Biases and complementarity in gut viromes obtained from bulk and virus-like particle-enriched metagenomic sequencing

Abstract

Due to varying sequencing strategies, current gut virome findings show significant variability. Specifically, bulk- and virus-like particle (VLP)-enriched metagenomic sequencing (termed bulk and VLP, respectively) present unique advantages and limitations, affecting viral genome discovery, taxonomic annotation, and community structure analysis. A comprehensive comparison of these strategies is crucial for thoroughly understanding the gut virome. This study comprehensively compared gut viromes identified from paired bulk and VLP data from 151 adult and 141 infant fecal samples. The VLP method showed superior performance to bulk in viral genome discovery in both data sets by recovering longer and more complete viral genomes, with higher sensitivity for low-abundant ones, resulting in a higher taxonomic annotation rate. However, we observed no correlations in the viral community structure (i.e., Shannon diversities) between bulk- and VLP-derived viromes, implying biases introduced during VLP enrichment. Such biases could be caused by the bacterial host features, such as the structural differences in cell walls and the prevalence and abundance of the viruses. Viruses that are of low prevalence, low abundance, or have Gram-positive bacteria as their hosts were enriched in VLP-derived viromes, in both the adult and infant data sets. Significant complementarity was observed between bulk and VLP viromes, with only about a quarter (26.7% in infants; 29.3% in adults) of VLP-viral genomes overlapping with bulk viruses. Together, our study identifies causal factors underlying the biases of bulk and VLP strategies in human gut virome studies and advocates the use of both strategies to enhance a comprehensive understanding of gut viromes.

Importance: The two mainstream gut phageome profiling strategies, namely bulk and virus-like particle (VLP), generated significantly overlapped results and have their own merits and drawbacks. Particularly, VLP exhibits higher efficiency in obtaining more, longer, and more complete viral genomes. However, VLP sequencing has the potential to alter the natural structure of viral communities, often resulting in the identification of viruses with lower prevalence and those specifically associated with Gram-positive bacterial hosts. While bulk metagenome features a more stable and diverse community, which can well reveal the interactions between viruses and bacteria. Nevertheless, bulk sequencing can suffer from lower coverage, leading to fragmented sequences and potentially missing some viral species. Therefore, it is essential to recognize that these methods are complementary rather than competitive in the comprehensive characterization of the gut phageome.

Keywords: bulk-metagenomic sequencing; gut health; gut phage; gut virus; viral-like particle metagenomic sequencing; virome.

Fig 1

Profiling of the human gut virome by bulk-metagenome sequencing (bulk) and viral-like particle-metagenome sequence (VLP). (A), Overall experimental design and process of the analysis. (B), Accumulation curve of vOTUs with sample number derived from the two methods. The right panel is infant data sets, and the left panel is adult data sets. (C) The distribution of different levels of completeness was determined by CheckV. The right panel is infant data sets and the left panel is adult data sets. (D) The left panel shows the comparison of the length in viral genomes by bulk and VLP. The right panel shows the length distribution of viral genomes divided into equal-length bins for both the bulk and VLP methods. The right panel is infant data sets, and the left panel is adult data sets. Color refers to groups: bulk (red) and VLP (blue). (E) The similarity of vOTUs of bulk and VLP by ANI > 95% (left for infants, and right for adults). Colors refer to groups: only (gray) and overlap (red).

Fig 2

Distinct community characteristics of the human gut virome by bulk-metagenome sequencing (bulk) and viral-like particle-metagenome sequence (VLP). (A), Comparison of within-sample Shannon diversity of vOTUs (left for infants, and right for adults). Paired Wilcoxon rank-sum test: ns.: P > 0.05; *P < 0.05; **P < 0.01, ***P < 0.001, ****P < 0.0001. Colors refer to groups: bulk (red) and VLP (blue). (B and C), Spearman’s rank correlation of alpha diversity (Shannon index, all of the virome on the left and the overlap of the virome on the right) between bulk-virome and VLP-virome with two sides. B refers to infants, and C refers to adults.

Fig 3

The distribution of abundance and prevalence of the human gut virome by bulk-metagenome sequencing (bulk) and viral-like particle-metagenome sequence (VLP). (A), The distribution of the relative abundance of vOTUs (all) (left for infants, and right for adults). (B), The relative abundance of the most abundant vOTUs in each sample (paired Wilcoxon rank-sum test, left for infants, and right for adults). (C and D) The comparison of the relative abundance of the vOTUs of bulk-only, bulk-overlap, VLP-overlap, and VLP-only (“only” refers to the vOTUs that only appeared in bulk or VLP; “overlap” refers to the vOTUs that appeared in both bulk and VLP). (E and F) The comparison of the prevalence of the vOTUs of bulk-only, bulk-overlap, VLP-overlap, and VLP-only. Wilcoxon rank-sum test: ns.: P > 0.05; *P < 0.05; **P < 0.01, ***P < 0.001, ****P < 0.0001. Colors refer to groups: bulk (red) and VLP (blue).

Fig 4

The consistency of taxonomy annotation for the human gut virome by bulk-metagenome sequencing (bulk) and viral-like particle-metagenome sequence (VLP). (A) The annotation rate in different levels of the vOTUs by bulk and VLP (left for infants, and right for adults). (B) Viral composition of the vOTUs by bulk and VLP at the phylum level (left for infants, and right for adults). (C and D) The correlation of the consistent rate of the two methods and prevalence in the genus level (the left panel is bulk and the right panel is VLP; C is infants and D is adults).

Fig 5

Enrichment and characterization of phages in bulk-metagenome sequencing (bulk) and viral-like particle-metagenome sequence (VLP) from infant and adult samples. (A) Enrichment of genera in bulk and VLP samples from infants and adults: The left panel shows that 11 genera are consistent between infant and adult data sets, while the remaining genera are not consistent. The middle panel illustrates the lifestyle characteristics of these genera, and the right panel depicts the host features associated with these genera. (B and C) Comparison of the abundance of phages targeting Gram-positive vs. Gram-negative bacteria in bulk and VLP samples. The left panel is infants, and the right panel is adults. (D and E) Effect size (i.e., explained variance or R²) explained by median abundance, prevalence, lifestyle, and host features as determined by PERMANOVA, using Bray-Curtis distances between all viral community profiles within each method group (VLP vs bulk) to assess the contribution (top for infants, and bottom for adults). FDR < 0.001 (***) refers to the significance of each factor, and FDR > 0.05 (ns.) refers to no relevance.