Human milk metagenome: a functional capacity analysis

Abstract

Background: Human milk contains a diverse population of bacteria that likely influences colonization of the infant gastrointestinal tract. Recent studies, however, have been limited to characterization of this microbial community by 16S rRNA analysis. In the present study, a metagenomic approach using Illumina sequencing of a pooled milk sample (ten donors) was employed to determine the genera of bacteria and the types of bacterial open reading frames in human milk that may influence bacterial establishment and stability in this primal food matrix. The human milk metagenome was also compared to that of breast-fed and formula-fed infants' feces (n = 5, each) and mothers' feces (n = 3) at the phylum level and at a functional level using open reading frame abundance. Additionally, immune-modulatory bacterial-DNA motifs were also searched for within human milk.

Results: The bacterial community in human milk contained over 360 prokaryotic genera, with sequences aligning predominantly to the phyla of Proteobacteria (65%) and Firmicutes (34%), and the genera of Pseudomonas (61.1%), Staphylococcus (33.4%) and Streptococcus (0.5%). From assembled human milk-derived contigs, 30,128 open reading frames were annotated and assigned to functional categories. When compared to the metagenome of infants' and mothers' feces, the human milk metagenome was less diverse at the phylum level, and contained more open reading frames associated with nitrogen metabolism, membrane transport and stress response (P < 0.05). The human milk metagenome also contained a similar occurrence of immune-modulatory DNA motifs to that of infants' and mothers' fecal metagenomes.

Conclusions: Our results further expand the complexity of the human milk metagenome and enforce the benefits of human milk ingestion on the microbial colonization of the infant gut and immunity. Discovery of immune-modulatory motifs in the metagenome of human milk indicates more exhaustive analyses of the functionality of the human milk metagenome are warranted.

Figure 1

Best hit analysis of 51 bp DNA sequences from human milk. DNA from human milk was sequenced using Illumina sequencing followed by alignment to known prokaryotic genomes. Sequences (75,521,216) were BLATed against 1,731 known prokaryotic genomes imported from NCBI (min 95% identity), with 1,331,996 sequences aligning to 370 prokaryotic genera. Other refers to genera each representing <0.1% of all sequences.

Figure 2

Best hit analysis of open reading frames within human milk. Assembled contigs (56,712) were submitted to MG-RAST for analysis. Contigs aligned to 194 known genomes at the genus level (maximum e-value of 1x10^-5, minimum identity of 60%, and minimum alignment length of 45 bp). Color denotes phylum and red bars indicate the number of positive alignments.

Figure 3

Functional categorization of open reading frames within human milk. The percent of ORFs assigned to each functional category is shown. Using the “Hierarchical Classification” tool within MG-RAST, 41,352 ORFs were submitted, 33,793 were annotated and assigned to a functional category (maximum e-value of 1x10^-5, minimum identity of 60%, and minimum alignment length of 15 aa). Three categories of genes (stress, virulence, carbohydrates) are expanded on the right to demonstrate the diverse capabilities of milk-derived DNA sequences.

Figure 4

Best hit comparison of bacterial phyla in human milk, infants’ feces and mothers’ feces. The percent of sequences assigned to each phyla according to MG-RAST (maximum e-value of 1x10^-5, minimum identity of 60%, and minimum alignment length of 45 bp) is shown. Breast-fed and formula-fed infant feces values are an average of five individuals, and mothers’ feces values are an average of three individuals. All subjects were unrelated. Other contains phyla each representing <1% of the contigs.

Figure 5

Functional category comparison of open reading frames within human milk versus infants’and mothers’feces. The percent of ORFs assigned to each functional category of genes is shown. Using the “Hierarchical Classification” tool within MG-RAST, ORFs within each metagenome were assigned to a functional category (maximum e-value of 1x10^-5, minimum identity of 60%, and minimum alignment length of 15 aa). Asterisk denotes that the proportion of ORFs within the category is significantly different from that in human milk (Student’s t-test, P < 0.05). Breast-fed and formula-fed infant feces values are an average of five individuals, and mothers’ feces values are an average of three individuals. All subjects were unrelated.

Figure 6

Pair-wise comparison of categorized open reading frames from human milk versus infants’and mothers’feces. Pair-wise comparisons for the human milk metagenome versus (A) breast-fed infants’ feces, (B) formula-fed infants’ feces and (C) mothers’ feces are shown. For comparison, a plot of breast-fed infants’ feces and formula-fed infants’ feces (D) is also shown. Each point represents a different SEED subsystem and its relative abundance within the human milk metagenome compared to the fecal metagenomes. Points lying on or near the dotted line have equal or similar abundance in both metagenomes. Points closer to the x-axis are more abundant in the feces metagenome, whereas points closer to the y-axis are more abundant in the human milk metagenome. Red dots signify those with significantly different proportions between the two metagenomes (Student’s t-test, P < 0.05). Breast-fed and formula-fed infants’ feces values are an average of five individuals, and mothers’ feces values are an average of three individuals. All subjects are unrelated.