We report a panel of 30 features, including 27 bacterial features with discriminatory ability to detect NAFLD-cirrhosis using a Random Forest classifier model. In a derivation cohort of probands, the model has a robust diagnostic accuracy for detecting NAFLDcirrhosis, confirmed in a validation cohort of relatives of proband with NAFLD-cirrhosis . This study provides evidence for a fecal-microbiome-derived signature to detect NAFLDcirrhosis.NAFLD is the most prevalent cause of chronic liver disease worldwide . NAFLDcirrhosis represents the most severe stage of the disease, carries a significant risk of hepatocellular carcinoma , and is consistently identified as the most important predictor of liver-related morbidity-mortality in NAFLD . However, non-invasive, accurate and easy-to-perform modalities for early detection of NAFLD-cirrhosis remain a major unmet need in the field. Overthe last decade, the gut-liver axis has emerged as a pivotal component of NAFLD and represents a potential source of non-invasive biomarkers for the detection and stage of liver disease . Limited data are available regarding the diagnostic accuracy of a stool microbiome derived signature for detecting NAFLD-cirrhosis. We previously demonstrated that first-degree relatives of probands with NAFLD-cirrhosis have a high risk of AF . However, factors associated with progression towards NAFLDcirrhosis among families remain obscure. Although earlier studies reported familial aggregation of NAFLD and NAFLD-related cirrhosis ,square pots and demonstrated that both liver steatosis and fibrosis are heritable , known genetic risk only accounts for ~10-30% of the variance observed in NAFLD .
This suggests an additional role for environmental factors, which predominate over genetic factors in shaping the human gut-microbiome . Heritability of gut-microbiome features has been reported in twins studies, but limited data exist regarding the similarity of gut-microbiome composition among family members, and whether similar microbiomes associate with disease traits especially in the entire spectrum of NAFLD including NAFLD-cirrhosis.Using a uniquely phenotyped twin and family study design including well-characterized participants with and without NAFLD, assessed using MRI-proton-density-fat-fraction for quantifying hepatic steatosis and MR-elastography for quantifying liver fibrosis , we examined familial correlation of gut-microbiome composition and tested whether a non-invasive stool-microbiome-derived signature accurately detects NAFLD-cirrhosis. This cross-sectional analysis included 203 well-characterized, prospectively recruited participants, encompassing the entire spectrum of NAFLD divided into three groups paired with their first-degree relatives. Subjects included 26 probands with NAFLD-cirrhosis and 37 of their first-degree relatives, 18 probands with NAFLD without advanced fibrosis and 17 of their first-degree relatives, and 54 non-NAFLD normal controls and 44 of their firstdegree relatives. The detailed derivation of the study cohort is shown in Figure 2.S1. The detailed demographic, biochemical, imaging data of the probands and first-degree relatives stratified by the metabolic and liver phenotype of the probands are provided in Supplementary Table 1 and Supplementary Table 21 , respectively.We identified a significant familial correlation of the gut-microbiome composition involving shared housing.
The gut-microbiome profile showed significant correlation within related pairs compared to random-unrelated pairs at the level of the phyla Figure 2.1a and at the level of 16S tag sequences Figure 2.1b. In our analyses at the phylum level, this familial correlation was mainly driven by significant correlation of Bacteroidetes and Actinobacteria between related individuals. Furthermore,phylogenetic dissimilarity assessed by unweighted UniFrac distances among related pairs was significantly lower than in random-unrelated pairs . When stratified by the liver phenotype of the proband, the phylogenetic dissimilarity remained significantly lower among nonNAFLD controls and relatives and probands with NAFLD without AF and relatives compared to unrelated pairs, while no significant difference was observed among probands with NAFLD-cirrhosis and relatives Figure 2.1c. These results suggest that familial gut-microbiome similarities are independent of mild/moderate liver phenotype but are impacted by severe stages of liver disease. Finally, related individuals with shared-housing had a lower phylogenetic dissimilarity than those who did not share housing Figure 2.1d. These results confirm a strong impact of the environment in the familial similarity of the gutmicrobiome and demonstrate that shared-housing is a major determinant that should be controlled for in study designs assessing the microbiome in liver disease. The gut-microbiome profile of NAFLD-cirrhosis was first assessed in a derivation cohort including the 3 groups of probands encompassing the entire spectrum of NAFLD. As shown in previous studies , α-diversity as measured by Faith’s phylogenetic diversity decreased with increase in liver damage severity . The β-diversity was lower among individuals with moderate liver damage compared to nonNAFLD controls , whereas it was higher among individuals with severe liver damage compared to probands with moderate liver damage Figure 2.2b.
This suggests an hourglass signature of disease severity in the gutmicrobiome, with an initial decrease in phylogenetic diversity associated with a moderate stage of the disease that progress towards a phylogenetic dispersion in individuals with severe stages of disease such as NAFLD-cirrhosis. Further investigations with larger sample sizes are needed to determine whether this phylogenetic dispersion reflects a distinct profile among NAFLD-cirrhotic patients, and whether it is associated with specific NAFLD-cirrhosis related outcomes.Several taxa were differentially abundant in NAFLD-cirrhosis compared to non-NAFLD controls. At the genus level, the NAFLD-cirrhosis group was enriched with Streptococcus and Megasphaera, whereas Bacillus and Lactococcus were enriched in the non-NAFLD controls Figure 2.2c. Source data are provided as a Source Data file2 . Species belonging to the family Enterobacteriaceae and the genera Streptococcus and Gallibacterium were the most enriched inNAFLD-cirrhosis, while Faecalibacterium prausnitzii and species belonging to the genus, Catenibacterium and the families Rikenellaceae, Mogibacterium, Peptostreptococcaceae were enriched in non-NAFLD controls. These results are consistent with the study performed by Ponziani and colleagues in an Italian cohort showing higher Enterobacteriaceae and Streptococcus in NAFLD-cirrhosis with and without HCC. In addition, it confirms a shift towards more Gramnegative microbes in advanced fibrosis stages, as previously reported in NAFLD . A stool-microbiome signature accurately detects NAFLD-cirrhosis. A Random Forest model comprised of 30 features identifies probands with NAFLD-cirrhosis. The bacterial features most important for predicting NAFLD-cirrhosis are shown in Figure 2.3. In a derivation cohort of probands, the model had a robust diagnostic accuracy, with an AUROC of 0.92 after cross-validation for detecting NAFLD-cirrhosis Figure 2.4a. The diagnostic accuracy of the model was then confirmed in a validation cohort of first-degree relatives of proband with NAFLD-cirrhosis with good diagnostic accuracy, with an AUROC of 0.87 for the detection of advanced fibrosis with a high negative predictive value of 91.6% Figure 2.4b. . In addition, we performed sensitivity analyses in another validation group enriched with mild to moderate stage of NAFLD including probands with NAFLD without AF. The diagnostic accuracy of the model was confirmed and yielded a very good diagnostic accuracy with an AUROC of 86% Figure 2.S2.In conclusion, using a unique twin and family study design including well-characterized participants with and without NAFLD, we identified a specific stool-microbiome-derived signature of NAFLD-cirrhosis that yielded a robust diagnostic accuracy for the detection of NAFLD-cirrhosis. Hence,large plastic pots this conveniently assessed microbial biomarker presents an adjunct tool to current invasive approaches to determine the stage of liver disease. We previously demonstrated that a microbial biomarker can detect AF in biopsy-proven NAFLD . The fundamental difference between the previous study and the present study is the clinical context of use of the gut-microbiome signature. In the present study, the clinical question is to accurately differentiate using a non-invasive gut-microbiome signature who among the first degree relatives have advanced form of NAFLD and who are unaffected in a general setting as opposed to a liver clinic setting. The context of use is critical for biomarker development as suggested by the BEST Guidelines by FDA. In order to address this clinically important question, this study leverages 2 distinct levels of innovations. 1. Study design: This innovative study leverages from a unique prospectively recruited case-control study design. This cross-sectional analysis included 203 well-characterized participants, encompassing the entire spectrum of NAFLD divided into three groups paired with their first-degree relatives. 2. Discovery of shared housing effect: the familial cohort design enabled us to discover the effect of shared housing on the gutmicrobiome signature related to NAFLD-cirrhosis. This unique familial cohort study design led us to a new discovery that shared housing had a dominant effect on microbiome. This novel effect would not be apparent from previous study in NAFLD among unrelated individuals. Hence, this study is novel due to its clinical context of use, study design, and co-housing effect on gut microbiome in families with NAFLD-cirrhosis. We acknowledge the following limitations of this study. This is a single-center study performed in a center with expertise in clinical investigation of NAFLD with advanced MRI-based phenotyping, and the generalizability of the findings in other clinical settings remains to be established. 16S rRNA sequencing may not have captured additional insights associated with the disease status available at the species or strain level. Finally, the association does not suggest causality, and additional studies are needed to assess whether these microbial species impact gut permeability and/or induce NAFLD progression through cross-talk between serum metabolites and the liver .
However, the strengths of the study include a prospective study design, detailed phenotyping of participants using the most accurate non-invasive imaging modalities available, and assessment of accuracy using AUROC in both a derivation and validation cohort. Further multi-center studies including a larger number of individuals are needed to validate the clinical utility of the proposed microbiome-derived signature to detect NAFLD-cirrhosis. This is a cross-sectional analysis of a prospective family cohort study of participants from the Familial Cirrhosis cohort and Twins and Family cohort who were participating in a biobank initiative and prospectively recruited at the University of California at San Diego NAFLD Research Center between December 2011 and December 2017. All participants underwent a standardized exhaustive clinical research visit including detailed medical history, physical examination, and testing to rule out other causes of chronic liver diseases , fasting laboratory tests at the University of California at San Diego NAFLD Research Center . Participants also underwent an advanced magnetic resonance examination including magnetic resonance imaging proton-density-fat-fraction and magnetic resonance elastography at the UCSD MR3T Research Laboratory forthe screening of NAFLD and advanced fibrosis . Participants from the Familial Cirrhosis cohort also underwent an ultrasound-based vibration controlled transient elastography assessment using a FibroScan. At the time of each research visit, patients provided stool samples. These were collected and immediately stored in a -80*C freezer. Written informed consent was obtained from every participant.Probands and first degree relatives had to be at least 18 years old. Probands were required to have a documented diagnosis of NAFLD-cirrhosis either by liver biopsy or by documented imaging evidence by a protocol-specified criterion. First-degree relatives with written informed consent who did not meet any exclusion criteria were included in the study. Exclusion criteria included: regular and excessive alcohol consumption within 2 years of recruitment ; use of hepatotoxic drugs or drugs known to cause hepatic steatosis; evidence of liver diseases other than NAFLD, including viral hepatitis , Wilson’s disease, hemochromatosis, alpha-1 antitrypsin deficiency, autoimmune hepatitis, and cholestatic or vascular liver disease; clinical or laboratory evidence of chronic illnesses associated with hepatic steatosis, including human immunodeficiency virus infection , celiac disease, cystic fibrosis, lipodystrophy, dysbetalipoproteinemia, and glycogen storage diseases; evidence of active substance abuse, significant systemic illnesses, contraindication to MRI, pregnant or trying to become pregnant, or any other condition which, in the investigator’s opinion, may affect the patient’s competence or compliance in completing the study.The study included 140 participants from the Twin and Family study corresponding to 100 twins and 40 siblings or parents-offspring. The non-NAFLD controls included 54 probands and 44 first degree relatives and the group with NAFLD without AF included 18 probands and 17 first-degree relatives of community-dwelling controls either twin, sib-sib or parent-offspring pairs. These twin, sib-sib, and parent-offspring pairs were prospectively recruited and they reside in southern California. Twins without evidence of NAFLD and advanced fibrosis were considered as non-NAFLD control and twins with evidence of NAFLD without evidence of advanced fibrosis and their twin pair were randomly assigned as proband or first-degree relatives. The study was approved by the UCSD Institutional Review Board number 111282. Patients were included if they were twins, siblings or parent-offspring at least 18 years old, willing and able to complete all research procedures and observations. For each twin pair, a detailed assessment of twin ship status or dizygotic was obtained. The majority of twin-pairs were diagnosed by their physician as either MZ or DZ by genetic testing. Furthermore, twin-ship status was confirmed by using a previously published questionnaire .