Human liver microbiota modeling strategy at the early onset of fibrosis

Camille Champion, Radu Mircea Neagoe, Maria Effernberger, Daniela Sala, Florence Servant, Jeffrey E. Christensen, María Arnoriaga‐Rodríguez, Jacques Amar, Benjamin Lelouvier, Fabrice Gamboa, Herbert Tilg, Massimo Federici, José Manuel Fernández‐Real, Jean Michel Loubes, Rémy BurcelinbioRxiv 2020 · DOI

Abstract

To understand the pathophysiological impact of liver microbiota on the early stages of fibrosis, we identified corresponding microbial sequences while overcoming the challenges of disparate group sizes and patient origins through adapted statistical approaches. Liver samples with low fibrosis scores (F0, F1, F2) were collected from a multi-centric European cohort (n=82). By accounting for frequency, sparsity, and unbalanced sample sizes, we utilized multivariate analyses—including adapted spectral clustering with L1-penalty fair-discriminant strategies—to identify taxonomic profiles. Our findings reveal that 50% of liver taxa consist of Enterobacteriaceae and Pseudomonadaceae, with specific families like Caulobacteraceae and Flavobacteriaceae discriminating between F0 and F1 stages. Furthermore, metabolic pathways involving preQ0 biosynthesis and glycogen degradation were negatively associated with fibrosis progression. These results suggest a significant role for bacterial translocation in early fibrosis and demonstrate a robust statistical framework for identifying microbial signatures across heterogeneous datasets.

Introduction

Non-alcoholic fatty liver disease (NAFLD) has emerged as a primary global health concern, closely linked to the rising prevalence of obesity and type 2 diabetes. While the gut-liver axis is a well-established pathway in disease progression, the "tissue microbiota"—the specific bacteria residing within the liver itself—remains poorly understood. Investigating this niche is technically demanding due to the extremely low bacterial biomass in liver tissue and the significant influence of environmental factors, such as diet and geography, which can create "batch effects" that obscure biological signals.

This study addresses these challenges by analyzing liver biopsies from patients at the earliest stages of fibrosis (F0–F2). By implementing a novel "fairness" statistical modeling strategy, we aimed to filter out geographical noise and identify a universal microbial signature associated with the onset of liver scarring.

Methods and Materials

Patient Cohort

The study analyzed liver biopsies from 82 Caucasian patients across four European countries to ensure a diverse geographical representation:

  • Romania (n=36)
  • Italy (n=19)
  • Spain (n=17)
  • Austria (n=10)

Patients were categorized by histological fibrosis scores: F0 (no fibrosis, n=34), F1 (mild, n=37), and F2 (moderate, n=11).

Sequencing and Statistical Framework

Bacterial profiles were generated using 16S rDNA sequencing. To manage high-dimensional data and cohort-specific bias, the researchers employed a multi-layered statistical approach:

  • L1-Spectral Clustering with Fairness: This robust variant of spectral clustering was used to detect natural structures in the data while ensuring results were not biased by the patient's country of origin.
  • sPLS-DA (Sparse Partial Least Square Discriminant Analysis): This method selected the most discriminative microbial features to classify samples by fibrosis stage.
  • TF-IDF (Term Frequency-Inverse Document Frequency): Borrowed from computational linguistics, this technique highlighted rare but biologically meaningful taxa that might otherwise be masked by dominant "core" bacteria.
  • Functional Prediction: PICRUSt2 was utilized to predict the metabolic pathways of the identified microbiota, providing insight into the functional shifts accompanying fibrosis.

Key Insights

Overcoming Geographical Noise

Initial Principal Component Analysis (PCA) revealed that patients grouped more strongly by their country of origin than by their disease state. This confirmed that environmental and geographical factors were the primary drivers of microbial variation. By applying "fairness" algorithms, the researchers successfully decorrelated the data from these influences, allowing the subtle biological signature of early fibrosis to emerge.

Taxonomic Composition

Unlike the gut microbiota, which is typically dominated by Firmicutes and Bacteroidetes, the human liver microbiota was found to be overwhelmingly composed of Proteobacteria (>75%). Specifically, the Enterobacteriaceae and Pseudomonadaceae families accounted for half of the total taxa. Families such as Caulobacteraceae, Flavobacteriaceae, and Propionibacteriaceae were identified as the primary discriminators between the F0 (healthy) and F1 (mild fibrosis) stages.

Metabolic and Functional Shifts

The transition from F0 to F1-F2 was marked by distinct metabolic changes. Early fibrosis was associated with a decrease in pathways related to preQ0 biosynthesis (a folate precursor) and glycogen degradation. Conversely, the absence of fibrosis (F0) was linked to bacteria capable of synthesizing Vitamin K2 (menaquinones), suggesting that certain microbial metabolites may exert a protective effect against liver inflammation.

Results and Discussion

The statistical models achieved an AUC of 0.81 when using predicted metagenomics to discriminate between F0 and F1/F2 stages, indicating high diagnostic potential even at the very onset of the disease. The presence of gut-associated bacteria like Escherichia-Shigella in the liver tissue supports the "leaky gut" hypothesis, where intestinal barrier dysfunction allows bacteria to translocate to the liver, potentially triggering the inflammatory response that leads to fibrosis.

While the study is limited by a relatively small sample size for stage F2 (n=11) and its observational nature, it provides a powerful proof-of-concept for tissue microbiota analysis. The ability to identify microbial signatures despite significant batch effects opens new doors for non-invasive diagnostics. If these liver-specific signatures are reflected in systemic circulation, they could eventually supplement or replace invasive biopsies for early-stage screening.

Conclusion

This research establishes a robust methodology for analyzing low-biomass tissue microbiota in multi-centric studies. By identifying specific bacterial families and metabolic pathways—such as Vitamin K2 synthesis—linked to liver health, this work lays the groundwork for future "pharmabiotic" interventions aimed at halting the progression of fibrosis before it reaches advanced, irreversible stages.

Summary updated

Abstract

Abstract To understand the pathophysiological impact of liver microbiota on the early stages of fibrosis we identified the corresponding microbiota sequences and overcome the impact of different group size and patient origins with adapted statistical approaches. Liver samples with low liver fibrosis scores (F0, F1, F2) were collected from Romania(n=36), Austria(n=10), Italy(n=19), and Spain(n=17). The 16SrDNA gene was sequenced. We considered the frequency, sparsity, unbalanced sample size between cohorts to identify taxonomic profiles and statistical differences. Multivariate analyses, including adapted spectral clustering with L1-penalty fair-discriminant strategies, and predicted metagenomics were used to identify that 50 % of liver taxa were Enterobacteriaceae and Pseudomonadaceae. The Caulobacteraceae, Flavobacteriaceae and Propionibacteriaceae discriminated between F0 and F1. The preQ0 biosynthesis and pathways involving glucoryranose and glycogen degradation were negatively associated with liver fibrosis F1-F2 vs F0. Altogether, our results suggest a role of bacterial translocation to the liver in the progression of fibrosis. This statistical approach can identify microbial signatures and overcome issues regarding sample size differences, the impact of environment, and sets of analyses.
BibTeX
@misc{champion2020human,
  author = {Camille Champion and Radu Mircea Neagoe and Maria Effernberger and Daniela Sala and Florence Servant and Jeffrey E. Christensen and María Arnoriaga‐Rodríguez and Jacques Amar and Benjamin Lelouvier and Fabrice Gamboa and Herbert Tilg and Massimo Federici and José Manuel Fernández‐Real and Jean Michel Loubes and Rémy Burcelin},
  title = {{Human liver microbiota modeling strategy at the early onset of fibrosis}},
  year = {2020},
  doi = {10.1101/2020.12.10.419051},
  howpublished = {bioRxiv},
}
Figure