A metagenomic study of diet-dependent interaction between gut microbiota and host in infants reveals differences in immune response

Abstract

Background: Gut microbiota and the host exist in a mutualistic relationship, with the functional composition of the microbiota strongly affecting the health and well-being of the host. Thus, it is important to develop a synthetic approach to study the host transcriptome and the microbiome simultaneously. Early microbial colonization in infants is critically important for directing neonatal intestinal and immune development, and is especially attractive for studying the development of human-commensal interactions. Here we report the results from a simultaneous study of the gut microbiome and host epithelial transcriptome of three-month-old exclusively breast- and formula-fed infants.

Results: Variation in both host mRNA expression and the microbiome phylogenetic and functional profiles was observed between breast- and formula-fed infants. To examine the interdependent relationship between host epithelial cell gene expression and bacterial metagenomic-based profiles, the host transcriptome and functionally profiled microbiome data were subjected to novel multivariate statistical analyses. Gut microbiota metagenome virulence characteristics concurrently varied with immunity-related gene expression in epithelial cells between the formula-fed and the breast-fed infants.

Conclusions: Our data provide insight into the integrated responses of the host transcriptome and microbiome to dietary substrates in the early neonatal period. We demonstrate that differences in diet can affect, via gut colonization, host expression of genes associated with the innate immune system. Furthermore, the methodology presented in this study can be adapted to assess other host-commensal and host-pathogen interactions using genomic and transcriptomic data, providing a synthetic genomics-based picture of host-commensal relationships.

Figure 1

Effect of diet on host transcriptional responses. Genes known a priori to be involved in intestinal biology or immunity and defense mechanisms were enriched for differential expression between BF and FF infants. (a-d) The distribution of P-values (a,b) and the distribution of q-values (c,d). (a,c) Intestinal biology: 459 genes known to be related to intestinal biology passed the quality control measures and were tested for differential expression between the BF and FF infants - 146/459 genes (32%) had FDR corrected q-values <0.2. (b,d) Immunity and defense: 660 genes known to be related to immunity and defense that passed the quality control measures and tested for differential expression between the BF and FF infants - 191/660 genes (29%) had FDR corrected q-values <0.2.

Figure 2

Effect of diet on infant microbiota. BF (breast-fed) infants (green) exhibited more heterogeneity than FF (formula-fed) infants (blue) with respect to phylogenetic composition. (a) Taxon assignment (phylum level) variability for BF and FF samples using 16S rRNA alignments to GreenGenes (see Materials and methods). A diet label permutation test using the statistic ∑_s |∑_iεBF p_is/6 - ∑_iεFF p_is/6|, where s indexes phylum and iεBF and iεFF denote that sample i is BF or FF infant, respectively, and p denotes the associated taxon proportion, rejected the null hypothesis that variability in phylogenetic composition was unrelated to BF/FF status with a P-value of 0.011. (b) Taxon assignments for all the shotgun reads (not just 16S rRNA homologs) using PhymmBL [17]. (c) Shannon-Weiner index for BF and FF infants, indicating alpha-diversity for each sample.

Figure 3

Functional analysis of metagenomic data. Top panel: SEED level 1 categories for which all BF or all FF samples had at least 200 reads mapped. At least 2% of the total number of mapped reads were tested for differences between BF (breast-fed) infants (green) and FF (formula-fed) infants (blue). A permutation test on the test statistic ∑_iεBF p_i/6 - ∑_iεFF p_i/6, where iεBF and iεFF denote that sample i is BF or FF infant, respectively, and p denotes the associated taxon proportion, was performed. The FDR corrected q-value for the virulence category was 0.058. Bottom panel: differences between BF and FF infants in the SEED level 2 virulence assignment (within the SEED level 1 virulence category) was assessed using a permutation test on the test statistic ∑_s |∑_iεBF p_is/6 - ∑_iεFF p_is/6|, where s indexes the SEED level 2 virulence categories, and P = 0.0140.

Figure 4

First and second canonical correlations between host gene sets and microbial virulence characteristics. Horizontal lines in the density plots are at 0.5, and the vertical lines are at 0.85. These cutoffs were chosen arbitrarily to emphasize enrichment in the upper-right quadrant of the plot that is suggestive of increased multivariate structure as identified by CCA. (a) First and second canonical correlations between triples of immunity and defense genes and virulence variables are shown. There are increased canonical correlations in the upper-right corner of the plot, suggesting an enriched multivariate relationship between the immunity and defense genes and microbiome virulence characteristics as compared to, for example, the set of random genes shown in (d). (b) Intestinal biology genes did not show the same level of enrichment of canonical correlations as the immunity and defense genes. (c) We analyzed 1,000 random sets each containing 660 genes in an analogous manner to the immunity and defense gene analysis (a). Of these, 969 random sets resulted in less than 12% of analyzed gene triples having first canonical correlation >0.85 and second canonical correlation >0.5. (d) An example random gene CCA plot. Additional examples are given in Additional file 6.

Figure 5

Frequency of host genes appearing in triples. Sets of gene triples were included when the first canonical correlation was at least 0.85 and the second canonical correlation was at least 0.65. These levels were chosen arbitrarily to represent strong multivariate structure as identified by CCA. Genes were ranked by their prevalence of top performing triples. This provided a qualitative profile to select genes that empirically show the strongest potential for being related to the virulence characteristics of the microbiome. (a,b) Genes related to immunity and defense far outperformed the other functional categories. For example, the best two performing intestinal biology genes were in fact also co-listed as immunity and defense genes. (c) In contrast, randomly selected genes did not display any strong multivariate structure with respect to the virulence characteristics of the microbiome.

Figure 6

Relative performance of the top 11 immunity and defense host genes and virulence characteristics. Data were assessed by the ranking described in Figure 5 with respect to multivariate association between annotated genes and the virulence characteristics of the microbiome. The size of the nodes reflects the number of triples of genes whose first canonical correlation was at least 0.85 and whose second canonical correlation was at least 0.5. The thickness of the edges connecting the nodes reflects the number of triples whose first canonical correlation was at least 0.85 and whose second canonical correlation was at least 0.5. This plot summarizes the potential relationships between genes with respect to the virulence characteristics of the microbiome. ALOX5, arachidonate 5-lipoxygenase; AOC3, amine oxidase, copper containing 3 (vascular adhesion protein); BPIL1, bactericidal/permeability-increasing protein-like 1; DUOX2, dual oxidase 2; IL1A, interleukin 1 alpha; KLRF1, killer cell lectin-like receptor subfamily F, member 1; NDST1, N-deacetylase/N-sulfotransferase (heparan glucosaminyl) 1; REL, v-rel reticuloendotheliosis viral oncogene homolog; SP2, Sp2 transcription factor; TACR1, tachykinin receptor 1; VAV2, vav 2 guanine nucleotide exchange factor.