Research ArticleMICROBIOLOGY

Fragile skin microbiomes in megacities are assembled by a predominantly niche-based process

See allHide authors and affiliations

Science Advances  07 Mar 2018:
Vol. 4, no. 3, e1701581
DOI: 10.1126/sciadv.1701581
  • Fig. 1 Variation among the skin microbiomes of 231 healthy young Chinese women.

    (A) Geographical locations of five large cities in China, where the subjects were recruited. These cities are at least 409 km away from each other. (B) City indices (population, density, GDP, and urbanization) of the five cities obtained from the China Statistical Yearbook 2013 showed various differences. (C) Mean relative abundance of bacterial phyla on human skin among the five cities. (D) Principal coordinates (PC1 and PC2) analysis of the skin microbiome derived from the unweighted UniFrac distance based on the 97% similar OTUs across the different cities. The percentage values indicated in parentheses denote the proportion of variation explained by each ordination axis. There was a significant difference in composition among cities [R2 = 0.47, P < 0.001, ANOSIM]. (E to G) Multidimensional plot of the proximity matrix calculated using random forest modeling of the bacterial composition data from the 231 subjects. We conducted random forest analysis for five groups (E), three groups (F), and two groups (G) and determined that it was most appropriate to divide all the samples into two groups based on the lowest OOB error rate of 0.43%.

  • Fig. 2 Assessment of the community assembly process using abundance-based β-null deviation measures and EAD.

    Abundance-based β-null deviations were measured for the five models of metacommunity assembly, ranging from niche-structured (type 1) to neutral (type 5) comprising 19,712 OTUs in 32 patches (microbiomes) by comparing the β diversity of the observed microbial community and the null-modeled community using the Canberra distance metric. (A) β-Null deviations measured for the five community types designed by Tucker et al. (B) Calculated deviations separated the community assembly processes into niche- or neutral-based models. (C) Standardized difference, in units of SDs (z score), between the observed and expected phylogenetic diversities (PD). The scatterplot and boxplot graph show the distribution of z scores in each region. The dashed line indicates the 95% CI. The black line and whiskers in the boxplot represent the median and range of the minimum and maximum z score among the communities, excluding outliers.

  • Fig. 3 Assessment of skin microbiome assembly process and network of skin microbes.

    (A) Fit of Sloan’s neutral model for analysis of microbial community assembly. The solid green line represents the best-fitting neutral model. The dashed line represents the 95% CIs around the best-fitting neutral model. OTUs within the CIs (black points) follow the neutral process, and OTUs that occur more frequently than predicted by the model are shown in red, whereas those that occur less frequently than predicted are shown in blue. (B) Network analysis of bacterial OTUs of the skin microbiome. Color markings indicate the major taxa at phylum level. The T and D values shown in the box represent the transitivity (clustering coefficient) and density, which was calculated using the ratio of the number of edges. Each node represents the OTUs, and the size of each node is proportional to the relative abundance.

  • Fig. 4 Interpretation of the skin microbiome signatures in five large cities in China.

    (A) Outline of the characteristics of the skin microbiome in five cities. Three complementary analyses and microbial network analysis were conducted to compare the robustness of the community assembly process and network. (B) Identification of the influencing factors that determine the characteristics of the skin microbiome for the β-null deviation values using random forest modeling. Asterisks indicate the significance level of each factor (**P < 0.001, *P < 0.05). The color or pattern of each box represents the following factors: red, value of α diversity; dark blue, atmospheric environmental factors; blue, climatic environmental factors; diagonal line pattern, city indices; and yellow, host-specific factors. (C) Heat map for the functional pathway of “environmental information processing,” which showed a significant difference between the two groups. UV index, ultraviolet index. PM, particulate matter; TEWL, transepidermal water loss; PTS, phosphotransferase system.

Supplementary Materials

  • Supplementary material for this article is available at http://advances.sciencemag.org/cgi/content/full/4/3/e1701581/DC1

    fig. S1. Information of the geographical location and characteristics of each city.

    fig. S2. Relative mean abundances of four predominant phyla.

    fig. S3. Mean relative abundance of bacterial genera in the skin microbiome among the five cities in China.

    fig. S4. Principal coordinate analysis with weighted UniFrac distances.

    fig. S5. Distance decay analysis.

    fig. S6. The β-diversity comparisons of the skin microbiome of the five cities.

    fig. S7. Distribution of PD.

    fig. S8. Analysis of MSE by random forest modeling.

    fig. S9. Comparison of predicted functional metagenomic profiling using PICRUSt between the megacity and non-megacity groups.

    fig. S10. The richness (observed OTUs) of the genus Staphylococcus in five cities.

    table S1. Sample metadata, number of sequences, and α diversity across samples based on rarefaction of 4860 reads per sample.

    table S2. Relative average abundances of the 15 most abundant genera.

    table S3. Environmental factors and city indices of five cities.

    table S4. Environmental factors and city indices of five cities.

    table S5. The observed species pool of five cities.

    data file S1. The information of OTUs in Fig. 3A.

  • Supplementary Materials

    This PDF file includes:

    • fig. S1. Information of the geographical location and characteristics of each city.
    • fig. S2. Relative mean abundances of four predominant phyla.
    • fig. S3. Mean relative abundance of bacterial genera in the skin microbiome among the five cities in China.
    • fig. S4. Principal coordinate analysis with weighted UniFrac distances.
    • fig. S5. Distance decay analysis.
    • fig. S6. The β-diversity comparisons of the skin microbiome of the five cities.
    • fig. S7. Distribution of PD.
    • fig. S8. Analysis of MSE by random forest modeling.
    • fig. S9. Comparison of predicted functional metagenomic profiling using PICRUSt between the megacity and non-megacity groups.
    • fig. S10. The richness (observed OTUs) of the genus Staphylococcus in five cities.
    • table S1. Sample metadata, number of sequences, and α diversity across samples based on rarefaction of 4860 reads per sample.
    • table S2. Relative average abundances of the 15 most abundant genera.
    • table S3. Environmental factors and city indices of five cities.
    • table S4. Environmental factors and city indices of five cities.
    • table S5. The observed species pool of five cities.
    • Legend for data file S1

    Download PDF

    Other Supplementary Material for this manuscript includes the following:

    • data file S1 (Microsoft Excel format). The information of OTUs in Fig. 3A.

    Files in this Data Supplement:

Navigate This Article