Transcriptomic Analyses Predict Enhanced Metabolic Activity and Therapeutic Potential of mTOR Inhibitors in Acne-Prone Skin

Abstract

Current acne therapies center on preventing new lesions in patients with acne. These therapies were historically found to be beneficial yet were chosen without knowledge of the specific changes in the skin that favor lesion development. A major challenge in developing new treatments is the incomplete understanding of nonlesional (NL), acne-prone skin's molecular characteristics. To address this, we compared RNA-sequencing data from NL skin of 49 patients with acne (denoted as NL acne [NLA]) with those from 19 healthy controls with no acne history. We found 77 differentially expressed genes in NLA (log fold change > 1; P < .05), including genes associated with innate immunity and epidermal barrier function. Notably, K RT 6C, K RT 16, S100A8, S100A9, and lactotransferrin were upregulated, and LCE4A, LCE6A, and CTSE were downregulated. Gene set enrichment analysis revealed that metabolic pathways were enriched in NLA skin, whereas keratinization was negatively enriched. To identify compounds that could shift the gene expression signature of NLA skin toward healthy control skin, we performed connectivity mapping with the Library of Integrated Network-Based Signatures. We identified 187 compounds, particularly mTOR inhibitors, that could potentially normalize the gene expression profile of acne-prone skin to that of healthy skin. Our findings indicate that NLA skin has distinct differences in epidermal differentiation, cellular metabolism, and innate immunity that may promote lesion formation and suggest that mTOR inhibitors could restore NLA skin toward a healthier state, potentially reversing the predisposition to lesion development.

Keywords: Acne; Metabolism; Nonlesional; Transcriptomics; mTOR.

Figure 1

Summarized demographic information of study participants and schematic of study design. (a) Whole skin punch biopsies (4 mm) were collected from the upper back of 49 individuals with acne (NLA) and 19 controls with no history of acne (HC). Biopsies from the NLA group were collected from nonlesional skin that was at least 5 cm away from any active lesion. The table shows the number of male and female participants, mean age, and mean acne Investigator’s Global Assessment score. (b) After RNA isolation from the biopsies, RNA-seq and subsequent analyses were performed. These included differential gene expression analysis, gene set enrichment analysis, TF activity prediction, and connectivity mapping to identify factors that distinguish NLA from HC skin and potential acne therapeutics. HC, healthy control; NLA, nonlesional acne; RNA-seq, RNA-sequencing; TF, transcription factor.

Figure 2

Differential gene expression analysis between individuals with acne (NLA) and HCs. (a) Plot showing the first 2 components from PC analysis colored by skin type: HC (blue) and NLA (red). (b) A volcano plot showing differentially expressed genes between NLA and HC, with the log₂ FC on the x-axis and the –log₁₀(adjusted P-value) on the y-axis. Red- and blue-shaded regions highlight genes that are the most significantly differentially expressed (absolute log FC > 1, adjusted P < .05). (c) Heatmap showing the Z-scored normalized counts for the most significantly differentially expressed genes (absolute log FC > 1, adjusted P < .05). Hierarchical clustering was performed on both the columns (patients) and rows (genes). Expr, expression level; HC, healthy control; log FC, log fold change; NLA, nonlesional acne; PC, principal component.

Figure 3

GSEA. (a) GSEA was performed using the Hallmark collection of gene sets from the molecular signatures database. Plot shows the significantly enriched (adjusted P < .01) gene sets ordered by their NES. Points are colored by –log₁₀ (adjusted P-value). (b) Significantly enriched Reactome gene sets (adjusted P < .01). (c) Individual enrichment plots for mTORC1 signaling, fatty acid metabolism, cholesterol homeostasis, and keratinization gene sets. GSEA, gene set enrichment analysis; NES, normalized enrichment score.

Figure 4

TF activity prediction suggests significant activation of SREBF1 and SREBF2 in NLA skin. (a) All TFs with predicted differential activity in NLA skin relative to those in healthy skin using the univariate linear model method from the decoupleR package (adjusted P < .01). (b) Volcano plots showing patterns of expression for SREBF2/SREBF1-regulated genes. Colors show the predicted effect of each TF–gene interaction from CollecTRI, with orange and blue indicating activation and repression of gene expression, respectively. NLA, nonlesional acne; TF, transcription factor.

Figure 5

Connectivity mapping identified mTOR inhibitors as potential therapeutic candidates to reverse acne skin signature. (a) Workflow for compound prioritization: the similarity of the acne-prone disease signature (genes differentially expressed between NLA and HC) and LINCS signatures for chemical perturbagens with an annotated mechanism of action was calculated. There were 187 compounds that could potentially shift NLA gene expression toward HC (reversing; adjusted P < .01, normalized connectivity score < –1) and 178 compounds that could potentially shift HC gene expression toward NLA (mimicking; adjusted P < .01, normalized connectivity score > 1). Additional enrichment analysis was performed to identify over-represented drug classes among reversing and mimicking compounds. (b, c) Plots showing the top 10 reversing and mimicking compounds. The y-axes show the unique perturbagen identifier with the common name in parentheses. Mechanisms of action are annotated on the plot. (d, e) Significantly enriched compound targets and drug classes (adjusted P < .01). (f) Comparative Hallmark GSEA results for this study (NLA) alongside those from public skin datasets GSE147950 (Rapa treated vs control mice) and GSE120783 (CBP treated vs untreated human skin). These public datasets are examples of a top reversing compound (mTOR inhibitor: rapa) or mimicking (glucocorticoid receptor agonist: CBP) compound. GSEA was performed separately for each indicted comparison. ∗P < .05, ∗∗P < .01, and ∗∗∗P < .001. CBP, clobetasol propionate; GSEA, gene set enrichment analysis; HC, healthy control; LINCS, Library of Integrated Network-based Cellular Signatures; NCSct, normalized connectivity score summarized across cell types; NES, normalized enrichment score; NLA, nonlesional acne; Rapa, rapamycin.

Figure 6

Diagram summarizing key findings and predicted follicular regions impacted.

Figure 7

Transcriptomic analyses of public skin datasets GSE147950 and GSE120783. (a) PCA plots (PC1–PC4) color coded by groups. Validation groups for connectivity mapping include Rapa-treated and Con mice from GSE147950, along with human skin samples from topical CBP or Unt controls from GSE120783. (b) No genes were differentially expressed between Rapa and Con mouse skin (left volcano plot). A total of 7454 genes were differentially expressed between topical CBP and Unt skin (log₂ fold change > 0; adjusted P < .05). (c, d) GSEA using Hallmark gene sets from the molecular signatures database revealed 20 and 21 significantly enriched gene sets for the Rapa and Con and CBP and Unt comparisons, respectively (adjusted P < .05). CBP, clobetasol propionate treatment; Con, control; GSEA, gene set enrichment analysis; NS, not significant; PC, principal component; PCA, principal component analysis; Rapa, rapamycin; Unt, untreated.