Dynamics of the viral community on the surface of a French smear-ripened cheese during maturation and persistence across production years
This article has been Reviewed by the following groups
Listed in
- Evaluated articles (Rapid Reviews Infectious Diseases)
- Evaluated articles (Arcadia Science)
Abstract
The surface of smear-ripened cheeses constitutes a dynamic microbial ecosystem resulting from the successive development of different microbial groups such as lactic acid bacteria, fungi and ripening bacteria. Recent studies indicate that a viral community, mainly composed of bacteriophages, also represents a common and substantial part of the cheese microbiome. However, the composition of this community, its temporal variations and associations between bacteriophages and their hosts remain poorly characterized. Here, we studied a French smear-ripened cheese by both viral metagenomics and 16S metabarcoding approaches to assess both the succession of phages and bacterial communities on the cheese surface during cheese ripening, and their temporal variations in ready-to-eat cheeses over the years of production. We observed a clear transition of the phage community structure during ripening with a decreased relative abundance of viral species (vOTUs) associated with Lactococcus phages, which were replaced by vOTUs associated with phages infecting ripening bacteria such as Brevibacterium, Glutamicibacter, Pseudoalteromonas and Vibrio . The dynamics of the phage community was strongly associated with bacterial successions observed on the cheese surface. Finally, while some variations in the distribution of phages were observed in ready-to-eat cheeses produced at different dates spanning more than 4 years of production, the most abundant phages were detected throughout. This result revealed the long-term persistence of the dominant phages in the cheese production environment. Together, these findings offer novel perspectives on the ecology of bacteriophages in smear-ripened cheese and emphasize the significance of incorporating bacteriophages in the microbial ecology studies of fermented foods.
IMPORTANCE
The succession of diverse microbial populations is critical for ensuring the production of high-quality cheese. We observed a temporal succession of phages on the surface of a smear-ripened cheese, with new phage communities showing up at the time when ripening bacteria start covering this surface. Interestingly, the final phage community of this cheese is also consistent over large periods of time, as the same bacteriophages were found in cheese products from the same manufacturer made over 4-years. This research highlights the importance of considering these bacteriophages when studying the microbial life of fermented foods like cheese.
Article activity feed
-
Celine Delbes
Review 1: "Dynamics of the Viral Community on the Cheese Surface during Maturation and Persistence across Production Years"
The reviewer highlighted the study's potential impact on the dairy sector, starter culture industry, and food policy makers, particularly in demonstrating the importance of bacteriophages in fermented food microbiology beyond lactic acid bacteria.
-
Celine Delbes
Review of "Dynamics of the Viral Community on the Cheese Surface during Maturation and Persistence across Production Years"
Reviewers: C Delbes (INRA) | 📘📘📘📘📘
-
-
-
You're right. Samples from the short-term study (dynamic) were collected directly by the manufacturer during the ripening process in 2019. Samples from the long-term study (persistence) were collected in the local retail shop, so after storage and transport. Marinomonas could therefore have developped during storage, which could explain the difference in terms of abundance between the two studies.
-
Thanks for the comment. Several statistical methods can be used on microbiome data for the detection of differentially abundant taxa, all having their own advantages and limitations. As one of them, DESeq2 (as well as other RNA-Seq based methods such as EdgeR) was shown to perform well for this purpose (ref: McMurdie and Holmes, 2014, https://doi.org/10.1371/journal.pcbi.1003531). Furthermore, DESeq2 has an official extension within the phyloseq package (https://joey711.github.io/phyloseq-extensions/DESeq2.html), which makes its use very convenient when working with phyloseq objects. Because we used phyloseq as the main R package in our analysis workflow, we naturally choose this solution for the differential abundance analysis. We added more information about this choice in the Methods section.
-
This information was partially available on the figure but not correctly explained. Indeed, NA (white color) in the “Group” column means that the corresponding vOTU doesn’t share enough sequence homology (< 30% identity x coverage Blat score) with a known phage. Similarly, “NA” in the “Host genus” and “Lifestyle” columns indicated that no prediction was available for the corresponding vOTUs. In the new version of the heatmaps (Figures 3 and 5), we replaced “NA” by “No hit” or “No prediction” to make the message more clear for the reader.
-
Unfortunately it is not possible to know what fraction of the total microbiome is viral since we purified/enriched the viral fraction from the original sample prior to sequencing. A shotgun metagenomic approach based on the total DNA isolated from our cheese samples would have been appropriate for this purpose. However, as indicated in the material and methods section, 332 (331 final vOTUs + 1 contig corresponding to PhiX) out of 3,122 dereplicated contigs >2 Kb were identified as viral (so ~10%). The percentage of reads mapping to the vOTUs was 91.8% in average (94.3% for the short-term dataset and 87.6% for the long-term dataset) indicating a low level of contamination with non-viral DNA. This information was added in Table 2 and mentioned in the text.
-
we added a new figure to illustrate some quality metrics about viral contigs (new Figure 2) and also a supplementary table with detailed annotations including completeness.
-
-
In the first venn diagramm we did not apply any threshold, so it could be one read. In the second diagram a vOTU was considered if it relative abundance was >0.005%.
-
yes, we analyzed three replicates per time point. This information was added at the end of the introduction, in the results and in Figure 1 which illustrates the experimental set-up.
-
Thank you for this positive comment. Lifestyle predictions in metagenomic datasets is not an easy task and can easily conduct to erroneous interpretation of the data. For exemple Lactococcus phage 949, one of the most abundant phage in our dataset, was detected as temperate while its true lifestyle is virulent (it indeed has an integrase gene encoded in its genome). So we added some statistics about the number of vOTUs detected as temperate phages versus virulent in the description of our dataset (first section of the results) but decided to not conduct further analysis in that way (abundance of virulent vs temperate in the different samples) for that reason.
-
at 56.6% of the 297 vOTUs present
Can you explain what the threshold is for a vOTU being present?
-
(Figure 2A).
assuming the three dots per timepoint are replicates? im sure it is mentioned in the methods but may be good to mention here
-
This result indicates that the cheese surface virome was mostly stable in terms of composition (presence/absence) and that dominant phages persist across productions years
This finding is really interesting! Especially since it looks like they are mostly virulent phages (it would be nice to quantify virulent vs temperate).
-
stability of the bacterial community composition over the three sampling campaigns
It seems like Marinomonas is consistently fairly abundant in these years compared to what you saw in the dynamic study. Any thoughts on why Marinomonas didn't show up in the dynamic study? Were the dynamic cheese samples taken in 2023?
-
We then applied the DESeq2 method to identify differentially abundant viral contigs
This is interesting - I'm not familiar with the use of DESeq for anything other than RNAseq data. Would it be possible, maybe in the methods section, to expand on how this is done and if there is anything to consider in using this on metagenomic contig abundance data vs RNAseq data? Or if there are any relevant references, include those?
-
These few vOTUs correspond to uncharacterized phages
It could be interesting to add another column to figure 2C that shows whether each vOTU is characterized or uncharacterized (not previously detected).
-
led to the production of a metavirome composed of 331 vOTUs >2 Kb
It would be great to know what fraction of the total microbiome the viruses represent, either in terms of # of contigs of viral vs non-viral contigs or % reads mapping to viral contigs vs non-viral.
-
metavirome composed of 331 vOTUs >2 Kb
It would be nice to have some more information here on the viral contigs that were assembled (e.g. the size distribution. did you see any particularly large genomes?). From the methods it looks like you did some assessments of genome quality/completeness, it would be nice to describe here also what proportion were low/high quality etc.
-