ALTERED GUT VIRAL ECOLOGY IN METABOLIC DYSFUNCTION-ASSOCIATED STEATOTIC LIVER DISEASE (MASLD)

Background:
Metabolic dysfunction-associated steatotic liver disease (MASLD) is highly prevalent and a leading cause of liver-related morbidity and mortality. Prior large scale studies have implicated the gut bacterial microbiome in MASLD etiopathogenesis, but few have assessed the role of gut viruses in this process.
Methods:
Our group recently developed a computational workflow to detect and quantify DNA and RNA viruses from host metagenomes and metatranscriptomes (BAQLaVa, Bioinformatic Application of Quantification and Labeling of Viral Taxonomy), which we applied to profile the gut metagenomes of 211 participants with MASLD and 502 age- and BMI- matched controls from a wider at-home stool collection nested within the Nurses’ Health Study II, yielding integrated host-microbiome-molecular profiles.
Results:
In total, we detected 6,385 unique DNA viral genome bins analogous to bacterial species-level taxa (VGBs). Notably, 224 (4%) were characterized or known VGBs (kVGBs), while 6,161 (96%) were unknown VGBs (uVGBs not found in current reference databases), quantifying the degree of so-called viral dark matter found even in frequently-studied environments such as the human gut. Collectively, these VGBs explained a small but statistically significant proportion of variance between MASLD cases and controls (R²=0.01, p-value=0.001). Similar to documented differences in overall bacterial ecology, MASLD was also associated with decreased viral alpha diversity compared to controls (Fig. A). We used multivariable linear modeling to evaluate differential abundance of viral taxa in MASLD vs. controls, finding 707 VGBs (11% of those detected) to be significantly enriched or depleted in cases (Fig. B). This high proportion is generally higher than most estimates of the number of distinguishing bacterial species. Unsurprisingly, the majority of these viral taxa were uncharacterized, though we did find significant enrichments of Eganvirus, Chivirus, and Punavirus in MASLD cases. Using a machine learning classifier, we found that using uVGBs as input resulted in more accurate discrimination between cases and controls (AUROC 0.65) than the subset of kVGBs (AUROC 0.60; Fig. C). We also found significant differences in the gut viromes of lean and non-lean MASLD cases compared to their respective referent counterparts with very little overlap in their respective viral signatures. This again represents a stark difference when compared to the generally similar bacterial alterations in lean and non-lean MASLD (Fig. D).
Conclusion:
Our work is among the most dense and complete evaluations of MASLD transkingdom ecology to-date. Next, we will interrogate relatively underexplored RNA viruses in MASLD, leveraging our sizable metatranscriptomic cohort, while more deeply investigating the interplay between the bacterial and viral constituents of the MASLD microbiome.