Commensal Cutibacterium acnes-derived indolelactic acid safeguards skin barrier function through the aryl hydrocarbon receptor

Summary

Skin microbial dysbiosis is a key characteristic of atopic dermatitis (AD). Although Cutibacterium acnes (C. acnes) is the most abundant bacterium in the skin microbiota, its precise functions in AD remain unclear. Here, whole-genome sequencing of C. acnes isolates reveals strain-level genomic and functional heterogeneity between AD patients and healthy individuals. Metabolomic profiling identifies reduced indolelactic acid (ILA), a tryptophan metabolite, in AD skin and in cultures of AD-associated C. acnes strains. In murine models of epidermal-barrier-damaged and AD-like dermatitis, topical ILA application activates the aryl hydrocarbon receptor (AhR) pathway in keratinocytes, upregulating tight junction proteins and suppressing interleukin (IL)-33 to restore barrier integrity and mitigate dermatitis. A proof-of-concept clinical trial confirms ILA’s efficacy in reducing transepidermal water loss and alleviating AD symptoms. Our study establishes C. acnes-derived ILA as a crucial regulator of skin barrier repair via AhR signaling, offering therapeutic potential for AD. Clinical trial registration: ChiCTR2400090988.

Graphical abstract

Highlights

• •AD skin has lower ILA and harbors C. acnes genomes with reduced ILA production
• •ILA activates AhR in keratinocyte, restoring tight junctions and suppressing IL-33
• •Topical ILA repairs epidermal barrier defects and mitigates dermatitis in AD mice
• •A proof-of-concept clinical trial confirms ILA alleviates AD in humans

Zhang et al. demonstrate that C. acnes isolated from the skin of AD patients have undergone genomic shifts that impair their ability to produce ILA. Topical application of ILA restores epidermal barrier function and mitigates AD symptom by activating AhR in both mice and humans.

Introduction

Atopic dermatitis (AD) is a chronic inflammatory skin disorder characterized by intense itching, recurrent eczematous dermatitis, and xerosis.¹ Skin microbiome dysbiosis, barrier dysfunction, and abnormal immune responses are core features of the pathogenesis of AD.^2,3 However, the role of the skin microbiome in AD, and the underlying mechanisms, are less well understood. In patients with AD, there are alterations in the diversity, composition, and function of the skin microbiota, and microbial diversity is negatively correlated with AD severity.⁴ Numerous studies have demonstrated that the abundance of the genus Staphylococcus is increased in AD, while that of the genus Cutibacterium is decreased.^5,6 Previous research has predominantly examined the pathogenic role of Staphylococcus aureus and regulatory function of Staphylococcus epidermidis in AD, with little attention to the role of Cutibacterium acnes—the most widely distributed bacterium on normal human skin.^6,7,8

Recent studies have highlighted the pivotal role of C. acnes in skin homeostasis.^9,10 For instance, C. acnes has been reported to decrease the surface pH of skin and promote the secretion of antimicrobial peptides,¹¹ both of which are crucial for skin health. The inhibitory effects of C. acnes against the growth and biofilm formation of S. aureus^11,12,13 as well as in conferring host resistance to S. aureus are also significant in this regard.¹⁴ Furthermore, different strains of C. acnes exhibit varying inflammatory modulatory potential and target markers of cutaneous immune responses, thereby differentially regulating innate immunity as well as inducing specific subpopulations of antimicrobial Th17 cells to modify adaptive immunity.^15,16,17,18 Additionally, C. acnes has been demonstrated to enhance the integrity of the skin barrier by inducing epidermal lipid synthesis.¹⁹ However, given the intricate complexity of the interactions between the host and the microbiota within the skin barrier, how C. acnes mediates multiple facets of skin barrier function has not been extensively studied.^20,21 Hair follicles and sebaceous glands are deemed as the main factors that cause variations in skin microbiota distribution across different regions of the skin.²² Patients with AD often exhibit reduced sebaceous gland secretion,²³ which might alter the relative abundance and function of C. acnes. At the strain level, we observed that C. acnes isolates from normal skin and AD lesions showed distinct genomic and functional characteristics, which is attributed to combined influence of disease status, individual factors, and site-specific characteristics.²⁴ Our recent study has demonstrated that dysregulation of the sebum-microbial metabolites-IL-33 axis serves as an initiating factor in inflammation in AD.²⁵ However, the specific mechanisms by which host factors influence the regulatory role of C. acnes in skin homeostasis remain to be elucidated.

Most of the functions of microbes are achieved through their metabolites, and skin microbial metabolites play important roles in skin homeostasis. Attenuated tryptophan (Trp) metabolism and fatty acid metabolism are found in patients with AD.^5,26,27 Our previous research has shown that IAId, a Trp metabolite from skin microbiota, negatively regulates skin inflammation in AD.²⁶ C. acnes is known to metabolize free fatty acids from sebum into short-chain fatty acids (SCFAs), which inhibit pathogenic Staphylococcus spp. while favoring other healthy commensal bacteria.¹² SCFAs from C. acnes also activate both a canonical and an epigenetic inflammatory response in human sebocytes²⁸ and induce the epidermal synthesis of lipids that are important for skin barrier function. Overall, the complex interplay between C. acnes and the host, with significant implications for health and disease, is an area of intense research. The integrity of the skin barrier is the first line of defense in determining skin homeostasis. Although C. acnes colonizes the skin surface, its impact on the skin barrier remains to be fully explored.

In this study, we isolated C. acnes strains from the skin of healthy controls (HCs) and patients with AD to explore the differences in subspecies colonization and metabolism. We found distinct dominant subspecies in HC and patients with AD, with significant functional gene heterogeneity. Metabolic profiling indicated that indolelactic acid (ILA) was downregulated in both in vitro C. acnes culture and on the skin surface of patients with AD. Functional study revealed that ILA repaired epidermal barrier damage and mitigate MC903-induced AD-like dermatitis by upregulating OCLN and CLDN1 and downregulating IL-33 through AhR pathway activation in keratinocytes (KCs). A proof-of-concept clinical trial demonstrated that topical ILA reduced transepidermal water loss (TEWL) and alleviated skin inflammation in patients with AD.

Results

C. acnes from patients with AD exhibits distinct genomic and metabolomic profiles

To investigate the strain-level heterogeneity of C. acnes across distinct skin conditions, two to three C. acnes strains were isolated from the facial skin of each participant. A total of nine HCs and nine patients with AD were enrolled for sample collection (Figure 1A). Following stringent quality control, a final set of 45 high-quality C. acnes isolates from eight HC and seven AD participants were subjected to whole-genome sequencing (WGS). Consistent with our previous studies,²⁴ single-locus sequence typing (SLST) identified different strain subgroups between HC and patients with AD (Figure 1B), with type K1 predominating in HC and type A2 in AD. Thus, WGS data indicated differential colonization of C. acnes subspecies in HC and patients with AD.

Figure 1.

Analysis of C. acnes genome signatures and metabolomic features in HC and AD patients

(A) Schematic diagram of the study design. After quality control, the WGS analytical cohort included 15 individuals (n = 8 HCs, n = 7 AD; 45 C. acnes isolates). Targeted Trp metabolite analysis used one representative strain per participant (n = 9 per group).

(B) SLST type composition of C. acnes strains in HC and AD groups (from the 45 WGS-analyzed isolates).

(C) Enriched KEGG pathways between HC and AD C. acnes isolates (colors indicating p values).

(D) Heatmap showing relative abundance patterns of differential Trp metabolites between HC and AD isolates (colors representing row Z scores).

(E) ILA concentration in C. acnes culture supernatants from HC and AD groups (n = 9 per group). 95% confidence interval (CI): −19.07 to −3.461.

(F) Major Trp metabolites levels on facial skin of HC and AD subjects (n = 30 per group). 95% CI: −13.83 to −0.1075 (IAA), −1.688 to −0.2258 (IEt), −46.94 to −0.8409 (ILA), and −692.0 to −42.01 (NA).

(G) TEWL on facial skin of HC and AD subjects (n = 30 per group). 95% CI: 6.00–10.00.

(H and I) Correlations between ILA levels and SCORAD (H) or TEWL (I) in AD patients. Spearman’s correlation coefficient (ρ), p value and 95% CIs are indicated.

Data are presented as mean ± SD (E). Data are median with interquartile range (IQR) (F and G). Significance: unpaired two-tailed t tests (E), Mann-Whitney test (F and G), and Spearman’s correlation tests (H and I). For (C), (E), and (F), exact p values shown in the figures are nominal (unadjusted).

For multiple comparisons, p values were adjusted using the Benjamini-Hochberg (BH) procedure, and BH-adjusted q values are reported for exploratory purposes in the Table S4. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001,∗∗∗∗p < 0.0001; ns, not significant. See also Figure S1.

Subsequently, to capture the genomic diversity of C. acnes within HC and patients with AD, Roary was used to assemble a comprehensive high-quality pan-genome from total coding sequences. Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis exposed distinct metabolic profiles between the HC and AD groups. HC isolates showed significant enrichment in multiple pathways, including starch and sucrose metabolism, amino sugar and nucleotide sugar metabolism, phosphotransferase system, glycan degradation, phenylalanine metabolism, Trp metabolism, and glycosaminoglycan degradation, among others. AD isolates, by contrast, exhibited a more limited metabolic enrichment, predominantly in starch and sucrose metabolism (Figure 1C).

Consistent with previous reports of attenuated Trp metabolism in the AD skin microbiota,²⁷ we next focused on this way. Pan-genome analysis revealed that HC-enriched isolates harbor genes encoding monoamine oxidase (aoFH) (responsible for converting tryptamine to indole-3-acetaldehyde), amidase (amiE) (responsible for converting the (indol-3-yl) acetamide to indole-3-acetate), and tryptophanase (tnaA) (responsible for converting the tryptophan into pyruvate), with two distinct genes mapping to amiE, underscoring the potential of multiple loci in C. acnes isolates from HC to drive Trp metabolism (Figure S1A). To identify the skin commensal microbiota involved in the production of Trp, targeted Trp metabolome analysis was conducted on the culture supernatants of several common skin commensal bacteria, including Staphylococcus hominis, Roseomonas mucosa, C. acnes, and Staphylococcus epidermidis. C. acnes was found to predominantly produce indole (IND) and ILA (Figure S1B). Then, we randomly selected one C. acnes isolate from each individual for targeted Trp metabolome analysis. The result showed that C. acnes isolates from HCs had a higher content of Trp metabolites than those from patients with AD (Figure 1D). Specifically, while the levels of ILA, IND, and 5-hydroxytryptophan (5-HTP) in the culture supernatants of C. acnes isolates all showed significant differences between patients with AD and HC, the reduction in ILA was the most statistically pronounced (Figure 1E; Figure S1C). To account for this difference, we integrated pan-genome profiling and transcriptome sequencing. ILA production correlated positively with the expression levels of aoFH and amiE (Figure S1D), and both genes were enriched at the pan-genome level and were responsible for generating precursors that fed directly into ILA synthesis. In contrast, no significant correlation was found between SLST typing and ILA production (Table S1). Additionally, we discovered that sebum concentration influenced the transcriptional characteristics and biological functions of C. acnes.²⁴ Upon culturing C. acnes strain ATCC6919 under increasing concentrations of artificial sebum (0.01%–0.25%), ILA production exhibited a non-linear pattern, with higher levels observed at 0.05% sebum compared to lower (0.01%) and higher (0.25%) concentrations (Figure S1E). Under the same conditions, C. acnes displayed comparable growth dynamics across sebum concentrations, as assessed by OD600 measurements over 48 h and at the 48-h endpoint (Figure S1F). These observations suggest that variations in sebum concentration may preferentially influence ILA production rather than overall bacterial growth under the conditions tested.

Measurements of Trp metabolites on non-lesional facial skin revealed reduced levels of ILA in patients with AD compared with HC, as depicted in Figure 1F. ILA levels were negatively correlated with disease severity, as assessed by scoring atopic dermatitis (SCORAD), and with TEWL in patients with AD (Figures 1G–1I). In contrast, no significant correlation was observed between ILA levels and stratum corneum hydration (SCH) (Figure S1G). Collectively, these findings suggest that a deficiency of C. acnes-derived ILA may contribute to the pathogenesis of AD.

ILA increases the expression of tight junctions in KCs

Given the pivotal role of KCs in the innate immune defense of the skin,^29,30 we investigated the effects of ILA on KCs using RNA sequencing (RNA-seq). Upon stimulation with 1 mM ILA for 24 h, a total of 126 differentially expressed genes (DEGs) were identified in KCs (32 downregulated and 94 upregulated). Gene set variation analysis (GSVA) revealed higher pathway-level activity in KCs incubated with ILA compared with vehicle controls, particularly in pathways related to barrier function, immune signaling, and metabolism, including tight junction (TJ), Trp metabolism, retinol metabolism, and xenobiotic metabolism by cytochrome P450 (Figure S2A). Next, to further examine the DEGs involved in key biological processes of KCs, we manually curated genes related to inflammation, various facets of the epithelial barrier (keratin assembly and differentiation and TJs), and xenobiotic processing genes (Figure 2A). Results showed that inflammation-related IL-24, TJs-related OCLN and CLDN4, and xenobiotic-process-related SULT2B1, CYP1B1, CYP1A1, ALDH7A1, and ALDH1A3 were differentially expressed (p adj <0.05, log2 fold change (FC) ≥ 1). However, the expression level of IL-33 in KCs upon ILA treatment showed a decreasing trend, although without statistical significance.

Figure 2.

ILA enhances tight junction expression in keratinocytes

(A) Heatmap showing log2FC of genes related to xenobiotic processing, epithelial barrier function, and inflammation in KCs (ILA vs. vehicle). Data from paired RNA-seq experiments with n = 3 per group. Asterisks denote BH-FDR-adjusted significance (q < 0.05).

(B) RT-qPCR of CLDN1, CLDN4, and OCLN mRNA expression in KCs treated with ILA for 24 h (n = 6 per group). 95% CI: 0.4019–0.6170 (CLDN1), 0.7372–1.403 (CLDN4), and 0.3599–1.440 (OCLN).

(C) Representative immunoblot images of CLDN1, CLDN4, and OCLN in KCs treated with ILA. Densitometric quantification of three independent experiments is shown without statistical analysis due to limited sample size (n = 3).

(D) Immunofluorescence staining of CLDN1, CLDN4, and OCLN in KCs treated with ILA for 24 h. Boxed regions indicate areas shown at higher magnification. Scale bars: 100 μm (main panels) and 25 μm (insets).

(E) RT-qPCR of CYP1A1 and AhR mRNA expression in KCs treated with ILA or vehicle for 24 h (n = 6 per group). 95% CI: 2.813–6.349 (CYP1A1) and −0.3365 to 0.03487 (AhR).

Data are presented as mean SD. Significance: unpaired two-tailed t tests. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001,∗∗∗∗p < 0.0001; ns, not significant. See also Figure S2.

We further examined the baseline effect of ILA on TJs expression in KCs. RT-qPCR analysis demonstrated that ILA treatment alone significantly upregulated the mRNA expression of CLDN1, CLDN4, and OCLN (Figure 2B). At the translational level, assessment by western blotting and immunofluorescence confocal microscopy visually corroborated these transcriptional changes, displaying a consistent trend toward elevated protein levels of CLDN1, CLDN4, and OCLN (Figures 2C and 2D). Furthermore, under a Th2-cytokine-driven microenvironment (IL-4, IL-13, and tumor necrosis factor alpha [TNF-α]), treatment with ILA was associated with elevated mRNA expression of CLDN1, CLDN4, and OCLN in KCs compared with Th2 stimulation alone (Figure S2B). Collectively, these results indicate a trend toward increased expression of CLDN1, CLDN4, and OCLN in the presence of ILA.

Among these xenobiotic processing genes, the pregnane X receptor, constitutive androstane receptor, peroxisome-proliferator-activated receptor-alpha (PPARa), and aryl hydrocarbon receptor (AhR) are key transcription factors in the skin.³¹ We showed that downstream target genes, such as cytochrome-p450 CYP1A1, and CYP1B1, which respond to AhR activation, were significantly upregulated in KCs treated with ILA (Figures 2A and 2E). However, no significant alteration in AhR expression was observed (Figure 2E).

Considering that AhR can activate multiple signaling pathways and mediate barrier repair and integrity,^32,33 the upregulation of several genes related to TJs and the AhR pathway suggests that ILA modulates epithelial barrier integrity, which is potentially mediated by the activation of AhR.

Topical ILA application fortifies the integrity of the skin barrier

Next, we explored the effects of ILA on epithelial barrier integrity by utilizing mouse models of acute barrier disruption. ILA or EtOH was topically applied to the dorsal skin of C57BL/6 mice for 7 days, followed by tape stripping on the 7th day (Figure 3A). Compared with the EtOH-treated group, mice receiving ILA showed reduced redness, visible scaling, and improved overall skin barrier scores (Figure 3B). In contrast, epidermal thickness, inflammatory cell infiltration, and cutaneous expression levels of Il-33 and Tslp were similar between groups (Figures S3A–S3C), indicating that topical ILA application does not elicit overt inflammatory responses under these conditions. Together, these observations suggest that ILA helps preserve skin barrier integrity without inducing overt inflammatory response. After tape-stripping injury, when TEWL reached approximately 20 g/m²/h, ILA-treated mice demonstrated a faster barrier repair rate over a 24-h period compared with the EtOH-treated controls (Figure 3C). Furthermore, ILA pre-treatment led to a significant upregulation of the expression levels of Cldn1, Cldn4, and Ocln in the dorsal skin of mice compared with that after EtOH treatment (Figure 3D).

Figure 3.

Topical ILA application enhances skin barrier function

(A) Experimental design for assessing the preventive effect of topical ILA application on skin barrier function in wild-type C57BL/6 mice.

(B) Representative gross appearance, H&E staining, and skin barrier scores of dorsal skin following preventive ILA treatment (n = 5 per group). Scale bars: 200 μm. 95% CI: 15.24–19.56 (skin barrier score).

(C) TEWL measured 24 h after skin barrier challenge in EtOH- and ILA-treated mice (n = 5 per group). Statistical analysis was performed at the 24 h time point. 95% CI: −11.37 to −9.105.

(D) RT-qPCR of Cldn1, Cldn4, and Ocln mRNA expression in dorsal skin following preventive ILA treatment (n = 5 per group). 95% CI: 1.555–4.468 (Cldn1), 0.3130–1.776 (Cldn4), and 0.8526–2.090 (Ocln).

(E) Experimental design for assessing the therapeutic effect of topical ILA on non-inflammatory barrier injury in wild-type C57BL/6 mice.

(F) Representative gross appearance and skin barrier score of dorsal skin following therapeutic treatment (n = 5 per group). 95% CI: 3.993–9.607 (control vs. EtOH) and −8.407 to −2.793 (EtOH vs. ILA).

(G) TEWL levels measured 24 h after tape stripping in EtOH and ILA groups (n = 7 per group). 95% CI of the paired difference (after vs. before): −10.86 to 0.8627 (EtOH group) and −11.60 to −3.447 (ILA group).

(H) RT-qPCR of Cldn1, Cldn4, and Ocln mRNA expression in dorsal skin following therapeutic treatment (n = 5 per group).

Data are presented as mean ± SD. Significance: unpaired two-tailed t tests (B, C, and D), one way-ANOVA followed by Tukey’s post hoc tests (F and H), paired two-tailed t tests (G), with 95% CIs (H) reported in Table S5. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001,∗∗∗∗p < 0.0001; ns, not significant. See also Figure S3.

Repeated exposure to desiccating agents and physical damage impair the skin barrier. To further assess the potential treatment effect of ILA on the epithelial barrier, we utilized a mouse model in which acetone/ethanol/water (AEW) exposure was combined with tape stripping, as previously described.^{34,35,36,37,38} After establishment of the model for 3 days, ILA or EtOH was applied topically to the dorsal skin of C57BL/6 mice for 7 days. During this period, the model was maintained concurrently, and TEWL was measured 24 h after the final modeling (Figure 3E). The result showed that ILA treatment significantly improved the skin barrier integrity and reduced water loss (Figures 3F and 3G). Additionally, ILA treatment resulted in upregulation of Cldn1, Cldn4, and Ocln, whereas EtOH vehicle alone selectively elevated Cldn4 but significantly reduced Cldn1 and Oldn expression (Figure 3H). These findings further indicate that ILA had therapeutic effects upon barrier damage caused by the downregulation of TJs.

ILA upregulates the expression of TJs in KCs via the AhR pathway

AhR is best known for mediating the toxic effects of xenobiotic substances, including environmental contaminants such as TCDD (dioxin), and for its role in physiological processes involving a range of endogenous ligands such as indoles and flavonoids of dietary origin.³⁹ ILA, an indole derivative from bacteria, acts as an agonist for AhR.^40,41,42 To further explore the interaction between ILA and AhR, we initially conducted molecular docking studies. Figure 4A depicts the optimal docking conformation of AhR with ILA, with a docking score of −8.7 kcal/mol. The three-dimensional (3D) and two-dimensional (2D) diagrams revealed that ILA forms two hydrogen bonds with Ser346 and one hydrogen bond with His337 of AhR. Additionally, the phenyl ring of ILA engages in two Pi-Pi stacked interactions with Phe295. The 2-pyrrolidinone ring of ILA forms two Pi-Pi T-shaped interactions with Phe351 and His291, and a Pi-Alkyl interaction with Val381. ILA also undergoes van der Waals interactions with non-bonded amino acids such as Leu353, Gly321, Ile325, Ala367, Ile349, Met340, Ser336, Cys333, and Ser365 at the active site of AhR. We empirically validated this in silico prediction using surface plasmon resonance (SPR), which confirmed a direct, dose-dependent physical interaction between ILA and AhR, with an equilibrium dissociation constant (K_D) of 4.23 μM (Figure S4A). Furthermore, using a HepG2-Lucia reporter cell line, we verified that ILA functions as a potent activator of the AhR pathway, eliciting a ∼26-fold induction of AhR-driven luciferase activity relative to the endotoxin-free control at 24 h (Figure S4B).

Figure 4.

ILA increases tight junction expression via AhR signaling in KCs

(A) Structural modeling and predicted ILA binding site of AhR, shown in three-dimensional (3D) and two-dimensional (2D) representations.

(B) Representative confocal microscopy images of AhR subcellular localization in KCs following ILA treatment at the indicated time points. Scale bars: 100 μm.

(C) Representative immunoblot images showing CLDN1 and OCLN protein expression in AhR-specific siRNA KCs and treated with ILA. Quantification of three independent experiments is shown without statistical analysis (n = 3).

(D) ChIP assay showing AhR occupancy at CLDN1 and OCLN promoters in KCs following ILA treatment (n = 3 independent experiments). Given the limited sample size, no statistical analysis was performed.

(E) Representative gross appearance of dorsal skin from K14-CreAhr^fl/fl and Ahr^fl/fl mice following topical ILA treatment for seven consecutive days.

(F) RT-qPCR of Cldn1 and Ocln mRNA expression in dorsal skin from K14-CreAhr^fl/fl and Ahr^fl/fl mice following topical ILA treatment (n = 5 per group).

(G) TEWL measured on day 7 following topical treatment in K14-CreAhr^fl/fl and Ahr^fl/fl mice (n = 5–6 per group). Data are presented as mean ± SD.

Significance: two-way ANOVA followed by Sidak’s multiple-comparisons test (F and G). Detailed 95% CIs are in Table S5. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001,∗∗∗∗p < 0.0001; ns, not significant. See also Figure S4.

Consistent with this receptor activation, utilizing fluorescence confocal microscopy, we found that ILA treatment facilitated the translocation of AhR from the cytoplasm to the nucleus (Figure 4B). To determine whether the ILA-mediated upregulation of TJs is dependent on AhR activation, we examined TJ protein expression in AhR-knockdown KC. We found that the protein expression of OCLN and CLDN1 in KCs was not increased after ILA treatment when AhR was knocked down (Figure 4C), whereas the reduction in CLDN4 expression was less pronounced (Figure S4C). To further examine the association between AhR and TJ gene regulation, a chromatin immunoprecipitation (ChIP) assay was employed to assess AhR occupancy at the promoter regions of CLDN1 and OCLN. Increased enrichment of CLDN1 and OCLN promoter DNA was observed in AhR immunoprecipitates from ILA-treated KCs compared with controls (Figure 4D). In contrast, no comparable enrichment was detected at the CLDN4 promoter (Figure S4D), suggesting that the regulation of CLDN4 may involve additional or district mechanisms.

Having established that AhR is essential for ILA-mediated TJ upregulation, we next asked whether this barrier-enhancing effect is a general consequence of AhR activation or a ligand-selective property of ILA. To address this, we compared the effects of ILA with those of the classical high-affinity AhR agonist FICZ and another well-known microbiota-derived indole derivative, indole-3-propionic acid (IPA).^43,44 Notably, while ILA, FICZ, and IPA all effectively activated classical AhR signaling, as indicated by CYP1A1 mRNA upregulation, consistent upregulation of TJs (CLDN1, CLDN4, and OCLN) was observed exclusively for ILA among the tested ligands (Figure S4E). This functional divergence demonstrates that general AhR activation is insufficient to enhance barrier function. Rather, the upregulation of epithelial TJs appears to be a ligand-specific response driven by metabolites such as ILA.

To further explore the contribution of AhR to skin barrier function, we generated mice in which the floxed Ahr allele (Ahr^fl/fl) was conditionally knocked out in KCs, using a Cre recombinase driven by the keratin-14 promoter (K14-Cre Ahr^fl/fl). In contrast to the Ahr^fl/fl controls, topical treatment with ILA failed to significantly upregulate the mRNA expression of Cldn1 and Ocln, nor did it lead to a significant reduction in TEWL in the K14-Cre Ahr^fl/fl mice (Figures 4E–4G). Notably, Cldn4 expression was unaffected by ILA treatment regardless of the genotype (Figure S4F). Together, these findings indicate that AhR signaling in KCs is required for the ILA-mediated enhancement of skin barrier function.

Topical ILA application attenuates MC903-induced AD-like dermatitis in mice

To investigate the role of ILA in ameliorating AD pathogenesis, we utilized a mouse model of MC903-induced AD-like dermatitis. MC903 was first applied to the ears of BALB/c mice, followed by topical application of ILA for 30 min daily for 12 consecutive days (Figure 5A). Compared with EtOH-treated AD mice, topical application of ILA markedly alleviated MC903-induced skin inflammation, as evidenced by reduced visible scaling and erythema (Figure 5B). Consistently, ear thickness measurements revealed a significant attenuation of ear swelling in ILA-treated mice over the course of treatment (Figure 5C). Histological analysis further demonstrated reduced epidermal thickness and decreased inflammatory cell infiltration in the ears of ILA-treated AD mice, as shown by H&E staining (Figures 5D and 5E). In addition, serum total immunoglobulin E (IgE) levels were significantly lower in mice receiving topical ILA treatment (Figure 5F). Moreover, we found that mRNA expression of TJs including Cldn1, Cldn4, and Ocln was elevated in the ILA-treated AD mice (Figure 5G). This suggests that ILA enhanced the integrity of the skin barrier. We also found that the levels of various inflammatory cytokines, including Il-33, Il-4, Il-13, and Il-6, were significantly lower in ILA-treated AD mice than in EtOH-treated AD mice (Figures 5H and 5I). However, no significant difference in Tslp expression was observed between ILA-treated AD mice and EtOH-treated AD controls (Figure S5A). Given that IL-33 and TSLP derived from KCs are recognized as key initiators of AD-associated inflammation,⁴⁵ we next examined the effect of ILA on IL-33 expression. RT-qPCR analysis showed a dose-dependent reduction in IL-33 expression in KCs treated with increasing concentrations of ILA (Figure S5B), which was consistent with the corresponding immunoblotting results (Figure S5C). Under a Th2-skewed inflammatory microenvironment induced by IL-4, IL-13, and TNF-α, ILA treatment was associated with reduced IL-33 expression in KCs, whereas no apparent changes were detected in other AD-related cytokines, including TSLP and IL-6 (Figure S5D). Furthermore, ChIP analysis revealed increased AhR occupancy at the IL-33 promoter region following ILA treatment, supporting a potential involvement of AhR in the regulation of IL-33 expression (Figure S5E).

Figure 5.

Topical ILA application attenuates MC903-induced AD-like dermatitis

(A) Experimental design of MC903-induced AD-like mouse model. MC903 was topically applied to the ears of BALB/c mice for consecutive 12 days to induce AD-like dermatitis. ILA was applied daily, with EtOH serving as the vehicle control.

(B) Representative gross appearance of mouse ears, H&E staining, and toluidine blue staining of ear sections on day 12. Scale bars: 250 μm (H&E) and 100 μm (toluidine blue).

(C) Dynamic changes in ear thickness measured on days 0, 4, 8, and 12 during MC903 treatment (n = 5–6 per group). Statistical analysis was performed on day 12. 95% CI: 0.1344–0.2135 (MC903+EtOH vs. MC903+ILA).

(D) Quantification of epidermis thickness in ear sections on day 12 (n = 5–6 per group). 95% CI: −94.21 to −51.12 (EtOH vs. MC903+EtOH), 22.96–66.05 (MC903+EtOH vs. MC903+ILA).

(E) Quantification of infiltrating mast cells in ear sections on day 12 (n = 5–6 per group). 95% CI: −44.10 to −25.13 (EtOH vs. MC903+EtOH), 15.13–34.10 (MC903+EtOH vs. MC903+ILA).

(F) Serum total IgE levels measured by ELISA on day 12 (n = 4 for naive mice as a baseline reference, n = 5 for treated groups). 95% CI: −1,960 to −1,002 (EtOH vs. MC903+EtOH) and 870.3–1828 (MC903+EtOH vs. MC903+ILA).

(G) RT-qPCR of TJ-related gene expression in ear tissues on day 12 (n = 3 for naive mice as a baseline reference, n = 5 for treated groups).

(H) RT-qPCR of cytokines gene expression in ear tissues (n = 5 per group).

(I) Representative immunoblot images and quantification of IL-33 protein expression in ear tissues on day 12 (n = 3–4 per group, no statistical analysis due to the limited sample size).

Data are presented as mean ± SD (C, E, G, H, and I), min to max (D and F). Significance: one-way ANOVA (C–H) followed by Tukey’s post hoc test, with 95% CIs (G and H) reported in Table S5. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001,∗∗∗∗p < 0.0001; ns, not significant. See also Figure S5.

In addition to evaluating the therapeutic effects of ILA, we also investigated its preventive potential in the MC903-induced AD model. Mice that received topical ILA treatment prior to disease induction exhibited attenuated skin inflammation, as evidenced by improved gross appearance, reduced ear thickness, lower serum IgE levels, and altered expression of Il-33 and TJs (Cldn1 and Cldn4) (Figures S5F–S5J). Collectively, these findings support that topical ILA application mitigates MC903-induced AD-like dermatitis, potentially through enhancement of TJs and modulation of IL-33 expression in KCs.

AhR is required for the protective effects of ILA

Mice with epithelial AhR deficiency have been reported to exhibit impaired skin barrier function and increased susceptibility to infection.³² Consistently, we found that K14-Cre Ahr^fl/fl mice failed to exhibit barrier repair in response to topical ILA treatment, as indicated by unchanged TJs and TEWL (Figures 4E–4G). To further evaluate the contribution of AhR to the anti-inflammatory effects of ILA in vivo, AD-like dermatitis was induced in K14-CreAhr^fl/fl mice using MC903. In contrast to Ahr^fl/fl control mice, topical ILA treatment did not result in apparent improvements in gross appearance (Figure 6A), histopathological features (Figure 6B), ear thickness (Figure 6C), or serum total IgE levels (Figure 6D) in K14-CreAhr^fl/fl mice. Moreover, ILA treatment failed to suppress the upregulation of Il-33 and Il-4 or to enhance the expression of TJs (Cldn1, Cldn4, and Ocln) in K14-CreAhr^fl/fl mice with MC903-induced AD-like dermatitis (Figure 6E; Figure S6A). Consistent with the findings obtained from the genetic model, pharmacological inhibition of AhR signaling using CH223191 similarly attenuated the protective effects of ILA in BALB/c mice. Mice treated with CH223191 in combination with ILA exhibited more pronounced skin inflammation, including increased erythema, scaling, and ear swelling (Figure S6B), accompanied by elevated ear thickness (Figure S6C) and serum total IgE levels (Figure S6D), compared with mice receiving ILA treatment alone.

Figure 6.

AhR is required for the protective effects of ILA

(A–E) K14-CreAhr^fl/fl and littermate control Ahr^fl/fl mice with MC903-induced AD-like dermatitis were treated topically with ILA or vehicle control (n = 5 per group).

(A) Representative gross appearance of mouse ears on day 12.

(B) Representative H&E staining of ear sections on day 12. Scale bars: 200 μm.

(C) Dynamic changes in ear thickness measured over the course of treatment.

(D) Serum total IgE levels measured on day 12.

(E) RT-qPCR of TJ-related genes and Il-33 mRNA expression in ear tissues on day 12. Data are presented as mean ± SD.

Significance: two-way ANOVA followed by Tukey’s multiple comparisons test (C–E). Detailed 95% CIs are in Table S5. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001,∗∗∗∗p < 0.0001; ns, not significant. See also Figure S6.

To contextualize this ILA-AhR axis within the native skin microenvironment, we next investigated whether physiological, strain-specific ILA production by live bacteria could drive barrier protection in vivo. Based on our in vitro screening, we selected a live C. acnes strain characterized by a high capacity for ILA production. We first evaluated its efficacy in an acute barrier disruption model induced by tape stripping in wild-type C57BL/6 mice. Strikingly, topical application of this live C. acnes strain (1 × 10⁷ colony-forming units [CFU]) significantly accelerated barrier recovery, as evidenced by a substantial reduction in TEWL at 12, 16, and 20 h post-injury (Figure S6E). Concurrently, the mRNA expression levels of epidermal TJs (Cldn1, Cldn4, and Ocln) were markedly elevated in the bacteria-treated group compared to vehicle controls (Figure S6F), confirming that live C. acnes functionally bolsters barrier integrity. Building on these findings in acute injury, we further evaluated the therapeutic potential of live C. acnes in the MC903-induced AD model. Application of C. acnes (1 × 10⁷ CFU) to MC903-treated mouse ears successfully attenuated inflammation, leading to reduced scaling, decreased erythema, and diminished ear thickness. Notably, the co-application of C. acnes with its metabolic substrates—either Trp or artificial sebum—resulted in a more pronounced attenuation of inflammatory responses (Figures S6G and S6H), highlighting the importance of substrate availability for microbial metabolite biosynthesis.

Collectively, these results demonstrate that epithelial AhR signaling is indispensable for the ILA-mediated modulation of IL-33 and TJs and that this protective immunoregulatory axis can be physiologically engaged by the skin commensal C. acnes through local metabolism.

Topical ILA application alleviates symptoms in patients with AD

To explore the therapeutic potential of topical ILA in patients with AD, we conducted a proof-of-concept split-body study involving 10 patients with mild-to-moderate AD. All participants were administered either a vehicle or ILA topically on their symmetrical lesional areas twice daily for 2 weeks, with each participant serving as their own control. Upon completion of treatment, areas treated with ILA showed greater clinical improvement compared with vehicle-treated areas in patients with AD (Figure 7A). Topical application of ILA also led to a decrease in regional SCORAD scores (the sum of local intensity and pruritus scores) (Figure 7B) and TEWL after the 2-week period (Figure 7C). These findings demonstrate the efficacy of topical ILA in ameliorating AD symptoms, highlighting its potential as a promising therapeutic strategy.

Figure 7.

Topical ILA application ameliorates clinical symptoms in AD patients

(A) Representative photographs of symmetrical target lesions in AD patients at baseline (week 0) and after 2 weeks of topical treatment with ILA or vehicle control.

(B and C) Changes in treated region-specific SCORAD scores (B) and TEWL (C) of target lesions before and after treatment with ILA (n = 10 per group).

Treated region-specific SCORAD scores were the sum of skin symptom intensity and pruritus scores. Statistical significance: paired two-tailed t tests, with 95% CIs reported in Table S5. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗∗p < 0.0001; ns, not significant.

Discussion

C. acnes is the predominant bacterium on the human skin, particularly in sebum-rich regions. The decline in relative abundance and clonal diversity of C. acnes is increasingly recognized as a significant contributor to skin disorders.⁴⁶ Previous studies have demonstrated that C. acnes, adapted to specific niches, plays a major role in maintaining local skin homeostasis and offer protection against AD.^46,47,48 However, details of the interaction between C. acnes and the host, as well as the mechanisms underlying its protective effects, remain incompletely understood. In the present study, we integrated WGS, pan-genome analysis, and metabolic profiling to uncover the functional heterogeneity of C. acnes between HCs and patients with AD at the strain level. ILA—a major tryptophan metabolite of C. acnes—has important functions in repairing skin barrier and attenuating AD inflammation through activation of the AhR pathway in KCs.

Previous metagenomic studies have revealed that C. acnes exhibits stable strain-level heterogeneity across different skin sites over time.⁴⁹ Our recent findings suggest that the genetic heterogeneity of C. acnes is influenced by both individual-level factors and the specific skin site in normal and AD skin.²⁴ To minimize confounding effects from different skin parts, we focused our analysis on C. acnes strains from the facial skin, a common site for AD and a predominant colonization site for C. acnes. Consistent with our prior findings, we found type K1 was enriched in HC, whereas type C6 was restricted to patients with AD. Further pan-genome analysis of functional diversity among disease-specific C. acnes isolates demonstrated distinct energy metabolism pathways, including Trp metabolism pathway, between HC and patients with AD, suggesting potential differences in metabolites. Our earlier studies characterized the differences in microbial Trp metabolites on the skin, revealing that IAId, derived from the skin microbiota, has a regulatory function in skin inflammation. In this study, we not only confirmed these previous findings, but also identified a decrease in ILA, a specific metabolite of C. acnes, on the skin surface of patients with AD. ILA is known for its anti-inflammatory properties, its role in maintaining gut homeostasis, and its potential to mitigate colorectal tumorigenesis.^{40,41,50,51,52} However, to date, the specific effects of ILA on the skin remain largely unexplored. Additionally, a previous study discovered two cutotypes on the human skin based on the dominance of one out of two species: C. acnes (referred to as “C-cutotype”) and Moraxella osloensis (referred to as “M-cutotype”). The two cutotypes differ markedly in functional modules related to their metabolic characteristics, indicating that host-dependent trophic chains might underlie the development of the skin microbiome.⁴⁸ In our in vitro experiment, changes in sebum concentration affected the production of ILA by C. acnes. Under conditions with lower or higher levels of sebum, the production of ILA declined. These results, along with previous studies, show that the functional and metabolic profiles of C. acnes are determined by the specific skin niche, and an appropriate sebum concentration and abundance of C. acnes might be prerequisites for its protective effect.

AhR, a transcription factor, is increasingly recognized as a key factor through which the microbiota interacts with the host to maintain homeostasis.^32,53,54,55 A previous study revealed that ILA significantly induced IL-22 production in primary CD4⁺ T cells via AhR, while also reducing pro-inflammatory IL-12p70 production in monocytes via both AhR and hydroxycarboxylic acid receptor 3 (HCA3), thereby modulating the immune response in humans.⁵⁶ AhR is predominantly expressed in KCs, which play a crucial role in sensing environmental stimuli, regulating the skin barrier, and mitigating inflammatory skin conditions.^32,39,57 In this study, we found that downstream genes, such as CYP1A1 and CYP1B1, which are responsive to AhR activation, were significantly upregulated in KCs treated with ILA, indicating the ILA-AhR interaction. TJs are now recognized for their crucial role in epidermal function and inflammatory skin conditions.^58,59 They regulate the permeability of intercellular spaces in response to environmental signals, controlling the movement of water, ions, proteins, and dendritic cell extensions.^60,61 Research has linked TJ defects in patients with AD to increased TEWL and found inverse correlations between CLDN1 expression and Th2 biomarkers.^62,63,64 Genetic variations in CLDN1 have also been associated with AD in different populations.^64,65 In this study, we uncovered that AhR could bind to the promoter regions of CLDN1, OCLN, and IL-33 following ILA stimulation, and upregulate the expression of TJs while downregulating the expression of IL-33, ultimately leading to the repair of barrier damage and attenuation of MC903-induced AD-like dermatitis. This study therefore provides mechanistic insights into how the ILA-AhR axis ameliorates AD pathogenesis. However, whether C. acnes has an anti-inflammatory effect in all inflammatory skin diseases needs further verification.

In conclusion, this study unveils the differences in the functional heterogeneity of C. acnes between HC and patients with AD, at the strain level. As a major tryptophan metabolite of C. acnes, ILA has important functions in skin barrier repair and attenuation of inflammation in AD by activating the AhR pathway in KCs. The results for the proof-of-concept clinical study further confirm the inhibitory effects of ILA against skin inflammation. Our research provides a theoretical and experimental basis for screening the most suitable C. acnes strains to alleviate skin barrier damage and inhibit inflammation in AD.

Limitations of the study

A primary limitation of our study is the restricted number of C. acnes strains examined and the use of samples from a single anatomical site. Consequently, the present findings may not fully capture the extensive strain-level diversity of the C. acnes community across distinct skin microenvironments. Furthermore, additional comprehensive genomic analyses are required to functionally characterize the key enzymes that govern ILA biosynthesis.

Resource availability

Lead contact

Further information and requests for resources and reagents should be directed to, and will be fulfilled by, the lead contact, Xu Yao (dryao_xu@126.com).

Materials availability

This study did not generate new unique reagents.

Data and code availability

• •All sequencing data generated in this study are publicly available as of the date of publication. WGS data (BioProject: PRJNA1245304), C. acnes transcriptome data (BioProject: PRJNA1369136), and keratinocyte transcriptome data (BioProject: PRJNA1220100) have been deposited in the NCBI BioProject database.
• •Metabolomics datasets generated in this study have been deposited in the OMIX database of the China National Center for Bioinformation/Beijing Institute of Genomics, Chinese Academy of Sciences (OMIX project: PRJCA052097) and are publicly available as of the date of publication.
• •This paper does not report original code. Any standard software and computational protocols used to analyze the data are detailed in the STAR Methods section.
• •Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Acknowledgments

This work was supported by the National Natural Science Foundation of China (82530099, 82273542, 82373489, 82304023, 82330098, 82404151, and 82522075), the National Key R&D Program of China (2022YFC3601800), the CAMS Innovation Fund for Medical Sciences (2021-I2M-1-059 and 2022-I2M-C&T-B-096), the Non-profit Central Research Institute Fund of Chinese Academy of Medical Sciences (2020-RC320-003 and 2021-RC320-001), and the Natural Fund of Jiangsu Province (BK20250265). We thank Prof. Yufeng Zhou for providing the AhR^fl/fl mice. The experiments were conducted in Jiangsu Provincial Key Laboratory of Dermatology.

Author contributions

Conceptualization, S.Z., W.L., and X.Y.; methodology, S.Z., X.X., and T.Y.; validation, S.Z., F.L., and X.X.; formal analysis, S.Z., F.L., and X.X.; investigation, X.X., Y. Zhou, Y. Zhang, and X.L.; resources, T.Y., Y.L., and Q.L.; data curation, S.Z., F.L., X.X., and X.F.; visualization, F.L., X.X., Y.L., and X.F.; funding acquisition, X.Y.,Y. Zhou, Y. Zhang, B.X., W.L.,Y.L., X.L., Q.L., and S.Z.; project administration, X.L., Y. Zhou, B.X., and Y. Zhang; supervision, X.L., W.L., and X.Y.; writing – original draft, S.Z., F.L., and X.X.; writing – review & editing, X.X., X.L., W.L., and X.Y.

Declaration of interests

A provisional patent related to this work is pending.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Purified Rat Anti-Mouse IgE (Clone R35-72)	BD Biosciences	Cat # 553413; RRID: AB_394846
Biotinylated Rat Anti-Mouse IgE (Clone R35-118)	BD Biosciences	Cat # 553419; RRID: AB_394850
AhR Rabbit mAb (Clone D5S6N)	Cell Signaling Technology	Cat # 83200S; RRID: AB_2800011
Claudin1-Polyclonal antibody	Proteintech	Cat # 13050-1-AP; RRID: AB_2079881
Claudin 4-specific Polyclonal antibody	Proteintech	Cat # 16195-1-AP; RRID: AB_2082969
Occludin Polyclonal antibody	Proteintech	Cat # 27260-1-AP; RRID: AB_2880820
IL-33	Proteintech	Cat # 66235-1-Ig; RRID: AB_2881624
TSLP	GeneTex	Cat # GTX85059; RRID: AB_10726419
β-actin	Cell Signaling Technology	Cat # 3700S; RRID: AB_2242334
Anti-mouse secondary antibodies	Cell Signaling Technology	Cat # 7076S; RRID: AB_330924
Anti-rabbit secondary antibodies	Cell Signaling Technology	Cat # 7074S; RRID: AB_2099233
Bacterial and virus strains
C. acnes isolates	This study	BioProject: PRJNA1245304
C. acnes ATCC 6919 strain	ATCC	ATCC 6919
High-ILA-producing C. acnes isolate	This study	NA
Biological samples
Skin swabs from healthy controls and AD patients	This study	N/A
Chemicals, peptides, and recombinant proteins
Columbia Blood Agar plates	BIO-KONT	Cat #P19003
Brucella Broth	BD Biosciences	Cat # 211088
Tryptophan	MedChemExpress	Cat # HY-N0623
Acetone	Sinopharm Chemical Reagent Co., Ltd.	Cat # 10000418
Ether	Sinopharm Chemical Reagent Co., Ltd.	Cat # 100092183
MC903 (calcipotriol)	Leo Pharma	Cat # H20150664
Indolelactic acid (ILA)	MedChemExpress	Cat # HY-113099
Dispase II	Roche	Cat # 04942078001
Trypsin	Gibco	Cat # 25200072
Keratinocyte Medium	Sciencell	Cat # 2101
Keratinocyte Growth Supplement	Sciencell	Cat # 2152
IL-4	Peprotech	Cat # 200-04
IL-13	Peprotech	Cat # 200-13
TNF-α	Peprotech	Cat # 300-01A
INTERFERin	Polyplus-transfection	Cat # 101000028
Squalene	Aladdin	Cat # S109119
Cholesterol	Nanjing Reagent	Cat # C0690800243
Cholesterol oleate	Aladdin	Cat # 113859
Paraffin wax	Nanjing Reagent	Cat # C0641510323
Palmityl palmitate	Aladdin	Cat # P100622
Oleic acid	Nanjing Reagent	Cat # C0680250224
Palmitic acid	Nanjing Reagent	Cat # C0680690223
Myristic acid	Macklin	Cat # M813385
Lauric acid	Nanjing Reagent	Cat # C0680740339
Olive oil	Macklin	Cat # O815210
Coconut oil	Macklin	Cat # C805404
Cottonseed oil	Macklin	Cat # C915369
Tween 80	Merck	Cat # P1754
Streptavidin-HRP	BD Biosciences	Cat # 554066; RRID: AB_2868972
Tetramethylbenzidine	Beyotime	Cat # P0215
Radioimmunoprecipitation assay buffer	Beyotime	Cat # P0013K
Protease and phosphatase inhibitors	MedChemExpress	Cat # HY-K0013
Chemiluminescence system ECL	ThermoFisher Scientific	Cat # 32106
CH223191	Selleck	Cat # S7711
Recombinant Human Aryl hydrocarbon receptor (AhR)	CUSABIO	Cat # CSB-EP001481HU3
Eagle’s minimal essential medium (EMEM)	MeilunBio	Cat # PWL104
Normocin	Invivogen	Cat # ant-nr-05
QUANTI-Luc™ 4 Reagent	InvivoGen	Cat # rep-qlc4Lg1
Indole-3-propionic acid (IPA)	MCE	Cat # HY-W015229
FICZ	MCE	Cat # HY-12451
Critical commercial assays
TIANamp Bacteria DNA kit	TIANGEN	Cat # DP302
SimpleChIP® Plus Enzymatic Chromatin IP Kit (Magnetic Beads)	Cell Signaling Technology	Cat # 9005
SteadyPure Universal RNA extraction kit	Accurate Biology	Cat # AG21017
HiScript® III RT SuperMix for qPCR (+gDNA wiper)	Vazyme	Cat # R323
AceQ qPCR SYBR Green Master Mix (High ROx Premixed #Q141)	Vazyme	Cat # Q141
BCA Protein Assay Kit	Beyotime	Cat # P0012
Series S Sensor Chip CM5	Cytiva	Cat # 29149603
Amine Coupling Kit	Cytiva	Cat # BR-1000-50
Deposited data
Whole genome sequencing Data of C.acnes	This study	BioProject: PRJNA1245304
Transcriptomic data of C.acnes	This study	BioProject: PRJNA1369136
Transcriptomic data of keratinocytes	This study	BioProject: PRJNA1220100
Metabolomic data	This study	OMIX project: PRJCA052097
Experimental models: Cell lines
HepG2-Lucia™ AhR reporter cells	InvivoGen	Cat # hpgl-ahr
Human primary epidermal keratinocytes (foreskin-derived)	Isolated in this study	NA
Experimental models: Organisms/strains
C57BL/6 mice	Nanjing Cavens Biotechnology Co., Ltd.	C57BL/6
BALB/c mice	Nanjing Cavens Biotechnology Co., Ltd.	BALB/c
Ahrflox/flox mice	provided by Yufeng Zhou’s laboratory	Ahrfl/fl
K14-Cre mice	Shanghai Model Organisms	K14-Cre
Oligonucleotides
AhR siRNA: sense: 5′-GUGACUUGUAC AGCAUAAUTT-3′, antisense: 5′-AUUAUGCUGUACAAGUCACTT-3′	This study	NA
NC siRNA: sense: 5′-UUCUCCGAACGUGUCACGUTT-3′, antisense: 5′-ACG UGACACGUUCGGAGAATT-3′	This study	NA
Real-time quantitativePCR primer sequences	This study	See Table S3
Software and algorithms
Fastp (v0.23.4)	Chen S et al.66	https://github.com/OpenGene/fastp
SPAdes(v3.11.1)	Bankevich et al.67	https://github.com/ablab/spades
QUAST (v5.2.0)	Gurevich et al.68	https://github.com/ablab/quast
CheckM (v1.1.3)	Parks et al.69	https://github.com/Ecogenomics/CheckM
GTDB-Tk (v2.1.0) andDatabase (R07-RS207)	Chaumeil et al.70 and Parks et al.71	https://github.com/ecogenomics/gtdbtk
SLST and SLST database	Scholz et al.72	http://medbac.dk/slst_server_script.html
BLASTn(v2.13.0+)	Johnson et al.73	http://www.ncbi.nlm.nih.gov/blast
Prokka (v1.14.6)	Seemann74	https://github.com/tseemann/prokka
Roary pipeline (v3.13.0)	Page et al.75	http://sanger-pathogens.github.io/Roary
glmmTMB (v 1.1.13)	Brooks et al.76	https://github.com/glmmTMB/glmmTMB
EggNOG-mapper (v2.1.10)	Cantalapiedra et al.77	https://github.com/eggnogdb/eggnog-mapper
AnnotationForge (v1.42.2)	Marc Carlson78	https://github.com/Bioconductor/AnnotationForge
ClusterProfiler (v4.2.2)	Wu et al.79	https://github.com/YuLab-SMU/clusterProfiler
Bowtie2 (v2.5.0)	Langmead et al.80	https://bowtie-bio.sourceforge.net/bowtie2/index.shtml
HISAT2 (v2.2.1)	Kim et al.81	https://github.com/DaehwanKimLab/hisat2
FeatureCounts (v2.0.1)	Liao et al.82	http://subread.sourceforge.net
DESeq2 (v1.34.0)	Love et al.83	http://www.bioconductor.org/packages/release/bioc/html/DESeq2.html
GSVA(v1.40.1)	Hanzelmann et al.84	https://bioconductor.org/packages/release/bioc/html/GSVA.html
MSigDB	Liberzon et al.85	https://www.gsea-msigdb.org/gsea/msigdb
Zeiss Zen software(v2.6)	Zeiss	https://portal.zeiss.com/download-center/softwares/mic
GraphPad Prism(v9.0)	GraphPad Software	https://www.graphpad.com
ImageJ	Schneider et al.86	https://imagej.net/ij/
Biacore T200 Evaluation Software	Cytiva	N/A
Other
Anaeropack system	Mitsubishi Gas Chemical Co., Inc.	Cat # C-1;Cat # C-31
C. acnes reference genome (NBRC 107605)	NITE Biological Resource Center (NRBC)	https://www.nite.go.jp/nbrc/catalogue/NBRCCatalogueDetailServlet?ID=NBRC&CAT=00107605
GPSkin	GOPOWER	N/A
Clinical trial	Chinese Clinical Trial Registry	ChiCTR2400090988

Experimental models and subject details

Participant characteristics and sample collection

The study protocols for human sample collection were approved by the Ethics Committee of the Institute of Dermatology, Chinese Academy of Medical Sciences. Written informed consent was obtained from all adult participants, as well as from all adolescent participants and their legal guardians. Both male and female participants were included in the observational cohorts to ensure the representativeness of the study population. Because the primary objective of these cohorts was to identify disease-driven microbial and metabolomic signatures (AD vs. HC), and because the sample sizes were not specifically powered to perform sex-stratified assessment of high-dimensional omics data, sex was not analyzed as an independent biological variable in these multi-omics analyses.

A total of 9 HC and 9 patients with AD enrolled in our previous study were included.²⁴ A 2ˆ2 cm² area of facial skin was scraped with an aseptic swab moistened by normal saline for 40 s. Each swab was promptly streaked onto Columbia Blood Agar plate (BIO-KONT) and incubated an anaerobic environment (Anaeropack system, Mitsubishi Gas Chemical Co., Inc.) at 37°C for 3–5 days. Colonies suggestive of C. acnes based on colony morphology were picked and identified by MALDI-TOF MS (Autof ms1000, Autobio). Pure cultures were obtained through two successive rounds of single-colony isolation on Columbia Blood Agar under anaerobic conditions (37°C, 3 days each). Consequently, 2–3 genetically distinct C. acnes strains per participant were harvested. Each isolate was resuspended in PBS and DNA extracted using the TIANamp Bacteria DNA Kit. Stringent quality-control filters were applied during library construction. Strains from one HC and two AD subjects failing preset quality control thresholds were excluded, yielding a final set of 45 high-quality isolates for downstream WGS.

For prokaryotic transcriptome analysis and targeted metabolomics, all isolates were retained. Unlike WGS, these analyses focus on gene expression and metabolic concentrations respectively, which can broader data variability and still provide valuable insights even when genetic precision is less precise. This approach ensures genetic reliability while maximizing biological information from transcriptional and metabolic perspectives. One randomly selected isolate per participant was inoculated into Brucella broth (BD Biosciences) supplemented with 10% tryptophan and incubated anaerobically at 37°C for 48h. After centrifugation at 4°C and 4000 rpm for 10 min the supernatants were stored at −80°C for targeted metabolomics, and pellets were washed with PBS and kept at −80°C for prokaryotic transcriptome analysis. In total, 18 strains were processed for metabolomics and transcriptome profiling. To investigate the impact of sebum concentration on C. acnes gene expression, we selected the standard strain ATCC6919 and introduced mixed artificial sebum⁸⁷ at concentrations of 0.01%, 0.05%, and 0.25%. We then conducted targeted metabolome mass spectrometry during the logarithmic growth phase of the strain.

An additional 60 volunteers aged 12–40 years (30 HC, AD 30) were recruited (Table S2) to explore the association between ILA and clinical manifestation. Exclusion criteria: systemic antibiotics or corticosteroids within 1 month, or topical antibiotics/corticosteroids within 2 weeks. AD severity was graded by SCORAD: 0–24 mild, 25–50 moderate, 51–103 severe. On the sampling day, subjects abstained from cleansers and moisturizers. A sterile saline-moistened swab was gently rubbed on non-lesional facial skin (2 × 2 cm²) for 40s to collect skin-surface metabolites, which were immediately stored at −80°C for targeted metabolomics. TEWL was measured on the forehead (0.5 cm above the eyebrow arch) using GPSkin^1,2 after a 12h withdrawal from topical products and 2h rest post-cleansing. Room temperature was maintained at 18°C–21°C with 40–60% relative humidity; subjects acclimated for 20–30 min before readings. SCH was recorded simultaneously.

Animal studies

Age-matched (6–8-week-old) male and female mice were utilized for all experimental procedures. Wild-type (WT) C57BL/6 and BALB/c mice were procured from Nanjing Cavens Biotechnology Co., Ltd. (Nanjing, China). AhR-floxed (Ahr^fl/fL) mice on a C57BL/6 background were graciously provided by Yufeng Zhou’s laboratory (Children’s Hospital of Fudan University, Shanghai, China), and K14-Cre mice were obtained from Shanghai Model Organisms (Shanghai, China). All mice were housed in a specific pathogen-free (SPF) facility within individually ventilated isolator cages, provided with ad libitum access to water and food, and subjetced to a standard 12-h light/dark cycle. All animal protocols were conducted with the strict approval of the Animal Welfare Ethics Review Committee at the Institute of Dermatology, Chinese Academy of Medical Sciences. To investigate skin barrier dynamics, an AEW dry skin model was established. Briefly, the dorsal skin was shaved 48h prior to treatment. A cotton ball saturated with a 1:1 mixture of acetone and ether was applied to the skin for 15s, immediately followed by the application of a distilled water-soaked cotton ball for 30s and subsequent tape-stripping. This procedure was repeated once daily for seven consecutive days, with control groups receiving distilled water only. For inflammatory studies, an AD-like model was induced by applying 4 nmol of MC903 (calcipotriol; Leo Pharma) topically to the ears for 12 days. To evaluate chemical intervention, 30 μL of 10 μg/mL ILA (MCE; dissolved in EtOH) was applied topically 30 min after each MC903 challenge, while model control mice received EtOH vehicle. Furthermore, to evaluate the in vivo therapeutic efficacy of topically applied live bacteria, the reference C. acnes strain ATCC 6919 was cultured anaerobically in Brucella broth at 37°C and applied to the AD model. Age-matched BALB/c mice were randomly assigned to five topical treatment groups: (1) 30 μL of artificial sebum (vehicle control); (2) live C. acnes alone (1 × 10⁷ CFU); (3) live C. acnes (1 × 10⁷ CFU) formulated in artificial sebum; (4) 30 μL of 10% tryptophan (Trp control); and (5) live C. acnes (1 × 10⁷ CFU) supplemented with 10% Trp. To directly validate the barrier-restoring capacity of strain-specific ILA production in vivo, an additional acute tape-stripping model was employed. A previously characterized high-ILA-producing C. acnes isolate was topically applied (1 × 10⁷ CFU) to the acutely disrupted dorsal skin. In this specific cohort, TEWL was dynamically monitored at 12h, 16h, and 20h post-disruption to assess acute barrier recovery. Throughout the studies, physiological metrics including TEWL, skin barrier scores (GPSkin), and full-thickness ear measurements using a micrometer were systematically recorded. At the experimental endpoints, mice were humanely euthanized via CO₂ inhalation followed by cervical dislocation. Target tissues were harvested, with portions formalin-fixed and paraffin-embedded for H&E histological evaluation, while fresh tissues were snap-frozen for subsequent RNA and protein extraction.

Cell culture and stimulation

Human primary KCs were isolated from skin samples of children’s circumcised foreskins, as described above.⁸⁸ The skin samples were cut into pieces and placed in 2.5 mg/mL Dispase II (Roche) solution at 4°C for about 20h. Then, the epidermis was torn off and digested with 0.025% trypsin for 8 min, and an equal volume of FBS was added. After filtering through a 70-μm cell strainer, the cells were cultured with Keratinocyte Medium with Keratinocyte Growth Supplement (Sciencell) in a 25-cm² culture flask at 37°C with 5% CO₂. For in vitro functional assays, primary KCs were seeded in 6-well plates and grown to ∼80% confluence prior to treatment. To evaluate baseline responses, cells were incubated with varying concentrations of ILA (0.05–2 mM) for 24h. To simulate AD-like inflammatory features, KCs were co-treated a pro-inflammatory cytokine cocktail comprising 100 ng/mL IL-4, 100 ng/mL IL-13, and 20 ng/mL TNF-α (all from Peprotech). Additionally, to assess the ligand-specific transcriptional regulation of TJs, KCs were stimulated for 24h with distinct AhR ligands: ILA (1 mM), IPA (100 μM), or the canonical agonist FICZ (100 nM).

For siRNA-mediated gene silencing, the AhR gene was knocked down in keratinocytes using siRNA purchased from GenePharma. SiRNA was transfected into keratinocytes using INTERFERin (Polyplus-transfection) following the manufacturer’s protocol. The cells were then collected for RNA or protein extraction. The siRNA sequences were as follows: AhR siRNA (sense: 5′-GUGACUUGUAC AGCAUAAUTT-3′, antisense: 5′-AUUAUGCUGUACAAGUCACTT-3′), and control siRNA (sense: 5′-UUCUCCGAACGUGUCACGUTT-3′, antisense: 5′-ACG UGACACGUUCGGAGAATT-3′).

Clinical trial registration and human study design

To explore the clinical effect of topical ILA in patients with AD, an exploratory, participant-blinded, vehicle-controlled split-body study was conducted as a proof-of-concept component of this work. The trial was registered at the Chinese Clinical Trial Registry under identifier ChiCTR2400090988. Ten patients with mild-to-moderate AD were recruited and enrolled at the Hospital for Skin Diseases, Institute of Dermatology, Chinese Academy of Medical Sciences and Peking Union Medical College, Nanjing. All participants provided written informed consent prior to enrollment. Both male and female patients were included. Due to the limited sample size of this proof-of-concept trial, the potential influence of sex on therapeutic outcomes was not analyzed separately and will be assessed in future larger-scale cohorts. Preparations containing 1 mM ILA or vehicle were applied to symmetrical lesional areas twice daily for 14 consecutive days, with one side receiving ILA and the contralateral side receiving vehicle. Data generated from this clinical trial component are presented in Figure 7.

Method details

Genome assembly and quality filtering

The extracted C. acnes DNA were sent to the NovaSeq Illumina PE150 (Novogene Bioinformatics Technology Co. Ltd. Beijing, China) to construct the library construction, quality control and sequencing. Sequencing reads were preprocessed using Fastp (v0.23.4)⁶⁶ to eliminate sequencing adapters and low-quality reads. Subsequently, genome assembly was performed utilizing SPAdes (v3.11.1)⁶⁷ with the default settings. Contigs shorter than 500 bp were excluded from the assembled genomes. The genome size, N50 value, and contig count were determined using QUAST (v5.2.0)⁶⁸ with default parameters. The completeness and contamination levels of the assembled genomes were evaluated with the “lineage_wf” workflow of CheckM (v1.1.3)⁶⁹ with default parameters. Taxonomic classifications of the assembled genomes were assigned through GTDB-Tk (v2.1.0)⁷⁰ with the Genome Database Taxonomy GTDB (R07-RS207).⁷¹ Draft genomes were deemed high-quality if they met the following criteria: size between 2.2 Mbps and 3.2 Mbps, completeness greater than 95%, contamination below 10%, N50 above 100000, and number of contigs fewer than 1000.

Phylotypes of C. acnes strains

SLST was conducted for all strains, using the C. acnes SLST database⁷² and blastn (version 2.13.0+)⁷³ for sequence alignment.

Pan-genome analysis

The gene coding sequences within the isolated bacterial genomes were identified using Prokka (v1.14.6)⁷⁴ with the kingdom parameter set to Bacteria. Subsequently, comprehensive pan-genome and core-genome analyses were conducted using the Roary pipeline (v3.13.0)⁷⁵ with default settings. To identify differential abundant genes between the groups, a linear mixed model, with disease status as a fixed effect and subjects as a random effect was performed on the gene content data derived from the Roary pipeline ouput using R package glmmTMB (v 1.1.13).⁷⁶ Genes passing the p-value <0.05 and FDR <0.25 threshold were processed for further analysis.

Functional characterization

The pan-genome reference, derived from the Roary pipeline, was annotated using EggNOG-mapper (v2.1.10).⁷⁷ Subsequently, the local KEGG databases for the pan-genome of C. acnes were established using the R package AnnotationForge (version 1.38.1)⁷⁸. Gene functional enrichment analysis of KEGG pathways was conducted with the R package clusterProfiler (v4.2.2)⁷⁹.

UHPLC-SRM-MS

The major metabolites of the Trp catabolism pathway were identified through targeted metabolomics analysis of the culture supernatant. The key metabolites—kynurenic acid and kynurenine—were procured from Sigma-Aldrich (Shanghai, China). Picolinic acid and indole ethanol/tryptophol were sourced from Vetec (Shanghai, China). Indolz, indole-3-carbonaldehyde, and indole-3-acetamide were obtained from Shyuanye (Shanghai, China). Tryptamine, indole-3-acetic acid, 5-hydroxyindoleacetic acid, indolelactic acid, and 5-hydroxytryptophan were all acquired from Aladdin (Shanghai, China). Nicotinamide, nicotinic acid, and L-tryptophan were purchased from Sinopharm Chemical Reagent Co., Ltd (Shanghai, China).

For the analysis, we employed UHPLC-SRM-MS. First, the Trp standard product was weighed and a single standard mother solution was prepared using 50% methanol. An appropriate amount of each mother solution was measured and mixed to form the standard product mother solution, which was then diluted with 10% methanol to achieve the desired concentration for each standard. Subsequently, an appropriate volume of the sample was transferred into a 2 mL centrifuge tube, to which 100 μL of an 80% methanol aqueous solution was added. The mixture was vortexed for 60 s, followed by the addition of 900 μL of 10% methanol to the sample, which was then vortexed again for 60 s. The sample was centrifuged at 12,000 rpm and 4°C for 5 min. An appropriate amount of the supernatant was taken and diluted 5-fold with 10% methanol. To this, 100 μL of the diluent was added to 100 μL of a 20 ppb Trp-d5 solution, and the mixture was vortexed for 10 s to ensure thorough mixing. Finally, 150 μL of the supernatant was pipetted into the assay vial for UHPLC-SRM-MS analysis.

Quantification of the targeted metabolites was performed using an EXion LC Liquid Chromatography system (AB SCIEX, USA) equipped with an ACQUITY UPLC®HSS T3 column (2.1 × 150 mm, 1.8 μm, Waters, USA) coupled to an AB6500 Plus (AB SCIEX, USA) for mass spectrometric detection. The gradient elution system was composed of mobile phase A: 0.1% formic acid in water and mobile phase B: 0.1% formic acid in methanol. The gradient elution conditions were as follows: 0–0.5 min, 10% B; 0.5–2 min, 10–30% B; 2–3 min, 60% B; 3–6 min, 60–98% B; 6–7.5 min, 98% B; 7.5–7.51 min, 98%–10% B; 7.51–9 min, 10% B. The flow rate was set at 0.25 mL/min, the injection volume was 5 μL, and the column temperature was maintained at 40°C. Mass spectrometry conditions were as follows: electrospray ionization source with positive ionization mode. The ion source temperature was 450°C, the ion source voltage was 4500 V, the collision gas was set at 10 psi and the curtain gas at 30 psi, and both the atomizing gas and auxiliary gas were at 50 psi. Multiple reaction monitoring was used for the scans. The metabolites were detected at Suzhou PANOMIX Biomedical Tech Co., Ltd and Shanghai Biotree Biomedical technology Co., Ltd.

Artificial sebum development

An artificial sebum formulation was produced based on the previously delineated components of artificial sebum as referenced in literature.⁸⁷ Subtle adjustments were made according to the accessibility and verifiability of specific constituents, such as the incorporation of palmitoleic acid; the substitution of myristic acid with stearic acid; and the replacement of olive oil, coconut oil, and cottonseed oil with glyceryl trioleate. The definitive composition used in this investigation comprised the following percentages: squalene (15%), cholesterol (1.2%), cholesterol oleate (2.4%), paraffin wax (10%), palmityl palmitate (15%), oleic acid (1.4%), palmitic acid (5%), myristic acid (2.5%), lauric acid (2.5%), olive oil (10%), coconut oil (10%), and cottonseed oil (25%). The constituents of the artificial sebum were combined in the aforementioned ratios and subjected to heating to achieve a molten state, followed by cooling and storage at 4°C. To enhance the solubility of the artificial sebum within the culture medium, a mixture was prepared by combining one part of artificial sebum with three parts of Tween80, which was then heated and melted until uniformly integrated, cooled, and stored at 4°C. This emulsion of artificial sebum and Tween80 was added to Brucella broth at final concentrations of 0.01%, 0.05%, and 0.25%, followed by sterilization via autoclaving. Thereafter, the Brucella broth enriched with artificial sebum was diluted to achieve the desired concentration gradient on a sterile work surface.

ELISA

Serum samples were collected and stored at −80°C prior to ELISA analysis. To determine total IgE in serum, a 96-well microplate was first coated with 100 μL/well of purified rat anti-mouse IgE (1 mg/mL) and incubated overnight at 4°C. The plate was then blocked with 200 μL/well of 10% FBS at room temperature for 30 min. Subsequently, diluted serum samples were added, followed by incubation at room temperature for 2h. Biotinylated rat anti-mouse IgE (2 mg/mL) was added next. The plate was further incubated with 100 μL/well of streptavidin-HRP. The reaction was developed by adding tetramethylbenzidine (TMB, Beyotime, Shanghai, China) and terminated with 50 μL/well of 2M H₂SO₄. The absorbance was measured at a wavelength of 450 nm.

ChIP

ChIP assay was carried out using SimpleChIP® Enzymatic Chromatin IP Kit (cat#: 9003, Cell Signaling Technology) according to the manufacturer’s protocol. Briefly, human primary keratinocytes grown in 15 cm dishes were fixed in 1% formaldehyde and quenched by adding glycine. Crosslinked chromatin complex was digested with Micrococcal Nuclease into 150–900 base pair fragments. Fragmented chromatin was then incubated with histone H3 (D2B12) XP Rabbit mAb (ChIP Formulated) (cat# 4620, Cell Signaling Technology), normal rabbit IgG antibodies (cat#:2729, Cell Signaling Technology), and AhR (D5S6H) Rabbit mAb (cat#:83200S, Cell Signaling Technology), together with protein G magnetic beads at 4°C overnight. After magnetic beads washing, reversal of cross-links, and DNA purification, RT-qPCR was used to measure the enrichment of immunoprecipitated DNA by using primers listed in Table. S3. The fold enrichment was calculated as 2^−ΔΔCt and expressed as the percentage of the total input chromatin (input%). Percent Input (input %) = 2% × 2(C [T] 2% Input Sample-C[T] IP Sample). C[T] = CT = Threshold cycle of PCR reaction.

RNA isolation and real-time quantitative PCR (RT-qPCR) assay

Total RNA was extracted from human primary keratinocytes or mouse tissues using SteadyPure Universal RNA extraction kit (Accurate Biology) following the manufacturer’s protocol. The concentration of the extracted RNA was measured with a NanoDrop 2000 spectrophotometer (Thermo Fisher Scientific) and reverse-transcribed to cDNA using the HiScript® III RT SuperMix for qPCR (+gDNA wiper) (cat#:R323, Vazyme). RT-qPCR was performed in a LightCycler 480 (Roche) using AceQ qPCR SYBR Green Master Mix (High ROX Premixed) (cat#:Q141, Vazyme) according to the manufacturer’s instructions. Relative gene expression (fold change) was analyzed by the ΔΔCt method and normalized to that of β-actin. The primers are listed in Table S3.

Western blotting

Human primary KCs were harvested and lysed with radio immunoprecipitation assay (RIPA) buffer (Beyotime) supplemented with protease and phosphatase inhibitors (MedChemExpress) for protein extraction. The mouse tissues were lysed in RIPA buffer and then ground in a grinder (cat#: KZ-III-F, Servicebio). The protein concentration was measured using the BCA Protein Assay Kit (Beyotime). Then, 1 × SDS loading buffer was added to the protein lysates followed by denaturing at 100°C for 10 min. Equal amounts of total proteins were run on 4–20% SmartPAGE™ Precast Protein Gel Plus (Smart Lifesciences) and transferred to 0.22 μm nitrocellulose membranes (Millipore Immobilon). The membranes were blocked with 5% nonfat powder milk for 2h at room temperature. Then, membranes were incubated with primary antibodies against CLDN1 (cat#:13050-1-AP, Proteintech), CLDN4 (cat#:16195-1-AP, Proteintech), OCLN (cat#27260-1-AP, Proteintech), IL-33 (cat#:66235-1-Ig, Proteintech), AhR (D5S6H) Rabbit mAb (cat#:83200S, Cell Signaling Technology), TSLP (cat#:GTX85059, GeneTex), and β-actin (cat#:3700S, Cell Signaling Technology) at 4°C overnight. On the following day, after three washes with TBST, the membranes were incubated with anti-mouse (cat#:7076S, Cell Signaling Technology) or anti-rabbit (cat#:7074S, Cell Signaling Technology) secondary antibodies for 1h at room temperature and washed three times with TBST. The protein blots were detected using the chemiluminescence system ECL (cat#:32106, ThermoFisher Scientific) and imaged with a Amersham ImageQuant™ 800 (Cytiva).

RNA-seq

The quality of RNA samples from C. acnes and keratinocytes was assessed using the Agilent 2100 Bioanalyzer. For both datasets, strand-specific libraries were prepared with the TruSeq Stranded mRNA LT Sample Prep Kit (Illumina) after rRNA depletion, and sequenced on the Illumina NovaSeq 6000 platform (PE150) by Novogene Bioinformatics Technology Co., Ltd (Beijing, China).

Raw reads underwent quality control using Fastp (v0.23.4)⁶⁶ to remove adapters and low-quality bases. For the C. acnes dataset, cleaned reads were aligned to the C. acnes reference genome (NBRC 107605) using Bowtie2 (v2.5.0)⁸⁰ in sensitive mode. For the keratinocyte dataset, cleaned reads were aligned to the human reference genome (GRCh38) using HISAT2 (v2.2.1)⁸¹. Gene-level read counts were obtained with FeatureCounts (v2.0.1)⁸² and normalized to FPKM values. DEGs were identified using the DESeq2 (v1.34.0)⁸³ package in R following the standard differential expression workflow, defining DEGs as those with adjusted q-value <0.05 and |log2foldchange|≥1.0. Functional enrichment analysis (GO and KEGG) was performed with ClusterProfiler (v4.2.2)⁷⁹. GSVA was conducted using the R package GSVA (v1.40.1)⁸⁴ on the log2-FPKM matrix with the “gsva” method (Gaussian kernel), gene-set size limits of 5–5000, and pathway signatures (c2.cp.kegg.v7.4.symbols.gmt) from MSigDB.⁸⁵ All analyses utilized high-quality clean data (Q30 > 90%, GC-content ∼50%).

AhR localization experiments

To determine the localization of AhR driven by ILA, KCs were stimulated with 1 mM ILA for for various durations (0, 6, 12, and 24 h). This was followed by incubation with the primary antibody AhR mAb (concentration 1:20) and nuclear staining with DAPI (Beyotime). Then, confocal images of the localization of AhR in keratinocytes at different time points were taken with LSM 800 confocal microscope (Zeiss). Image processing was performed with the Zeiss Zen software (v2.6) and Adobe Photoshop software.

Surface plasmon resonance (SPR) analysis

The direct binding affinity between ILA and AhR was evaluated using a Biacore T200 SPR system (Cytiva). Purified recombinant human AhR was immobilized onto a CM5 sensor chip via standard amine coupling chemistry, achieving a final immobilization level of approximately 16,000 response units (RU). A reference channel without AhR was similarly treated for background subtraction. For the multi-cycle binding kinetics assay, the ILA analyte was serially diluted in the running buffer (10 mM HEPES, 150 mM NaCl, 3 mM EDTA, 0.005% Tween 20, 1% DMSO, pH 7.4) to generate a 7-point concentration gradient (0, 3.9, 15.6, 62.5, 125, 250, and 500 μM). Analytes were injected sequentially over the sensor surface at a constant flow rate of 30 μL/min with a 120 s association phase and a 200 s natural dissociation phase. No surface regeneration was required between cycles due to complete analyte dissociation. Data were double-referenced (subtracting reference channel and blank buffer signals), and the K_D was calculated by fitting the sensorgrams to a 1:1 Langmuir binding model using the Biacore T200 Evaluation Software.

AhR reporter luciferase assay

To evaluate AhR activation, the HepG2-Lucia™ AhR reporter cell line (InvivoGen) was cultured in EMEM supplemented with 10% heat-inactivated FBS, 1× NEAA, and 100 μg/mL Normocin. For the assay, cells were seeded into 96-well plates (2×10⁴ cells/well) in Normocin-free medium and stimulated with endotoxin-free water (vehicle), 1 mM ILA, or 3.5 μM FICZ (positive control) for 24h. Secreted Lucia luciferase activity in the supernatant was quantified using the QUANTI-Luc™ 4 Reagent (InvivoGen) on a microplate luminometer. Raw relative light units (RLUs) were normalized as fold-induction relative to the vehicle control, with a fold-induction ≥5 predefined as biologically significant AhR activation per the manufacturer’s criteria. The cell line was confirmed to be mycoplasma-negative by the manufacturer.

Evaluation of the clinical effects of ILA application

Clinical characteristics were assessed by two investigators to treatment-side assignment, as described previously.⁸⁹ Regional SCORAD scores were determined using standardized approaches.⁹⁰ The intensity of skin symptoms in the treated area was rated based on dryness, erythema, edema, oozing, excoriation, and lichenification, each on a scale of 0–3. Participants also provided a subjective itchiness score for the treated region, ranging from 0 to 10. Region-specific SCORAD scores were calculated as the sum of skin symptom intensity scores and pruritus scores. TEWL was measured with a GPSKIN device. All assessments were performed under controlled temperature (23 ± 1°C) and relative humidity (50 ± 10%). Given the small sample size and exploratory nature of this prespecified pilot component, these findings should be considered preliminary and warrant confirmation in the subsequent randomized, double-blind, placebo-controlled phase.

Quantification and statistical analysis

Statistical analyses were performed as described in the corresponding figure legends, and all data are presented as mean ± SD unless otherwise indicated. Western blot band intensities were quantified by densitometry using ImageJ software(NIH).⁸⁶ To ensure strict statistical rigor, for experiments with extremely small sample sizes (n < 5), no formal inferential statistical testing was performed; instead, these data are exclusively presented descriptively to illustrate biological magnitude and visual trends rather than to support formal statistical inference. Conversely, for adequately powered routine experiments (n ≥ 5), a two-tailed Student’s t test or the nonparametric Mann–Whitney U test was used to compare two independent groups, depending on the normality of data distribution. One-way or two-way ANOVA followed by Tukey’s or Sidak’s post hoc tests was applied for comparisons involving more than two groups, where appropriate. Correlations were assessed using Spearman’s rank correlation test. To ensure absolute computational transparency, routine statistical visualization and the aforementioned baseline comparisons were performed using GraphPad Prism (version 9.0). Furthermore, to ensure rigorous statistical control across multiple comparisons, R software (4.0.2; R Foundation for Statistical Computing) was exclusively employed to execute precise Benjamini–Hochberg False Discovery Rate (FDR) adjustments utilizing the native ‘stats’ package, as well as to conduct all multi-omics bioinformatic processing.

Additional resources

The human study component presented in Figure 7 was included within a registered clinical trial and is reported here as an exploratory proof-of-concept analysis. The trial was registered at Chinese Clinical Trial Registry under identifier ChiCTR2400090988: https://www.chictr.org.cn/showproj.html?proj=214988.

Published: June 23, 2026