B, and 1 Child C. None of the patients had a history of hepatic encephalopathy. All the cirrhotic patients were detected with esophageal and gastric varices. A total of 12 cirrhotic patients had received previous endoscopy varices sclerotherapy or ligation. One biopsy sample was taken from the distal duodenum of each individual. All participants provided written informed consent prior to entering the study. The study conformed to the ethical guidelines of the 1975 Declaration of Helsinki. The Institutional Review Board of the First Affiliated Hospital of Zhejiang University approved the study protocol on November 11 th , 2014.
Scientific RepoRts | 6:34055 | DOI: 10.1038/srep34055 DNA extraction from the mucosa. Total genomic DNA was extracted from the biopsy samples by using a combination of the QIAamp DNA isolation kit (Qiagen, Valencia, CA, USA) and a bead-beating method. Briefly, the biopsy sample was lysed in 180 μ l of QIAamp ATL buffer and 20 μ l of proteinase K for 1 h at 56 °C. Glass beads of different diameters (0.1 mm, 0.5 mm and 1 mm, Sigma, St. Louis, MO, USA) were added, and samples were homogenized in a FastPrep FP120 bead beater (Bio 101, Morgan Irvine, CA, USA) for 30 sec at 4 m/s and incubated for an additional hour at 56 °C. Four ml of RNase A (100 mg/ml) and 200 μ l of AL buffer were added to the lysate, and samples were incubated for 30 min at 70 °C. After the addition of 200 μ l absolute ethanol, lysates were purified over a QIAamp column as specified by the manufacturer. Samples were eluted in 200 μ l of AE buffer.
16S rRNA gene pyrosequencing. PCR amplification of the bacterial 16S rRNA gene V1-V3 region was performed using universal primers (27F 5′ -AGAGTTTGATCCTGGCTCAG-3′, 533R 5′-TTACCGCGGCTGCTGGCAC-3′ ). Details for PCR conditions, DNA purification were described previously 38 . Massive partial reads of 16S rRNA gene generated by the 454 GS FLX Titanium sequencer were initially trimmed for quality using standard the software tools from Roche/454. Bioinformatics and statistics. The 16S rRNA reads were processed and compared using QIIME pipeline 1.7.0 39 . The raw reads were trimmed using a minimum read length of 200 bp and an average quality score of 25. Two mismatches were allowed along the primer sequences. The number of hompolymers authorized in sequences was limited to 6. OTUs were picked using denovo OTU picking protocol with a 97% similarity threshold. The most abundant sequence of each OTU was selected as representative reads, and compared to the RDP classifier (cutoff = 0.5) taxonomy assignments of OTUs. The chimera identification sequence was performed by USEARCH 40 .
The QIIME results were imported into Phyloseq, an R package, for manipulation, subsampling normalization, and graph visualization 41 . The software PICRUSt was used to make functional gene content predictions based on 16S rRNA gene data present in the Greengenes database 42 . LEfSe was used to elucidate taxa and genes associated with healthy or diseased states. Both PICRUSt and LEfSe are freely available online in the Galaxy workflow framework 43 . To compare the alpha diversity and weighted unifrac distance between groups, the Two-sided student's t-test was used for significance test. PLS-DA with the variable importance in the projection parameter was used to explore key OTUs contributing to differentiation between groups. Variable with the variable importance in the projection parameter > 1.5 is defined as key OTUs associated with differentiation. Mann-Whitney U test was applied to compare genera abundances between different groups, with multiple testing correction whenever applicable (adjustment for false discovery rate). Adjusted p values < 0.05 were considered significant. The t-test, non-parametric test, and PLS-DA projection were conducted in R (V.3.1.3) with package "mixOmics", "plyr" and "reshape 2".