Study of the viral and microbial communities associated with Crohn's disease: a metagenomic approach

Abstract

Objectives: This study aimed to analyze and compare the diversity and structure of the viral and microbial communities in fecal samples from a control group of healthy volunteers and from patients affected by Crohn's disease (CD).

Methods: Healthy adult controls (n=8) and patients affected by ileocolic CD (n=11) were examined for the viral and microbial communities in their feces and, in one additional case, in the intestinal tissue. Using two different approaches, we compared the viral and microbial communities in several ways: by group (patients vs. controls), entity (viruses vs. bacteria), read assembly (unassembled vs. assembled reads), and methodology (our approach vs. an existing pipeline). Differences in the viral and microbial composition, and abundance between the two groups were analyzed to identify taxa that are under- or over-represented.

Results: A lower diversity but more variability between the CD samples in both virome and microbiome was found, with a clear distinction between groups based on the microbiome. Only ≈5% of the differential viral biomarkers are more represented in the CD group (Synechococcus phage S CBS1 and Retroviridae family viruses), compared with 95% in the control group. Unrelated patterns of bacteria and bacteriophages were observed.

Conclusions: Our use of an extensive database is critical to retrieve more viral hits than in previous approaches. Unrelated patterns of bacteria and bacteriophages may be due to uneven representation of certain viruses in databases, among other factors. Further characterization of Retroviridae viruses in the CD group could be of interest, given their links with immunodeficiency and the immune responses. To conclude, some methodological considerations underlying the analysis of the viral community composition and abundance are discussed.

Figures

Figure 1

Taxonomic classification and relative abundance of the viral communities from the fecal CD samples (labeled as C1 to C10), fecal healthy control volunteer's samples (labeled as V1 to V8), and intestinal CD sample (labeled as IC1). The results are presented at the family level for (a) non-assembled and (b) assembled reads. Only those families with a presence ⩾1.0% in the global count are displayed in the legend, ranged by decreasing abundance.

Figure 2

Comparative distribution of viruses. Comparison of the relative distribution of prophages, bacteriophage families, and other viruses between the Crohn's disease group and the control group, in non-assembled (a) and assembled reads (b).

Figure 3

Taxonomic classification and relative abundance of the microbial communities from the fecal CD samples (labeled as C1 to C10), fecal healthy control volunteer's samples (labeled as V1 to V8), and intestinal CD sample (labeled as IC1). Three taxonomic levels are shown: (a) phylum, (b) class, and (c) order.

Figure 4

Cladogram representing the features that are discriminative with respect to the classes “Crohn” and “Volunteer” using the linear discriminant analysis model results on the bacterial hierarchy.

Figure 5

Diversity within samples based on the abundance of viral (left) and microbial (right) operational taxonomic units within the community, plotted by rarefaction curves. Three different diversity metrics are used: (a) Observed number of species, (b) Chao1 estimator, and (c) Shannon diversity index. Viral CD groups and individual samples are plotted on the first and second columns, respectively. Microbial CD and control groups and individual samples are plotted on the third and fourth columns, respectively.

Figure 6

Diversity between samples. Jackknifed replicate Principal Coordinate Analysis (PCoA) two-dimensional plots obtained for the (a) microbiota, and for the (b) assembled reads virome using from the top to the bottom, the Bray–Curtis (a1 and b1), Canberra (a2 and b2), and Manhattan (a3 and b3) distance matrices. Only comparisons of P1 vs. P2 axes are shown. Red dots represent the samples from the CD group; blue dots represent the samples from the control group. The circle indicates the intestinal tissue sample. Black arrows indicate samples V5 and V7.

Figure 7

Cluster tree with jackknifing support for bacteria (a1) and for viruses (a2) obtained with Bray–Curtis dissimilarity matrix on the left, and with Canberra distance matrix (b1) and (b2) on the right. Jackknifing support values of the nodes are represented as well as the color scale for them. Yellow stars represent CD samples; green stars, control samples; and red stars, samples that are grouped within the other group. Ten thousand replicates were carried out.

Figure 8

16S rDNA and bacteriophage-based bacterial composition and abundance. Comparison of the composition and relative abundance of bacterial orders based on the 16S rDNA sequences and on the bacterial hosts of the bacteriophages, in the (a) overall data set and (b) with the distinction between the Crohn's disease and the control group. Only bacterial orders represented over 0.1% in at least one group are displayed.

Figure 9

Functional annotation of the assembled viral metagenomic sequences of each sample, grouped by TIGRfam main roles. An additional category: “Pfam-only domains”, which includes those reads without homology to the TIGRfam functional database but with hits against the Pfam database, is included. For more details on this category see Supplementary Table S3 in the Supplementary Materials and Methods.