Oral Colonisation of the Gut May Explain the Cirrhosis Virome Signature

To the Editor,

1.

I read with interest the study by Chang et al. [1] who reanalysed publicly available faecal metagenomes, mainly the PRJEB6337 cirrhosis cohort of Qin et al. [2], and reported reduced gut viral richness, a virus‐bacteria correlation network of 130 rather than 509 edges, and random forest classifiers reaching an area under the curve of 0.938 internally and 0.773 on external validation [1]. What the analysis did not address is that its discriminatory signal is carried by organisms that do not normally reside in the gut. The 6 highest‐ranking features of the combined model were Veillonella parvula , Veillonella dispar , Veillonella atypica , Haemophilus parainfluenzae , Streptococcus salivarius and Streptococcus parasanguinis , all oral commensals, and Streptococcus and Escherichia appeared as predicted hosts only among the cirrhosis‐enriched vOTUs.

In another study by Jin et al. [3], next‐to‐identical Veillonella and Streptococcus strains from paired saliva and faeces in advanced chronic liver disease were isolated, with an average nucleotide identity above 99.98%. The study found that these translocators uniquely encode a collagenase‐like prtC gene and showed that oral isolates gavaged into carbon tetrachloride‐treated mice worsened gut barrier disruption and hepatic fibrosis [3]. Applied to the same PRJEB6337 samples, cumulative faecal prtC abundance was 20.6‐fold higher in cirrhosis and separated patients from controls with an area under the receiver operating characteristic curve of 0.93. One bacterial gene therefore matches what a 98‐feature virome‐bacteriome model achieves in the same cohort, which suggests that the present classifier gives a lower‐resolution interpretation of oral colonisation of the gut.

This is not a reason to set the virome aside but to ask a different question. A phage recovered from stool carries the signature of the niche in which it last replicated, so phages of oral hosts could date that colonisation instead of merely detecting it. Pursuing this needs paired salivary and faecal sampling with strain‐level tracking, annotation of lytic against temperate lifestyle, and medication histories. Van Espen et al., using purified viral‐like particles, found phage diversity rising within patients at the onset of acute‐on‐chronic liver failure (p = 0.023) together with a non‐significant move towards temperate phages [4], while Bajaj et al. found phage Shannon diversity indistinguishable between cirrhosis and controls (p = 0.79) and saw linkages centred on urease‐producing Streptococcus collapse after 8 weeks of rifaximin without any change in individual taxa [5]. PRJEB6337 offers neither saliva nor drug exposure, and I believe that is the gap the next study should close.

Funding

The author has nothing to report.

Disclosure

Use of large language models, AI and machine learning tools: None.

Conflicts of Interest

The author declares no conflicts of interest.

Linked Articles

This article is linked to Chang et al. paper. To view this article, visit https://doi.org/10.1111/liv.70836.

Data Availability Statement

Data sharing not applicable to this article as no datasets were generated or analysed during the current study.