reAnalyze #2 - Skin Ageing

What is reAnalyze?

Recent advancements in sequencing technology and database development have provided new insights into microbial communities. Notably, many microbial genomes have been sequenced from metagenomes without official descriptions. With up-to-date databases, especially those incorporating genomospecies, and improved software, we can revisit existing datasets to uncover previously overlooked organisms.

Our goal with these reAnalyze blog posts is to utilize EzBioCloud’s comprehensive databases and proprietary pipeline to explore public metagenomic datasets and share relevant findings.

Revisiting how skin microbiomes change with age

Let's reAnalyze: Skin microbiome attributes associate with biophysical skin aging by Zhou et al., 2023. This paper explores the bidirectional effects of skin environments on microbiome taxonomic and functional profiles. They report how ageing is a major determinant in a changing skin environment, a longitudinal and inexhaustible force that, given time, draws wrinkles on us all. External and internal stressors can lead to changes in skin tissue. These changes can be measured in parameters such as dermis water content, skin capacitance, collagen quality and quantity. Skin microbiomes respond to these changing environments, not only at the species level; the authors observed strain and gene level variations. Thus, they studied shotgun metagenomics and biophysical measurements of a uniform study group between older and younger healthy adult females.

Three undescribed species prevalent in the skin microbiome

Figure 1. Group composition of older and younger skin microbiome cohorts. Highlighted are the three genomospecies: g__QFNR01 MSSCM01094130_s, g__QFNR01 MSSCM01094112_s, and Lawsonella MSSCM00283847_s.

Our first reAnalyze finding was three undescribed taxa in the five most prominent species across both cohorts. So what is with the names of these taxa? These aren't the progeny of Elon Musk. They are genomospecies.

Genomospecies deserve species status and are supported by genomic data (e.g. ANI). However, they have not been officially named so the EzBioCloud team gives them unique names, usually derived from the accession numbers of INSDC databases. For example, the phylotype CP013274_s is represented by a genome sequence deposited to INSDC as a strain of Bacillus thuringiensis but showed <95% ANI to all known Bacillus species. Therefore, it is assigned as a new phylotype (equivalent to species). A genomospecies in the EzBioCloud database is always represented by an accurate 16S sequence.

MAGs (Metagenome-Assembled Genomes): Genomes that have been assembled from metagenomic samples, which can be used to identify new genomospecies.

SGBs (Species-Level Genome Bins): Clusters or bins of similar genomic content that represent species-level groupings, focusing less on whole genomes and more on clustered sequences that can represent genomospecies.

Genomospecies: A species-level classification based on genomic data, which can include well-characterized genomes used as references for identifying and organizing similar genomic sequences.

Utilising metagenomic sequences to expand on official species descriptions, we can attain Metagenomic Associated Genomes (MAGs) and Species-level Genome Bins (SGBs).

Cutibacterium acnes clearly dominate the majority of samples which, with Staphylococcus epidermidis, Zhou et al. explore intra- and inter-individual differences at strain and gene levels. Yet the three ubiquitous genomospecies indicate commonly found but poorly understood skin microbiota.

Perhaps they are difficult to isolate, depend on complex communities, or require particular environmental conditions. This leaves a gap in our knowledge which can be filled with genomic data: what genes are present and active, how their 16S rRNA sequences align with others, and the size and GC content of their genomes, for instance. To understand their roles in skin microbiomes without physically isolating and testing each species, we extrapolate information from gene orthologs identified in their genomes and plasmids. These can include virulence factors and antimicrobial resistance genes. With better complete genomes, we can gather more information.

More information on the three genospecies identified in the skin samples can be found below (only Laswonella MSSCM00283847_s has a representative genome in NCBI):

g__QFNR01 MSSCM01094130_s

g__QFNR01 MSSCM01094112_s

Lawsonella MSSCM00283847_s

Furthermore, comparing the top eight species with the reAnalyze results: other than the three missing genomospecies, Corynebacterium kroppenstedtii, which they discussed could be the candidate for previous Corynebacterium associations with ageing (Kim et al., 2022 and Shibagaki et al., 2017), is missing a closely related species, that is differentially abundant in older skin microbiomes (Tables 1a. and 2a), Corynebacterium pseudokroppenstedtii, and their paper includes Malassezia restricta, a fungal commensal and potential pathogen.

Skin microbiomes display wide taxonomic variation but share characteristics between old and young

Figure 2. Principal coordinate analysis plots comparing the skin microbiomes of younger and older individuals.

The principal coordinate analysis (PCoA) plots reveal a significant amount of overlap between the samples, which may be attributed to the strong variance explained by the principal coordinates. However, distinct clusters that could clearly differentiate the groups are not readily apparent. To better understand these observations, it would be important to investigate whether there are dominant taxa driving these patterns. Potential factors to consider include differences in microbiomes, the physical conditions of the skin, or other variables that might correlate with group composition. These hypotheses could be further explored through differential abundance analysis.

The observed patterns and associations may also be influenced by the fact that these samples were collected from individuals with similar health status, geographical region, ethnicity, and sex. The primary confounding factors to consider are individual variation and age.

“To reduce data dimensionality and control for confounding factors of the skin microbiome composition, we confine the investigation to only the facial skin microbiome of healthy Caucasian women in Paris area."

These insights from the PCoA plots can be further explored to understand the specific microbial taxa contributing to the observed differences and how they relate to age-associated changes in skin microbiome composition.

Several microbial species associated with either young or old skin microbiomes

The beta diversity analysis showed overlapping of microbiomes but are there species-level influencers? Also, recall the three genomospecies present in both cohorts; are they biomarkers of ageing skin microbiomes? The differential abundance analysis (DAA) ensemble tool can provide insight.

Although microbiomes are complex communities comprised of countless individuals responding to each other and their environments, it can take just one species or strain to disrupt that system. Biomarkers can be indicative of such perturbers of the peace or beneficial microbes or, simply, commensals associated with a particular cohort. Below are reAnalyze results using seven differential abundance analysis tools to find consensus over candidate biomarkers.

Figure 3. Candidate biomarkers in older (red) and younger (blue) skin microbiome cohorts. Upset Plot: rank: species, alpha: 0.05, other parameters: default.

There's a lot happening in this plot with many of the DAA tools indicating potential biomarkers, some in agreement and others independently (Figure 3). Seven differential abundance tools are in consensus over four biomarkers associated with 'young' microbiomes (blue, far left). Combinations of six differential abundance tools are in consensus over eight biomarkers associated with 'old' microbiomes (red, left). SIAMCAT found 49 biomarker candidates associated with 'old' microbiomes but are not in consensus with other tools (red, far right).

Now, let's find out which taxa they are!

Table 1a. Candidate biomarker table excerpt displaying top taxa. Positive scores indicating assocations with old skin microbiomes.

Table 1b. Candidate biomarker table excerpt displaying bottom taxa. Negative scores indicating assocations with young skin microbiomes.

There is strong consensus among the seven different tools used, as indicated by the consistent scores. Taxa with a score of 6 or 7 (positive or negative) are likely to be reliably associated with either an increase or decrease in older skin microbiomes.

The enrichment of certain taxa in older skin microbiomes (e.g., Lactococcus lactis, Streptococcus spp.) could suggest changes in skin environment or immunity with age, possibly linked to skin ageing or disease susceptibility.

The depletion of other taxa (e.g., Cutibacterium acnes, Staphylococcus capitis) in older individuals could indicate shifts in the skin microbiome that affect the skin's barrier function, microbial diversity, or susceptibility to pathogens. At the top of this list, do you recognise Lawsonella MSSCM00283847_s? That's one of the three genomospecies dominant in the skin microbiota samples (Figure 1). All seven DAA tools are in consensus that L. MSSCM00283847_s is associated with young skin microbiomes.

"These findings propose a mechanistic link between aging and skin microbiome: aging decreases the production of collagen, which could result in a decrease in C. acnes abundance and consequently an increase in species diversity."

Arguably, any of these candidates could be true biomarkers of young or old skin microbiomes, whether they have a score of seven, supported by all the DAA tools or a score of one, supported by a single tool. To fully understand the implications of these findings, further analysis should explore the functional roles of these differentially abundant taxa, as well as potential correlations with clinical or physiological data from the study population.

Stricter profiling standards provide similar key biomarkers

The former dataset was created using the 'shallow sequencing' setting, made for when little sequencing data is available, such as for skin microbiomes. Using stricter taxonomic profiling parameters, we can create a new dataset from the same sequencing data. If we run that through the DAA ensemble tool we will see a reduced number of species. Let's compare any differences in candidate biomarkers found.

Figure 4. Candidate biomarkers from a strict taxonomic profiling dataset in older (red) and younger (blue) skin microbiome cohorts. Upset Plot: rank: species, alpha: 0.05, other parameters: default.

This upset plot is far less busy with fewer candidate biomarkers.

Table 2a. Strict candidate biomarker table excerpt displaying top taxa. Positive scores indicating assocations with old skin microbiomes.

Table 2b. Strict candidate biomarker table excerpt displaying bottom taxa. Negative scores indicating assocations with young skin microbiomes.

Anaerococcus nagyae, Corynebacterium pseudokroppenstedtii, and Cutibacterium granulosum are the top three taxa that are highly indicative of the older age group, with strong support across multiple statistical methods. These three had a score of four in the 'shallow sequencing' dataset. Inversely, Lactobacillus crispatus, Cutibacterium acnes, and, our favourite genomospecies, Lawsonella MSSCM00283847_s are all associated with young skin microbiomes. This is also in agreement with the previous dataset. Other Corynebacterium species also show consistent associations with the older age group, indicating a pattern within this genus.

Microbiomes are complex entities that interact with equally complex environments. Even when we simplify the data (e.g. reduce dimensions), it remains challenging to directly link specific microbiome profiles to physical health conditions. While the microbiome is an important area of study, focusing on infectious diseases may offer a more straightforward path forward. It is often easier and more definitive to identify and prove that a particular species or strain is causing a problem in the human body, even if that species triggers a cascade of other effects.

Skin microbiomes follow counterintuitive diversity patterns

Figure 5. Alpha diversity analyses of samples between oldrer and younger skin microbiomes. *s indicate statistical significance.

Most microbiome patterns suggest that the more diverse and balanced the microbiota in a region, the healthier and less prone to dysbiosis the sample site is. However, in the case of the skin microbiome, it appears that older skin is more diverse. Whether this increased diversity corresponds to greater disease resistance remains to be tested. This observation aligns with the findings of the paper, even though some notable species differ. The authors partially attribute this difference to the prevalence of C. acnes in younger skin microbiomes.

Beta diversity follows the same patterns whether a single cheek is represented or both cheeks

Figure 6. Beta diveristy incorportating both cheeks (above) and one cheek (below) samples.

For the majority of their analyses, they averaged the community composition profiles between left and right cheeks to have individuals represented only once. Combining left and right cheeks increases robustness and follows similar beta diversity patterns as using only one cheek side.

---

References

Kim, H. J., Oh, H. N., Park, T., Kim, H., Lee, H. G., An, S., & Sul, W. J. (2022). Aged related human skin microbiome and mycobiome in Korean women. Scientific reports, 12(1), 2351.

Shibagaki, N., Suda, W., Clavaud, C., Bastien, P., Takayasu, L., Iioka, E., ... & Hattori, M. (2017). Aging-related changes in the diversity of women’s skin microbiomes associated with oral bacteria. Scientific reports, 7(1), 10567.

Zhou, W., Fleming, E., Legendre, G., Roux, L., Latreille, J., Gendronneau, G., ... & Oh, J. (2023). Skin microbiome attributes associate with biophysical skin ageing. Experimental dermatology, 32(9), 1546-1556.