T0070907

A long‑wave UVA filter avobenzone induces obesogenic phenotypes in normal human epidermal keratinocytes and mesenchymal stem cells

Sungjin Ahn1,2 · Seungchan An1,2 · Moonyoung Lee1,2 · Eunyoung Lee1,2,3 · Jeong Joo Pyo1,2 · Jeong Hyeon Kim1 · Min Won Ki1,2 · Sun Hee Jin1,2 · Jaehyoun Ha3 · Minsoo Noh1,2

Abstract

Avobenzone is the most commonly used ultraviolet (UV) A filter ingredient in sunscreen. To investigate the biological activ- ity of avobenzone in normal human epidermal keratinocytes (NHEKs), the genome-scale transcriptional profile of NHEKs was performed. In this microarray study, we found 273 up-regulated and 274 down-regulated differentially expressed genes (DEGs) in NHEKs treated with avobenzone (10 μM). Gene Ontology (GO) enrichment analysis showed that avobenzone significantly increased the DEGs associated with lipid metabolism in NHEKs. In addition, avobenzone increased the gene transcription of peroxisome proliferator-activated receptor γ (PPARγ) and fatty acid binding protein 4 in NHEKs, implicating that avobenzone may be one of the metabolic disrupting obesogens. To confirm the obesogenic potential, we examined the effect of avobenzone on adipogenesis in human bone marrow mesenchymal stem cells (hBM-MSCs). Avobenzone (EC50, 14.1 μM) significantly promoted adipogenesis in hBM-MSCs as its positive control obesogenic chemicals. Avobenzone (10 μM) significantly up-regulated mRNA levels of PPARγ during adipogenesis in hBM-MSCs. However, avobenzone did not directly bind to PPARγ and the avobenzone-induced adipogenesis-promoting activity was not affected by PPARγ antago- nists T0070907 and GW9662. Therefore, avobenzone promoted adipogenesis in hBM-MSCs through a PPARγ-independent mechanism. This study suggests that avobenzone functions as a metabolic disrupting obesogen.

Keywords Avobenzone · Normal human epidermal keratinocytes · Human bone marrow mesenchymal stem cells · Obesogen

Introduction

Avobenzone, also known as butyl methoxydibenzoylmethane, is extensively used in sunscreen formulations as a long-wave UV A filter to protect human skin from UV A irradiation-induced skin damage (Uter et al. 2014). In the sunscreen regulation of the United States Food and Drug Administration (US FDA), avobenzone is approved within the concentration of up to 3% in over-the-counter sunscreen formulations (Wang and Lim 2011). In the European Union and Australia, the concentration limit of avobenzone is 5% in sunscreen products (Jansen et al. 2013). Due to its exten- sive use in sunscreen products as a long-wave UV A fil- ter ingredient, avobenzone enters the aquatic environment via various waste streams, and is significantly detected in freshwater from rivers and lakes (Poiger et al. 2004). In this regard, avobenzone may indirectly affect human health. Avobenzone is easily excited by UV irradiation or other biological processes inducing oxygen-free radical produc- tion (Chatelain and Gabard 2001; Sayre et al. 2005). Excited avobenzone derivatives may result in protein or DNA dam- age. Avobenzone was reported to cause contact dermatitis (Motley and Reynolds 1989); however, it has been consid- ered relatively safe. In addition, photodegraded avobenzone derivatives have been suspected to be the main cause of sunscreen-induced toxic responses in human skin (Afonso et al. 2014; Jansen et al. 2013). In an acute oral toxicity study with the Sprague–Dawley rat, the no-observed adverse effect level (NOAEL) of avobenzone was 450 mg/kg/day (ECHA, 2013). Although a twenty-one-day repeated dose dermal toxicity study with the rabbit revealed that the NOAEL of avobenzone was 360 mg/kg/day, the highest concentration tested, a minimum toxic dose of avobenzone in human skin has not been established.
Epidermal keratinocytes are the primarily affected cells by environmental chemicals like avobenzone (Choi et al. 2017; Lee et al. 2016, 2018a; Piao et al. 2018). The main physiological function of keratinocytes is to form the epi- dermal permeability barrier (Feingold et al. 2007). In the stratum corneum of the human epidermis, keratinocytes terminally differentiate into corneocytes, forming a hydro- philic cornified cell envelope structure, surrounded by lipo- philic multilayered lamellar lipid sheets (Matsui and Amagai 2015). The hydrophilic nature of the cornified envelope is primarily associated with cornified envelope proteins, such as filaggrin, loricrin, involucrin, and keratins. Because extra- cellular lipids in the stratum corneum consist of ceramides, cholesterol, and free fatty acids in an approximately equi- molar ratio, the balanced regulation of lipid metabolism in keratinocytes is essential to the integrity of the normal epidermal permeability barrier (Van Smeden et al. 2014). When the epidermal permeability barrier function is dis- rupted, environmental chemicals easily penetrate through the epidermis, and can trigger cutaneous inflammatory reactions in human skin (Lin et al. 2017). In this regard, the biologi- cal response of epidermal keratinocytes should be primarily investigated to understand the potentially adverse cutane- ous effects of various environmental toxic stimuli or skin personal care products (Kim et al. 2018a). Considering that avobenzone is virtually the only one long-wave UV A filter of choice and its chronic long-time exposure to human skin, the molecular and cellular effects on keratinocytes should be well defined. However, the cellular effects of avoben- zone on human epidermal keratinocytes have not been fully understood.
This study was first aimed to elucidate the genome-scale transcriptional profile of normal human epidermal keratino- cytes (NHEKs) in response to avobenzone by an oligonu- cleotide-based microarray study. In the functional module analysis based on the Gene Ontology (GO) biological pro- cess (BP), cholesterol biosynthetic process-associated genes were identified as the most significantly up-regulated biolog- ical process in the avobenzone-treated NHEKs. Generally, avobenzone increased the transcription of genes regulating lipid metabolism. To confirm the lipid metabolism modi- fying activity, we examined whether avobenzone affected adipogenesis in human bone marrow mesenchymal stem cells (hBM-MSCs). Avobenzone promoted adipogenesis in hBM-MSCs, like metabolic disrupting obesogens.

Materials and methods

Culture of NHEKs and cell viability tests

NHEKs primarily cultured from neonatal foreskins were purchased from Lonza (Basel, Switzerland). NHEKs were cultured in Keratinocyte Basal Medium (KBM) supple- mented with human epidermal growth factor, insulin, bovine pituitary extract, epinephrine, hydrocortisone, transferrin, and gentamicin/amphotericin B (Lonza). NHEKs were sub- cultured at 90% confluence for cell expansion up to the third passage for the study.
Avobenzone was purchased from Sigma-Aldrich (St. Louis, MO, USA). Cell viability in avobenzone-treated NHEKs was evaluated using a Cell Counting Kit-8 assay (CCK-8, Dojindo, Kumamoto, Japan). Avobenzone was treated to NHEKs cultured in 24-well plates up to 100% con- fluence immediately before experiments. To measure their viability, NHEKs were incubated for 24 h. After washing 3 times with phosphate-buffered saline (PBS), 2-(2-meth- oxy-4-nitrophenyl)-3-(4-nitrophenyl)-5-(2,4-disulfophe- nyl)-2H-tetrazolium (WST-8) solution was treated to the avobenzone-treated NHEKs. After 2 h of WST-8 treatment, the absorbance at 450 nm was measured with a microplate reader (BioTek, Winooski, VT, USA).

Total RNA isolation, a genome‑scale microarray experiment, and Gene Ontology enrichment analysis

Total RNA was prepared using Trizol reagent (Invitrogen, Carlsbad, CA, USA). RNA integrity was determined with a BioAnalyzer 2100 (Agilent Technologies, Santa Clara, CA, USA). Total RNA (2 μg) from each sample was transcribed using SuperScript™ reverse transcriptase (Invitrogen) for studying the transcriptional profile with microarrays. Affym- etrix Human Genome U133 Plus 2.0 GeneChip arrays were used for the microarray study (Affymetrix, Santa Clara, CA, USA). Microarray experiments and GO enrichment analysis of the DEGs in the avobenzone-treated NHEKs were performed as described (Lee et al. 2016, 2018b). GO enrichment analysis of the avobenzone-induced DEGs in NHEKs was performed by comparing the frequency of genes annotated as the same GO biological process (BP) term in a group of DEGs with that in the entire set of genes in the Affymetrix Human 133 2.0 GeneChip array. The GO annotation files were downloaded from the Gene Ontology Consortium webpage (http://www.geneontology.org), and the August 2016 version of the GO BP terms were used in the analysis. Due to the redundancy in representing a specific gene in the Affymetrix Human 133 2.0 GeneChip array, 22,482 was considered as the total number of genes analyzed in the GO enrichment analysis. A 2 × 2 contingency matrix was constructed to compare the frequency of a spe- cific GO BP term in the DEG set with the number of genes annotated as the GO BP term in the total 23,624 genes. The 2 × 2 contingency matrix was analyzed for the calculation of p values by χ2 test (frequency ≥ 5) or Fisher’s exact test (frequency < 5), using SPSS for Windows (SPSS Science, Chicago, IL, USA). Quantitative real‑time reverse transcription‑polymerase chain reaction The expression levels of target mRNAs were quantified by quantitative real-time reverse transcription-polymerase chain reaction (Q-RT-PCR) using an AB7500 Real Time PCR System (Applied Biosystems, Foster City, CA, USA). The TaqMan RT-PCR primer sets used in the Q-RT-PCR were: insulin-induced gene 1 (INSIG1), Hs00356479_g1; 3-hydroxy-3-methylglutaryl-CoA reductase (HMGCR), Hs00168352_m1; 3-hydroxy-3-methylglutaryl-CoA syn- thase 1 (HMGCS1), Hs00940429_m1; aldo–keto reductase family 1 member C2 (AKR1C2), Hs04194036_gH; solute carrier family 2 member 12 (SLC2A12), Hs00376943_m1; metallothionein 1H (MT1H), Hs00823168_g1; filaggrin (FLG), Hs00856927_g1; loricrin (LOR), Hs01894962_s1; keratin 1 (KRT1), Hs00196158_m1; transglutaminase 1 (TGM1), Hs00165929_m1; peroxisome proliferator- activated receptor α (PPARα), Hs00231882_m1; PPARγ, Hs00234592_m1; PPARδ, Hs04187066_g1; fatty acid bind- ing protein 4 (FABP4), Hs00609791_m1; osteoprotegerin (OPG, also known as TNF receptor superfamily member 11b, TNFRSF11B). Human GAPDH (4333764F, Applied Biosystems) was used as a control gene. Quantification of relative expression levels was performed using a mathemati- cal model developed by Pfaffl (2001). Evaluation of obesogenic potential during adipogenesis in hBM‑MSCs hBM-MSCs were purchased from Lonza (Walkersville, MD, USA), and cultured in Dulbecco Modified Eagle’s Medium (DMEM; glucose 1 g/L) supplemented with 10% fetal bovine serum (FBS) and 1% streptomycin–penicillin (Invitrogen). When adipocyte differentiation was induced in hBM-MSCs, culture media were replaced by 10% FBS-DMEM (4.5 g/L of glucose) containing an adipogenesis-inducing cocktail consisting of 10 μg/mL insulin, 0.5 μM dexamethasone, and 0.5 mM isobutylmethyxanthine (IBMX). Dexamethasone, insulin, IBMX, pioglitazone, and other chemicals were pur- chased from Sigma-Aldrich (St. Louis, MO, USA). Oil Red O (ORO) staining was performed to visualize lipid droplets in differentiated adipocytes. For ORO staining, cells were washed twice with PBS, and then fixed with 10% formalin in PBS for 15 min. Formalin-fixed cells were washed with 60% isopropanol. Lipid droplets were stained with 0.2% ORO in 60% isopropanol for 10 min at 25 °C, and then washed with tap water three times. To quantify the lipid accumulation, adsorbed Oil Red O stain hBM-MSCs samples were dis- solved with 100% isopropanol, and absorbance was meas- ured at 540 nm using a spectrophotometer. Hematoxylin reagent was further incubated for 1 min to stain the nucleus. The differentiated cells were observed and photographed using an Eclipse TS100 inverted microscope (Nikon Co., Tokyo, Japan). For quantitative measurement of adiponec- tin and osteoprotegerin (OPG) in cell culture supernatants, Quantikine™ immunoassay kits (R&D Systems, Minneapo- lis, MN, USA) were used. Adiponectin and OPG concentra- tions were determined as described (Noh 2012). PPARγ–ligand binding assays The time-resolved fluorescence resonance energy transfer (TR-FRET)-based receptor binding assay was performed using Lanthascreen™ competitive binding assay kits (Invit- rogen) to evaluate the binding of ligand to PPARγ. The assay measurements were performed using CLARIOstar (BMG LABTECH, Ortenberg, Germany), with instrument settings as described in the TR-FRET manufacturer’s instructions. Statistical analysis Statistical analysis was conducted using SPSS® for Windows (SPSS Science, Chicago, IL, USA). Student’s t test was used for a comparison between a single treated group and the experimental control. One-way analysis of variance was fol- lowed by Tukey’s post hoc tests for multiple comparisons. Results Effects of avobenzone on the whole genome transcriptional responses of NHEKs The sub-cytotoxic concentration of avobenzone in NHEKs was determined for a whole genome transcriptional study, because the cell death of NHEKs may not occur at the con- ventional exposure levels of avobenzone. The cell viability of NHEKs was unaffected up to 10 μM of avobenzone, and the half maximal cell viable concentration (CV50 value) was 34.9 μM (Supplementary Fig. S1). Despite no statistical significance, 83% of viable cells were measured in NHEKs treated with 20 μM of avobenzone compared to the vehicle- treated condition. In the whole genome transcriptional anal- ysis, 10 μM of avobenzone was treated for 24 h in NHEKs, and the oligonucleotide microarray was performed. The oli- gonucleotide array data in the Affymetrix Cel file format are available in the Gene Expression Omnibus database from the National Center for Biotechnology Information (NCBI GEO), and their accession number is GSE122901. Among 22,482 genes represented by 54,000 probes in the oligonu- cleotide array, 273 up-regulated and 274 down-regulated genes were selected as differentially expressed genes (DEGs) in the avobenzone-treated NHEKs (Supplementary Tables 1 and 2). To understand the avobenzone-induced biological changes, GO-based biological process (BP) was analyzed by comparing the frequencies of GO BP terms between DEGs and whole gene set transcripts (Lee et al. 2016). Among 12,297 GO BP terms in whole gene sets of the oligonucleo- tide array, DEGs represented 2998 GO BP terms. In the GO BP-based enrichment analysis of the avobenzone-induced DEGs in NHEKs, 2998 contingency matrices were generated to perform Fisher’s exact test- or χ2 test-based statistic calcu- lations. In the avobenzone-treated NHEKs, cholesterol bio- synthetic process (GO:0,006,695) was the most significantly enriched GO BP term in the up-regulated DEGs (Table 1). In 22,482 genes in the oligonucleotide array, 39 genes were associated with the GO BP cholesterol biosynthetic process. In comparison to the total gene set, 13 genes in the 273 up-regulated DEGs were annotated as the cholesterol bio- synthetic process, resulting in the most significant statistical probability, less than 10−10 of the p value. The cholesterol metabolism-associated DEGs, 3-hydroxy-3-methylglutaryl- CoA reductase (HMGCR), 3-hydroxy-3-methylglutaryl- CoA synthase 1 (HMGCS1) and farnesyl-diphosphate farnesyltransferase 1 (FDFT1), also contributed to the sig- nificant enrichment of the isoprenoid biosynthetic process (GO:0,008,299) and the cellular lipid metabolic process (GO:0,044,255) (Table 1). Metallothionein module-contain- ing transcription factors such as metallothionein 1E (MT1E), MT1F, MT1G, MT1H, MT1X, and MT2A were associated with the up-regulation of two biological processes: negative regulation of growth (GO:0,045,926), and cellular response to cadmium ion (GO:0,071,276). The other significantly enriched GO BP terms were oxidation–reduction process (GO:0,055,114), blood vessel development (GO:0,001,568), progesterone metabolic process (GO:0,042,448), hexose transmembrane transport (GO:0,035,428), and atrioventricu- lar valve morphogenesis (GO:0,003,181) in the up-regulated DEGs, compared to the total gene set (Table 1). Consider- ing that avobenzone can be easily degraded by free radical producing cellular mechanisms (Chatelain and Gabard 2001; Sayre et al. 2005), the functional enrichment of the oxida- tion–reduction process in the up-regulated DEGs of NHEKs in response to avobenzone was an expected outcome. In the 274 avobenzone-induced down-regulated DEGs, the four epidermal differentiation marker genes, filaggrin (FLG), loricrin (LOR), keratin 1 (KRT1), and transglu- taminase 1 (TGM1), contributed to the identification of three BP: the establishment of skin barrier (GO:0,061,436), cornification (GO:0,070,268), and keratinocyte differen- tiation (GO:0,030,216). This result suggested that avoben- zone may affect the integrity of the skin permeability bar- rier structure. In addition, avobenzone also down-regulated a variety of DEGs associated with the cellular response to extracellular stimulus (GO:0,031,668), blood vessel morphogenesis (GO:0,048,514), retinol metabolic pro- cess (GO:0,042,572), cellular response to organic cyclic compound (GO:0,071,407), transforming growth factor beta receptor signaling pathway (GO:0,007,179), regula- tion of growth (GO:0,040,008), and liver development (GO:0,001,889) (Table 1). Validation of the GO BP Enrichment Analysis on the Avobenzone‑induced DEGs in NHEKs The GO BP-based enrichment analysis showed that avoben- zone up-regulated cellular processes to regulate cholesterol and lipid metabolism. To confirm the avobenzone-induced up-regulation of DEGs associated with cholesterol metabo- lism in NHEKs, the gene transcriptional changes of insulin- induced gene 1 (INSIG1), HMGCR, and HMGCS1 were investigated at 4, 24, and 48 h after the avobenzone treat- ment by Q-RT-PCR (Figs. 1a–c). Consistent with the micro- array results, the mRNA levels of INSIG1, HMGCR, and HMGCS1 were up-regulated in NHEKs at 24 h after the avobenzone treatment. At 48 h, transcription of both INSIG1 and HMGCS1 remained in the state of up-regulation, whereas no significant change was observed for the HMGCR gene transcription. The avobenzone-induced up-regulation of progesterone metabolism-associated AKR1C2, hexose transmembrane transport-related SLC2A12, and growth- associated transcription factor MT1H was also validated in NHEKs by Q-RT-PCR (Figs. 1d–f). In NHEKs, the GO BP-based enrichment analysis on the avobenzone-induced down-regulated DEGs, suggesting that avobenzone may affect the homeostasis of the skin permeability barrier devel- opment. When the transcription of major skin permeability barrier-associated genes was investigated for validating their microarray expression, mRNA levels of FLG, LOR, KRT1, and TGM1 were down-regulated in NHEKs in response to avobenzone (Fig. 2). Therefore, the major transcriptomics signature that was identified by the GO BP-based enrich- ment analysis for both up-regulated and down-regulated genes was experimentally confirmed in the independently prepared NHEKs treated with avobenzone. Effects of avobenzone on the gene transcription of peroxisome proliferator‑activated receptors in NHEKs In the GO BP-based enrichment analysis of the avobenzone- induced up-regulated DEGs, avobenzone affected cellular cholesterol and lipid metabolism in NHEKs. Peroxisome proliferator-activated receptor α (PPARα), PPARγ, and PPARδ play a regulatory role in the cholesterol and lipid metabolic pathways in NHEKs (Varga et al. 2011). Although these PPARs had not been selected as the avobenzone- induced DEGs in the microarray study, the microarray expression of PPARα and PPARγ was up-regulated by 63 and 72%, respectively, compared to the control in the avobenzone-treated NHEKs. In contrast, there was no sig- nificant change in the PPARδ transcription in the microarray study. It is possible that the sensitivity of microarray probes for PPARs is not sufficient to detect differential expression. Therefore, the mRNA levels of PPARs were measured by Q-RT-PCR (Fig. 3). The gene transcription of both PPARα and PPARγ was significantly increased in NHEKs at 24 h after avobenzone treatment, compared to those of the vehi- cle-treated control (Fig. 3a, b). The change in the mRNA levels of both PPARα and PPARγ was not significant at 48 h after the avobenzone treatment. Consistent with the micro- array result, the transcription of PPARδ was unaffected in NHEKs in response to the avobenzone treatment. Therefore, avobenzone significantly increased the gene transcription of both PPARα and PPARγ in NHEKs, suggesting that avoben- zone has the potential to affect the lipid metabolic pathway of mammalian cells. Investigation of avobenzone as a potential metabolic disrupting obesogen The adipogenic differentiation model of human bone mar- row mesenchymal stem cells (hBM-MSCs) was used to investigate the metabolic disrupting potential of avoben- zone. When inducing adipocyte differentiation of murine preadipocytes like 3T3-L1 preadipocytes, the addition of insulin, dexamethasone, and isobutylmethylxanthine (IDX condition) can differentiate virtually all cells to have adipo- cyte phenotypes (Han et al. 2018; Ruiz-Ojeda et al. 2016). In contrast, the IDX condition is not sufficient to achieve notable adipocyte differentiation in hBM-MSCs (Chen et al. 2016; Shin et al. 2009). To increase adipogenesis in hBM- MSCs, PPARγ agonists are usually supplemented to the IDX condition (Ahn et al. 2018; Scott et al. 2011). In general, obesogenic metabolic disrupting chemicals, such as para- bens, phthalates, polychlorinated biphenyls, and organotins, promote the lineage commitment of hBM-MSCs into adipo- cytes (Veiga-Lopez et al. 2018). To investigate the obesogenic potential, the effect of avobenzone on adipogenesis in hBM-MSCs was evalu- ated (Fig. 4). As reference obesogens, currently prescribing PPARγ anti-diabetic pioglitazone and well-defined obeso- gens, such as bisphenol A, butylparaben, benzyl butyl phtha- late, and tributyltin, were simultaneously evaluated during adipogenesis in hBM-MSCs (Muscogiuri et al. 2017). In addition, bis(2-ethylhexyl) phthalate was included in the study, because the obesogenic potential of bis(2-ethylhexyl) phthalate in in vitro culture systems has been controversial (Pereira-Fernandes et al. 2014; Veiga-Lopez et al. 2018). Avobenzone and other obesogens increased the number and size of lipid droplets in differentiated adipocytes during adi- pogenesis in hBM-MSCs (Fig. 4). When the level of ORO staining was quantitated, 10-μM pioglitazone increased the ORO accumulation by 288%, compared to that of the IDX control (Fig. 4c). When hBM-MSCs were treated with 20 μM of avobenzone, bisphenol A, butylparaben, and benzyl butyl phthalate, the ORO accumulation was signifi- cantly increased by 58.2, 43.3, 51.5, and 109.6%, respec- tively (Fig. 4j). The ORO level was increased by 24.9%; however, the change induced by bis(2-ethylhexyl) phthalate (DEHP) was not statistically significant. Tributyltin exhib- ited significant cytotoxicity to hBM-MSCs at 20 μM (65.5% cell viability at 1 μM). Notably, non-cytotoxic tributyltin (0.1 μM) significantly increased the ORO level by 260.1% during adipogenesis in hBM-MSCs. The level of adiponec- tin, an adipocytokine primarily produced in mammalian adi- pocytes, was measured in culture supernatants to confirm the avobenzone-induced up-regulation of adipogenesis (Fig. 5a). Compared to the IDX control, adiponectin secretion was significantly up-regulated by 193.8, 91.5, 163.1, 236.7, 104.3 and 240.2% in adipogenesis-induced hBM-MSCs in response to avobenzone, bisphenol A, butylparaben, ben- zyl butyl phthalate, DEHP, and tributyltin, respectively. In contrast to the ORO result, DEHP significantly increased adiponectin production during adipogenesis in hBM-MSCs compared to the IDX control. Considering that the adiponec- tin production significantly correlates the level of adipogen- esis in hBM-MSCs (Yu et al. 2017), it could be concluded that DEHP has an obesogenic potential. In addition, the level of osteoprotegerin (OPG) production was measured in par- allel because OPG is endogenously produced during osteo- genesis and suppressed during adipognenesis in hBM-MSCs (Noh, 2012). As expected, the level of OPG was signifi- cantly down-regulated in response to avobenzone and other reference obesogens during adipogenesis in hBM-MSCs (Fig. 5b). When the maximum adiponectin-secretion activity of PPARγ agonists was used as a 100% response to calculate the half-maximum concentration (EC50), the EC50 values of pioglitazone and rosiglitazone were 0.23 and 0.006 μM in the concentration–effect analysis, respectively (Fig. 5c). Although the EC50 value of tributyltin was lower than that of pioglitazone, tributyltin did not induce the adiponectin production up to the maximum level as high as PPARγ full agonists in hBM-MSCs. Avobenzone also promoted adi- ponectin secretion in a concentration-dependent manner (EC50, 14.1 μM) but its maximum activity in non-cytotoxic concentrations was not as potent as those of PPARγ full agonists. To confirm the avobenzone-induced up-regulation of adipogenesis, mRNA levels of PPARα, PPARγ, adiponec- tin, FABP4, and osteoprotegerin (OPG) were measured by Q-RT-PCR during adipogenesis in hBM-MSCs (Fig. 5d–h). Compared to the IDX control, avobenzone significantly up-regulated the gene transcription of PPARα, PPARγ, adiponectin, and FABP4 (Fig. 5d–g). The mRNA level of OPG was down-regulated similar to the change of its protein level (Fig. 5h). Taken together, avobenzone induced major obesogenic phenotypes during adipogenesis in hBM-MSCs, similar to other reference obesogens. Avobenzone‑induced obesogenic phenotypes were induced by PPARγ‑independent mechanisms In mammalian tissue, PPARγ plays pivotal roles in adi- pocyte differentiation and its functional regulation (Kim et al. 2018b). To examine whether avobenzone affected cellular PPARγ pathway, TR-FRET-based PPARγ binding assays were performed (Fig. 6). In analysis, 20-μM benzyl butyl phthalate and 1-μM tributyltin significantly inhibited the PPARγ binding of the labeled GW1929 by 36.6, and 68.0%, respectively (Fig. 6a). However, avobenzone and other control obesogens like bisphenol A, butylparaben, and DEHP, did not bind to PPARγ up to 20 μM. In a con- centration–effect analysis, avobenzone did not affect the ligand–PPARγ binding activity up to 60 μM. Interestingly, tributyltin significantly replaced the labeled ligand bind- ing to PPARγ in a concentration-dependent manner (IC50, 0.062 μM). The PPARγ binding activity of tributyltin was more potent than that of pioglitazone, although pioglitazone was stronger to promote adipocyte differentiation in hBM- MSCs. These results suggest that tributyltin may affect other cellular mechanisms to inhibit the PPARγ-dependent mecha- nism, which requires further research to identify molecular targets. To confirm that the obesogenic activity of avobenzone was mediated by PPARγ-independent mechanisms, the effect of PPARγ antagonist T0070907 on the avobenzone- induced up-regulation of adiponectin production in hBM- MSCs was examined (Fig. 6c). T0070907 inhibited the IDX- induced adipogenesis in hBM-MSCs, which was an expected outcome, because PPARγ regulates adipocyte differentiation and its functions (Lee et al. 2002). T0070907 significantly antagonized the pioglitazone-induced up-regulation of adi- ponectin production. Compared to the level of inhibition against the pioglitazone-induced activity during adipogen- esis in hBM-MSCs, the avobenzone-induced obesogenic effect on adiponectin production was relatively unaffected by T0070907 during adipogenesis in hBM-MSCs. Notably, GW9662, a well-known PPARγ antagonist, functioned as a PPARγ partial agonist during adipogenesis in hBM-MSCs (Fig. 6d). Compared to the IDX control, GW9662 signifi- cantly promoted adiponectin production during adipogenesis in hBM-MSCs in a concentration-dependent manner. Par- tial agonists inhibit agonist-induced effects in general (Ahn et al. 2019). As expected, GW9662 significantly inhibited the pioglitazone-induced up-regulation of adiponectin pro- duction. In contrast, GW9662 had no effect on the avoben- zone-induced promotion of adipogenesis and adiponectin production (Fig. 6d). The avobenzone-induced obesogenic phenotypes were not affected during adipogenesis in hBM- MSCs treated with a PPARγ antagonist. Therefore, avoben- zone induced the metabolic disrupting obesogenic pheno- types via PPARγ-independent mechanisms. Discussion Avobenzone is extensively prescribed in over-the-counter or cosmetic sunscreen products, because of its filtering efficacy against long-wave UVA (Jansen et al. 2013; Wang and Lim 2011). The GO BP-based enrichment analysis of the avobenzone-induced up-regulated DEGs in NHEKs suggested that avobenzone can modify cellular lipid meta- bolic processes. For example, avobenzone significantly up-regulated the gene transcription of INSIG1, HMGCR and HMGCS1 in NHEKs. The expression of these lipid metabolic enzymes is generally regulated by the PPARγ- dependent pathway (Wahli and Michalik 2012). In time- course analysis, the gene transcription of INSIG1 and HMGCS1 was up-regulated in NHEKS in response to avobenzone at both 24 h and 48 h. In contrast, the HMGCR gene transcription was increased only at 24 h and no signifi- cant change was detected at 48 h in the avobenzone-treated NHEKs. Interestingly, INSIG1 protein can bind to HMG CoA reductase, and this binding promotes the degrada- tion of HMG CoA reductase (Sever et al. 2003), suggest- ing that INSIG1-dependent negative feedback mechanism may be involved in the HMGCR gene transcription at 48 h in NHEKs. Avobenzone significantly up-regulated the PPARγ gene transcription in NHEKs, although it did not directly bind to PPARγ. Because the chemical-induced up- regulation of PPARγ is one of the major obesogenic pheno- types (Veiga-Lopez et al. 2018), the obesogenic potential of avobenzone was investigated with the adipogenesis model of hBM-MSCs. Avobenzone significantly increased both lipid accumulation and adiponectin production during adipogen- esis in hBM-MSCs compared to the control, which was simi- lar to reference obesogenic chemicals, such as bisphenol A, butylparaben, benzyl butyl phthalate, DEHP, and tributyltin. Therefore, avobenzone, a US FDA approved long-wave UVA filter, is a metabolic disrupting obesogen. Avobenzone has been massively used, because it is the only option for long- wave UVA filters in some countries (Wang and Lim 2011; Wang et al. 2017). A significant amount of avobenzone has been detected in the aquatic environment (Poiger et al. 2004; Ramos et al. 2015). In this regard, avobenzone exposure to human population is inevitable, regardless of its topical application to skin. The toxicity of avobenzone has been focused on photo-induced irritation or photosensitization (Afonso et al. 2014; Chatelain and Gabard 2001; Wang and Lim 2011). The metabolic modifying activity of avobenzone in NHEKs and hBM-MSCs supported that obesity should be considered as one of the major toxicological outcomes induced by avobenzone. Considering that the availability of long-wave UVA filters is limited to a few compounds, avobenzone usage may require extensive risk–benefit anal- ysis in terms of environmental perspectives on metabolic disrupting obesogens. The GO BP-based enrichment analysis of the down-regulated DEGs in NHEKs showed that avobenzone signifi- cantly affected the transcription of gene sets associated with keratinocyte differentiation, important in the establishment of skin barrier. The skin permeability barrier plays a role as a physical defense system against diverse environmental toxic chemicals. FLG, LOR, and KRT1 are essential components in the cornified envelope of the stratum corneum structure (Samuelov and Sprecher 2014). It is well established that genetic polymorphisms or mutations in the cornified enve- lope proteins, FLG, LOR, and KRT1, are associated with many pathologic skin conditions, such as atopic dermatitis and ichthyosis vulgaris (Agrawal and Woodfolk 2014; Candi et al. 2005). It was recently reported that cytoskeletal inter- mediate filament components, keratin 1 and keratin 10, are associated with the linkage between cutaneous inflammation and the skin permeability barrier function (Roth et al. 2012). The avobenzone-induced down-regulation of cornified envelope-associated genes may disrupt the integrity of the skin permeability barrier. It is possible that avobenzone may promote the transepidermal penetration of diverse environ- mental toxic chemicals, due to the disruption of the homeo- stasis of skin barrier development. Avobenzone-containing sunscreen products are directly applied onto human skin. In this regard, future study should be directed to characterize the toxicological outcomes in terms of skin permeability homeostasis at the avobenzone exposure level during daily sunscreen usage. Currently, the risk assessment of chemicals for various health issues has been directed to reduce, replace, or refine animal experiments. Regarding the evaluation of obesogenic potential, reliable in vitro testing assays are required for assessing the obesogenic risk of various chemicals. When evaluating the obesogenic potential of avobenzone, the adi- pogenesis in hBM-MSCs was exploited in this study. During adipogenesis in hBM-MSCs, avobenzone and its reference obesogens, such as bisphenol A, butylparaben, benzyl butyl phthalate, and tributyltin, significantly promoted both lipid accumulation and adiponectin production. DEHP was likely to increase lipid droplet formation during adipogenesis, in spite of no statistical significance. Notably, DEHP increased the adiponectin production over twofold, compared with the IDX control in hBM-MSCs, supporting the conclusion that DEHP is an obesogen. In a human epidemiological study, the significant association of the childhood exposure of DEHP with obesity was reported (Harley et al. 2017). Animal studies also strongly supported that DEHP induces obesogenic phenotypes (Lv et al. 2016). However, the obe- sogenic effects of DEHP on the 3T3-L1 murine preadipo- cyte system have been controversial, which is the first-line selection assay model for metabolic disrupting obesogens (OECD 2014). For example, DEHP was regarded as a non- obesogen in the 3T3-L1 murine preadipocyte model, which was evaluated with both the transcriptomics analysis, and the measurement of lipid accumulation (Pereira-Fernandes et al. 2014). A significant proportion of 3T3-L1 murine preadipo- cytes spontaneously accumulates lipids in their lipid droplets in long-term culture without obesogen or IDX treatments. When IDX was treated to 3T3-L1 cells, virtually all cells quickly differentiated into adipocyte phenotypes within a few days (Ruiz-Ojeda et al. 2016). In contrast, the IDX condition is not enough to induce the adipocyte differen- tiation of hBM-MSCs. The addition of obesogenic chemi- cals to IDX medium requires increase of adipogenesis in hBM-MSCs, as shown in this study. From the standpoint of DEHP, the adipogenesis model of hBM-MSCs may be efficient to use as an alternative in vitro assay for testing the obesogenic potentials of environmental chemicals. The dif- ferentiation process of hBM-MSCs into adipocytes includes the lineage commitment of stem cells into preadipocytes, and terminal adipocyte differentiation as well (Cristancho and Lazar 2011). Therefore, human MSCs may provide addi- tional mechanistic coverage associated with obesogenic out- comes, compared with that of the lineage-committed 3T3-L1 preadipocyte cell line. In addition, the use of hBM-MSCs in testing obesogenic activity is advantageous over the 3T3-L1 cell line, because of the use of human origin cells. Further studies should be directed to determining the efficiency, sensitivity, and predictability of obesogenic potentials by exploiting hBM-MSCs or other human MSCs. Conclusions In conclusion, avobenzone is an obesogen. The obesogenic potential of avobenzone was first supported by the gene tran- scription signature of NHEKs. Importantly, the GO func- tional module analysis of the avobenzone-induced DEGs indicated that avobenzone may disrupt skin permeability barrier homeostasis, which may further aggravate toxicolog- ical outcomes in human skin. Second, avobenzone promoted lipid droplet accumulation and adiponectin production dur- ing adipogenesis in hBM-MSCs. Therefore, the effective long-wave UVA filter avobenzone will require risk–benefit analysis, by balancing its sunscreen activity and potential obesogenic activity. References Afonso S, Horita K, Sousa e Silva JP, Almeida IF, Amaral MH, Lobao PA, Costa PC, Miranda MS, Esteves da Silva JCG, Sousa Lobo JM (2014) Photodegradation of avobenzone: stabilization effect of antioxidants. J Photochem Photobiol B 140:36–40. https://doi.org/10.1016/j.jphotobiol.2014.07.004 Agrawal R, Woodfolk JA (2014) Skin barrier defects in atopic derma- titis. Curr Allergy Asthma Rep 14:433. https://doi.org/10.1007/ s11882-014-0433-9 Ahn S, Lee M, An S, Hyun S, Hwang J, Lee J, Noh M (2018) 2-For- myl-komarovicine promotes adiponectin production in human mesenchymal stem cells through PPARγ partial agonism. Bioorg Med Chem 26:1069–1075. https://doi.org/10.1016/j. bmc.2018.01.019 Ahn S, Ma CT, Choi JM, An S, Lee M, Le THV, Pyo JJ, Lee J, Choi MS, Kwon SW, Park JH, Noh M (2019) Adiponectin-secretion- promoting phenylethylchromones from the agarwood of aquilaria malaccensis. J Nat Prod 82:259–264. https://doi.org/10.1021/acs. jnatprod.8b00635 Candi E, Schmidt R, Melino G (2005) The cornified envelope: a model of cell death in the skin. Nat Rev Mol Cell Biol 6:328–340. https://doi.org/10.1038/nrm1619 Chatelain E, Gabard B (2001) Photostabilization of Butyl methoxy- dibenzoylmethane (avobenzone) and ethylhexyl methoxycin- namate by Bis?ethylhexyloxyphenol methoxyphenyl triazine (Tinosorb S), a new UV broadband filter. Photochem Photobiol 74:401–406. https://doi.org/10.1562/0031-8655(2001)0740401POBMAA2.0.CO2 Chen Q, Shou P, Zheng C, Jiang M, Cao G, Yang Q, Cao J, Xie N, Velletri T, Zhang X, Xu C, Zhang L, Yang H, Hou J, Wang Y, Shi Y (2016) Fate decision of mesenchymal stem cells: adipo- cytes or osteoblasts? Cell Death Differ 23:1128–1139. https://doi. org/10.1038/cdd.2015.168 Choi M, Park M, Lee S, Lee JW, Cho MC, Noh M, Lee C (2017) Estab- lishment of immortalized primary human foreskin keratinocytes and their application to toxicity assessment and three dimensional skin culture construction. Biomol Ther 25:296–307. https://doi. org/10.4062/biomolther.2017.043 Cristancho AG, Lazar MA (2011) Forming functional fat: a growing understanding of adipocyte differentiation. Nat Rev Mol Cell Biol 12:722–734. https://doi.org/10.1038/nrm3198 ECHA (2013) Registered substances CAS 70356-09-1. https:// echa.europa.eu/registration-dossier/-/registered-dossier/14835. Accessed 9 Apr 2019 Feingold KR, Schmuth M, Elias PM (2007) The regulation of perme- ability barrier homeostasis. J Invest Dermatol 127:1574–1576. https://doi.org/10.1038/sj.jid.5700774 Han MH, Kim HJ, Jeong JW, Park C, Kim BW, Choi YH (2018) Inhi- bition of adipocyte differentiation by anthocyanins isolated from the fruit of vitis coignetiae pulliat is associated with the activa- tion of AMPK signaling pathway. Toxicol Res 34:13. https://doi. org/10.5487/TR.2018.34.1.013 Harley KG, Berger K, Rauch S, Kogut K, Henn BC, Calafat AM, Huen K, Eskenazi B, Holland N (2017) Association of prenatal urinary phthalate metabolite concentrations and childhood BMI and obe- sity. Pediatr Res 82:405–415. https://doi.org/10.1038/pr.2017.112 Jansen R, Osterwalder U, Wang SQ, Burnett M, Lim HW (2013) Pho- toprotection: part II sunscreen: development, efficacy, and con- troversies. J Am Acad Dermatol 69:867. https://doi.org/10.1016/j. jaad.2013.08.022 Kim HJ, Lee E, Lee M, Ahn S, Kim J, Liu J, Jin SH, Ha J, Bae IH, Lee TR, Noh M (2018a) Phosphodiesterase 4B plays a role in benzo- phenone-3-induced phototoxicity in normal human keratinocytes. Toxicol Appl Pharmacol 338:174–181. https://doi.org/10.1016/j. taap.2017.11.021 Kim SO, Han Y, Ahn S, An S, Shin JC, Choi H, Kim HJ, Park NH, Kim YJ, Jin SH, Rho HS, Noh M (2018b) Kojyl cinnamate esters are peroxisome proliferator-activated receptor α/γ dual agonists. Bioorg Med Chem 26:5654–5663. https://doi.org/10.1016/j. bmc.2018.10.010 Lee G, Elwood F, McNally J, Weiszmann J, Lindstrom M, Amaral K, Nakamura M, Miao S, Cao P, Learned RM, Chen JL, Li Y (2002) T0070907, a selective ligand for peroxisome proliferator-activated receptor g, functions as an antagonist of biochemical and cellular activities. J Biol Chem 277:19649–19657. https://doi.org/10.1074/ jbc.M200743200
Lee E, Kim HJ, Lee M, Jin SH, Hong SH, Ahn S, Kim SO, Shin DW, Lee ST, Noh M (2016) Cystathionine metabolic enzymes play a role in the inflammation resolution of human keratinocytes in response to sub-cytotoxic formaldehyde exposure. Toxicol Appl Pharmacol 310:185–194. https://doi.org/10.1016/j.taap.2016.09.017
Lee H, Cheong KA, Kim JY, Kim NH, Noh M, Lee AY (2018a) IL-1 Receptor antagonist reduced chemical-induced keratinocyte apop- tosis through antagonism to IL-1α/IL-1β. Biomol Ther 26:417–423. https://doi.org/10.4062/biomolther.2017.167
Lee M, Lee E, Jin SH, Ahn S, Kim SO, Kim J, Choi D, Lim KM, Lee ST, Noh M (2018b) Leptin regulates the pro-inflammatory response in human epidermal keratinocytes. Arch Dermatol Res 310:351–362. https://doi.org/10.1007/s00403-018-1821-0
Lin TK, Zhong L, Santiago J (2017) Anti-inflammatory and skin barrier repair effects of topical application of some plant oils. Int J Mol Sci 19:70. https://doi.org/10.3390/ijms19010070
Lv Z, Cheng J, Huang S, Zhang Y, Wu S, Qiu Y, Geng Y, Zhang Q, Huang G, Ma Q, Xie X, Zhou S, Wu T, Ke Y (2016) DEHP induces obesity and hypothyroidism through both central and periph- eral pathways in C3H/He mice. Obesity 24:368–378. https://doi. org/10.1002/oby.21359
Matsui T, Amagai M (2015) Dissecting the formation, structure and bar- rier function of the stratum corneum. Int Immunol 27:269–280. https://doi.org/10.1093/intimm/dxx024
Motley RJ, Reynolds AJ (1989) Photocontact dermatitis due to iso- propyl and butyl methoxy dibenzoylmethanes (eusolex 8020 and parsol 1789). Contact Dermatitis 21:109–110. https://doi. org/10.1111/j.1600-0536.1989.tb04708.x
Muscogiuri G, Barrea L, Laudisio D, Savastano S, Colao A (2017) Obe- sogenic endocrine disruptors and obesity: myths and truths. Arch Toxicol 91:3469–3475. https://doi.org/10.1007/s00204-017-2071-1
Noh M (2012) Interleukin-17A increases leptin production in human bone marrow mesenchymal stem cells. Biochem Pharmacol 83:661–670. https://doi.org/10.1016/j.bcp.2011.12.010
OECD (2014) Detailed review paper on the state of the science on novel in vitro and in vivo screening and testing methods and endpoints for evaluating endocrine disruptors, OECD Series on Testing and Assessment, No. 178, OECD Publishing, Paris. https://doi. org/10.1787/9789264221352-en
Pereira-Fernandes A, Vanparys C, Vergauwen L, Knapen D, Jorens PG, Blust R (2014) Toxicogenomics in the 3T3-L1 cell line, a new approach for screening of obesogenic compounds. Toxicol Sci 140:352–363. https://doi.org/10.1093/toxsci/kfu092
Pfaffl MW (2001) A new mathematical model for relative quantifica- tion in real-time RT–PCR. Nucleic Acids Res 29:e45. https://doi. org/10.1093/nar/29.9.e45
Piao MJ, Ahn MJ, Kang KA, Ryu YS, Hyun YJ, Shilnikova K, Zhen AX, Jeong JW, Choi YH, Kang HK, Koh YS, Hyun JW (2018) Par- ticulate matter 2.5 damages skin cells by inducing oxidative stress, subcellular organelle dysfunction, and apoptosis. Arch Toxicol 92:2077–2091. https://doi.org/10.1007/s00204-018-2197-9
Poiger T, Buser HR, Balmer ME, Bergqvist PA, Muller MD (2004) Occurrence of UV filter compounds from sunscreens in surface waters: regional mass balance in two Swiss lakes. Chemosphere 55:951–963. https://doi.org/10.1016/j.chemosphere.2004.01.012
Ramos S, Homem V, Alves A, Santos L (2015) Advances in analytical methods and occurrence of organic UV-filters in the environment-a review. Sci Total Environ 526:278–311. https://doi.org/10.1016/j. scitotenv.2015.04.055
Roth W, Kumar V, Beer HD, Richter M, Wohlenberg C, Reuter U, Thier- ing S, Staratschek-Jox A, Hofmann A, Kreusch F, Schultze JL, Vogl
T, Roth J, Reichelt J, Hausser I, Magin TM (2012) Keratin 1 main- tains skin integrity and participates in an inflammatory network in skin through interleukin-18. J Cell Sci 125:5269–5279. https://doi. org/10.1242/jcs.116574
Ruiz-Ojeda FJ, Ruperez AI, Gomez-Llorente C, Gil A, Aguilera CM (2016) Cell models and their application for studying adipogenic differentiation in relation to obesity: a review. Int J Mol Sci 17:1040. https://doi.org/10.3390/ijms17071040
Samuelov L, Sprecher E (2014) Peeling off the genetics of atopic dermati- tis–like congenital disorders. J Allergy Clin Immunol 134:808–815. https://doi.org/10.1016/j.jaci.2014.07.061
Sayre RM, Dowdy JC, Gerwig AJ, Shlelds WJ, Lioyd RV (2005) Unex- pected photolysis of the sunscreen octinoxate in the presence of the sunscreen avobenzone. Photochem Photobiol 81:452–456. https:// doi.org/10.1111/j.1751-1097.2005.tb00207.x
Scott MA, Nguyen VT, Levi B, James AW (2011) Current methods of adipogenic differentiation of mesenchymal stem cells. Stem Cells Dev 20:1793–1804. https://doi.org/10.1089/scd.2011.0040
Sever N, Yang T, Brown MS, Goldstein JL, DeBose-Boyd RA (2003) Accelerated degradation of HMG CoA reductase mediated by bind- ing of insig-1 to its sterol-sensing domain. Mole cell 11:25–33. https://doi.org/10.1016/S1097-2765(02)00822-5
Shin JH, Shin DW, Noh M (2009) Interleukin-17A inhibits adipocyte differentiation in human mesenchymal stem cells and regulates pro-inflammatory responses in adipocytes. Biochem Pharmacol 77:1835–1844. https://doi.org/10.1016/j.bcp.2009.03.008
Uter W, Goncalo M, Yazar K, Kratz EM, Mildau G, Liden C (2014) Coupled exposure to ingredients of cosmetic products: III. Ultra- violet filters. Contact Dermat 71:162–169. https://doi.org/10.1111/ cod.12245
Van Smeden J, Janssens M, Gooris GS, Bouwstra JA (2014) The impor- tant role of stratum corneum lipids for the cutaneous barrier func- tion. BBA-Mol Cell Biol L 1841:295–313. https://doi.org/10.1016/j. bbalip.2013.11.006
Varga T, Czimmerer Z, Nagy L (2011) PPARs are a unique set of fatty acid regulated transcription factors controlling both lipid metabolism and inflammation. BBA-Mol Basis Dis 1812:1007–1022. https://doi. org/10.1016/j.bbadis.2011.02.014
Veiga-Lopez A, Pu Y, Gingrich J, Padmanabhan V (2018) Obesogenic endocrine disrupting chemicals: identifying knowledge gaps. Trends Endocrinol Metab 29:607–625. https://doi.org/10.1016/j. tem.2018.06.003
Wahli W, Michalik L (2012) PPARs at the crossroads of lipid signaling and inflammation. Trends Endocrinol Metab 23:351–363. https:// doi.org/10.1016/j.tem.2012.05.001
Wang SQ, Lim HW (2011) Current status of the sunscreen regulation in the United States: 2011 food and drug administration’s final rule on labeling and effectiveness testing. J Am Acad Dermatol 65:863–869. https://doi.org/10.1016/j.jaad.2011.07.025
Wang SQ, Xu H, Stanfield JW, Osterwalder U, Herzog B (2017) Compar- ison of ultraviolet A light protection standards in the United States and European union through in vitro measurements of commercially available sunscreens. J Am Acad Dermatol 77:42–47. https://doi. org/10.1016/j.jaad.2017.01.017
Yu J, Ahn S, Kim HJ, Lee M, Ahn S, Kim J, Jin SH, Lee E, Kim G, Cheong JH, Jacobson KA, Jeong LS, Noh M (2017) Polypharmacol- ogy of N 6-(3-Iodobenzyl) adenosine-5′-N-methyluronamide (IB- MECA) and related A3 adenosine receptor ligands: peroxisome proliferator activated receptor (PPAR) γ partial agonist and PPARδ antagonist activity suggests their antidiabetic potential. J Med Chem 60:7459–7475. https://doi.org/10.1021/acs.jmedchem.7b00805

Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.