The SZ95 immortalized human sebaceous gland cell line [33] was cultured at 37°C in a humidified atmosphere containing 5% (v/v) CO₂ in Sebomed Basal Medium® (Biochrom, Cambridge, UK) supplemented with 10% fetal bovine serum ([FBS], Biowest, Nuaillé, France), 1 mM CaCl₂, 1% penicillin/streptomycin (Sigma-Aldrich, St. Louis, MO, USA) and 5 ng/ml epidermal growth factor ([EGF], Sigma-Aldrich). When reaching approximately 80% confluence, the medium was replaced with Sebomed Basal Medium® containing 0.5% FBS, 1 mM CaCl₂, 1% penicillin/streptomycin, lacking EGF for 24 hours, followed by treatments with 1 μg/ml PAM3CSK4 (TLR1/2 activator; dissolved in sterile water; Cat. no.: TLRL PMS, InvivoGen, San Diego, CA, USA), 1 μg/ml LPS (TLR4 activator; derived from Escherichia coli; dissolved in sterile water; Cat. no.: L4391, Sigma-Aldrich) or 50 μM arachidonic acid (dissolved in 98% ethanol; Cat. no.: A3611, Sigma-Aldrich).

In order to determine if the TLR1/2 and 4 activators (PAM3CSK4 and LPS) were efficient in the treated samples, levels of IL-6 and CXCL-8 were measured (S1 Fig) with DuoSet ELISA Development Kit (R&D Systems, Minneapolis, MN, USA) according to the manufacturers’ instructions. SZ95 sebocyte supernatants were collected at 24 hours after PAM3CSK4 and LPS treatments, and were aliquoted and stored at –20°C until further analyses. Samples were measured in triplicates for each cytokine. One-way ANOVA and Dunnett post-hoc test were used in the analyses of ELISA data.

SZ95 sebocytes were cultured in the presence of TLR1/2 and 4 ligands (PAM3CSK4 and LPS, respectively) or vehicle for 6 and 24 hours as described previously. Total RNA was isolated using TRI Reagent (MRC, Cincinnati, OH, USA) according to the manufacturer’s protocol and quantified by using NanoDrop 2000 (Thermo Fisher Scientific, Walthman, MA, USA).

For RNA sequencing (RNA-Seq) libraries were generated from 1 μg total RNA using TruSeq RNA Sample Preparation Kit (Illumina, San Diego, CA, USA) according to the manufacturer’s protocol. Briefly, poly-A tailed RNAs were purified by oligodT-conjugated magnetic beads and fragmented on 94°C for 8 min, then 1st strand cDNA was transcribed using random primers and SuperScript II reverse transcriptase (Life Technologies, Carlsbad, CA, USA). Following this step second strand cDNA was synthesized, the double-stranded cDNA was end-repaired, 3’ ends were adenylated then Illumina index adapters were ligated. After adapter ligation enrichment PCR was performed to amplify adapter ligated cDNA fragments. Fragment size distribution and molarity of libraries were checked on Agilent BioAnalyzer DNA1000 chip (Agilent Technologies, Santa Clara, CA, USA). Concentrations of RNA-Seq libraries were diluted to 10 nM and 5 libraries were pooled together before sequencing. Single read 50 bp sequencing run was performed on Illumina HiScan SQ instrument (Illumina). Each library pool was sequenced in one lane of sequencing flow cell, 16–18 million reads per sample was obtained.

CASAVA software (Illumina) was used for pass filtering and demultiplexing process. Sequenced reads were aligned to Human Genome v19 using TopHat and Cufflinks algorithms and bam files were generated. Further statistical analyses were executed using NGS modul of GeneSpring 12.6 software (Agilent Technologies). Relative mRNA expression levels were calculated with DESeq algorithm. Aligned sequencing data have been deposited into the NCBI SRA database under accession no.: SRP126626.

Cells were washed in PBS and lysed in RIPA buffer containing a protease inhibitor mix (aprotinin, leupeptin, pepstatin, bestatin) (Sigma Aldrich). Proteins (20 μg) were separated by electrophoresis using appropriate polyacrylamide gel and transferred to nitrocellulose membrane (Bio-Rad, Hercules, CA, USA). After blocking, membranes were probed with anti-SAA1+SAA2 (Cat.no.: ab207445; Abcam, Cambridge, UK) antibody. The Ag–Ab complexes were labeled with appropriate HRP-conjugated secondary antibody (Bio-Rad) and visualized by WesternBright ECL HRP substrate (Advansta, CA, USA).

Anonymized formalin-fixed and paraffin embedded (FFPE) sections of human skin from the tissue archive of the Department of Dermatology, University of Debrecen were acquired after the approval of the Regional and Institutional Ethics Committee, University of Debrecen. At least 5 different FFPE samples of each condition (healthy skin, papulopustular acne, papulopustular rosacea) were evaluated. Samples were collected from the back of acne, from the face of rosacea, and from matching skin areas of healthy individuals [34] who underwent surgery with histologically later confirmed benign melanocytic lesions.

Slides from FFPE samples were deparaffinized and rehydrated. Endogenous peroxidase activity was blocked by treatment with 3% (v/v) H₂O₂ in distilled water for 15 min. Heat mediated antigen retrieval was performed with Tris-EDTA buffer (Tris Base, 1mM EDTA solution, 0.05% Tween 20, pH 9). Nonspecific binding was blocked with 1% bovine serum albumin (BSA) in PBS for 30 min at room temperature. Anti-SAA1+SAA2 (Cat. no.: ab 207445; Abcam) antibody was diluted 1:200 in PBS and sections were incubated for 1 h at room temperature in humidity chambers. For negative controls, the appropriate non-immune control sera (rabbit IgG; Vector Laboratories, Burlingame, CA, USA) were used instead of the primary antibody. HRP-conjugated secondary antibody was used in accordance with the manufacturer’s instructions (SuperSensitive One-step Polymer-HRP Detection System, BioGenex Laboratories, Fremont, CA, USA). Immunoreaction was visualized by Vector VIP Kit (Vector Labs.) Sections were counterstained with methylene green. Images were acquired with a Leica DM2000 LED microscope (Leica Microsystems, Wetzlar, Germany).

SZ95 sebocytes were cultured on coverslips in the presence of TLR1/2 and 4 ligands (PAM3CSK4 and LPS, respectively), arachidonic acid or vehicle for 24 and 48 hours as described previously. Cells were fixed in 4% paraformaldehyde (Sigma-Aldrich) for 10 min. Coverslips were placed in 100% propylene glycol (Amresco, Solon, OH, USA) for 1 min, followed by a rinse with distilled water. Cells were stained with 0.7% Oil Red O (Sigma-Aldrich) solution for 7 minutes then rinsed with 85% propylene glycol solutions. Nuclei were counterstained with hematoxylin and slides were covered using Mount Quick Aqueous mounting medium (Bio Optica, Milano, Italy).

To identify the transcriptional changes in sebocytes upon activation of TLR1/2 and TLR4 pathways, we treated SZ95 sebocytes with the specific, selective activators PAM3CSK4 (TLR1/2 pathway activator) and LPS (TLR4 pathway activator). By choosing a 6-hour time point for our studies we focused on the initial changes, while the 24-hour time point allowed us to address the ability of sebocytes to maintain their altered gene expression profile and thus their commitment towards inflammation (Fig 1A). Identifying the transcriptomes that reached a significance in all biological triplicates at a given time point clearly fulfilled our criteria to define genes with a statistically relevant altered expression.

As a first step, we focused on the changes at 6 hours and found a significant expression change of approximately 500 genes in response to the administered TLR1/2 and TLR4 activators (505 in response to PAM3CSK4 and 437 in response to LPS). Next, we determined the changes at 24 hours and found a further increase in the number of the differentially expressed genes (611 in response to PAM3CSK4 and 558 in response to LPS). Interestingly, 43.3% (in case of PAM3CSK4) and 60% (in case of LPS) of the genes from the 6-hour time point were still having significantly changed expression levels at 24 hours (Fig 1B).

To assess the changes that were found at 6 but not at 24 hours we performed functional clustering of the relevant genes also presented in a heat map form (S2A Fig). Interestingly, we could not confirm a clear signature or function for these genes (S2B Fig) that could mark a biologically relevant temporary effect of the applied stimuli.

These results altogether showed that the response of sebocytes to the used TLR activators is not a transient state but induces long term changes in the genetic programs.

To detect how close the two signaling pathways are in their effects on differential gene expression profiles, we compared the gene lists of both conditions of TLR-activated sebocytes from the relevant time points. We found that the differentially expressed genes showed a prominent overlap (320 out of 437 [LPS] and 505 [PAM3CSK4]) at 6 hours, while at 24 hours the gene expression profile was almost identical (over 80% in case of PAM3CSK4 and 90% in case of LPS) (Fig 2A).

Hierarchical clustering of the genes that were expressed and changed between any conditions in unstimulated, TLR1/2- and TLR4-activated sebocytes exhibited changes mostly due to up-regulation, as only approximately 20% of the genes were down-regulated. This prominent induction in the number of regulated genes suggests that sebocytes respond to the activation of the TLR 1/2 and TLR4 pathways by adding a new profile to their transcriptome rather than losing the one from their untreated state. This was further supported by pathway analysis of the down-regulated genes that could not sort them into any functional category.

The generated heat map also showed that of the genes reaching the level of significance at 24 but not at 6 hours, a very high number were already showing tendency of up-regulation at 6 hours, supporting that most of the changes observed at 24 hours occurred directly through the applied TLR stimuli and that sebocytes were already committed at this early time point towards their new profile (Fig 2B).

To assess the biological response and functions that TLR1/2- and TLR4-activated sebocytes could gain at the level of gene expression, we extended our studies with gene clustering and pathway analysis of the altered genes at 6 and at 24 hours. Notably, as it was expected from the previously detailed similarities in the genes regulated by TLR1/2 and TLR4 pathways, the functional analysis also resulted in the same clusters in both treatments. This was confirmed by using two independent softwares (Panther and Cytoscape pathway analysis programs).

Analysis of the altered genes from the 6-hour time point revealed that the most significant changes were related to genes involved in immune system/defense processes underpinning that TLR 1/2 and TLR4 pathways are also fully active in sebocytes and are a potential link between the inflammatory environment and the gene expression profile also in this cell type. The gene clustering provided many interesting clusters, as well as novel ones, to sebocyte research: such as a possible involvement in wound healing, regulation of cell proliferation, chemotaxis and a more complex cytokine and chemokine production than what was previously known besides the widely used detection of IL-6 and CXCL-8.

The analysis of the 24-hour samples found the same clusters as at 6 hours, which was in line with the gene expression data showing that the changes at 6 hours were also consistent in the 24-hour samples. However, the cluster of cholesterol and steroid metabolic processes, which represents the key function of sebocytes to produce lipids, was only affected at the 24-hour time point (Table 1) with a significant up-regulation in the expression of the cluster forming genes but with the exception of ABCG1 that was down-regulated (S1 Table). Importantly, this also provides a possible explanation for our observations that no change was detected in the lipid body formation of TLR1/2- and TLR4-activated sebocytes, which is the morphological hallmark for an altered lipid metabolism (S3 Fig).

Altogether, these findings were surprising, since a cell type with a primary role to produce lipids was not promptly changing its lipid profile on the level of gene expression or lipid body formation. On the other hand, the experiments delivered interesting data on how rapidly sebocytes could add the inflammatory status to their profile and accommodate to a new environment.

The available microarrays from whole tissue biopsies of acne samples (GSE53795) [2], histological analysis and in vitro experiments all suggested a role for the TLR1/2 and TLR4 pathways [9, 23] in the pathogenesis of acne. However, no cell type-specific association with disease development and inflammation was addressed in response to these stimuli. Therefore, as a next step, a meta-analysis was performed using available gene expression data of acne whole tissue samples [2] and the list of the up-regulated genes in TLR-stimulated SZ95 sebocytes to determine possible genes and pathways with which sebocytes could integrate into the TLR2/TLR4-acne cascade.

Using a Venn diagram to visualize the up-regulated genes in the acne samples compared to healthy ones and in the TLR1/2- and TLR4-stimulated SZ95 sebocytes at 24 hours, a possible contribution of sebocytes with 92 of the 900 significantly altered transcripts in acne could be detected (Fig 3B).

To assess how sebocytes may contribute to acne pathogenesis, further analysis was performed using only the overlapping 92 genes (Fig 3A). The genes were grouped into functions such as immune response, defense response, response to wounding, response to stress and chemotaxis (Fig 3C).

Next, we aimed to analyze and describe markers for TLR 1/2- and TLR4-mediated signaling in sebocytes. The candidate genes, which showed the most robust induction upon PAM3CSK4 and LPS treatments, compared to non-stimulated sebocytes at both 6 and 24 hours were selected from the RNAseq results. The list of 207 genes were then visualized in a Venn diagram with the genes that were differentially regulated in acne samples when compared with healthy control (Fig 4A). The overlapping 56 genes were subjected to hierarchical clustering to identify the ones with the highest change of expression in all conditions and to choose the ones whose expression showed a further increase from 6 to 24 hours in the stimulated SZ95 sebocytes (Fig 4B). As a result Serum Amyloid A 1/2 (SAA1/2) was identified to be suitable marker to identify activated sebocytes at the level of gene expression (Fig 4C). Furthermore, correlating with the mRNA data, the protein levels of SAA1/2 were also induced in TLR1/2 and 4 activated SZ95 sebocytes as revealed by Western blotting (Fig 4D).

To assess if the detection of SAA1/2 could also be applied and therefore be routinely used to mark activated sebaceous glands at the protein level in histological specimens, we performed immunohistochemistry using specific antibodies on normal as well as pathological skin samples such as papulopustular acne and papulopustular rosacea. SAA1/2 was successfully detected, with prominent differences in the staining intensities of sebaceous glands in papulopustular acne and papulopustular rosacea samples when compared with healthy skin. Importantly, SAA1/2 positivity had a characteristic distribution within the sebaceous glands in both diseases, localizing exclusively at the basal cell layers, revealing a well-defined functional architecture of sebaceous glands regarding its immune-competence (Fig 5).