Integration of skin phenome and microbiome

reveals the ­key role of bacteria in human skin aging

Jingjing Xia 
Fudan University
Zhiming Li 
Fudan University
Qian Zhong 
Fudan University
Qingzhen Wei 
Fudan University
Liuyiqi Jiang 
Fudan University
Cheng Duan 
Fudan University
Huijue Jia 
Fudan University
Yimei Tan 
Shanghai Skin Disease Hospital
Lianyi Han 
Fudan University
Jiucun Wang 
Fudan University
Xiao Liu 
(

liuxiao@sz.tsinghua.edu.cn
)
Tsinghua University

Research Article

Keywords: Shotgun metagenomic sequencing, Skin microbiome, Skin phenome, Aging, Moraxella
osloensis, Bacterial isolation, Keratinocyte , Fibroblast

Posted Date: March 1st, 2023

DOI: https://doi.org/10.21203/rs.3.rs-2629420/v1

License:


This work is licensed under a Creative Commons Attribution 4.0 International
License.
 
Read Full License
Page 1/23
Page 2/23
Abstract
Background: Despite the complexity, distinct ecological niches are believed to primarily drive the skin
microbiome composition. Meanwhile, skin aging is a dynamic process with a spectrum of phenotypical
changes, making it an attractive model for studying microbiome-phenotype interactions. Although a large
number of studies confirmed the impact of chronological age in skin bacterial communities, the
understanding of cross-kingdom microbiome variation with skin aging remains minimal. And this is not
trivial because one’s skin condition or perceived age may deviate largely from their actual age as skin
aging is a complex process combining chronological and extrinsic aging.

Results: To this end, 822 facial microbial samples and skin phenotypes from the corresponding area were
assessed in a Chinese cohort, the largest population size to date for skin shotgun metagenomic profiling.
Our data revealed that among 14 measured variables, porphyrin and chronological age explained the
most significant microbial variability. Consistent with previous studies based on 16S rRNA gene
sequencing, we revealed increased biodiversity with aging and further specified age-associated species
across kingdoms. While the abundance of most bacteria increased with age, two species, Cutibacterium
acnes and Aeromicrobium choanae, declined. Microbiome undergoes active function selection from
energy demands/growth to stress adaptation along aging. In addition, we characterized microbial
changes in skin aging, asa combined consequence of both intrinsic and extrinsic reasons and reflecting
the actual dynamic of niche conditions rather than chronological age. Using the multiple linear regression
model, we predicted premature-aging/delayed-aging-related microbial species, mainly localizing to
Moraxella osloensis and C. acnes. Furthermore, we validated the biological functions in vitro of some
host-microbe interactions predicted by the microbiome-skin phenome association network. M. osloensis
regulated collagen metabolism, extracellular matrix assembly and promoted cell senescence in human
keratinocyte and fibroblast cells.

Conclusions: We presume that application of both culture-independent and culture-dependent

approaches can advance a good understanding of  microbiome-phenotype interactions. Our study is of
significance for designing interventions against aging-related skin conditions.

Introduction
The skin is colonized by millions of diverse microorganisms, including bacteria, fungi, and viruses,
collectively referred to as the human skin microbiome [1]. These commensals are considered important in
maintaining skin homeostasis, and disruptions in the skin microbiome are associated with various
clinical conditions of some most frequent skin diseases [2].

Even in healthy human beings, the skin microbiome was proven to be highly individualized and dynamic.
Numerous variables were characterized to influence skin bacterial communities, including host factors
(such as race, gender, and age) and environmental factors due to occupation or lifestyle (e.g., diet,
hygiene, skin product and medication) or other exposure (e.g., climate, geographical location, pollution,

Page 3/23
ultraviolet and other radiation) [4, 5]. Nevertheless, from the perspective of classic ecology, given that skin
is perceived as an ecosystem [3], the selection pressures shaping the ecosystem should be able to be
simplified into three main elements: 1) resource availability (presence of nutrients); 2) environmental
conditions (temperature, geographical access) and 3) biological factors (the occupation of the ecological
niches through self-adaptation or microbe-microbe interactions predators and pathogens) [6, 7]. Previous
evidence also strongly suggested that distinct niches' physical and chemical landscape primarily drives
the local microbial composition (niche selection) [3, 8–10]. In turn, the growth and metabolism of these
micro-inhabitants modify the local niches, such as pH or moisture state, and reshape host skin
phenotypes [3, 7, 11]. However, understanding skin microbiome-phenotype interaction is still ambiguous
and needs systemic characterization.

Skin aging is a dynamic process with a spectrum of age-related phenotypical changes in skin
appearances and functions, making skin aging an ideal model for studying microbiome-phenotype
interactions. Although a large number of studies already confirmed the impact of chronological age in
skin bacterial communities [12–15], the understanding of cross-kingdom microbiome variation with skin
aging remains minimal. And this is not trivial. Although, in most conditions, age correlates well with skin
aging phenotypes or vice versa, one’s perceived age may deviate largely from their actual chronological
age because skin aging is a complex process combining chronological aging and extrinsic aging [16].

In order to explore cross-kingdom microbiome interaction with skin phenotypes, 296 healthy individuals
were recruited in Shanghai, by far the largest population size for skin metagenomic profiling in the
Chinese population. Skin microbiome from three facial sites, i.e., forehead (FH), cheek (CK), and the back
of the nose (NS), were collected. In total, 822 samples were assessed by shotgun metagenomic
sequencing, allowing for a more precise survey across all kingdoms (bacteria, fungi, and viruses).
Multiple potential variables of the host, including age, gender, and 12 skin phenotypes were cross-
analyzed to elucidate the host-microbiota interactions in healthy Chinese. In contrast to other studies, the
current study took a step further. In addition to addressing the aging-microbiome correlation, we used
two-dimensional (2D) cultures to validate some associations. The general workflow of this study is
demonstrated in Fig. 1a.

Results
Chronological age plays a crucial role in skin phenotypes and explains a large skin bacterial variability

In parallel to the skin microbiome, skin phenome (a collection of skin traits) from the corresponding
sampling areas was assessed (Fig. 1a). Due to anatomic restriction, some parameters from the back of
the nose were not collected. As expected, age demonstrated a central impact on skin appearance and
physiology (Fig. 1b, Fig. S1a-c). The scores of lentigines, telangiectasia, and pores area increased with
age, whereas the sebum content, porphyrins, hydration, and skin elasticity declined with age (Fig. 1b),
consisting of our recognition on skin aging. In addition, skin color presented more dullness (L*), yellowish
(b*) and reddish (a*) with increased age, which was also in line with previous findings [17, 18](Fig. 1b). To

Page 4/23
our surprise, the pH was not significantly associated with age, and the Trans Epidermal-Water-Loss
(TEWL) reflecting skin barrier function, was unexpectedly decrease with age. These inconsistencies may
further imply the existence of deviation in skin aging from chronological age. These associations were
generally of a high degree of consistency in the three facial sites.

We then assessed the skin microbial variations attributed to different host features by PERMANOVA. 14
factors were identified as significantly correlated with skin microbiota (adjusted P < 0.05, Fig. 1c). Overall,
porphyrin explained the highest microbiome variance, especially in the bacterial community, which
consistent with previous findings in North American volunteers [8]. It was followed by age and hydration
level. Interestingly, for the fungal community, age explained very marginally while gender showed a
prominent effect on the variation (Fig. 1c). It is interesting to observe that many skin phenotypes, such as
skin color (L*, a*), telangiectasia, pH and TEWL explained more variability in the fungal community over
the bacterial community. The variability of the viral community, which was commonly considered
transient members of the skin microbiota [19], was also attributed to porphyrin, age and other skin traits
(adjusted P < 0.05, Fig. 1c). Co-inertia analysis revealed a significant relationship between the viral and
bacterial communities (Permutation Tests, P < 0.01), and the abundance of bacteria and their
corresponding bacteriophages were highly correlated (Figure S2a,b), suggesting that across-kingdoms
microbe-microbe interaction underlie the community balance of the ecosystem.

Age-related skin microbiome dynamics

Besides porphyrin, age was the most influential factor in the bacterial variance (Fig. 1c). To specify age-
associated species, we explored the bacterial dynamics across different age groups: young group (20 to
35, N = 74), middle-aged group (36 to 50, N = 131), and old group (above 50, N = 89) (Fig. 2a).

Overall, bacterial diversities from all tested facial sites increased with age (Fig. 2b), which were consistent
with previous findings [13, 20–24]. Distance-based redundancy analysis (dbRDA) demonstrated a clear
separation between the microbial species of different age groups (Fig. 2d). In addition, beta diversity
between groups showed that the greater the age difference, the more significant the change of skin
bacterial microbiome (Fig. 2c). The comparison of bacterial species across different age groups identified
58 age-associated bacterial species (Fig. 2e). While most species increased with age, especially
Moraxella spp. (mainly Moraxella osloensis), Chryseobacterium spp. (mainly Chryseobacterium
taklimakanense), Elizabethkingia spp. and Paracocus spp., two species decreased progressively with age,
namely Cutibacterium acnes and Aeromicrobium choanae (Fig. 2e). Of note, these two species showed a
synergism effect that was mutually positive-correlated with each other while negatively correlated with
other age-increasing species (Fig. 2f). Besides skin bacteria, we also found three viruses associated with
age (Spearman's p < 0.05), including Betapapillomavirus 3, Staphylococcus virus phiETA and
Streptococcus phage IPP18 (Fig. S2).

The skin microbiome is highly correlated with skin aging features

Page 5/23
Skin aging is a dynamic process with a series of phenotypical changes in appearance and physiological
parameters. These phenotypes can reflect the niche conditions, such as pH and hydration, or be ultimate
manifestations of microbial metabolic activities [7].

To specify phenotype-related microbial dynamics, we performed a cross-kingdom Spearman correlation

analysis between the microbiome and these aging phenotypes (Fig. 3). Of the top 20 most abundant
bacterial species, C. acnes and M. osloensis displayed an opposite trend. The abundance of C. acnes was
positively correlated with the level of porphyrin, sebum, TEWL, hydration, and pore area, but it was
negatively correlated with age and skin yellowish color (b*). In contrast, the abundance of M. osloensis
was prone to aging traits, such as reduced hydration, increased lentigines and increased yellowish skin
color (Fig. 3a). Another famous skin commensal Staphylococcus epidermidis showed a similar tendency
towards C. acnes-associated patterns, however, except for TEWL, the rest associations were not
significant (Fig. 3a, Fig. S3).

Although fungi account for a minority of the microbial community, their cell sizes are typically a hundred-
fold larger than bacterial cells, and many fungi are considered influential in shaping host phenotypes
through their unique metabolism [25]. Therefore, we further evaluated skin phenotypes-associated fungal
members. Indeed, many fungal members were associated with distinct skin traits (Fig. 3b). Notably, many
Malassezia species were positively correlated with skin sebum level, consistent with their lipophilicity
properties.

Machine learning-based skin-age modeling to locate aging/delayed-aging-related microbial species

Although chronological age is central to most skin aging phenotypes in population, one’s perceived age,
or following called “skin-age” may deviate from their chronological age. Here, the applied skin-age
algorithm was developed based on machine learning for assessing one’s perceived age. The algorithm
has been previously trained on 200,000 images using chronological age as training tag and finetuned in
perceived age cohort (N = 5,768). The correlation coefficient between the model-predicted value and the
means of facial age perceived by investigators reached 0.97.

Multiple linear regression (MLR) model was used to predict skin-age. The residual standard error (RAE)
achieved by the MLR model configuration was 3.639 years, with 0.7665 of R2. To identify aging/anti-
aging related microbial species, we filtered individuals whose Δage (difference between predicted and
chronological age) was greater than RAE and divided them further into “premature-aging group” and
“delayed-aging group” (Fig. 4a). We sequentially compared the differences in skin microbial composition
between these two groups. Bacteria species such as M. osloensis, Bacillus coagulans, Chryseobacterium
greenlandense, Pseudomonas thivervalensis were more enriched in the premature aging group. In
contrast, C. acnes was enriched in the delayed-aging group (Fig. 4b). To explore the potential function of
aging-related species, we investigated the bacterial KEGG KOs between the groups. Interestingly, we
found that the microbiome from the premature aging group demonstrated marked gene module
enrichment in the two-component regulatory system, drug resistance and drug efflux transporter/pump,

Page 6/23
which may imply a selection of species capable of adapting themselves to the environmental stresses
along aging (Fig. 4c). In contrast, microbiome from delayed-aging group enriched the phosphotransferase
system responsible for transporting carbohydrates [26] and central carbohydrate metabolism, which may
represent an active growth and energy demands from these commensals.

M. osloensis regulates collagen metabolism and extracellular matrix assembly in human keratinocyte
and fibroblast cells

M. osloensis and C. acnes, also known as cutotype-determinant species in Chinese skin [27], were
significantly correlated with age and skin aging phenotypes (Fig. 2e, Fig. 3a, Fig. 4b). While C. acnes
negatively correlate with age and aging traits, such as increased sebum/porphyrin production, higher
elasticity and hydration levels, and increased lentigines; M. osloensis exhibited the opposite correlation
and was prone to aging traits. We also assessed phenotype association with S. epidermidis, a well-known
dominant skin commensal, and found that it had a rather neutral role during skin aging (Fig.S3).

The correlation between skin traits and bacteria might be attributed to skin-niches-driven bacterial
variance, or vice versa. To further explore the host-microbe interactions potentially underlying these
associations, we performed RNA-seq analysis of human epidermal keratinocyte (HaCaT) and primary
dermal fibroblast with supernatant from the above commensals [28](Fig. 5a). The results revealed
marked accumulation of differentially expressed genes (DEGs) implicated in cell senescence according to
Senequest [29], a database of cell senescence (http://Senequest.net) when human dermal fibroblast
incubated with supernatant from M. osloensis, but not C. acnes nor S. epidermidis. With M. osloensis-
conditioned medium, fibroblast cells were characterized by altered gene expression in cell-cycle arrest,
senescent-associated secretory phenotype (SASP), macromolecular damage, and deregulated metabolic
profile (Fig. 5d). These cellular changes may have deleterious effects on the tissue microenvironment and
contribute to a wide range of age-related pathologies [30]. GO enrichment analysis of HaCaT also showed
a clear enrichment of DEGs in regulating collagen catabolic process, extracellular matrix disassembly,
and collagen metabolic process upon treatment of M. osloensis-conditioned medium (Fig. 5b). RNA
expression of many MMPs (matrix metalloproteinases) from M. osloensis treated HaCaT group were
significantly upregulated, in line with the association of MMPs and skin aging, as these enzymes are
known to be major contributors to collagen degradation and skin wrinkling (Fig. 5c)[31, 32].
Correspondingly, HaCaT cells treated with the supernatant of S. epidermidis and C. acnes did not show
upregulation of MMPs (Fig. 5e).

Discussion
The driving force for the establishment of various human commensal systems is so far unclear. Whether
it is a random process or involves selection or which selection, if any, is among the most concerning
issues in the field. Previous studies suggested that neutral drift with niche selection from the host could
give rise to biodiversity and biogeography of human microbiomes [10, 33]. It was proposed that distinct
skin niches drive the local microbial composition [3, 8, 9]. In turn, skin commensals are able to modify

Page 7/23
local physiology, such as pH or hydration, by converting host-derived nutrients into acidic metabolites [7,
34–38].

Interestingly, in this study, we found that porphyrin among all measured variables, explained the highest
microbial variability, which is in line with the finding in a recent large-cohort study with North American
populations [8]. These results strongly suggested that, regardless of the ethnical population, porphyrin-
producing species play a central role in maintaining the stability of the facial skin ecosystem. It is known
that several Propionibacterium, such as C. granulosum, Cutibacterium avidum, and Cutibacterium
modestum (previously, "Propionibacterium humerusii") were able to produce porphyrin [39]. However, C.
acnes is the main contributor [40, 41]. C. acnes is one of the most essential symbiotic microorganisms in
human skin, promoting the accumulation of oleic acid in the skin and metabolizing fatty acids or other oil
components into propionic acid and acetic acid, thereby eliminating harmful microorganisms and
protecting human skin [42]. This species is also prone to a sebum-rich environment associated with lower
biodiversity. Given these results, we propose that C. acnes is the foundation species in the skin
ecosystem, a species that control the biological diversity of associated species, modulates critical
ecosystem processes, and often has essential cultural values and resonance [43].

Skin aging is associated with changes in cutaneous physiology, including interactions with a skin
microbial community. Many 16S rRNA gene surveys have reported the impact of age on the composition
of the skin microbiome in different ethnic populations. Despite significant disparities in species in
different populations, a higher alpha diversity/species richness of bacteria in the elderly was observed in
both Asian [44–47] and western European [48]. Specifically, the aged population exhibited a significant
increase in the abundance of Corynebacterium but a decrease in Propionibacterium [48, 49]. This result
contradicts the fact that higher biodiversity correlated with better eco-stability (healthier skin), and
declined biodiversity was often linked to fragile stability and skin conditions. As the elderly suffer more
from skin conditions, we proposed that the increased diversity in the elderly is a passive consequence due
to the decline of foundation species of the ecosystem, namely C. acnes. This also complies with the fact
that skin sebum and porphyrin levels go down during aging; the decline of C. acnes can make room for
the survival of other species.

A series of studies also suggested other microbiome-phenotype associations. For example, a study on the
skin microbiome of Koreans found that Lawsonella had a negative correlation with skin moisture and
brown spots, Staphylococcus and Corynebacterium both had negative correlations with the number of UV
spots and positive correlations with TEWL, and Staphylococcus aureus had a negative correlation with
skin moisture parameters [50]. However, most of these studies need more direct evidence with function
validations. In this study, we characterized aging-related species, e.g. M. osloensis. This species was
previously rarely studied, only been scattered case-reported as a pathogen in immunocompromised
adults [51]. However, this species was proven to be an essential member of the human skin commensals
[52, 53] and emerged as the second most abundant species on the skin of Han Chinese [27]. Here, we
confirmed the association of M. osloensis with premature-aging traits and further validated its capability

Page 8/23
of promoting skin cell senescence, which explained the premature aging phenotype in Chinese and other
elderly populations [54].

Despite the rapid progress being made toward deciphering the human virome, several roadblocks remain
to its full characterization and utilization [55]. Due to methodological difficulties [56], the human skin
virome is the least-known community in the skin microbiome. It still needs to be determined if the viruses
are actually a normal part of the human skin commensal or if they mutually benefit the host [56]. In our
cohort, numerous phages were found, similar to a previous study that reported a strong core phagosome
presence in the North American population [57]. Notably, the abundance of bacteriophages and their
corresponding bacteria were highly correlated. Our data further revealed a potential from-phenome
interaction as skin traits such as porphyrins, age, skin color, pores area, and lentigines explained a certain
level of virome variability. Given their low abundance and what was learned from gut microbiome [58, 59],
these virome-phenotype associations were likely due to the virus-bacteria interactions.

Conclusions
In this study, Interpreting results from shotgun metagenomic sequencing, accompanied by culture-
dependent approaches, guided us a good understanding of the composition ,diversity and variability of
the human skin microbiota, even give us some relevant hints about functional and mechanistic attributes.
In the present study it allowed us (i) to profile age/phenome-related skin microbiome dynamics, (ii) to
predicte premature-aging/delayed-aging-related microbial species, mainly M.osloensis and C. acnes and
(iii) to validate the biological functions in vitro of some host-microbe interactions.

Altogether, current research has moved beyond the question of which microbes are present on the skin
but to assess the link to their function and outcome on the skin. We systematically characterized the skin
microbiome-phenome association network, especially on the aging phenotypes. Furthermore, we
validated some predicted associations with in vitro function experiments as a deep understanding of the
molecular pathways underpinning the aging process is critical in searching for products that fertilize skin
microbiota to improve skin phenotypes. Therefore, our study is significant for designing interventional
strategies against various aging-related skin conditions.

Materials And Methods

Skin phenotypes

A collection of skin traits were determined from 296 healthy volunteers, including surface pH, hydration
status, sebum and porphyrin level, which reflect the ecological niche environment; skin barrier function
TEWL, elasticity and a series of skin appearance features, such as constitutive skin color (L*, a*, b*),
lentigines, telangiectasia and pore area, were assessed by ImageJ based on photos obtained from the
VISIA-CR (Canfield Scientific Inc, Fairfield, NJ). Porphyrins, lentigines and telangiectasia were assessed
manually using UV fluorescent, normal light, and RBX RED images, respectively. Images were scored by

Page 9/23
three trained scorers independently with a standardized protocol based on a photo reference scale. The
score ranged from 0 (hardly any present) to 3 (severe). The final score of each parameter used for further
study is the summation of three scorers.

Permutational multivariate analysis of variance

The permutational multivariate analysis of variance (PERMANOVA) was used to assess the effect of
different covariates, such as age, sex, physicochemical index, and skin image information on all types of
profiles. We performed the analysis using the method implemented in R package (vegan), and 1000 times
permutations to obtain the permuted P value.

Correlation analysis

Spearman’s correlation coefficients were computed for relationships between the relative abundance of
identified species and skin phenotypes or other species. A scaled heatmap was constructed for the
correlation matrix. Visual presentation of multiple omics correlations was performed using the Corrplot
package in R (v3.5.3). FDR-adjusted P values of less than 0.05 were used as the detection threshold for
significance.

Multivariate analysis

Multivariate statistical analyses (PCA, PCOA) were applied to assess the skin fungus within individuals.
Principle component analysis (PCA) was performed on the three facial sites and two genders, as
previously described. Principle coordination analysis (PCOA) was performed based on the Jensen-
Shannon distance (JSD)/Bray Curtis distance on the skin fungal composition and functional profile using
the ade4 package.

Machine learning-based skin-age modeling

The applied skin-age algorithm was developed by Meitu Eve based on deep learning model MobileNet V3
for assessing one’s perceived age. The algorithm has been trained on 200,000 images using
chronological age as training tag and finetuned in perceived age cohort (N = 5,768). The correlation
coefficient between the model predicted value and the means of facial age perceived by investigators
reached 0.97.

Bacterial isolation

Five healthy volunteers with no visible acne on the face and no recent topical antibiotic medication were
sampled to isolate three skin commensal bacteria (Moraxella osloensis, Cutibacterium acnes,
Staphylococcus epidermidis). Volunteers ranged in age from 23 to 56, with two males and three females.
Samples were taken from the forehead and cheek using soaked swabs in a solution of 0.15 M NaCl. The
sampling regions were swabbed 20 times each. Facial swab specimens were sequentially used to
inoculate R2A agar and Tryptic Soy Agar (TSA) with 5% sheep blood, placed in aerobic and anaerobic

Page 10/23
growth conditions. Colonies were picked based on phenotypic diversity, followed by primary screening
using specific primers M.osloensis-F(5’-GGTGAAAGGGGGCGCAAGC-3’) and M.osloensis-R (5’- GTGCTTA
TTCTGCAGGTAACGTCTAATC-3’),C.acnes - F(5’-GGATAACTTCAGGAAACTGGGGC T-3’) and C.acnes- R (5’-
TACCCATTACCGTCACTCACGC-3’), S. epidermidis -F (5’- TCTTCAAGAAGCTCACTCAGAG-3’) and S.
epidermidis - R(5’- ATTCCCACGCTAAC GATTTACG-3’ ), and finally characterized by 16S rRNA gene
sequencing.

Bacterial growth conditions and supernatant preparation

M.osloensis were streaked on R2A plates and incubated overnight at 30°C. The following day, individual
colonies of this strain were grown in liquid R2A medium at 30°C overnight, shaking at 220 RPM. Cultures
from the activated strain were inoculated into fresh R2A medium at 2%. After 24h of further incubation
(OD 6000.8-1.0), bacteria were spun and culture supernatants were filtered twice using 0.22 µm Spin-X
centrifuge tube filters (Corning)[60]. And the supernatants stored at -80°C for the following experiments.
Collect the supernatants of C. acnes(RCM medium, 37°C, anaerobic) and S. epidermidis(LB medium,
37°C, aerobic, 220rpm) similarly. The difference is that their supernatants were collected 72h and 12h
after inoculation, respectively, to reach the desired OD600.

Cell culture

Normal human skin fibroblast cultures were established from a punch biopsy obtained from healthy male
donors undergoing circumcision surgery[61]. Briefly, skin specimens were immersed in normal saline and
washed three times with PBS. The skin specimens were cut into 1 mm2 sections and laid onto the surface
of a 6-cm Petri dish(epidermis upward and dermis downward).DMEM supplemented with 10% (v/v) fetal
bovine serum (FBS) and 1% (v/v) penicillin-streptomycin (all from Gibco, USA) were added to each dish
and medium was changed every 2–3 days. A dense outgrowth of cells appeared after 7–14 days, which
were passaged using 0.25% trypsin EDTA[62, 63]. All experiments used primary fibroblasts within 5
generations. The immortalized human keratinocytes cell line HaCaT was purchased from the China
Center for Type Culture Collection (Wuhan, China). All cells were incubated in standard incubators at 37°C
with 5% CO2.

Bacterial supernatant stimulation

HaCaT cells and normal skin primary fibroblasts were plated on 6-cm Petri dishes respectively and grown
to 80% confluence. The filtered supernatants of M. osloensis, C. acnes and S. epidermidis were diluted in
DMEM (1:25 ratio) and further used to stimulate the cells for 24 hours[60]. The cells stimulated by the
bacterial culture medium (R2A, RCM or LB) served as the control group.

RNA-seq

Total RNA from HaCaT cells and primary fibroblasts was extracted using Trizol reagent (Thermo Fisher,
USA) according to the manufacturer's instructions. RNA integrity was assessed using the RNA Nano 6000

Page 11/23
Assay Kit of the Bioanalyzer 2100 system (Agilent Technologies, CA, USA). cDNA was generated with use
of M-MuLV Reverse Transcriptase and DNA Polymerase I. The library fragments were purified with
AMPure XP system (Beckman Coulter, Beverly, USA). Then PCR was performed with Phusion High-Fidelity
DNA polymerase, Universal PCR primers and Index (X) Primer. PCR products were purified (AMPure XP
system) and library quality was assessed on the Agilent Bioanalyzer 2100 system. The clustering of the
index-coded samples was performed on a cBot Cluster Generation System using TruSeq PE Cluster Kit v3-
cBot-HS (Illumia). After cluster generation, the library preparations were sequenced on an Illumina
Novaseq platform and 150 bp paired-end reads were generated.

Raw data (raw reads) of fastq format were firstly processed through in-house perl scripts. In this step,
clean data (clean reads) were obtained by removing reads containing adapter, reads containing ploy-N
and low-quality reads from raw data. Index of the reference genome was built using Hisat2 v2.0.5 and
paired-end clean reads were aligned to the reference genome using Hisat2 v2.0.5. FeatureCounts v1.5.0-
p3 was used to count the reads numbers mapped to each gene. And then FPKM of each gene was
calculated based on the length of the gene and reads count mapped to this gene. Differential expression
analysis of two groups was performed using the DESeq2 R package (1.20.0). Genes with an adjusted P-
value < 0.05 found by DESeq2 were assigned as differentially expressed. Gene Ontology (GO) enrichment
analysis of differentially expressed genes was implemented by the cluster Profiler R package, in which
gene length bias was corrected. GO terms with corrected P value < 0.05 were considered significantly
enriched by differential expressed genes.

qPCR

For real-time qPCR analysis, cDNA was synthesized using HiScript III RT SuperMix (+ gDNA wiper)
(Vazyme Biotech, China) and amplified with QuantiFast SYBR Green PCR Kit (QIAGEN, Germany) on a
QuantStudioTM 7 Flex Real-Time PCR System (Life Technologies, USA). Gene expression levels were
normalized to that of GAPDH[64]. The candidate genes were amplified by qPCR using primers MMP-1-
F(5’-GGGGCTTTGATGTACCCTAGC-3’)and MMP-1-R༈5’- TGTCACACGCTTTTGGGG TTT-3’༉; MMP-10-F༈5’-
CTCTGGA GTAA TGTCACACCTCT − 3’༉and MMP-10-R༈5’-TGTTGGTCCACCTTTCATCTTC-3’༉;MMP-12-
F༈5’-CATGAACCGTGAGGATGTT GA-3’༉and MMP-12-R༈5’-GCAT GGGCTAGGATTCCACC-3’༉༛ MMP-13-F༈5’-
CCAGACTTCACGATGGCATG-3’༉and MMP-13-R༈5’-GGCATCTCCTCCATAATT TGGC-3’༉

Declarations
Ethics approval and consent to participate

This study was ethically approved by the Ethics Committee, School of Life Sciences, Fudan University,
China. All study subjects provided written informed consent before participation.

Consent for publication

Not applicable

Page 12/23
Availability of data and material

The sequencing data from this study have been deposited in the CNSA (https://db.cngb.org/cnsa/) of
CNGBdb with accession number CNP0000635 and NODE (https://www.biosino.org/node/index) with
accession number OEP001168. A website (https://db.cngb.org/microbiome/genecatalog/genecatalog/?
gene_name=Human%20Skin%20(10.9M)) has been set up to better visualize the annotation information
of the gene catalog and guide researchers who are interested in using our data set and downloading
specific sets of data.

Competing interests

The authors declare that they have no competing interests.

Funding

This work was partially supported by the Shanghai Municipal Science and Technology Major Project
(2017SHZDZX01), National Natural Science Foundation of China (31521003, 81703097, 81770066),
Research Unit of population genetics on skin and skin diseases and new technologies for treatment and
prevention of dermatological diseases, Chinese Academy of Medical Sciences (2019-I2M-5-066), and 111
Project (B13016).

Authors' contributions

Conceptualization: XL, JW and JX; JX, LJ and YT assessed skin phenome; Data Curation: JX, ZL, QZ and
QW; Function Experiment: QZ, CD and LJ; Writing-Original Draft Preparation: JX, ZL,QZ and QW; Writing-
Review & Editing: JX, JW; Supervision: XL, JW and JX; We also thank LH,HJ for their helpful discussions.
All authors read and approved the final manuscript.

Acknowledgements

We thank Guangdong Provincial Key Laboratory of Genome Read and Write, Shenzhen Key Lab of
Neurogenomics (BGI-Shenzhen), for support in sequencing and analysis.

References
1. Kong, H.H., Skin microbiome: genomics-based insights into the diversity and role of skin microbes.
Trends In Molecular Medicine, 2011. 17(6): p. 320–328.
2. Byrd, A.L., Y. Belkaid, and J.A. Segre, The human skin microbiome. Nature reviews. Microbiology,
2018. 16(3): p. 143–155.
3. Chen, Y.E., M.A. Fischbach, and Y. Belkaid, Skin microbiota-host interactions. Nature, 2018.
553(7689): p. 427–436.
4. Grice, E.A. and J.A. Segre, The skin microbiome. Nat Rev Microbiol, 2011. 9(4): p. 244–53.

Page 13/23
 5. Harris-Tryon, T.A. and E.A. Grice, Microbiota and maintenance of skin barrier function. Science, 2022.
376(6596): p. 940–945.
6. Williams, G.C., Natural Selection, Ecology, and Morphogenesis, in Adaptation and Natural Selection.
1996, Princeton University Press: Princeton. p. 56–91.
7. Chen, H., et al., Skin Microbiome, Metabolome and Skin Phenome, from the Perspectives of Skin as
an Ecosystem. Phenomics, 2022.
8. Dimitriu, P.A., et al., New Insights into the Intrinsic and Extrinsic Factors That Shape the Human Skin
Microbiome. mBio, 2019. 10(4).
9. Oh, J., et al., Biogeography and individuality shape function in the human skin metagenome. Nature,
2014. 514(7520): p. 59–64.
10. Wei, Q., et al., Shotgun metagenomic sequencing reveals skin microbial variability from different
facial sites. Front Microbiol, 2022. 13: p. 933189.
11. Lynch, J.B. and E.Y. Hsiao, Microbiomes as sources of emergent host phenotypes. Science, 2019.
365(6460): p. 1405–1409.
12. Oh, J., et al., Shifts in human skin and nares microbiota of healthy children and adults. Genome
Medicine, 2012. 4(10): p. 77.
13. Shibagaki, N., et al., Aging-related changes in the diversity of women's skin microbiomes associated
with oral bacteria. Scientific Reports, 2017. 7(1): p. 10567.
14. Jugé, R., et al., Shift in skin microbiota of Western European women across aging. Journal of Applied
Microbiology, 2018. 125(3): p. 907–916.
15. Li, Z., et al., New Insights Into the Skin Microbial Communities and Skin Aging. Frontiers In
Microbiology, 2020. 11: p. 565549.
16. Krutmann, J., et al., Environmentally-Induced (Extrinsic) Skin Aging: Exposomal Factors and
Underlying Mechanisms. J Invest Dermatol, 2021. 141(4S): p. 1096–1103.
17. de Rigal, J., et al., The effect of age on skin color and color heterogeneity in four ethnic groups. Skin
Research and Technology: Official Journal of International Society For Bioengineering and the Skin
(ISBS) [and] International Society For Digital Imaging of Skin (ISDIS) [and] International Society For
Skin Imaging (ISSI), 2010. 16(2): p. 168–178.
18. Wu, Y., et al., Objective measurement and comparison of human facial skin color in East Asian
females. Skin Research and Technology: Official Journal of International Society For Bioengineering
and the Skin (ISBS) [and] International Society For Digital Imaging of Skin (ISDIS) [and] International
Society For Skin Imaging (ISSI), 2020. 26(4): p. 584–590.
19. Oh, J., et al., Temporal Stability of the Human Skin Microbiome. Cell, 2016. 165(4): p. 854–866.
20. Zhai, W., et al., Profile of the skin microbiota in a healthy Chinese population. The Journal of
Dermatology, 2018. 45(11): p. 1289–1300.
21. Kim, H.-J., et al., Aged related human skin microbiome and mycobiome in Korean women. Scientific
Reports, 2022. 12(1): p. 2351.

Page 14/23
22. Kim, J.-H., et al., Taxonomic profiling of skin microbiome and correlation with clinical skin
parameters in healthy Koreans. Scientific Reports, 2021. 11(1): p. 16269.
23. Kim, M., et al., Investigation of Age-Related Changes in the Skin Microbiota of Korean Women.
Microorganisms, 2020. 8(10).
24. Zhu, T., et al., Age and Mothers: Potent Influences of Children's Skin Microbiota. The Journal of
Investigative Dermatology, 2019. 139(12).
25. Limon, J.J., J.H. Skalski, and D.M. Underhill, Commensal Fungi in Health and Disease. Cell Host
Microbe, 2017. 22(2): p. 156–165.
26. Deutscher, J., C. Francke, and P.W. Postma, How phosphotransferase system-related protein
phosphorylation regulates carbohydrate metabolism in bacteria. Microbiol Mol Biol Rev, 2006. 70(4):
p. 939–1031.
27. Li, Z., et al., Characterization of the human skin resistome and identification of two microbiota
cutotypes. Microbiome, 2021. 9(1): p. 47.
28. Moura-Alves, P., et al., Host monitoring of quorum sensing during infection. Science (New York, N.Y.),
2019. 366(6472).
29. Gorgoulis, V., et al., Cellular Senescence: Defining a Path Forward. Cell, 2019. 179(4): p. 813–827.
30. Coppe, J.P., et al., The senescence-associated secretory phenotype: the dark side of tumor
suppression. Annu Rev Pathol, 2010. 5: p. 99–118.
31. Panwar, P., et al., Aging-associated modifications of collagen affect its degradation by matrix
metalloproteinases. Matrix biology: journal of the International Society for Matrix Biology, 2018. 65:
p. 30–44.
32. Bonnans, C., J. Chou, and Z. Werb, Remodelling the extracellular matrix in development and disease.
Nature reviews. Molecular cell biology, 2014. 15(12): p. 786–801.
33. Ma, Z.S., Niche-neutral theoretic approach to mechanisms underlying the biodiversity and
biogeography of human microbiomes. Evol Appl, 2021. 14(2): p. 322–334.
34. Watabe, A., et al., Sweat constitutes several natural moisturizing factors, lactate, urea, sodium, and
potassium. J Dermatol Sci, 2013. 72(2): p. 177–82.
35. McGrath, J.A., Filaggrin and the great epidermal barrier grief. Australas J Dermatol, 2008. 49(2):
p. 67–73; quiz 73 – 4.
36. Caspers, P.J., et al., In vivo confocal Raman microspectroscopy of the skin: noninvasive
determination of molecular concentration profiles. J Invest Dermatol, 2001. 116(3): p. 434–42.
37. Cui, L., et al., Advancements in the maintenance of skin barrier/skin lipid composition and the
involvement of metabolic enzymes. J Cosmet Dermatol, 2016. 15(4): p. 549–558.
38. Pappas, A., Epidermal surface lipids. Dermatoendocrinol, 2009. 1(2): p. 72–6.
39. Barnard, E., et al., Porphyrin Production and Regulation in Cutaneous Propionibacteria. MSphere,
2020. 5(1).

Page 15/23
40. Spittaels, K.J., et al., Porphyrins produced by acneic Cutibacterium acnes strains activate the
inflammasome by inducing K(+) leakage. iScience, 2021. 24(6): p. 102575.
41. Shu, M., et al., Porphyrin metabolisms in human skin commensal Propionibacterium acnes bacteria:
potential application to monitor human radiation risk. Current Medicinal Chemistry, 2013. 20(4):
p. 562–568.
42. Marples, R.R., D.T. Downing, and A.M. Kligman, Control of free fatty acids in human surface lipids by
Corynebacterium acnes. J Invest Dermatol, 1971. 56(2): p. 127–31.
43. Ellison, A.M., Foundation Species, Non-trophic Interactions, and the Value of Being Common.
iScience, 2019. 13: p. 254–268.
44. Zhai, W., et al., Profile of the skin microbiota in a healthy Chinese population. J Dermatol, 2018.
45(11): p. 1289–1300.
45. Somboonna, N., et al., Bacterial communities on facial skin of teenage and elderly Thai females.
Arch Microbiol, 2017. 199(7): p. 1035–1042.
46. Wilantho, A., et al., Diversity of bacterial communities on the facial skin of different age-group Thai
males. PeerJ, 2017. 5: p. e4084.
47. Leung, M.H., D. Wilkins, and P.K. Lee, Insights into the pan-microbiome: skin microbial communities
of Chinese individuals differ from other racial groups. Sci Rep, 2015. 5: p. 11845.
48. Juge, R., et al., Shift in skin microbiota of Western European women across aging. J Appl Microbiol,
2018. 125(3): p. 907–916.
49. Shibagaki, N., et al., Aging-related changes in the diversity of women's skin microbiomes associated
with oral bacteria. Sci Rep, 2017. 7(1): p. 10567.
50. Kim, J.H., et al., Taxonomic profiling of skin microbiome and correlation with clinical skin parameters
in healthy Koreans. Sci Rep, 2021. 11(1): p. 16269.
51. Tabbuso, T., et al., Moraxella osloensis infection among adults and children: A pediatric case and
literature review. Archives de Pediatrie: Organe Officiel de La Societe Francaise de Pediatrie, 2021.
28(4): p. 348–351.
52. Cosseau, C., et al., from the human skin microbiota: Species-level diversity and hypotheses. One
Health (Amsterdam, Netherlands), 2016. 2: p. 33–41.
53. Lim, J.Y., et al., Complete Genome Sequences of Three Strains Isolated from Human Skin. Genome
Announcements, 2018. 6(3).
54. Larson, P.J., et al., Associations of the skin, oral and gut microbiome with aging, frailty and infection
risk reservoirs in older adults. Nature Aging, 2022. 2(10): p. 941–955.
55. Delwart, E., A roadmap to the human virome. PLoS Pathog, 2013. 9(2): p. e1003146.
56. Schommer, N.N. and R.L. Gallo, Structure and function of the human skin microbiome. Trends
Microbiol, 2013. 21(12): p. 660–8.
57. Oh, J., et al., Temporal Stability of the Human Skin Microbiome. Cell, 2016. 165(4): p. 854–66.

Page 16/23
58. Erickson, A.K., et al., Bacteria Facilitate Enteric Virus Co-infection of Mammalian Cells and Promote
Genetic Recombination. Cell Host Microbe, 2018. 23(1): p. 77–88 e5.
59. Barr, J.J., et al., Bacteriophage adhering to mucus provide a non-host-derived immunity. Proc Natl
Acad Sci U S A, 2013. 110(26): p. 10771–6.
60. Moura-Alves, P., et al., Host monitoring of quorum sensing during Pseudomonas aeruginosa
infection. Science (New York, N.Y.), 2019. 366(6472).
61. Bottai, G., et al., 17β-estradiol protects human skin fibroblasts and keratinocytes against oxidative
damage. Journal of the European Academy of Dermatology and Venereology: JEADV, 2013. 27(10):
p. 1236–1243.
62. Eltwisy, H.O., et al., Pathogenesis of Staphylococcus haemolyticus on primary human skin fibroblast
cells. Virulence, 2020. 11(1): p. 1142–1157.
63. Park, I.-H., et al., Reprogramming of human somatic cells to pluripotency with defined factors. Nature,
2008. 451(7175): p. 141–146.
64. Yamamoto, Y., et al., Long-term expansion of alveolar stem cells derived from human iPS cells in
organoids. Nature methods, 2017. 14(11): p. 1097–1106.

Figures

Figure 1
Page 17/23
(a) The overall workflow of this study. (b) Spearman’s correlation between phenotype parameters and
age. FH, forehead. CK, cheek. NS, the back of the nose. Blue, negative correlation. The color bar represents
correlation values. The significance levels in the Spearman correlation are: *, p< 0.05; **, p < 0.01; ***,
p<0.001. (c) Bar charts illustrating the effect size of variables on bacterial, fungal and viral compositions.
Variables found to be significant with species were shown. Adjusted R2 values represent the fractions of
microbial composition variation explained by the variables. Different colors represent different kingdoms.
Partially created with BioRender.com.

Page 18/23
Figure 2

Dynamic changes of bacterial community during aging. (a) Different age groups. Young group (20 to 35,
N=74), middle-aged group (36 to 50, N=131), and old group (above 50, N=89). (b) Alpha diversity of the
bacterial community increased slightly during aging. Wilcox test is used to determine significance. (c)
The beta diversity of the bacterial community was different between the three age groups. Wilcox test is
used to determine significance. (d) Distance-based redundancy analysis (dbRDA) depicting the variations
Page 19/23
of bacterial community composition between different age groups. Each point represents a single
sample. Colors represent age groups. Shapes represent different sites. Lines represent different ages. (e)
Associations between bacterial taxa and age. Colors represent three sites: Green, CK(cheek); Orange,
FH(forehead); Purple, NS(back of the nose). (f) Co-occurrence network of age-related species. The nodes
represent age-related species. The colors of nodes represent the phylum. Blue lines represent positive
correlations. Red lines represent negative correlations.

Figure 3

Correlation between bacteria, fungi and phenotype parameters. Heatmap of the Spearman's correlation
between the top 20 bacteria (a) and top 10 fungi (b) in abundance and skin phenotype parameters on the
forehead. Blue, negative correlation; Red, positive correlation. The significance levels in the Spearman
correlation are: *, p< 0.05; **, p < 0.01; ***, p<0.001.

Page 20/23
Figure 4

(a)Multiple linear regression (MLR) model was used to predict skin-age. Red, belonging to premature-
aging group; blue, belonging to delayed-aging group. (b)The differences in microbial composition
between the young skin group and the elderly skin group on the FH, CK and NS. Red, bacteria that
enriched in the elderly skin group. Blue, bacteria that enriched in young skin group. (c) Features of
bacterial functions between premature-aging and delayed-aging group on the FH, CK and NS. Bar-chart
showing the relative abundance of differential KEGG-modules. Red, modules enriched in the premature-
aging group. Blue, modules enriched in the delayed-aging group.

Page 21/23
Figure 5

Genes expressed by keratinocytes and fibroblasts co-cultured with three skin bacteria. (a) Scheme of skin
cells stimulated by bacterial supernatant. RNA-seq analysis of HaCaT cells and primary dermal fibroblast
upon 24 hours of stimulation with supernatant from M.osloensis, C. acnes and S.epidermidis grown in
R2A, RCM or LB medium, at OD 600 0.8-1.0(n = 3). The cells stimulated by the bacterial culture medium
(R2A, RCM or LB) served as the control group. (b) GO enrichment showing the top 10 enrichment
Page 22/23
pathways of HaCaT cells treated with bacterial supernatant from M. osloensis (p < 0.05). (c) Gene
expression analysis of several MMPs in HaCaT cells upon 24 hours of stimulation with supernatants of
M. osloensis, C. acnes and S. epidermidis, individually (*, p< 0.05; **, p < 0.01; ***, p<0.001. data pooled
from n = 3 independent experiments). Heatmap of the results of the Spearman's correlation between the
three species and aging-related genes in fibroblasts (d) and in HaCaT cells (e) (adjusted P < 0.05).

Supplementary Files
This is a list of supplementary files associated with this preprint. Click to download.

FigureS1.pdf
FigureS2.pdf
FigureS3.pdf

Page 23/23