Skin colonization with S. aureus may be a crucial factor in AD pathogenesis. To identify the aberrantly expressed miRNAs in keratinocytes by S. aureus colonization, we established miRNA expression profiles for keratinocytes stimulated with heat-killed S. aureus (HKSA). The significance analysis of microarrays algorithm was used to analyze the miRNA profiling data. We identified 9 significantly differentially expressed miRNAs (fold change >1.5, p-value <0.05) between HKSA-stimulated keratinocytes and controls ( Figure 1A ). Interestingly, hsa-miR-939–5p was the top upregulated miRNA in HKSA-stimulated keratinocytes ( Figure 1A ). To validate our profiling data, we measured miR-939 expression in keratinocytes stimulated with different doses of HKSA by qRT-PCR. HKSA significantly induced miR-939 expression in keratinocytes in a dose-dependent manner ( Figure 1B ). To study the effects of other Staphylococcal strains on the expression of miR-939, we used heat killed Staphylococcal epidermidis (HKSE) and heat killed Staphylococcal hominis (HKSH) to treat keratinocytes. We found neither HKSE nor HKSH had any effect on miR-939 expression ( Figure 1C ). To examine the expression pattern of miR-939 in AD, RNA fluorescence in situ hybridization (FISH) was carried out with miR-939-specific probes on skin lesion sections obtained from 3 AD patients and 3 healthy individuals. Consistent with the upregulation of miR-939 in HKSA-stimulated keratinocytes, miR-939 expression was significantly increased in the epidermis of the AD samples ( Figure 1D ).

Pathogens secrete pathogen-associated molecular patterns to activate pattern recognition receptors on host cells to activate downstream signaling cascades (20). To determine which receptor in keratinocytes is required for S. aureus-induced miR-939 expression, we treated keratinocytes with different TLR ligands, including flagellin (TLR5 ligand), LPS (TLR4 ligand), poly(I: C) (TLR3 ligand), Zymosan (TLR2 ligand), and PamCSK4 (TLR2 ligand) (21, 22) ( Figure 1E ). Pam3CSK4 significantly induced miR-939 expression in a dose-dependent manner ( Figure 1F ). Taken together, these results likely indicate that S. aureus activates TLR2 to induce miR-939 expression in AD.

To study the functional relevance of miR-939 expression under inflammatory conditions induced by HKSA in keratinocytes, we performed a global transcriptomic analysis in keratinocytes with miR-939 overexpression followed by HKSA stimulation. Principal Component Analysis (PCA) indicated that the clustering of samples was categorized into 4 distinct groups designated miRCtrl, miR-939, miRCtrl+HKSA, and miR-939+HKSA, indicating a clear difference in the transcriptome ( Figure 2A ). Subsequently, a two-by-two differential expression analysis of the four groups was performed to evaluate the differentially expressed genes (DEGs) regulated by miR-939 or HKSA in keratinocytes ( Figures 2B, C ; Supplementary Figures S1A, B ). Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis revealed that cytokine-cytokine receptor interaction, TNF signaling pathway, and IL-17 signaling pathway were enriched in the HKSA-stimulated groups ( Supplementary Figures S1C, D ), which indicates the successful induction of an inflammatory response in keratinocytes by HKSA. Interestingly, these inflammation-related pathways were further enriched in upregulated genes by miR-939 overexpression; however, few enriched pathways were associated with the downregulated genes ( Figures 2D–G ). Therefore, we focused on 15 genes associated with inflammation-related signaling pathways upregulated by miR-939, most of which were upregulated by both miR-939 and HKSA ( Figure 2H ). These results indicate that miR-939 performs a regulatory role in amplifying the HKSA-induced inflammatory response in keratinocytes.

To determine the potential interactions among 15 key genes, we constructed a functional protein association network using STRING network analyses ( Figure 3A ). Moreover, the plugin MCODE identified the four most important hub genes, including three matrix metalloproteinases (MMP1, MMP3, and MMP9) and one cell surface adhesion receptor gene, ICAM1 ( Figure 3B ), which regulates leukocyte recruitment from the circulation (23). To validate the results obtained from the above bioinformatics analysis, we treated keratinocytes with HKSA at different times. The qRT-PCR results confirmed that the expression of MMP1, MMP3, MMP9, and ICAM1 was significantly upregulated by HKSA in a time-dependent manner ( Figures 3C–F ). Moreover, overexpression of the miR-939 significantly amplified the expression of MMP1, MMP3, MMP9, and ICAM1 following an 8-hour treatment of keratinocytes with HKSA ( Figures 3G–J ). To confirm these results, we further performed immunostaining of MMP1 and MMP9 after miR-939 transfection followed by HKSA stimulation. We also found that miR-939 significantly upregulate HKSA-induced MMP1 and MMP9 protein expression ( Figures 3K–M ). To determine whether the HKSA-induced endogenous miR-939 can act as positive feedback in keratinocytes, we transfected the keratinocytes with miR-939 blocking oligonucleotides that hybridize to mature miRNAs (inhibitors) followed by HKSA stimulation. Inhibition of miR-939 is not able to change the expression of MMPs at basal level. This may be due to the low expression level of miR-939 in psychological conditions. However, miR-939 inhibitor significantly decreased HKSA-induced MMP1, MMP3 and MMP9 expression, demonstrating that HKSA-induced miR-939 expression acts as positive feedback in keratinocytes ( Figures 3N–Q ). Together, these data suggest that miR-939 enhances the HKSA-induced inflammatory responses in human primary keratinocytes.

We further explored the physiological relevance of miR-939 in human keratinocytes and AD mice. We mixed either the miR-939 agomir or the control agomir with in vivo-jetPEI transfection reagent and administered it through intradermal injection into the mice’s skin followed by topical application of S. aureus ( Figure 4A ). Consistent with a previous study (9), S. aureus colonization was successfully established as an AD-like phenotype in mouse back skin ( Figures 4B, C ). The expression of AD-related cytokines, e.g. IL-13, TSLP and IL-6, were increased, while the antimicrobial peptide expression, e.g. mBD4, mBD14 and CAMP were deceased in S. aureus colonized skin ( Supplementary Figure S2 ). Skin redness and epidermal thickness were significantly increased in the S. aureus colonization group ( Figures 4B, D ). Moreover, we also observed increased neutrophil infiltration in the skin dermis by S. aureus ( Figures 4B, E ). After injecting miR-939, S. aureus had a greater ability to colonize the skin of mice compared with the control mice ( Figure 4F ). This was evident by increased skin redness, epidermal thickness, and neutrophil infiltration in the miR-939+S.aureus group ( Figures 4B–F ). The increased S. aureus colonization and AD phenotype may be attributed, in part, to upregulated MMP1 and MMP9 expression in the skin ( Figures 4G–J ). Together, these data demonstrate that miR-939 amplifies the S. aureus-induced AD phenotype in vivo.

Three AD and three normal healthy samples were obtained from biobank of Institute of Dermatology, Chinese Academy of Medical Sciences, Jiangsu Biobank of Clinical Resources (BM2015004). All patients had signed informed consent for donating their samples. The study was approved by the Research Ethics Committee of the Hospital of Dermatology, Chinese Academy of Medical Sciences & Peking Union Medical College (2023-KY-033).

Seven-week-old male BALB/c mice were purchased from GemPharmatech (Nanjing, China) and acclimatized for one week. The mice were depilated on their backs and allowed to rest for one day to restore the skin barrier. The hsa-miR-939–5p agomir and corresponding negative control (GenePharma, Suzhou, China) were transfected into the mouse skin using in vivo jetPEI via intradermal injection, which is an in vivo transfection reagent (Polyplus), three times per week according to the manufacturer’s protocol. During this period, the mice were externally colonized with S. aureus by placing ten million CFU of S. aureus onto sterile gauze, which was then immobilized on the dorsal skin of the mice with a transparent bio-occlusive dressing (Tegaderm; 3M). On the 8th day, the mice were photographed and euthanized. The skin was collected for various analyses, including qRT-PCR, H&E staining, immunofluorescence, and S. aureus enumeration. A disease score was determined by the cumulative scores of erythema, scale, edema and lichenosis in skin inflammation, with each being designated as none (0), mild (1), moderate (2), or severe (3). The animal studies were conducted under approved protocols by the Animal Use and Care Committee of the Hospital of Dermatology, Chinese Academy of Medical Sciences & Peking Union Medical College (2023-DW-010).

S. aureus (USA 300 LAC) was a gift from Professor Yuping Lai (East China Normal University). It was cultured with TSB (Solarbio, Beijing, China) in a shaker at 37°C and 220 rpm. The mixed S. aureus bacterial solution was diluted in a 1:10 concentration gradient. Different concentrations of bacterial solution as well as TSB were added to 96-well enzyme-labeled plates at 200 µl per well and the absorbance was read at 600 nm using a microplate reader (BioTek, Synergy™ H1). In addition, 10 µl of different concentrations of the bacterial solution was inoculated into the TSA (Solarbio, Beijing, China) plate and incubated at 37°C for 24 h to count CFUs. Finally, a standard growth curve of S. aureus was plotted based on the OD600 values and CFU counts for the different concentrations of S. aureus. To measure the concentration of S. aureus, one CFU was placed in 10 ml of a TSB in a shaker at 37°C and 220 rpm for 24 h. The next day, 10 µl of TSB was collected and incubated in 5 ml of fresh TSB for 3–4 h. The absorbance was measured and the concentration of S. aureus was calculated from the standard growth curve. To determine the number of S. aureus in colonized skin, skin of the same size was collected, homogenized in cold PBS, diluted, and placed onto TSA plates. The number of CFUs was determined after incubating at 37°C for 24 h.

Frozen mouse skin was homogenized with a Tissuelyser (JingXin, Shanghai, China). Total RNA was extracted from tissue homogenates and cells using RNAiso Plus (Takara). RNA quality and concentration were determined with a NanoDrop™ One/OneC (Thermo Fisher Scientific). Complementary DNA (cDNA) was synthesized using PrimeScript™ RT Master Mix (Takara) based on the manufacturer’s instructions. qRT-PCR was performed using TB Green® Premix Ex Taq™ II (Takara). For miRNA quantification, Bulge-loop miRNA qRT-PCR Primer Sets (one RT primer and a pair of qPCR primers for each set) specific for hsa-miR-939–5p were designed by RiboBio (Guangzhou, China). The expression of mRNA or miRNA was normalized to Gapdh or U6, respectively. The relative mRNA/miRNA expression was calculated using the 2^−ΔΔCt method.

FISH was done using a miRNA FISH Kit with FISH probes (GenePharma, Suzhou, China). Briefly, paraffin sections were deparaffinized, digested with proteinase K, denatured at 73°C, hybridized overnight at 37°C, washed, and stained for nuclei. All images were collected at a magnification of 400X using a fluorescence confocal microscope (Olympus, BX53, Tokyo, Japan). DAPI (Invitrogen, Thermo Fisher Scientific) and streptavidin Cy3 channels were used for signal detection.

Mouse skin tissues were fixed in formalin, embedded in paraffin, sectioned, and stained with hematoxylin and eosin. To calculate the average epidermal thickness of the mouse sections, we randomly selected five epidermal area on one image to measure the thickness by using ImageJ software. We totally analyzed four images from four mice. For immunofluorescence, fluorescence detection of neutrophil surface marker proteins in paraffin sections of mouse dorsal skin was done using an anti-Ly-6G/Ly-6C antibody (mouse monoclonal antibody; Thermo Fisher Scientific), followed by an anti-mouse AF488-conjugated secondary antibody (Invitrogen A32723). The fluorescent signal was observed using a fluorescence confocal microscope.

NHEKs (Cat: FC0025, Lot 09213, Lifeline Cell Technology) were cultured in DermaLife Basal Medium (Lifeline Cell Technology) supplemented with DermaLife K lifefactor kit and 1× antibiotics [penicillin (100 U/ml), streptomycin (100 U/ml); Thermo Fisher Scientific] at 37°C and 5% CO₂. This donor is a 56-year-old Caucasian male. To study the effect of TLRs ligands on miR-939 expression, Flagellin (0.5ug/ml, InvivoGen), LPS (5ug/ml, InvivoGen), poly(I:C) (300ng/ml, InvivoGen), Zymosan (50ug/ml, InvivoGen) and Pam3CSK4 (10ug/ml, InvivoGen) was used to treat keratinocyte for 24 hours. The expression of miR-939 was measured by qRT-PCR. Before stimulating keratinocytes with HKSA (Thermo Fisher Scientific) or TLR ligands (Thermo Fisher Scientific), 20 nmol/L of hsa-miR‐939 mirVana miRNA mimic or mirVana miRNA mimic negative control #1 (Invitrogen, Thermo Fisher Scientific) was transfected into keratinocytes using Lipofectamine RNAiMAX (Thermo Fisher Scientific) when the confluence reached 50% to 60%. The final concentration of HKSA was 1 x 10⁸ CFU/ml for RNA-sequencing and in vitro studies, whereas the final concentration of various TLR ligands was based on the manufacturer’s instructions. NHEKs were only used for experiments between passages three and five.

To determine the relationship between the samples, we performed PCA using the R packages “factoextra” and “FactoMineR”, based on the normalized FPKM values for each sample. Raw Count values were used to perform a DEG analysis of the RNA-sequencing data using the R package “Deseq2”. Briefly. the first step was to construct a DESeqDataSet object using the function DESeqDataSetFromMatrix. Then, the function DESeq was used for the difference analysis. To acquire DEGs regulated by HKSA or miR-939, genes with parameters of |logFC| > 0.5 and P-value <0.05 were considered significantly changed. The package “ggplot2” was used to generate a volcano map. We used an online tool (https://maayanlab.cloud/Enrichr/) for KEGG and Elsevier enrichment analysis of the DEGs, and the top five enriched pathways are shown as bar graphs. The heatmap was generated with DEGs from inflammation-related signaling pathways upregulated by miR-939 using the “ComplexHeatmap” package. Protein–protein interactions were determined using STRING (https://cn.string-db.org/) and visualized by Cytoscape (version 3.9.0). Four hub genes were identified using the plugin MCODE for Cytoscape.

RNA-sequencing was done using HKSA-stimulated keratinocytes transfected with hsa-miR-939–5p mimic or the corresponding negative control. Briefly, total RNA was extracted and RNA quality was assessed. Then, mRNA was enriched using Oligo(dT), fragmented, and a cDNA Library was constructed using the NEB Next Ultra RNA Library Prep Kit for Illumina (NEB #7530, New England Biolabs, Ipswich, MA, USA). The resulting cDNA library was sequenced using an Illumina Novaseq6000 by Gene Denovo Biotechnology Co. (Guangzhou, China). The reads were subjected to quality control, sequence alignment, and gene abundance quantification by the FPKM and Count methods. The FPKM method eliminates the effects of varying gene lengths and sequencing data volume on gene expression calculations. Thus, the calculated gene expression may be used directly for subsequent bioinformatics analyses.

Statistical analysis was performed with GraphPad Prism version 8.0 (GraphPad). All data were presented as means ± SEM. Statistical significance between groups was determined using either a two-tailed Student’s t-test or ANOVA analysis, by using GraphPad Prism 8 (GraphPad software Inc, California, USA). P values <0.05 were considered statistically significant.