The balance of metagenomic elements shapes the skin microbiome in acne and health

Abstract

Studies have emphasized the importance of disease-associated microorganisms in perturbed communities, however, the protective roles of commensals are largely under recognized and poorly understood. Using acne as a model disease, we investigated the determinants of the overall virulence property of the skin microbiota when disease- and health-associated organisms coexist in the community. By ultra-deep metagenomic shotgun sequencing, we revealed higher relative abundances of propionibacteria and Propionibacterium acnes phage in healthy skin. In acne patients, the microbiome composition at the species level and at P. acnes strain level was more diverse than in healthy individuals, with enriched virulence-associated factors and reduced abundance of metabolic synthesis genes. Based on the abundance profiles of the metagenomic elements, we constructed a quantitative prediction model, which classified the clinical states of the host skin with high accuracy in both our study cohort (85%) and an independent sample set (86%). Our results suggest that the balance between metagenomic elements, not the mere presence of disease-associated strains, shapes the overall virulence property of the skin microbiota. This study provides new insights into the microbial mechanism of acne pathogenesis and suggests probiotic and phage therapies as potential acne treatments to modulate the skin microbiota and to maintain skin health.

Figure 1. Bacteria dominate the skin follicular microbiome.

(A) A box-and-whiskers plot comparing the relative abundances of bacteria, fungi, and P. acnes phage in the follicular microbiota of acne patients (n = 38), age-matched healthy individuals (n = 30), healthy individuals over 55 (H55 + ; n = 4), and all healthy individuals combined (n = 34). (B) A few fungal organisms were found in the follicle. Sequencing reads pooled from all subjects mapped to six fungal species, with less than 1X coverage for any species. (C) The relative abundance of P. acnes phage in all the samples suggests an increased prevalence and abundance of P. acnes phage in healthy individuals and a trend of increased phage abundance with age. (D) The relative abundances of bacterial species in the follicle. Each column represents the relative abundances of the bacterial species found in each individual. P. acnes was the dominant skin bacterium in all but one individual. On average P. acnes accounted for 91% of the bacterial taxa identified. An increase in the average relative abundances of P. acnes and P. granulosum was observed in the healthy individuals, whereas an increase in the average relative abundances of minor taxa was observed in the acne group. Five major skin bacterial species (P. acnes, P. humerusii, P. avidum, P. granulosum, and S. epidermidis) are shown separately from the phyla that they belong to.

Figure 2. Differences in the relative abundances of P. acnes OGUs between acne patients and healthy individuals.

(A) A heat map showing the relative abundances of the OGUs in P. acnes loci 1, 2, and 3 in acne patients and healthy individuals. Each column represents an OGU, ordered based on the genomic location of the OGUs. OGUs 101–200 in the pan-genome were plotted to show locus 1, flanking OGUs, and locus 2. The 74 OGUs from locus 3, which is a plasmid, are also shown. Each row represents an individual. Acne patients (n = 38) and healthy individuals (n = 34, including those with age over 55) were compared. Individuals within each group were clustered based on the average relative abundance of locus 2 OGUs. Ribotype composition and past and current acne treatments are indicated on the right. Multiple treatments are depicted by more than one color. (B) Fold changes in relative abundance of the OGUs in loci 1, 2, and 3 between acne patients and healthy individuals. Acne-associated OGUs had a fold change >1, while health-associated OGUs had a fold change <1. (C) Prevalence ratio of the OGUs in loci 1, 2, and 3 between acne patients and healthy individuals. The presence of an OGU in a sample is defined as an OGU with at least 1X coverage after normalization.

Figure 3. The relative abundances of acne- and health-associated metagenomic elements in acne and healthy individuals.

(A) The relative abundances of 62 P. acnes OGUs, including 25 acne- and 37 health-associated OGUs, and three organisms associated with healthy skin, P. acnes, P. granulosum, and P. acnes phage, were plotted for each individual to illustrate the importance of a balance between these metagenomic elements in health and acne. Each column represents an individual, and each row represents an OGU or an organism. The top ten ribotype composition and acne severity score (acne patients only) of each individual are also shown. (B) The prediction score of each individual based on the relative abundances of the 45 metagenomic elements is shown, where red indicates acne and green indicates healthy skin. The classification of the clinical states had an overall accuracy of 85%.

Figure 4. Class prediction accuracy using leave-one-out cross-validation and weighted gene-voting.

(A) For the training sample set (n = 72), using the clinically defined acne and healthy individual grouping, the classifier correctly assigned the clinical states of 34 of the 49 assigned samples (69% accuracy) using a prediction strength threshold of 0.25. This result is statistically significant (P = 0.001), because only one of the 1,000 permutated groupings had a higher accuracy. However, that particular grouping (accuracy of 72%) had fewer samples assigned than the clinical grouping (n = 39 vs 49). This demonstrates that the differences in the relative abundances of the metagenomic elements between acne patients and healthy individuals can be used to predict the clinical states of the skin. When we used the refined set of 45 metagenomic elements, including 43 P. acnes OGUs, P. acnes locus 2, and P. granulosum, we further improved the prediction accuracy of the training sample set to 85%. (B) Consistent with the prediction accuracy on the training sample set, for the independent sample set (n = 10), the 45 metagenomic elements were able to assign 70% of the samples with 86% accuracy.