Thank you for visiting Nature. The browser version you are using has limited support for CSS. For the best experience, we recommend that you use a newer version of the browser (or turn off the compatibility mode in Internet Explorer). At the same time, to ensure continued support, we will display sites without styles and JavaScript.

Scientific Reports Volume 11, Article Number: 18302 (2021) Cite this article

Environmental pollution is composed of many factors, namely particulate matter (PM2.5, PM10), ozone, and ultraviolet (UV), etc. The first tissue exposed to these substances is the skin epidermis. It has been confirmed that exposure to air pollution can aggravate various skin diseases such as eczema, acne, freckles and wrinkles. Although pollutants can interact with the skin surface, ultrafine particles or polycyclic aromatic hydrocarbons (PAH) may contaminate deep skin layers because they are present in the blood and hair cortex. The molecular mechanism of skin dysfunction caused by pollution exposure has been seldom explored in humans. In addition to various host skin components, the skin microbiome is another target of these environmental aggressors, and can actively promote visible clinical manifestations such as wrinkles and aging. This study aims to investigate the association between pollution exposure, skin microbiota, metabolites, and skin clinical symptoms of women from two cities with different pollution levels. Analysis of non-targeted metabolomics and targeted proteins from D-Squame samples of healthy women (n = 67 per city), 25-45 years old, living in Baoding, China for at least 15 years (used as a pollution area model) And Dalian (a control area with a lower degree of pollution). Additional swab samples were collected from the cheeks of the same population, and the microbiome was analyzed using bacterial 16S rRNA and fungal ITS1 amplicon sequencing and metagenomics analysis. The level of pollution exposure was individually assessed by analyzing polycyclic aromatic hydrocarbons (PAH) and their metabolites in hair samples collected from each participant. All participants in the study received an assessment of the clinical parameters of the skin (acne, wrinkles, stains, etc.). Women from two cities (polluted and less polluted) showed different metabolic characteristics and changes in skin microbiota. Analytical data from 350 identified metabolites, 143 microorganisms, and 39 PAHs were used to characterize biochemical events related to pollution exposure. Finally, using the multi-block data analysis method, we obtained a potential molecular map composed of multiple omics features that are related to the existence of skin pigmentation disorders in individuals living in polluted environments. In general, these characteristics point to changes in macromolecules caused by pollution, which may manifest as clinical symptoms of early skin pigmentation and/or other defects.

The Lancet Pollution and Health Committee reported in 2017 that pollution is still the world’s greatest environmental threat to human health. In 2017, it caused 15% of all deaths worldwide and 275 million disability-adjusted life years. Ambient air pollution caused more people around. People die globally more than any other form of pollution (https://gahp.net/pollution-and-health-metrics/). Epidemiological and clinical studies have shown that short-term and long-term exposure to particulate matter (PM2.5: particulate matter with an aerodynamic diameter of less than 2.5 microns, PM10: particulate matter with an aerodynamic diameter of less than 10 microns) and ozone can increase respiratory and cardiovascular disease Rate and lead to the development of certain cancers1,2,3. Of all organs, the skin is the most obvious, and more and more scientific evidence shows that air pollution plays an important role in external aging4,5. The skin is a multi-layered tissue consisting of the uppermost stratum corneum, which is in direct contact with atmospheric pollutants, while the deeper layers, such as the dermis, are exposed to various xenobiotics through systemic pathways. One of the main mechanisms by which PM exerts its harmful effects is through the production of oxidative stress6,7, which is an important factor leading to external skin aging8,9. PM2.5 is usually related to toxic chemicals such as heavy metals or polycyclic aromatic hydrocarbons (PAHs). Some photoreactive PAHs can cause strong oxidative stress under UVA exposure10. Clinical studies conducted in China and elsewhere have shown phenotypes of premature skin aging, such as pigmented spots and wrinkles11,12. In previous studies using in vitro skin models and non-targeted proteomics, we have shown that exposure to PAH(s) leads to the destruction of several cellular processes6. However, the molecular mechanisms linking exposure and its effects on clinical manifestations remain to be deciphered.

The current work is part of a multi-parameter study in which non-invasive facial skin samples (D-Squame) were collected from 134 healthy women living in two cities with different pollution levels in China13. According to a verification method that provides information on average internal doses of chemicals, the exposure was determined by quantifying 39 parent PAHs and their monohydroxy metabolites in hair samples collected from the same person. A set of dermatological skin parameters for each person was also evaluated. The microbiome analysis of all subjects from the two cities gave us the first insight into the changes in the microbiome caused by pollution exposure14. In order to explore the mechanistic link between pollution exposure and its clinical manifestations, we considered several molecular features. We first apply non-targeted metabolomics analysis to skin samples to characterize pollution-dependent biochemical events, and then perform targeted proteomics analysis to supplement the biochemical characteristics generated by metabolomics. Finally, we supplemented multiple omics data with 16S and ITS amplicon sequencing. PAH quantification and clinical evaluation of the same individual complete this global characterization. These data sets are calculated in a block structure multivariate analysis to propose a molecular map of human skin, linking pollution exposure to skin dysfunction.

In order to assess the effects of long-term exposure to pollution on skin exposure areas, we conducted 35 tests on 67 women living in Baoding (the most polluted area) and 67 women living in Dalian (the most polluted area). Dermatological clinical evaluation of facial parameters. Contaminated area) in China. These women have lived in their respective cities for at least 15 years and are between 25 and 45 years old. The clinical parameters analyzed for these women are shown in the supplement. Table 1. In short, 10 main facial clinical themes/clusters for each person were evaluated, including 35 sub-conditions. Pigment disorders and wrinkles are the two clinical themes with the most significant increase in facial skin of individuals in contaminated areas. Shiny and dull skin was also adjusted in the second instance (Supplementary Table 1). Previously, we reported that people from two cities had significantly different levels of PAH measured in their hair shafts, indicating different levels of exposure to outdoor pollution13. These initial clinical evaluations were performed in a larger population of 204 women, and this difference is still significant for the 134 study subgroup (Supplementary Figure 1). Therefore, we calculated the PAH score (described in the "Materials and Methods" section) as a supplementary factor for analysis along with the other data sets introduced in this article (clinical, metabolites, microbiome, etc.). As shown in Table 1, macules on the forehead and cheeks were significantly more common among women in Baoding and Dalian, the most polluted cities (25.4% vs. 6% and 56.7% vs. 35.8%, respectively). For women with PAH scores higher than the median of the study population, a similar pattern was observed in the prevalence of diffuse macula on the forehead and cheeks, while women with PAH scores below the median (22.4% vs. 9% and 53.7 % Vs. 38.8% respectively). Previous studies have shown that among people with skin types III-IV living in India and Southeast Asia, the incidence of facial hyperpigmentation diseases (such as melasma, spots, and freckles) is the highest15, which are also geographical areas with a heavy burden of environmental PAH ( https://gahp.net/pollution-and-health-metrics/). Our research results show a similar trend. Spreading macula (SP) is a facial pigment disease that is little known and has been previously described in South Asian populations16 living in large cities in India, which are usually the most polluted cities.

Non-targeted metabolomics analysis on facial skin tape strips allowed the identification of a total of 350 metabolites. We first performed pathway enrichment analysis to explore whether long-term exposure to pollution would lead to accumulation or disturbance of specific metabolic pathways. At higher levels, increased enrichment was observed in samples from Baoding, including amino acid and fatty acid metabolism (Figure 1A and Supplementary Table 2). Compared with Dalian, dietary sources of xenobiotics seem to have decreased. Further analysis showed increased levels of N-acetyl amino acids (Figure 1B), γ-glutamyl amino acids (Figure 1C) and urea cycle intermediates. N-acetyl amino acids are derived from proteins that have undergone post-translational acetylation or free amino acids that react with acetyl groups. Gamma-glutamyltransferase (GGT) is an enzyme that transfers the gamma-glutamyl portion of glutathione to a receptor, which can be an amino acid or peptide17. Therefore, the GGT system plays an important role in transporting amino acids and dipeptides into cells and regulating the exchange of intracellular and extracellular glutathione. The higher levels of γ-glutamyl amino acids in the contaminated group indicate that the GGT system is up-regulated, which may be to provide cellular pyrrolidone carboxylic acid (PCA or 5-oxoproline), which is the natural moisturizer released by the GGCT enzyme after GGT activation Factor (NMF) 18. NMF is an amino acid or its derivatives (PCA and urocanic acid, together with lactic acid, urea, citrate, and sugar). They are produced by the hydrolysis of filaggrin and are only found in the stratum corneum (SC). The levels of free amino acids (Figure 2A), 5-oxoproline and lactic acid in Baoding and Dalian increased (Figure 2B), indicating that the proteolytic pathway of filaggrin was up-regulated. To further explore the latter, we conducted targeted protein analysis on the proteins involved in the skin barrier function. Surprisingly, we observed decreased levels of CASP14, PADI1, TGM3 and GGCT. Afterwards, γ-glutamyltransferase is involved in the last step of PCA synthesis, and due to the increase in PCA levels, this observation indicates another mechanism that may lead to an increase in PCA. Interestingly, KLK7, the main enzyme in peeling, was reduced by 2 times, indicating that the surface renewal rate was slowed down, which may result in the accumulation of PCA and other amino acids. KLK5, another peeling activating enzyme, can be considered unchanged, with a 1.2-fold increase. To further support this accumulation hypothesis, filaggrin and filaggrin 2 as the main NMF providers remained unchanged in skin samples from Baoding (Supplementary Figure 2). In conclusion, protein and metabolite analysis showed that the disturbance of molecular entities in Baoding skin samples contributes to skin barrier function and enhanced repair/compensation. It seems that long-term exposure to pollution causes the skin to be in a dynamic state to manage damage and repair at the same time. We previously reported the effect of pollution on skin barrier function using in vitro models and non-targeted proteomics.

Differentially regulated metabolic pathways in individual skin samples from the polluted city Baoding. (A) The log2 heat map of the multiple change of Baoding (more pollution) and Dalian (less pollution). A positive value indicates that Baoding is higher, and a negative value indicates that Dalian is higher. (B) Focus on N-acetyl amino acids, and box plots of their log-normalized values, colored by site. (C) Focus on γ-glutamyl amino acids, box plots with log-normalized values, colored by site. For a given metabolite, a red star indicates a significant difference in the average value between the two cities (q value <0.05, using the Benjamini-Hochberg method to adjust for multiple testing using t-test). The figures were created using R software: R Core Team (2017), R Foundation for Statistical Computing, Vienna, Austria. URL https://www.R-project.org/.

(A) Box plot of logarithmic conversion values ​​of free amino acids (B) Box plot of logarithmic conversion values ​​of NMF. The red star indicates that for a given metabolite, the average difference between Baoding (more polluted) and Dalian (less polluted) is significant (q value <0.05, t-test is adjusted using multiple tests). Jemini-Hochberg method). The figures were created using R software: R Core Team (2017), R Foundation for Statistical Computing, Vienna, Austria. URL https://www.R-project.org/.

As a next step, in order to more accurately quantify the differences in metabolomics characteristics between the two cities, without considering the classification as sub-pathway or super-pathway, we calculated the log2-Fold of all metabolites in Baoding as the reference population Change and express it as the cumulative distribution in Suppl_Fig. 3. In addition, in order to test the significance of the fold change, we performed a t-test on each of them using the volcano map visualized in Figure 3A.

(A) The volcano graph shows log2(FC) on the x-axis and -log10(q − value) on the y-axis. Add the threshold 0.6 of log2(FC) as the vertical black line, and add the threshold of -log10(0.05) as the horizontal black line of the q value. For presentation purposes, the y-axis scale is compressed 5 times between 5 and 12. If the metabolite's |log2(FC)|> 0.6 and q value <0.05 (red and marked as graph). (B) Box plot of significantly regulated metabolites, collected by route, Baoding (more pollution) and Dalian (less pollution). The figures were created using R software: R Core Team (2017), R Foundation for Statistical Computing, Vienna, Austria. URL https://www.R-project.org/.

In order to further study the relationship between metabolite profile and city membership, we also performed a multivariate analysis using a random forest classification model (for more details, please refer to the "Methods" section). The variable importance value is shown in Suppl_Fig. More details of the RF model are given in Suppl_Fig. 4A. 4B, C.

When regrouping by pathway (Figure 3B), we found the most important members of the different metabolite categories mentioned above, so we can gain a deeper understanding of the mechanisms most significantly related to pollution exposure. In the lipid group, oleamide, oleoyl and linoleyl ethanolamide are fatty acid amides, which are biologically active lipid signaling molecules. It has been reported that these are related to a wide range of physiological reactions in various tissues, including skin diseases such as atopic dermatitis and contact dermatitis20. Particulate matter leads to the secretion of pro-inflammatory cytokines, such as tumor necrosis factor (TNF)-α, IL-1α, IL-8, and up-regulation of matrix metalloproteinases 1, 2 and 921,22. Biologically active fatty acids are likely to be caused by pollution Inflammatory stress response. Among other metabolites, 4-imidazole acetate, N-acetylarginine, and unknown X-13737 were significantly up-regulated, while p-cresol sulfate was down-regulated (Figure 3B). The detection of increased levels of 4-imidazole acetate indicates a disorder of histidine metabolism, which will be discussed in the next section. Elevated levels of kynurenate indicate a disorder of tryptophan metabolism, and a focused analysis showed that indoxyl-3-sulfate decreased in individuals in Baoding (Supplementary Figure 5). The kynurenine pathway (KP) of tryptophan metabolism accounts for most of the tryptophan that is not involved in protein synthesis, including compounds that are active in the nervous and immune systems. Kynurenine acts on the aromatic hydrocarbon receptor (AhR), affects the metabolism of exogenous substances and affects carcinogenesis. Limited research has focused on the effects of acute ultraviolet radiation and the induction of KP in human skin-derived fibroblasts and keratinocytes. It is known that ultraviolet radiation can trigger inflammatory components in skin cells. These cells may induce KP under ultraviolet radiation, and this KP metabolite may be a mediator of inflammation and anti-inflammatory reactions23. We have observed a decrease in indoxyl-3-sulfate in individuals living in the most polluted areas. Several studies have identified the tryptophan metabolite indole as the main extracellular metabolite produced by gut bacteria (such as E. coli). Interestingly, indoxy-3-sulfate has been characterized as AhR agonist剂24. A decrease in p-cresol sulfate was found in skin samples from Baoding (Figure 2A). P-cresol has previously been reported as a biomarker of healthy aging25.

The other top mark in the volcano map is the unknown metabolite "X-13737" (Figure 3A, B). After further investigation, it was determined that the MS/MS spectrum of "X-13737" actually contained a mixture of ions from two metabolites, one is (S)-a-amino-ω-caprolactam, and the other is not yet identified Molecule (Suppl_Fig. 6). Although the exact identity and contribution of this additional co-elution whose signal "X-13737" is unknown has not been fully verified, the data shows that it is the main participant in distinguishing the metabonomic characteristics of individuals living in one city from another city. One of them (Suppl_Fig. 4). The novelty of this discovery is that caprolactam itself is a moderately toxic irritant, which was previously considered a harmful air pollutant. Although it is not known how (S)-a-amino-omega-caprolactam is formed from caprolactam (or if possible), but the fact that caprolactam is a known air pollutant, it is worth noting that the metabolomics data is generated and Compared with other women, the skin of women exposed to pollutants has an increased signal of (S)-α-amino-ω-caprolactam.

From the above analysis, we found that there are significant differences in the skin metabolomics characteristics of individuals living in polluted and non-polluted environments. To take it a step further and establish a direct correlation between exposure and its molecular effects on the skin, we used sparse canonical correlation analysis (sCCA) to analyze PAHs concentration (measured in a single hair sample) and skin SC metabolites to identify the correlation Linked polycyclic aromatic hydrocarbons and metabolite groups. Hair analysis is increasingly used to assess exposure, and some studies have shown that the concentration of pollutants in the hair represents a body burden26,27,28. Hair analysis can provide comprehensive information on chronic exposure for several months (considering an average increase of one centimeter per month), and allows the detection of maternal pollutants and their metabolites that are contrary to biological fluids26. Among the different correlations observed between PAH and skin metabolites, the positive correlation with N-acetylamino acids is interesting (Figure 4A). Acetylation is a post-translational protein modification that has many effects on cellular proteins and metabolites. Acetyl-CoA provides an acetyl group that can be linked to the alpha amino group of the N-terminal of the protein or the epsilon amino group of a lysine residue after translation. This reaction is catalyzed by N-acetyltransferase (NAT), which is involved in a variety of signaling pathways that affect a variety of cellular functions. The correlation between N-acetyl aa and PAH is interesting. Previous studies have shown a link between exposure to particulate matter and histone acetylation30. Other studies have shown that NAT and cotinine levels are related to secondhand smoke exposure31. Interestingly, we observed a strong correlation between skin metabolites and 2OH naphthalene (a PAH metabolite related to smoking)32. In short, the relationship between N-acetyl amino acids and polycyclic aromatic hydrocarbons needs further study. Another interesting pathway is histidine metabolism, as a correlation between cis-urocamsate and 4-imidazole acetate and PAH was observed (Figure 4A). Previous studies have shown that benzo[a]pyrene can disrupt histidine metabolism in human lung epithelial cells33. We observed elevated cis-urate levels in individuals from contaminated cities (Figure 4B). The level of trans uricate between the two cities remained constant, leading to an increase in the cis/trans ratio of the polluted city Baoding (Figure 4B). Histidine is deaminated by histase to form trans-uric acid salt. Although trans-uronic acid salt acts as a photoprotective agent, cis-uronic acid salt is related to suppressing the immune response, leading to ultraviolet-induced immunosuppression.

(A) Heat map of cross-correlation between metabolites selected by sparse canonical correlation analysis (sCCA) and PAH. The value in the heat map corresponds to the Pearson correlation coefficient. (B) Boxplots using log-normalized values ​​focus on cis- and trans-urocanate metabolites. We also express the ratio of cis/trans uric acid esters. The red star indicates that there is a significant difference in the average value between Baoding (more pollution) and Dalian (less pollution) (q value <0.05, using the Benjamini-Hochberg method to adjust the multiple test t-test). The figures were created using R software: R Core Team (2017), R Foundation for Statistical Computing, Vienna, Austria. URL https://www.R-project.org/.

On the surface of the skin, there are complex interactions between host metabolism and its microbiota. It cannot be ruled out that a subset of the above-mentioned metabolites is directly produced or altered by the skin microbiota. In fact, the skin microbiota plays an important role in maintaining skin health, and the skin microbiota constantly adapts to internal and external factors. Environmental and pollution exposure may affect the skin microbial community, and bacteria isolated from human skin have been shown to degrade polycyclic aromatic hydrocarbons and related exogenous compounds33.

In previous studies based on the same cohort, we have shown that the increase in Shannon's diversity is related to the PAH level score. In the metagenomics analysis of a limited number of individuals (n = 32), pathways to host-microbe interactions and aromatic compound degradation were determined, which further supports this observation. The PAH levels in the hair of these individuals are very high. High and very low samples 14. Here, in order to explore the effect of the microbiome on skin metabolites and vice versa, we used the PLS prediction model and Pearson's correlation between metabolites and microbiome for all 134 individuals. In previous studies on the gut, it was reported that the fecal metabolome accounted for 68% of the difference in the composition of the gut microbiome, but this comparison has not yet been explored for the skin to a large extent. The persistence of metabolites derived from skin care products35 or urbanization-related chemicals36 or higher-level 3D modeling of skin surface components37 has been concerned. In our research, we analyzed the endogenous metabolites and microbial diversity at the same time for the first time.

Using the predictive model, the metabolome of the skin samples from the two cities accounted for 35% of the variance of bacterial diversity. Using the predictive model, the metabolome of skin samples accounted for only 14% of the variance of fungal skin diversity, indicating that bacterial diversity contributes higher to the composition of skin metabolites than fungi. Pearson's correlation highlights 42 metabolites that are significantly related to bacterial diversity, mainly amino acids (n = 19; including kynurenate), peptides (n = 10; including γ-glutamylglutamate) ) And lipids (n = 11; including maleate and oleamide) (Supplementary Table 3). Two acetylated amino acids, N6-acetyllysine and N-δ-acetylornithine, were determined to be related to the bacteria Shannon The strongest metabolite associated with reduced diversity. For fungal diversity, except for three metabolites, Pearson's correlation is very weak (< 0.25): trans-4-hydroxyproline; linoleate (18:2n6) and nicotine (Suppl_Table 3). Our observations indicate that bacterial diversity contributes more to the composition of metabolites than fungi.

To study this analysis in more depth, we performed sparse CCA ("Methods" section). As shown in Figure 5A, in addition to the environmental bacteria Bacillus sp. B52, sCCA mainly recovered symbiotic bacterial groups (Propionibacterium B1, B19 and B8535; Staphylococcus B107 and B4; Corynebacterium B21). It is mainly positively correlated with lipid metabolites (6/11 metabolites); lauric acid and myristic acid are the metabolites with the strongest correlation with Staphylococcus B107. As mentioned earlier, the presence of free fatty acids can reflect the hydrolysis of triglycerides by microbial lipase 37, 38. The second group of metabolites is positively correlated with OTU. Three acylcarnitines (C18:1; C14:1; C16:1) are collected. They are not synthesized by bacteria, but can be used as a source of carnitine. Carnitine is one A known nutrient and osmoprotectant, for bacteria 39, suggests that these acylcarnitines/carnitines are beneficial to skin symbiosis on the skin. This result is in contrast with the intestinal metabolome, in the case of intestinal dysbiosis, acylcarnitine is enriched. Only two amino acids are positively correlated with Propionibacterium and Staphylococcus OTU, which correspond to the same metabolites related to bacterial diversity: N6-acetyllysine and the arginine metabolite N-δ-acetylornithine. Although staphylococcus and other skin bacteria have the ability to produce these amino acids, as far as we know, there is no existing data to support their contribution to the skin metabolome content, nor is it related to skin dysbiosis. In addition, two metabolites are positively correlated with bacterial taxa: histamine and phenyllactic acid30. Histamine is a well-known mediator of allergic reactions41. Histamine is not only synthesized in mast cells, but also produced by symbiotic microorganisms in the intestine under physiological conditions42,43. We can assume a similar situation on the skin, because Propionibacterium and Staphylococcus also have the ability to secrete histamine44. PLA is a biological preservative produced by propionic acid bacteria such as Propionibacterium, Lactobacillus (LAB) 45 and Paenibacillus. In view of the symbiotic relationship between Paenibacillus and stingless honeybees, which is related to the antibacterial activity of PLA, our observation puts forward the possibility that PLA produced by skin symbionts can protect the skin from external pathogens46. The skin metabolome analysis introduced in the previous sections showed that the tryptophan pathway is regulated in individuals living in the polluted city Baoding (Supplementary Figure 5). Specifically, we have observed that indoxyl-3-sulfate, an aromatic hydrocarbon receptor (AhR) agonist, is reduced in individuals living in heavily polluted cities, and is reported to be significantly reduced in the skin of AD patients. The current analysis did not show a significant correlation between indoxyl-3-sulfate and bacterial taxa. Metagenomics is being used to conduct additional analyses at the functional level to study the link between tryptophan metabolism and the microbiome.

The cross-correlation heat map between OTU abundance and metabolites selected by sparse canonical correlation analysis (sCCA), adjusted for urban confounding factors. The value in the heat map corresponds to the Pearson correlation coefficient. (A) Bacterial OTU. (B) Fungus OTU. The figures were created using R software: R Core Team (2017), R Foundation for Statistical Computing, Vienna, Austria. URL https://www.R-project.org/.

Most of the fungal taxa selected with sCCA correspond to Malassezia OTU (Figure 5B). Compared with bacteria with amino acids/peptides (arginine, alanine, γ-glutamyl histidine), long-chain sphingosine bases, sphingosine, sphingosine and sphingosine, and unknown metabolites , They have obvious correlation. Phytosphingosine, sphingosine and sphingosine are the main components of stratum corneum ceramide. In addition to their role in barrier function, these molecules are reported to be beneficial to the growth of Malassezia. It also inhibits the growth of opportunistic pathogens such as Candida albicans; this may explain the correlation observed in this study.

Comparison of skin microbiota and skin metabolome highlights the close relationship between sebum degradation and bacterial taxa (Propionibacterium and Staphylococcus), which also reveals that these taxa are related to carnitine, histamine and PLA present on the skin The potential connection between this may represent a new symbiosis of factors-host homeostasis. Due to the stronger correlation between bacterial diversity and metabolites, we decided to focus on bacterial taxa for further analysis.

The results presented so far allow us to characterize individual parameters such as clinical, metabolites, microbiome and their correlation with pollution exposure. This analysis revealed certain mechanisms related to barrier function, as shown by metabolites and protein and/or sebum degradation in microbiome analysis. In order to investigate the possible links between these different biological parameters, exposures, and visible and perceptible clinical skin dysfunction, we conducted a multi-block data analysis (see the "Methods" section for details), which involves the analysis of observations under the same conditions Analysis of multiple sets of variables (blocks). personal. It is necessary to jointly analyze different data sets to discover their main sources of covariance, and it is necessary to use very specific calculation methods, such as the method recently proposed by Tenenhaus et al. in the generalized canonical correlation analysis framework49. This method can obtain unique representations of individuals and descriptors in the consensus space, which is an important step for searching specific profiles/clusters based on classification49,50.

According to the method described in the method section, the entire population is divided into 4 different clusters, as shown in Figure 6. Cluster 1 and cluster 3 are the largest, with 93.4% (n = 46) and 92.8% (n = 42) of Baoding being the population of Dalian. Cluster 3 (n = 15) and cluster 4 (n = 29) are smaller and have mixed populations. Each cluster has its own unique molecular spectrum, consisting of a series of metabolites, microorganisms, PAH and its derivatives. Each of these clusters is characterized by using a v-test to characterize variables that are differentially expressed in the cluster relative to the entire study population (see the "Methods" section for more details). We then obtained a subset of meaningful/relevant clinical parameters for each cluster. Interestingly, among the 46 people in group 1, 26 (57%) and 11 (25%) had scattered macula on the cheeks and forehead respectively, and this group also had the highest severity score for pigmentation disorders ( Figure 7).

(A) A cluster dendrogram of hierarchical clustering performed on the consensus space constructed by the MAXVAR-A model, which includes PAHs, metabolites, and bacterial OTUs selected by sCCA. Individuals are colored by city. Based on the height of the gap between the two consecutive levels of the dendrogram, we choose to construct 4 individual clusters. (B) The representation of hierarchical clustering on the MAXVAR-A consensus space, with colored convex hulls around the clusters. The two first parts of the consensus space represent 45.9% of the total variability. The table on the upper right of the figure represents the redistribution of individuals in each cluster in Baoding (more polluted) and Dalian (less polluted). The figures were created using R software: R Core Team (2017), R Foundation for Statistical Computing, Vienna, Austria. URL https://www.R-project.org/.

(A) The relative proportion of spread severity scores in the consensus cluster. 1 point corresponds to "non-spreading macula", 2 points to "weakly spreading macula", 3 points to "moderately spreading macula", and 4 points to "severely spreading macula". (B) Characterize the consensus cluster by using the Spread Macules severity score of v-test statistics. We represent the consensus cluster in the row and the location of the diffuse spots on the face (cheek or forehead) in the column. A positive v-test value (green) indicates that the severity score accounts for a higher proportion of the cluster compared to the global population. Negative v-test values ​​(red) indicate that severity scores account for a higher proportion of the cluster compared to the global population. If the v-test value reaches threshold 2 (the black vertical dashed line in the figure), it is considered significant. The figures were created using R software: R Core Team (2017), R Foundation for Statistical Computing, Vienna, Austria. URL https://www.R-project.org/.

On the other hand, groups 2, 3, and 4 did not show diffuse macula as the main condition. In fact, these clusters exhibit mixed clinical characteristics (data not shown), so it is impossible to classify them into precise clinical parameter categories. Figure 8 shows the v-test values ​​of variables that are significantly regulated (underexpressed or overexpressed) in cluster 1, consisting of 30 metabolites, 14 bacteria, and 9 PAH and its derivatives. We observe that the metabolites described earlier in this manuscript overlap very well and are most significantly regulated between polluted and less polluted cities. These include metabolites of the tryptophan and histidine pathways (kynurate, cis-uronate, and 4-imidazole acetate), peptides (γ-glutamylglycine), and fatty acids (oleamide). The characteristics of cluster 1 are negatively correlated with the commensal bacterial taxa (Propionibacterium B1; Staphylococcus B4 and Corynebacterium B2 and B21), and are observed to be inversely related to the two genera (Neisseria B119) previously observed on aging skin. ; Rothia B39) 51 and PAH (Brevibacterium B3714 and Paracoccus B5252). This analysis allows us to propose a potential molecular ID or roadmap to study the association between long-term exposure to pollution and the appearance of pigment diseases. It is too early to speculate on the mechanisms and effects of the different entities (metabolites, microbiome and PAH) responsible for this clinical condition. These are some of the questions we have raised in follow-up studies through extensive metagenomics analysis and other clinical studies.

Use v-test statistics to focus on the characterization of consensus cluster 1 by PAH, metabolites, and bacterial OTU. A positive v-test value (green) indicates that the average value of the variable in the cluster is higher than the global population. Negative v-test value (red indicates that the average value of the variable in the cluster is lower than the global population. It only indicates that the variable has a significant v-test value. The y-axis text is colored by variable type (blue represents metabolites, green represents Polycyclic aromatic hydrocarbons, yellow represents bacterial OTU). The data was created using R software: R Core Team (2017), R Foundation for Statistical Computing, Vienna, Austria. URL https://www.R-project .org/.

In conclusion, our research provides for the first time an in-depth understanding of potential molecular perturbations and skin phenotypic changes caused by long-term exposure to PAH. It also provides a roadmap for biological and clinical measurements and computational tools that can be applied to other similar studies, including larger Queue.

This multi-parameter clinical study was conducted by the Sino-German Cosmetics Research Institute (Beijing) in accordance with the recommendations of the Declaration of Helsinki and was approved by the local ethics committee (Study No. 2015-033-DY-024, CAIQ Cosmetics Technology Center, Beijing, China, 2015 August 19). Prior to any research-related procedures, informed written consent was obtained from all participants. The details of subject recruitment are described extensively elsewhere in 13, 14, and 28. In short, the research volunteers live in two Chinese cities because they are similar in terms of UV exposure, geographic location, and population lifestyle, but different pollution levels, which are measured by the monitoring station as an air quality index (Supplementary Figure 9). The selection is based on data collected by China's National Air Reporting System over the past few months. These data show different air pollution conditions in the two cities, but at comparable latitudes and altitudes, and with similar climatic conditions53. 67 women were recruited in Dalian, a less polluted city, and 67 women were recruited in Baoding, a more polluted city. The age of the subjects is restricted to 25-45 years old to avoid the main influence of hormonal changes during adolescence or menopause. Exclusion criteria include pregnancy, use of drugs (such as antibiotics or antifungals), skin diseases, and current smoking. All participants provided information about their health, medical history and daily habits. Each subject filled out a self-management questionnaire about skin care habits and outdoor time. Suppl_Table 4 provides information about personal demographics and lifestyle. Participants also received a clinical evaluation by a dermatologist, including a score for skin symptoms and phenotype. For microbiome analysis, cheek skin sampling was performed in a climate-controlled room at 22 °C and 60% humidity as described earlier. Use a dry swab with a sterile cotton swab to rub vigorously on the cheek for 60 seconds to cover a surface area of ​​2 cm2. The cotton swab is then placed in a coded microcentrifuge tube, immediately snap-frozen in liquid nitrogen, and stored at -80°C before further analysis. For metabolomics and proteomics, collect D-Squame samples from the cheek area according to the manufacturer's instructions. D-SQUAME is the registered name of the round strip used for stratum corneum sampling, produced by CuDerm Corporation (Dallas, TX, USA). For PAH analysis, as described in 13, 28, stainless steel scissors were used to cut hair samples from the occipital region of each subject. Each sample contains only the first 12 cm of hair from the scalp, corresponding to hair growth approximately one year before sampling. The hair sample was transferred to aluminum foil and stored at room temperature until analysis.

Non-targeted metabolomics analysis was performed at Metabolon, Inc (Morrisville, NC). The sample preparation and analysis methods described in this section were performed at Metabolon, Inc (Morrisville, NC) and have been previously described elsewhere54. All samples are stored at -80 °C until processing. Sample preparation: Prepare samples using Hamilton Company's automated MicroLab STAR system. For QC purposes, several recovery criteria were added before the first step of the extraction process. In order to remove the protein, dissociate the small molecules bound to the protein or trapped in the precipitated protein matrix, and recover the chemically diverse metabolites, precipitate the protein with methanol under vigorous shaking (Glen Mills GenoGrinder 2000) for 2 minutes (Glen Mills GenoGrinder) 2000), and then centrifuged. The resulting extract is divided into five parts: two parts for analysis by two separate reversed-phase (RP)/UPLC-MS/MS methods using positive ion mode electrospray ionization (ESI), and one for analysis by RP /UPLC-MS/MS uses negative ion mode to analyze ion mode ESI, one is used for HILIC/UPLC-MS/MS and negative ion mode ESI analysis, and the other sample is reserved for future use. All methods use Waters ACQUITY (UPLC) and Thermo Scientific Q-Exactive high-resolution/accurate mass spectrometers, connected to a heated electrospray ionization (HESI-II) source and Orbitrap mass analyzer, with a mass resolution of 35,000. MS analysis uses dynamic exclusion to alternate between MS and data-dependent MS scans. The scanning range of different methods is slightly different, but covers 70–1000 m/z. Compounds are identified by comparison with library entries based on retention time, molecular weight, preferred adducts and in-source fragments, and related MS spectra of purified standards or recurring unknown entities, and quality is visually inspected using software developed by Metabolon Control. Proteomics analysis by targeting MRM/SRM: Protein extraction, digestion and mass spectrometry analysis are performed by the proteomics platform of the CHU de Quebec Research Center in Quebec, Quebec, Canada. The sample (one D-Squame per subject) was resuspended in 550ul extraction buffer (50 mM ammonium bicarbonate (ABC), 0.5% sodium deoxycholate, 50 mM DTT, protease inhibitor cocktail and 1uM pepsin inhibitor) Vegetarian) in. After continuous filtration and centrifugation, the precipitate was dissolved in 50 µL of 50 mM ABC/1% DOC solution for protein quantification by Bradford protein assay.

Trypsin digestion: heat about 10ug of each sample at 95°C for 5 minutes, reduce protein with 0.2 mM DTT at 37°C for 30 minutes, and use 0.8 mM IAA (iodoacetamide) in the dark at room temperature Alkylation for 30 minutes. Then digest the protein with trypsin (1 ug) and incubate overnight at 37 °C.

SRM optimization: Three peptides/proteins of interest are selected through the three strongest transitions with the best collision energy. The selected peptides for SRM analysis were synthesized by ThermoFisher Scientific (Germany) and have heavy isotopes on C-ter amino acids ([13C6]-Lys and [13C6]-Arg). Just before the injection, 2.8 µL of the sample was mixed with 2.8 µL of the relabeled synthetic peptide standard solution.

Analyze 1 µg of each sample on the Eksigent NanoLC 400 chromatography system (Sciex, Concord, Ontario), which is coupled online with a 6500QTRAPTM (ABSciex, Concord, Ontario) mass spectrometer with a nanospray ion source. The peptide was eluted with a linear gradient of 5% to 40% solvent B (A: 0.1% FA, B: acetonitrile, 0.1% FA) within 30 minutes, and then 40% was eluted at 300 nL/min within 10 minutes To 95% B. The samples are randomly injected. Import the original file into Skyline v3.6 software for peak integration.

The results of Skyline are processed with excel. Simply put, two stronger conversions of two peptides per protein are used for quantification. The peak area was normalized based on the heavily labeled synthetic peptide. Calculate the normalization factor and apply it to the endogenous peptide. For each group, calculate the average of the sum of the normalized areas for each condition, and then apply the ratio and t test.

The details of the methods used to analyze hair PAH and metabolites have been previously reported13, here we only briefly describe the main steps. First, the hair sample is washed and purified to remove any compounds on the surface of the hair, without removing those compounds that are incorporated into the bulk matrix through biological mechanisms and representative doses present in the body. Then, according to the mature method based on gas and liquid chromatography combined with tandem mass spectrometry (GC-MS/MS and LC-MS/MS), the hair samples were crushed, hydrolyzed, extracted and analyzed. PAH, PAH metabolites, and nicotine/cotinine were analyzed separately and quantified for epiisotope-labeled analogs. This method allows the analysis of 15 parent PAHs (based on the US-EPA priority list), 56 PAH monohydroxy metabolites (all commercially available standards), nicotine and cotinine. The detection limit was also evaluated.

The detailed method of skin microbiome analysis has been reported before. In short, the PowerSoil DNA Isolation Kit (MO BIO Laboratories, Carlsbad, CA, USA) was used to extract gDNA from skin samples according to the manufacturer’s instructions and modifications as described above, and each gDNA sample was passed against bacteria 16S The primers of the rRNA gene were subjected to three repetitions of PCR in the V1-3 region and the fungal ITS1 region. After amplicon purification and index-PCR, the library was prepared and paired-end sequencing of bacteria and fungi was performed on the Illumina Miseq platform of SeqMatic LLC (Fremont, California, USA). Use USEARCH (v9.2.64) and QIIME (v1.9) for processing and bioinformatics analysis of 16S rRNA gene and ITS sequences. After clustering with 97% sequence identity using the USARE algorithm in USEARCH, bacterial OTU was obtained for representative sequences of SILVA database (version 128) and fungus OUT, and against the fungal database previously designed for skin microbiome analysis Inquiry 55 was made. After quality control and removal of unwanted readings, 9,656,916 and 14,649,172 bacterial and fungal readings were retained, respectively. As previously reported, use the separation (v4.0) package in R v3.5.1 to estimate the Shannon diversity (or alpha diversity) within the sample. The sCCA method was then used to determine the correlation between metabolites and bacterial and fungal taxa.

This article is divided into several parts, each part has a specific statistical method, as described below.

Association of clinical symptoms with city and PAH: Descriptive statistics of clinical symptoms have been tabulated by city and PAH group. The Wilcoxon test was used to compare clinical scores between cities or PHA groups. The PAH score is the best univariate summary of all 39 PAH measurements. It is the first principal component of principal component analysis (PCA) based on log-normalized PAH measurements.

The study is observational, the p-value must be interpreted in an exploratory rather than confirmatory way, and the association may be related to confounding factors.

Metabolites and cities: A pathway enrichment analysis was performed. A heat map using -log10 q value (using the Benjamini-Hochberg multiple test adjustment method to correct the p value) and a box plot of the city normalized by the selected logarithm are displayed in individual skin samples from polluted and less polluted cities Differentially regulated metabolite pathway descriptors.

In order to evaluate the differences between the metabonomics characteristics of the two cities, whether classified as sub-paths or super-paths, the cumulative distribution of Log2-Fold changes was calculated using Baoding, the most polluted city, as the reference population (Supplementary Figure 3). After removing outliers whose values ​​were greater than or equal to 3 times the interquartile range, the fold change was calculated. The t-test is used to test the significance of these multiple changes, and the results are represented by a volcano graph, which represents log2(FC) on the x-axis and −log10(q value) on the y-axis. This bivariate representation allows the most relevant metabolites to be highlighted, namely those with a q value <0.05 and |log2(FC)|> 0.6.

Complementary multivariate methods using random forest classification models were also used to confirm the robustness of the previous results (Supplementary Figure 4A). Random forest allows to model city membership as a non-linear multivariate function of metabolites and extract the importance of variables, that is, the metabolites that contribute the most to the model's prediction accuracy. We use the rfsrc function in the Random Forest SRC R package to execute the random forest classification model. More detailed information about the model is given in Suppl_Fig. 4B. We also evaluated the model's predictive performance on Out-Of-Bag (OOB) samples, and the accuracy results are given in Suppl_Fig. 4C. In order to calculate the value of variable importance (VIMP), we use a method of prediction error, which involves "noising" each variable in turn. The VIMP of the variable Xj is the difference between the prediction error when Xj is disturbed by randomly arranging its values ​​and the prediction error under the observed value 21. Since VIMP is the difference between the OOB prediction errors before and after permutation, a larger VIMP value indicates that incorrect assignment will reduce the prediction accuracy of the variables in the forest. VIMP close to zero indicates that the variable does not contribute to the prediction accuracy, and a negative value indicates that the prediction accuracy improves when the variable is specified incorrectly. For clarity, we only represent the 20 metabolites with the highest VIMP values ​​in Suppl_Fig. 4A.

Metabolites and polycyclic aromatic hydrocarbons: In order to directly compare exposure and its molecular effects on the skin, we performed sparse canonical correlation analysis (sCCA) to determine the set of common changes of polycyclic aromatic hydrocarbons and metabolites. sCCA is a regularized version of Canonical Correlation Analysis (CCA), which is used to study the relationship between two data sets while selecting only significant correlations17. If we have two random variable data sets X and Y, and there is a correlation between the variables, then sCCA will find a sparse linear combination of X and Y (the component with a weight of zero for uncorrelated variables), and the correlation between them The largest. The sparse parameters of sCCA are estimated using the permutation scheme (nperms = 500) and the MultiCCA.permute function from the PMA R package,]18. The permutation process gives an ap value of 0.004, and the correlation coefficient between the first sparse components of the two data sets is 0.539. The first sparse component has 20 and 7 variables, respectively, with non-empty weights of metabolomics and PAH data sets. The cross-correlation between selected metabolites and PAH is visualized using a heat map representation.

Metabolites and skin microbiome: Use the Pearson correlation coefficient to estimate the correlation between the Shannon index of alpha bacterial or fungal diversity and metabolites, and rank them according to significance.

Use metabolites as independent variables to predict the Shannon index of bacterial or fungal diversity, using SIMCA software version 16.0, Umetrics, Umea, Sweden using partial least squares regression. Then sCCA was performed to explore the global correlation between bacterial and fungal OTU and the relative abundance of metabolites (Figure 5A, B). Use the permutation scheme of the MultiCCA.permute function from the PMA R package (nperms = 500) to estimate the sparsity parameters of sCCA. The sCCA replacement program between bacterial OTU and metabolites gives an ap value <0.001, and the correlation coefficient between the first sparse component of the two data sets is 0.618. For the metabolomics and bacterial OTU datasets, the first sparse component has 16 and 7 variables with non-empty weights, respectively. The replacement program of sCCA between fungal OTU and metabolite gave an ap value of 0.206, and the correlation coefficient between the first sparse component of the two data sets was 0.436. For the metabolomics and fungal OTU datasets, the first sparse component has 25 and 5 variables with non-empty weights, respectively. The cross-correlation between selected metabolites and OTU is visualized using a heat map representation.

Multi-block analysis: It is necessary to jointly analyze different data sets to discover their main sources of covariance, and it is necessary to extend the sCCA method to more than two groups. In this case, we use a sparse version of the regularized generalized canonical correlation analysis [ref] called Sparse Generalized Canonical Correlation Analysis (SGCCA) [ref]. S/RGCCA is a general component-based framework for integrated exploration of multi-modal and high-dimensional data sets. S/RGCCA classifies many multi-block component methods as special cases, including MAXVAR-A (see Tenenhaus et al. 49 for details). MAXVAR-A allows to visualize the relationship between variables belonging to different blocks in a single space (the so-called consensus space). Therefore, we used MAXVAR-A to combine the metabolomics, PAH, and bacterial OTU data sets. Based on the first two components of the consensus space, we conducted hierarchical clustering and divided the population into 4 clusters (see Figure 6A and B). The selection of the 4 clusters is based on the visual evaluation of the tree diagram in Figure 6A, more specifically, by looking at the height of the gap between 2 consecutive levels of the hierarchy. Then use v-test20 (a program available from the catdes function of the Factominer R package) to characterize each cluster by clinical symptoms, metabolites, PAH, and bacterial OTU. v The test value allows to determine whether the variable is significantly overrepresented or underrepresented in the subgroup compared to the total population. For continuous variables, we test whether the mean of a particular subgroup is different from the mean of the population. For discrete/qualitative variables, we test whether the proportion of a certain modality in a particular subgroup is overexpressed or underexpressed in the subgroup compared with the entire population. The results of v-test are shown in Figure 7B, which is used to spread the clinical symptoms of macula, and Figure 8 shows metabolites, PAH, and bacterial OTU, which are significantly regulated in cluster 1.

Gonzales, MC, Yu, P. & Shiao, SP MTHFR gene polymorphism mutations and air pollution as risk factors for breast cancer: a meta-predictive study. Nurse. Reservoir 66(2), 152–163 (2017).

PubMed PubMed Central article Google Scholar 

Heroux, ME etc. Quantifying the health effects of ambient air pollutants: recommendations from the WHO/Europe project. internationality. J. Public Health 60(5), 619–627 (2015).

Masiol, M., Agostinelli, C., Formenton, G., Tarabotti, E. & Pavoni, B. Urban air pollution is monitored hourly for 13 years: potential sources, trends, cycles, and effects of car-free days. science. Total environment. 494-495, 84-96 (2014).

ADS PubMed article CAS Google Scholar 

Araviiskaia, E. etc. The effect of air pollution on the skin. J. Euro. Akkad. skin disease. Venereal disease. 33(8), 1496–1505 (2019).

CAS PubMed PubMed Central Article Google Scholar 

Schikowski, T. & Huls, A. Air pollution and skin aging. Curry environment. Health Representative 7(1), 58–64 (2020).

PubMed article CAS Google Scholar 

Rajagopalan, P. et al. How does long-term exposure to cigarette smoke affect human skin? Global proteomics study of primary human keratinocytes. OMICS 20(11), 615–626 (2016).

CAS PubMed article Google Scholar 

Vattanasit, U. etc. Oxidative DNA damage and inflammation in cultured human cells and humans exposed to traffic-related particles. internationality. J. Hagrid. environment. Health 217(1), 23–33 (2014).

CAS PubMed article Google Scholar 

Vierkotter, A. etc. Exposure of particles in the air and external skin aging. J. Investment. skin disease. 130(12), 2719–2726 (2010).

PubMed article CAS Google Scholar 

Krutmann, J. Premature skin aging caused by ultraviolet radiation and other environmental hazards. Molecular basis. Hautarzt 54(9), 809–817 (2003).

CAS PubMed article Google Scholar 

Marrot, L. Pollution and sun exposure: harmful synergies. Mechanisms and opportunities for skin protection. Curry medicine. Chemistry 25(40), 5469–5486 (2018).

CAS PubMed article Google Scholar 

Lefebvre, MA etc. The impact of urban pollution on skin conditions. A comparative study in Shanghai. internationality. J. Cosmetics. science. 38(3), 217–223 (2016).

Lefebvre, MA etc. Assessing the impact of urban pollution on skin quality: a multicenter study in Mexico. internationality. J. Cosmetics. science. 37(3), 329–338 (2015).

CAS PubMed article Google Scholar 

Palazzi, P. etc. Hair analysis revealed the exposure of women living in Baoding and Dalian, China to polycyclic aromatic hydrocarbons. environment. internationality. 121(Pt 2), 1341–1354 (2018).

CAS PubMed article Google Scholar 

Liang, MHY, etc. Changes in human skin microbiota after long-term exposure to polycyclic aromatic hydrocarbon pollutants. Microbiome 8(1), 100 (2020).

PubMed PubMed Central article Google Scholar 

KrupaShankar, DS, etc. A cross-sectional, multi-center clinical epidemiological study of melasma in India. skin disease. that. (Heidelb.) 4(1), 71–81 (2014).

Hourblin, V., Nouveau, S., Roy, ​​N. & de Lacharriere, O. The skin complexion and pigmentation of the facial skin of 1204 women in 4 cities in India. J. Dermatol, India. Venereal disease. leprosy. 80(5), 395–401 (2014).

Whitfield, JB Gamma Glutamyltransferase. Crit. Clinical chaplain. laboratory. science. 38(4), 263–355 (2001).

CAS PubMed article Google Scholar 

Rawlings, AV & Harding, CR moisturizing and skin barrier function. skin disease. that. 17(Suppl 1), 43–48 (2004).

Rajagopalan, P. et al. Changes in the proteome of primary skin keratinocytes exposed to diesel particulate extract-the role of antioxidants in skin health. J. Skin diseases. science. 91(3), 239–249 (2018).

CAS PubMed article Google Scholar 

Biro, T., Toth, BI, Hasko, G., Paus, R. & Pacher, P. The endocannabinoid system of the skin in health and disease: new perspectives and therapeutic opportunities. Trends in pharmacy. science. 30(8), 411–420 (2009).

CAS PubMed PubMed Central Article Google Scholar 

Kim, KE, Cho, D. & Park, HJ Air pollution and skin diseases: The adverse effects of airborne particles on various skin diseases. life sciences. 152, 126–134 (2016).

CAS PubMed article Google Scholar 

Mancebo, SE & Wang, SQ recognize the impact of ambient air pollution on skin health. J. Euro. Akkad. skin disease. Venereal disease. 29(12), 2326–2332 (2015).

CAS PubMed PubMed Central Article Google Scholar 

Stone, TW, Stoy, N. & Darlington, LG The target range of tryptophan kynurenine metabolites continues to expand. Trends in pharmacy. science. 34(2), 136–143 (2013).

CAS PubMed article Google Scholar 

Hubbard, TD, etc. Adaptation of human aromatic hydrocarbon receptors to microbiota-derived indole. science. Representative 5, 12689 (2015).

ADS CAS PubMed PubMed Central Article Google Scholar 

Tenenhaus, A., & Tenenhaus, M. Regularized generalized canonical correlation analysis. Psychometrika 76(2), 257 (2011).

MathSciNet MATH article Google Scholar 

Appenzeller, BMR etc. Hair analysis for biological monitoring of pesticide exposure-comparison with blood and urine in a rat model. arch. poison. 91(8), 2813–2825 (2017).

CAS PubMed article Google Scholar 

Grova, N., Hardy, EM, Fays, F., Duca, RC & Appenzeller, BMR Hair analysis for biological monitoring of PAH exposure: comparison with urine metabolites and DNA adducts in a rat model. arch. poison. (2018) (in press).

Naudin, G. etc. Human pollution exposure is related to accelerated degradation of the ultrastructure of hair fibers. The process of the national team Akkad. science. United States 116(37), 18410–18415 (2019).

CAS PubMed PubMed Central Article Google Scholar 

Aksnes, H., Ree, R. and Arnesen, T. Co-translational, post-translational and non-catalytic effects of N-terminal acetyltransferase. Mole. Cell 73(6), 1097-1114 (2019).

CAS PubMed PubMed Central Article Google Scholar 

Leclercq, B. etc. Genetic and epigenetic changes in the bronchial epithelial cells of normal and sensitive COPD patients who have been repeatedly exposed to PM2.5, a source of air pollution. environment. Pollution. 230, 163–177 (2017).

CAS PubMed article PubMed Central Google Scholar 

LeMasters, GK, etc. N-acetyltransferase 1 polymorphism increases cotinine levels in Caucasian children exposed to secondhand smoke: CCAAPS birth cohort. Pharmacogenomics. J. 15(2), 189–195 (2015).

Hagdorn, HW et al. The urinary excretion of phenolic polycyclic aromatic hydrocarbons (OH-PAH) in non-smokers and smokers has different ISO tar production. J. Anus. poison. 33(6), 301–309 (2009).

CAS PubMed article PubMed Central Google Scholar 

Smith, sir, etc. Benzo[a]pyrene disrupts mitochondria and amino acid metabolism in lung epithelial cells, and has a similar correlation with the metabolic changes in human serum. J. Occupation. environment. medicine. 61 (Supplement 12), S73–S81 (2019).

CAS PubMed PubMed Central Article Google Scholar 

Zierer, J. et al. The fecal metabolome serves as a functional reading of the gut microbiome. Nat. Gene. 50(6), 790–795 (2018).

CAS PubMed PubMed Central Article Google Scholar 

Bouslimani, A. etc. The effect of skin care products on skin chemistry and microbiome dynamics. BMC Biology. 17(1), 47 (2019).

PubMed PubMed Central article Google Scholar 

McCall, Lee et al. Family chemistry and microbial transformation in the process of urbanization. Nat. microorganism. 5(1), 108–115 (2020).

CAS PubMed article PubMed Central Google Scholar 

Bouslimani, A. etc. 3D molecular mapping of the surface of human skin. The process of the national team Akkad. science. United States 112(17), E2120-2129 (2015).

CAS PubMed PubMed Central Article Google Scholar 

Xie, W., Khosasih, V., Suwanto, A. & Kim, HK Characterization of Staphylococcus aureus and Staphylococcus epidermidis lipase isolated from human facial sebaceous skin. J. Microorganisms. Biotechnology. 22(1), 84–91 (2012).

CAS PubMed article Google Scholar 

Meadows, JA and Wargo, MJ The role of carnitine in bacterial physiology and metabolism. Microbiology (reading) 161(6), 1161–1174 (2015).

Lloyd-Price, J. etc. Multi-omics of the gut microbial ecosystem in inflammatory bowel disease. Nature 569 (7758), 655–662 (2019).

ADS CAS PubMed PubMed Central Article Google Scholar 

Thurmond, RL, Gelfand, EW & Dunford, PJ The role of histamine H1 and H4 receptors in allergic inflammation: the search for new antihistamines. Nat. Priest drug discovery. 7(1), 41–53 (2008).

CAS PubMed article Google Scholar 

Barcik, W. etc. The microbes that secrete histamine in the intestines of adults with asthma are increased. J. Allergy clinical. Immunology. 138(5), 1491–1494 (2016).

Sudo, N. Biogenic amines: a signal between symbiotic microbiota and gut physiology. front. endocrine. (Lausanne) 10, 504 (2019).

Allaker, RP, Greenman, J. & Osborne, RH Propionibacterium acnes and other skin organisms produce inflammatory compounds. Brother J. Dermatopathy. 117(2), 175–183 (1987).

CAS PubMed article Google Scholar 

Zhang, X., Zhang, S., Shi, Y., Shen, F. & Wang, H. Comparative analysis of a new high-producing phenyllactic acid Lactobacillus plantarum IMAU10124 and lactate dehydrogenase gene. FEMS microorganisms. Wright. 356(1), 89–96 (2014).

ADS CAS PubMed article Google Scholar 

Menegatti, C. et al. The Bacillus polymyxa related to the stingless bee Melipona scutellaris produces antibacterial compounds against insect pathogens. J. Chemistry. Ecology. 44(12), 1158–1169 (2018).

CAS PubMed article Google Scholar 

Yu J. et al. The tryptophan metabolites of the skin microbiota reduce inflammation in patients with atopic dermatitis through aromatic hydrocarbon receptors. J Allergy clinical immunology. 143(6), 2108–2119.e12. https://doi.org/10.1016/j.jaci.2018.11.036 (2019).

CAS Article PubMed Google Scholar 

Nenoff, P. & Haustein, UF In vitro activity of phytosphingosine against Malassezia furfur and Candida albicans. Journal of Dermatology. Venereal disease. 82(3), 170–173 (2002).

CAS PubMed article Google Scholar 

Tenenhaus A. et al. Variable selection of generalized canonical correlation analysis. Biostatistics 15(3), 569–583 (2014).

PubMed MATH article Google Scholar 

Garali, I. etc. Multimodal data integration strategy: the application of biomarker identification in spinocerebellar ataxia. briefly. Bioinformatics. 19(6), 1356–1369 (2018).

Shibagaki, N. etc. Aging-related changes in the diversity of female skin microbiota associated with oral bacteria. science. Representative 7(1), 10567 (2017).

ADS PubMed PubMed Central Article CAS Google Scholar 

Sowada, J. et al. The toxicity of symbiotic bacteria from human skin to polycyclic aromatic hydrocarbons. arch. poison. 91(6), 2331–2341 (2017).

CAS PubMed PubMed Central Article Google Scholar 

Rohde, RA & Muller, RA Air Pollution in China: Concentration and Source Map. PLoS ONE 10(8), e0135749 (2015).

PubMed PubMed Central article CAS Google Scholar 

Linde, L. et al. The characteristics of plasma metabolomics are different from liver fat, visceral fat tissue and pancreatic fat. J. Clinical. endocrine. Yuan. 106(1), e118–e129 (2021).

Findley, K. etc. The topographical diversity of fungal and bacterial communities in human skin. Nature 498 (7454), 367–370 (2013).

ADS CAS PubMed PubMed Central Article Google Scholar 

Thanks to Eric Arbey of L'Oréal Research and Innovation in France and Brian Keppler of Metabolon Inc in the United States for additional analysis to determine the unknown metabolites mentioned in this study.

Research and Innovation, L'Oréal SA, Aulnay Sous Bois, France

Namita Misra, Cécile Clavaud, Florent Guinot, Nasrine Bourokba, Stephanie Nouveau, Sakina Mezzache, Philippe Bastien, Luc Aguilar and Nükhet Cavusoglu

Human Biomonitoring Research Group, Luxembourg Institute of Health, Strassen, Luxembourg

Paul Palazzi and Brice MR Appenzeller

CentraleSupelec Laboratoire des Signaux et Systemes, Université Paris-Saclay, CNRS, Gif-sur-Yvette, France

Institute of Brain and Spine, Paris, France

School of Energy and Environment, City University of Hong Kong, Hong Kong Special Administrative Region, Kowloon, China

Marcus HY Leung & Patrick KH Lee

You can also search for this author in PubMed Google Scholar

You can also search for this author in PubMed Google Scholar

You can also search for this author in PubMed Google Scholar

You can also search for this author in PubMed Google Scholar

You can also search for this author in PubMed Google Scholar

You can also search for this author in PubMed Google Scholar

You can also search for this author in PubMed Google Scholar

You can also search for this author in PubMed Google Scholar

You can also search for this author in PubMed Google Scholar

You can also search for this author in PubMed Google Scholar

You can also search for this author in PubMed Google Scholar

You can also search for this author in PubMed Google Scholar

You can also search for this author in PubMed Google Scholar

You can also search for this author in PubMed Google Scholar

NM, NC ​​ and LA conceived and designed this study. NB and SN designed clinical protocols and conducted research. PP, BMRA and SM conducted experiments to measure the content of PAH in hair fibers. NM, CC conducted metabolomics and proteomics analysis. MHYL, PKHL and CC conducted microbiome and metagenomics analysis. AT, PB, and FG developed calculation strategies for this study and performed statistical analysis on the data. NM, NC, CC, FG, and PB wrote the manuscript, and all authors reviewed the manuscript.

The author declares no competing interests.

Springer Nature remains neutral on the jurisdiction claims in the published maps and agency affiliates.

Open Access This article has been licensed under the Creative Commons Attribution 4.0 International License Agreement, which permits use, sharing, adaptation, distribution and reproduction in any media or format, as long as you appropriately indicate the original author and source, and provide a link to the Creative Commons license And indicate whether any changes have been made. The images or other third-party materials in this article are included in the article’s Creative Commons license, unless otherwise stated in the material’s credit line. If the article’s Creative Commons license does not include the material, and your intended use is not permitted by laws and regulations or exceeds the permitted use, you need to obtain permission directly from the copyright owner. To view a copy of this license, please visit http://creativecommons.org/licenses/by/4.0/.

Misra, N., Clavaud, C., Guinot, F. etc. Multi-omics analysis deciphers the molecular link between long-term exposure to pollution and human skin dysfunction. Scientific Representative 11, 18302 (2021). https://doi.org/10.1038/s41598-021-97572-1

DOI: https://doi.org/10.1038/s41598-021-97572-1

Anyone you share the following link with can read this content:

Sorry, there is currently no shareable link in this article.

Provided by Springer Nature SharedIt content sharing program

By submitting a comment, you agree to abide by our terms and community guidelines. If you find content that is abusive or does not comply with our terms or guidelines, please mark it as inappropriate.

Scientific Report (Sci Rep) ISSN 2045-2322 (online)