Transcriptome Analysis of Particulate Matter 2.5-Induced Abnormal Effects on Human Sebocytes

Particulate matter 2.5 (PM2.5), an atmospheric pollutant with an aerodynamic diameter of <2.5 μm, can cause serious human health problems, including skin damage. Since sebocytes are involved in the regulation of skin homeostasis, it is necessary to study the effects of PM2.5 on sebocytes. We examined the role of PM2.5 via the identification of differentially expressed genes, functional enrichment and canonical pathway analysis, upstream regulator analysis, and disease and biological function analysis through mRNA sequencing. Xenobiotic and lipid metabolism, inflammation, oxidative stress, and cell barrier damage-related pathways were enriched; additionally, PM2.5 altered steroid hormone biosynthesis and retinol metabolism-related pathways. Consequently, PM2.5 increased lipid synthesis, lipid peroxidation, inflammatory cytokine expression, and oxidative stress and altered the lipid composition and expression of factors that affect cell barriers. Furthermore, PM2.5 altered the activity of sterol regulatory element binding proteins, mitogen-activated protein kinases, transforming growth factor beta-SMAD, and forkhead box O3-mediated pathways. We also suggest that the alterations in retinol and estrogen metabolism by PM2.5 are related to the damage. These results were validated using the HairSkin® model. Thus, our results provide evidence of the harmful effects of PM2.5 on sebocytes as well as new targets for alleviating the skin damage it causes.


Introduction
Atmospheric pollutants from global industrialization cause serious human health problems. In 2005, the World Health Organization (WHO) published air quality guidelines regarding the major health-damaging air pollutants, including particulate matter (PM), ozone, nitrogen dioxide, and sulfur dioxide. In 2019, more than 90% of the world's population lived in areas with concentrations of particulate matter 2.5 (PM 2.5 ) exceeding the air quality guideline (10 µg/mL) [1].
Particulate matter (PM) is classified according to its aerodynamic diameter; PM 2.5 has an aerodynamic diameter of <2.5 µm. PM 2.5 contains various chemicals, including metal elements (arsenic, cadmium, lead, etc.), ionic components (ammonium, nitrate, sulfite ions, etc.), and carbonaceous fractions (organic carbon, elemental carbon, polycyclic aromatic hydrocarbons [PAHs], etc.) [2,3]. The small size of PM 2.5 allows them to enter bronchioles, alveoli, and blood, through which they can further affect other organs [4,5]. PM can penetrate the barrier-disrupted epidermis by tape-stripping and can enter hair follicles regardless of barrier damage [6,7].
The human sebaceous gland, a skin appendage, together with the associated hair follicle and arrector pili muscle, collectively form the pilosebaceous unit [8,9]. The main function of the sebaceous gland in humans is the production and holocrine secretion of

Differentially Expressed Genes (DEGs) and Gene Ontology (GO)-Based Gene-Set Enrichment Analysis of SZ95 Sebocytes after PM 2.5 Treatment
Based on the results of the cell viability test ( Figure S1), we determined the optimal treatment concentration of PM 2.5 . From RNA-sequencing (RNA-seq) data, we identified that the expression of 381 genes (−2 > fold changes > 2; p-value < 0.05) was significantly changed in PM 2.5 -treated experimental groups compared with the control SZ95 sebocytes. Treatment with PM 2.5 resulted in the upregulation of 188 genes and the downregulation of 193 genes.
The top 10 enriched GO terms for biological processes and molecular functions are shown in Figure S2. The most highly enriched molecular functions were related to binding, including protein binding, ion binding, cation binding, and metal ion binding. In addition, oxidoreductase activity was significantly enriched. Oxidoreductases such as aldo-keto reductase (AKR) and cytochrome P450 (CYP450) are related to the metabolism of chemicals, including steroids [20]. Thus, the GO enrichment analysis indicated that PM 2.5 -induced functional changes are caused by various components of heterogenic PM 2.5 .

Kyoto Encyclopedia of Genes and Genomes (KEGG) Terms and Canonical Pathway Prediction Analysis
In total, 55 KEGG pathways were enriched (false discovery rate [FDR] < 0.05), with the top 20 KEGG terms shown in Figure 1a. In addition, 18 canonical pathways were identified (−2 > z-score > 2) by Ingenuity Pathway Analysis (IPA), with the top 10 listed in Table 1. The highly predicted KEGG pathways were metabolic pathways related to steroid hormone biosynthesis, retinol metabolism, pentose and glucuronate interconversions, and the metabolism of xenobiotics by CYP450 (Figure 1a). Xenobiotics produced by CYP450 are known to be involved in the toxicity of PM 2.5 [16,21]. Steroid hormone biosynthesis and retinol metabolism regulate the functions of sebaceous glands, such as lipid synthesis [8,22]. Inflammatory response-related pathways, such as cytokine-cytokine receptor interaction and the IL-17 signaling pathway, were enriched. In addition, cell adhesionrelated pathways, including tight junctions and cell adhesion molecules, were significantly enriched, implying that PM 2.5 affects barrier function. The transforming growth factor beta (TGFβ) and PI3K-Akt signaling pathways were predicted to be enriched by PM 2.5 . Via canonical pathway analysis, cholesterol biosynthesis-related pathways, including the superpathway of cholesterol biosynthesis, three cholesterol biosynthesis (I, II, III), and  superpathway of geranylgeranyl diphosphate biosynthesis I, were significantly enriched  (Table 1). The xenobiotic metabolism aryl hydrocarbon receptor (AHR) signaling pathway and estrogen-dependent breast cancer signaling pathway related to steroid hormone metabolism were enriched, consistent with the KEGG analysis results. Moreover, endothelial nitric oxide synthase (eNOS) signaling, which is associated with oxidative stress, was also enriched. Thus, we identified that PM 2.5 mainly alters the levels of genes associated with xenobiotic metabolism, inflammatory responses, oxidative stress, lipid metabolism, steroid hormone biosynthesis, retinol metabolism, and cell adhesion. Among the DEGs (−2 > fold changes > 2, and p-value < 0.05), the genes related to the enriched pathways are summarized in Table S1. The expression of significantly altered genes was confirmed by real-time quantitative reverse-transcription polymerase chain reaction (RT-qPCR), Western blotting analysis, and enzyme-linked immunosorbent assay (ELISA) (Figures 1b-d and S3). synthesis [8,22]. Inflammatory response-related pathways, such as cytokine-cytokine receptor interaction and the IL-17 signaling pathway, were enriched. In addition, cell adhesion-related pathways, including tight junctions and cell adhesion molecules, were significantly enriched, implying that PM2.5 affects barrier function. The transforming growth factor beta (TGFβ) and PI3K-Akt signaling pathways were predicted to be enriched by PM2.5. Via canonical pathway analysis, cholesterol biosynthesis-related pathways, including the superpathway of cholesterol biosynthesis, three cholesterol biosynthesis (I, II, III), and superpathway of geranylgeranyl diphosphate biosynthesis I, were significantly enriched (Table 1). The xenobiotic metabolism aryl hydrocarbon receptor (AHR) signaling pathway and estrogen-dependent breast cancer signaling pathway related to steroid hormone metabolism were enriched, consistent with the KEGG analysis results. Moreover, endothelial nitric oxide synthase (eNOS) signaling, which is associated with oxidative stress, was also enriched. Thus, we identified that PM2.5 mainly alters the levels of genes associated with xenobiotic metabolism, inflammatory responses, oxidative stress, lipid metabolism, steroid hormone biosynthesis, retinol metabolism, and cell adhesion. Among the DEGs (−2 > fold changes > 2, and p-value < 0.05), the genes related to the enriched pathways are summarized in Table S1. The expression of significantly altered genes was confirmed by real-time quantitative reverse-transcription polymerase chain reaction (RT-qPCR), Western blotting analysis, and enzyme-linked immunosorbent assay (ELISA) (Figures 1b-d and S3).

Upstream Regulator Analysis and Disease and Biological Function Prediction by IPA
We analyzed upstream regulators by IPA; the top 10 predicted activated and inhibited upstream regulators in SZ95 sebocytes by PM 2.5 are listed in Table 2. As a result, sterol regulatory element binding transcription factor (SREBF, which encodes sterol regulatory element binding protein, SREBP) signaling pathway-associated factors were predicted as upstream regulators. SREBF1, SREBF2, and SREBP cleavage-activating protein (SCAP) were activated, and INSIG1, which inhibits SREBP-mediated signaling pathways, was inhibited. Additionally, mitogen-activated protein kinase (MAPK)-mediated signaling pathway-related factors MAPK7 and MAP2K5 were predicted to be activated by upstream regulators. SREBPs regulate lipid homeostasis and inflammation [23,24]. Moreover, MAPK7 and its upstream activator, MAP2K5, are also known to control lipid metabolism in small cell lung cancer cells [25]. In contrast, TGFβ signaling pathway-related factors, such as TGFβ1 and SMAD3, and forkhead box O3 (FoxO3) were all predicted to be inhibited by upstream regulators. TGFβ-SMAD and FoxO3 signaling were associated with the differentiation and lipid synthesis of sebocytes [26,27]. These results imply that PM 2.5 affects sebocyte differentiation, lipid synthesis, and inflammation by altering various signaling pathways. Furthermore, IPA predicted diseases and biological functions in PM 2.5treated SZ95 sebocytes (Table 3). Lipid metabolism and organismal survival were predicted to be PM 2.5 -associated biological functions, and dermatological diseases and cancer were predicted to be associated diseases. Table 2. Upstream regulators of DEGs. The top upstream regulators of DEGs (|fold changes| > 1.5, p-value < 0.01) in PM 2.5 -treated SZ95 sebocytes were analyzed with IPA. Activation z-score indicates the predicted activation (positive value) or inhibition (negative value) of the upstream regulator. The p-value of overlap indicates the statistical significance of the overlap of the DEGs and genes regulated by an upstream regulator. p-values were calculated by right-tailed Fisher's exact tests.   We validated the signaling pathways of the predicted upstream regulators. The processed mature forms of SREBP1 and SREBP2 were found to be higher in the nucleus of PM 2.5 -treated SZ95 sebocytes than those in the control group (Figures 2a and S4a). Moreover, PM 2.5 treatment increased the phosphorylation of MAP2K5 and MAPK7 (Figures 2b and S4b) and inhibited TGFβ-induced phosphorylation of SMAD3 compared with the control (Figures 2c and S4c). The expression level of FoxO3a was decreased by PM 2.5 (Figures 2d and S4d). These results indicate that alterations in SREBP-, MAPK-, TGFβ-SMAD-, and FoxO3a-signaling activity by PM 2.5 exposure may contribute to the induction of alterations in SZ95 sebocytes.

Effect of PM 2.5 on Lipid Production, ROS Generation, and Lipid-Peroxidation in SZ95 Sebocytes
According to the lipid staining results, lipid content was increased by PM 2.5 (Figure 3a). Furthermore, using liquid chromatography-tandem mass spectrometry (LC-MS/MS), we found that PM 2.5 induced lipid compositional changes, with increased levels of cholesterol and free fatty acids (FAs) and decreased levels of squalene and triglycerides compared with the control (Figure 3b). We also tested whether PM 2.5 increases ROS generation and lipid peroxidation. We found that PM 2.5 increased intracellular ROS generation and lipid peroxidation (Figure 3c,d). These results indicate that PM 2.5 can cause damage, including increased lipid synthesis, ROS generation, lipid peroxidation, and altered lipid composition in sebocytes.

Confirmation of PM 2.5 Effects on Human Skin Tissue
Immunohistochemical analysis was performed on human tissue (HairSkin ® ) to confirm the PM 2.5 -induced alterations. The expression levels of CYP1B1, SERBP1, IL1B, and interleukin 6 (IL6) were increased, and the expression of claudin1 was decreased by PM 2.5 ( Figure 4). Cytokeratin 7 and mucin 1 were used as markers for the sebaceous glands.

Effect of PM2.5 on Lipid Production, ROS Generation, and Lipid-Peroxidation in SZ95 Sebocytes
According to the lipid staining results, lipid content was increased by PM2.5 ( Figure  3a). Furthermore, using liquid chromatography-tandem mass spectrometry (LC-MS/MS), we found that PM2.5 induced lipid compositional changes, with increased levels of cholesterol and free fatty acids (FAs) and decreased levels of squalene and triglycerides compared with the control (Figure 3b). We also tested whether PM2.5 increases ROS generation and lipid peroxidation. We found that PM2.5 increased intracellular ROS generation and lipid peroxidation (Figure 3c,d). These results indicate that PM2.5 can cause damage, including increased lipid synthesis, ROS generation, lipid peroxidation, and altered lipid composition in sebocytes.

Confirmation of PM2.5 Effects on Human Skin Tissue
Immunohistochemical analysis was performed on human tissue (HairSkin ® ) to confirm the PM2.5-induced alterations. The expression levels of CYP1B1, SERBP1, IL1B, and interleukin 6 (IL6) were increased, and the expression of claudin1 was decreased by PM2.5 (Figure 4). Cytokeratin 7 and mucin 1 were used as markers for the sebaceous glands.

Discussion
In this study, we demonstrated the harmful effects of PM 2.5 on SZ95 sebocytes and human skin tissue through transcriptome analysis. PAHs, which are components of PM 2.5 , regulate the expression of xenobiotic-metabolizing enzymes (XMEs) through the AHR/AHR nuclear translocator (ARNT) [21]. XMEs are classified into phase I, adding functional groups such as an oxide and epoxide into the substrate, and phase II, conjugating with moieties such as glucuronide, acetate, sulfate, glutamine, thiocyanate, and glucoside to facilitate excretion [28]. In our transcriptome analysis, the expression of type I XMEs, such as CYP450 (CYP1A1 and CYP1B1), aldehyde dehydrogenase (ALDH1L1, ALDH3A1, and ALDH1A1), and prostaglandin-endoperoxide synthases (PTGS1), and type 2 XMEs, such as UDP glucuronosyltransferases (UGT1A6 and UGT1A3), glutathione S-transferases, mu 2 (GSTM2), and sulfotransferases (SULT1A2 and SULT1E1), were highly altered by PM 2.5 .
The expression of ALDH1A1, a well-known enzyme for all-trans-retinoic acid (ATRA), was downregulated by PM 2.5 . We also found that the expression of retinol dehydrogenase 10 (RDH10), an enzyme that oxidizes retinol into retinal aldehyde, and retinoic acid receptor beta (RARB), an ATRA receptor, was decreased by PM 2.5 (Figure 1b) [22,29]. As ATRA reduces lipid synthesis in the sebaceous gland, a low concentration of ATRA is important for the function of sebaceous glands that produce lipids. Additionally, we found that the expression of hydroxysteroid 17-beta dehydrogenase 1 (HSD17B1), which reduces estrone to 17β-estradiol (the most active form), was downregulated in PM 2.5 -treated SZ95 sebocytes (Figure 1b) [30]. Furthermore, CYP1A1 was capable of catalyzing the 2-hydroxylation of estradiol to produce less active estrogen [31]. Estrogen (17β-estradiol) inhibited lipogenesis in the sebaceous glands [8]. Thus, these results indicate the possibility that PM 2.5 causes the downregulation of ATRA-mediated signaling and the reduction of active estrogen; the altered retinol metabolism and steroid hormone metabolism subsequently affect sebocyte functions, such as lipid synthesis.
There are differences in the sebum of subjects in areas with and without severe pollution. Increased sebum excretion rates and lower concentrations of squalene were observed in highly polluted regions [32,33]. Liao et al. reported that PM 2.5 disrupts barrier functions by upregulating cholesterol synthesis in human keratinocytes and a three-dimensional epidermis tissue model [34]. Kwack et al. showed that PM 10 (100 µg/mL) increases the expression of inflammatory cytokines and ROS production in sebocytes and upregulates sebum production and inflammatory nodule production in Cutibacterium acnes-induced mice [35]. The expression of various genes related to lipid metabolism changed in our RNAseq analysis, and the activation of signaling pathways related to cholesterol biosynthesis was analyzed using IPA. As a result, PM 2.5 increased lipid synthesis and peroxidation. Moreover, it altered the lipid composition by decreasing squalene and triglyceride levels and increasing cholesterol and free FA levels. In addition, the expression of stearoyl-CoA desaturase (SCD), an FA desaturase enzyme, was elevated in PM 2.5 -treated SZ95 sebocytes (Figure 1b,c). Increased sebum production, alterations in lipid composition, increased sebum oxidation particularly for squalene, and the altered ratio between saturated and unsaturated FAs in skin diseases such as acne have been reported [36][37][38][39]. Triglycerides are catalyzed to free FAs by resident and microflora lipases, such as Propionibacterium acnes, leading to the activation of inflammasome [40]. Moreover, free FAs, such as palmitic acid, induce lipid synthesis and increase the secretion of IL6 and IL8 [41]. Therefore, alterations in lipid metabolism and the increase in lipid peroxidation by PM 2.5 may result in the dysfunction of sebaceous glands and the development of skin diseases. In our analysis, SREBP-, TGFβ-SMAD-, MAPK-, and FoxO3a-related signaling pathways were predicted with PM 2.5 -induced phenomena-related signaling pathways. SREBPs regulate lipid homeostasis by activating the genes required for lipid synthesis, including cholesterol, FAs, and triacylglycerol as transcription factors [24]. Furthermore, PM induces inflammasomes by activating SREBP1 in human pulmonary fibroblasts [23]. TGFβ signaling maintains sebocytes in an undifferentiated state by decreasing the expression of lipogenesis-related genes [27]. MAPKs regulate a wide range of cellular homeostatic processes, including proliferation, differentiation, survival, apoptosis, and responses to stress conditions. Suppression of MAP2K5-MAPK7 signaling downregulates SREBP target genes and decreases lipid synthesis in small cell lung cancer cells [25]. The transcription factor FoxO3 regulates biological processes such as apoptosis, cell cycle, oxidative stress resistance, and inflammation [42]. Melnik reported that ATRA promoted apoptosis by upregulating the expression of FoxO3a and its direct target, tumor necrosis factor (ligand) superfamily member 10 (TN-FSF10), thereby downregulating sebum secretion [26,43]. In our results, PM 2.5 decreased the expression of TNFSF10 and FoxO3a (Figures 1b and 2d). Although the changes in activity were not validated in this study, wnt-β-catenin signaling pathway-related factors, such as catenin beta 1 (CTNNB1), WNT3A, and low-density lipoprotein receptor-related protein 6 (LRP6), were predicted by IPA to be inhibited upstream regulators ( Table 2). The wnt-β-catenin signaling pathway plays a role in regulating the differentiation of sebaceous glands [12,44,45]. Thus, wnt-β-catenin signaling may be associated with PM 2.5 -induced sebocyte damage.
In summary, this study investigated the effects of PM 2.5 on SZ95 sebocytes via transcriptome analysis. We showed that PM 2.5 increased lipid synthesis, lipid peroxidation, inflammatory cytokine expression, and oxidative stress; moreover, it altered lipid composition and expression of cell barrier-related proteins in SZ95 sebocytes. We also identified the signaling pathways related to the harmful effects of PM 2.5 and validated the altered activity of the signaling pathways in PM 2.5 -exposed sebocytes and human skin models. Further studies are required to elucidate the causal relationship between skin irritant reaction by PM 2.5 through hair follicles and skin disease aggravations. Nevertheless, the elucidated molecular alterations in the expression levels of DEGs and analysis of the associated mechanisms provide novel biomarkers and therapeutic targets to relieve the symptoms of skin damage caused by PM 2.5 .

PM 2.5 Preparation and Analysis
Ambient PM 2.5 was collected on a polytetrafluoroethylene filter (Zefluor™; Pall Life Science, Mexico City, Mexico) using a high-volume sampler machine (TE6070; Tisch Environmental Inc., Cleves, OH, USA) equipped with a PM 2.5 selective-inlet head at a flow rate of 1.13 m 3 /min. Sample collection of PM2.5 was carried out on the rooftop of the Amorepacific Corporation R&I building located in Yongin, Korea (37 • 15 N, 127 • 06 E), from January to April 2018. PM 2.5 was extracted from the collected samples and resuspended in 20% ethanol. The detailed methods were described in a previous study [3,7].

RT-qPCR Analysis
Total RNA was isolated using TRIzol reagent (Ambion, Carlsbad, CA, USA) according to the manufacturer's protocol and quantified using a NanoDrop spectrophotometer (Thermo Scientific, Waltham, MA, USA). cDNA was generated from 1 µg of RNA using the RevertAid First Strand cDNA Synthesis Kit (Invitrogen, Carlsbad, CA, USA) according to the manufacturer's protocol. RT-qPCR was performed in a solution containing 1 µL of cDNA, TaqMan Universal Master Mix II (Applied Biosystems, Foster City, CA, USA), and TaqMan probes using a 7500 Fast Real-Time PCR system (Applied Biosystems, Foster City, CA, USA). The parameters and TaqMan probes that were used are summarized in the Supplementary Material.

Western Blotting of Cell Lysates and ELISA
Sub-confluent cells were treated with PM 2.5 or TGFβ (5 ng/mL) for various durations. To measure the IL6 secretion level, the medium was spun down at 2000 rpm for 5 min and analyzed using an ELISA kit (#BMS213HS; Invitrogen, Vienna, Austria) according to the manufacturer's protocol. To detect nuclear proteins, fractionation of cytosolic and nuclear proteins was performed using a hypotonic buffer. For Western blot analysis, cell lysates were resuspended in a sodium dodecyl sulfate (SDS) sample buffer containing a reducing buffer, resolved by SDS-polyacrylamide gel electrophoresis (PAGE), and transferred to a polyvinylidene difluoride membrane. The antibodies that were used are listed in the Supplementary Material.

Measurement of Lipid Production
SZ95 sebocytes were seeded on glass coverslips coated with collagen I and were then treated with PM 2.5 (100 µg/mL) for 48 h. Following incubation, the cells were washed with phosphate-buffed saline (PBS) and treated with Bodipy 493/503 (2 µM; Invitrogen, Eugene, OR, USA) for 30 min at 37 • C. Subsequently, the cells were washed thrice with PBS and fixed with 4% paraformaldehyde. A mounting medium containing DAPI (P36962; Invitrogen, Eugene, OR, USA) was used for nuclear staining. The cells were observed under a confocal microscope (LSM980; Zeiss, Oberkochen, Germany).

Measurement of ROS and Lipid Peroxidation
SZ95 sebocytes were seeded on glass coverslips coated with collagen I. For ROS measurements, the cells were treated with PM 2.5 for 2 h and with CellROX Green (5 µM; Life Technologies, Carlsbad, CA, USA) for the last 30 min. The cells were then washed thrice with PBS and fixed with 4% paraformaldehyde. A mounting medium containing DAPI was used for nuclear staining. For detection of lipid peroxidation, the SZ95 sebocytes were treated with PM 2.5 for 24 h and then 5 µM diphenyl-1-pyrenylphosphine (DPPP; Invitrogen, Eugene, OR, USA) for 30 min at 37 • C. The cells were observed under a confocal microscope (LSM980; Zeiss, Oberkochen, Germany).

Human Skin Tissue Model and Immunohistochemistry
HairSkin ® models purchased from Genoskin (Toulouse, France) were cultured in a humidified atmosphere of 5% CO 2 at 37 • C for 24 h, and PM 2.5 (20 µg/cm 2 ) was applied on the surface of the HairSkin ® models for 0, 2, 4, or 6 days. HairSkin ® models were fixed with 10% formalin (Sigma-Aldrich, St. Louis, MO, USA) for paraffin-embedded sectioning. The paraffin-embedded tissues were sliced into 3-µm sections and stained with hematoxylin and eosin, along with the specific antibodies listed in the supplementary materials. The stained sections were visualized under an optical microscope (Olympus, Tokyo, Japan).

Funding:
The authors declare that no funds, grants, or other support were received during the preparation of this manuscript.

Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable. Data Availability Statement: All NGS data files are available from the National Center for Biotechnology Information Gene Expression Omnibus (NCBI GEO, https://www.ncbi.nlm.nih.gov/geo/), accession number GSE213620.

Conflicts of Interest:
The authors declare no conflict of interest.