Gut mycobiome maturation and its determinants during early childhood: a comparison of ITS2 amplicon and shotgun metagenomic sequencing approaches

Abstract

Introduction: Microbial colonization of the gut in early life is important for the development of metabolism, immunity, and the brain. Fungi and bacteria both colonize the human infant gut. The relatively smaller contribution of fungi to the gut microbiome, as compared to bacteria, has posed technical challenges for the precise characterization of fungal communities (mycobiomes) and limited the ability to longitudinally examine mycobiome development.

Background: The aims of this study were to (1) characterize mycobiome maturation and identify clinical determinants of mycobiome compositional variation during the first 2 years of life and (2) compare two sequencing approaches (ITS2 amplicon and whole genome metagenomics) for characterizing mycobiome maturational features. Longitudinal fecal samples and associated clinical metadata were obtained from subjects enrolled as part of the MAGIC (Microbiome, Antibiotics and Growth Infant Cohort) study.

Results: Overall, fungal richness increased and mycobiome composition changed in a similar ordered pattern during the first 2 years of life utilizing either amplicon or metagenomic sequencing approaches. Less resolution of taxa to species and genera levels was observed for the metagenomic dataset. The predominant taxa identified by both sequencing approaches, Candida albicans, Saccharomyces/S. cerevisiae, and Malassezia restricta, each exhibited similar dynamics in abundances and prevalences over the first 2 years of life, irrespective of sequencing approach. Antibiotic exposure and breastfeeding status contributed to time-specific mycobiome compositional variation, results that were consistent for both types of sequence datasets. Candida albicans exhibited altered abundance dynamics in association with perinatal antibiotic exposure and birth mode for both sequencing approaches. Post hoc analyses suggested that the birth mode association could be driven by exposure to perinatal antibiotics in children delivered by cesarean section rather than by birth mode itself.

Discussion: In summary, amplicon and metagenomic sequencing approaches provide generally similar results with respect to mycobiome maturational dynamics and the contribution of clinical variables to variation. Differences in taxa identification by the two approaches likely due to sequence database differences, primer/genome sequence variation, and/or sequencing depth should be taken into consideration.

Keywords: ITS2 amplicon; childhood; gut mycobiome; longitudinal variation; whole genome shotgun metagenomics.

Figure 1

Dot plots of fungal alpha diversity over the first ~2 years of life using amplicon (left panels) and metagenomic (right panels) sequences. Alpha diversity measures: richness (top panels), Shannon index (bottom panels). Each dot represents an individual sample, and all samples are organized by time of collection as a continuous variable. Random sampling was used in linear statistical models to account for multiple samples per infant. Linear regression line of best fit, red; gray shading, 95% confidence intervals.

Figure 2

Principal coordinates analysis (PCoA) plots of fecal fungal taxonomic compositions (beta-diversity) by time cluster and sequencing approach. Ellipses show the standard error (SE) confidence limit set at 0.2, and crosshatches show the standard deviation (SD) with confidence limit set at 0.95. In the case of multiple samples from an infant in a time cluster, only the earliest sample was used in the analysis. Statistical comparisons of fungal beta diversity among time clusters and number of samples (n) within each time cluster for each sequencing dataset are found in Supplementary Table 4.

Figure 3

LOESS regression curves of relative abundances and prevalences of dominant fungal taxa during the first 2 years of life for amplicon and metagenomic sequence datasets. LOESS regression curves for individual taxa indicated by different colors; gray shading, 95% confidence intervals. Dashed line not detected but included for comparison.

Figure 4

Bar plots of explained variance of six potential mycobiome covariables modeled by EnvFit for amplicon and metagenomic sequence data. Horizontal bars represent the amount of variance (r²) explained by each covariable in the model. Covariables that contributed significant differences in mycobiome composition (p-value <0.05 with q-value of <0.25) are represented in bold font. Asterisk denotes statistical significance within a time point. For amplicon dataset: time cluster 3, infant race (p = 0.017, q = 0.05) and infant sex (p = 0.015, q = 0.05); time cluster 4, childhood antibiotics (p = 0.005, q = 0.03); time cluster 6, breastfeeding status (p = 0.027, q = 0.16). For metagenomic dataset: time cluster 2, breastfeeding status (p = 0.039, q = 0.23); time cluster 5, childhood antibiotics (p = 0.001, q = 0.006). NA, not able to perform statistical analysis due to inadequate number of subjects within one of the comparison groups (e.g., not breastfeeding and exposed to childhood antibiotics for time cluster 1).