Transcriptomic evidence for T cell‐fibroblast‐keratinocyte axis via IL‐13‐periostin‐integrin in atopic dermatitis

To the Editor,

Increasing evidence suggests that POSTN encoding periostin plays important roles in atopic dermatitis (AD). For instance, serum periostin levels were positively correlated with the severity of AD. ¹ In a mouse model of AD, Th2 cell‐producing IL‐4/IL‐13 stimulated fibroblasts to produce periostin, which interacted with integrin receptors on keratinocytes, inducing proinflammatory cytokines, which then accelerated Th2‐type responses. ² , ³ However, human evidence supporting the hypothesis that periostin links Th2‐type responses to keratinocyte activation in AD remains largely lacking.

This study aimed to validate the hypothesis by applying cell–cell interaction analysis to single‐cell RNA sequencing (scRNA‐seq) datasets from AD. We combined scRNA‐seq data from skin biopsy samples of healthy controls, patients with AD, and patients with psoriasis from four published datasets (GSE222840, GSE147424, GSE173706, and GSE221648). After quality control, a total of 221,014 cells from 20 healthy, 9 AD, and 22 psoriasis patients are available (Table S1).

We identified 17 clusters of major cell types in human skin (Figure 1A,B). Comparing the proportion of each cell type, proliferating and suprabasal keratinocytes were high in AD and psoriasis compared to healthy skin (Figure 1C, Table S2). In addition, T cells, proliferating T cells, dendritic cells (DCs), and macrophages were high in AD compared to psoriasis and healthy skin (Figure 1C, Table S2). These findings based on 4 combined scRNA‐seq datasets confirmed previous reports. ⁴

FIGURE 1.

scRNA‐seq analysis of healthy, atopic dermatitis, and psoriasis samples. (A) UMAP of cell clusters integrated from healthy skin (n = 20), lesional skin from AD (n = 9), and lesional skin from psoriasis (n = 22) samples according to similarity of transcriptome. In total, seventeen cell clusters were identified. (B) Dot plot displaying average expression and frequency of canonical cell‐type markers for each cluster. (C) Boxplots represent the proportion of keratinocytes, fibroblast, and immune cell clusters (calculated from Table S2) in healthy, AD, and psoriasis skin. *p ≤ .05, **p ≤ .01, ***p ≤ .001, ****p ≤ .0001 Kruskal‐Wallis one‐way ANOVA with Dunn's post‐hoc test. (D) Dot plot displaying average expression and frequency of POSTN, ITGAV, ITGB3, ITGB5, IL‐4, IL‐13, and IL‐17 for each cluster. (E) Violin plot displaying expression levels for POSTN, ITGAV, ITGB3, and ITGB5 for fibroblast and keratinocytes. (F) UMAP for subcluster analysis of fibroblasts (left panel). Dot plot (middle panel) and violin plot for (right panel) the expression of POSTN, COL6A5, and MFAP5 in the subclusters of fibroblasts.

We then examined the expression of POSTN and its receptors consisting of an alpha integrin (ITGAV) and a beta integrin (ITGB3 or ITGB5) ⁵ among the identified cell types. In line with previous research, POSTN and its receptors were predominantly expressed in keratinocytes and fibroblast (Figure 1D). In healthy skin or psoriasis, POSTN expression was highest in basal keratinocytes or fibroblasts, respectively. In AD, all subpopulations of keratinocytes and fibroblasts expressed higher POSTN than those in healthy skin and psoriasis (Figure 1E, Table S3). Fibroblast subpopulation analysis revealed that a subset of POSTN ⁺ fibroblasts expressed COL6A5, which was specific for AD (Figure 1F, Figure S1). ⁶ The expressions of ITGAV and ITGB5 were comparable in keratinocytes and fibroblasts among normal skin, AD, and psoriasis, but ITGB3 was undetected (Figure 1E).

We performed cell–cell communication analysis (Appendix S1) focusing on periostin and other signalings that could be upstream of periostin including IL‐4/IL‐13 and TGF‐β, and also IL‐17 due to its specificity for psoriasis (Table S4). The communication probability via periostin was significantly higher in AD and psoriasis fibroblasts (total communication probability = 0.025 for AD and 0.026 for psoriasis) than in healthy fibroblasts (0.003). However, the communication probability via periostin in AD keratinocytes (0.026) was comparable with healthy keratinocytes (0.022) (Table S4). In addition, the communication probability via IL‐4/IL‐13 was specifically significant in AD T cells and proliferating T cells (0.006) (Table S4). The communication probability via IL‐17 was specifically significant in psoriasis cytotoxic/NK T cells and proliferating T cells (0.0001) (Table S4). The communication probability involving fibroblasts via TGF‐β was significantly high in psoriasis (total communication probability = 0.014) compared to that in AD (0.001) and healthy skin (0.003) (Table S4).

We then visualized the network of the above‐mentioned signaling pathways. In AD and psoriasis, the highest communication via periostin was from fibroblast to fibroblast and to proliferating keratinocytes, whereas, in healthy skin, it was from basal keratinocytes to fibroblast, differentiated, and proliferating keratinocytes (Figure 2A, Table S4). Highest communication via IL‐4/IL‐13, specific for AD, was from T cells and proliferating T cells to fibroblasts, keratinocytes, DCs, and macrophages (Figure 2B). The highest communication via IL‐17, specific for psoriasis, was from cytotoxic/NK T cells and proliferating T cells to keratinocytes (Figure 2C). These results are consistent with the expression of IL‐13 and IL‐17A in T cells, proliferating T cells, and cytotoxic/NK T cells (Figure 1D). In psoriasis, the highest communication via TGF‐β was from T cells to fibroblasts, suggesting that periostin in the fibroblast might be induced by TGF‐β alone (Figure S2), while in AD, it might be the combined effect of IL‐13 and TGF‐β.

FIGURE 2.

Cell–cell communication analysis for keratinocytes, fibroblast, and immune cells from the integrated scRNA‐seq. (A) Circle plot demonstrates the connections via POSTN‐ITGAV/ITGB5 (upper panel) and heatmap shows the relative importance (lower panel) of each cell group for the Periostin signaling network in healthy, AD, and psoriasis. In circle plot, each cell group is assigned a color together with edge color to illustrate the signal started from that cell group. Edge width represents the relative communication probability. (B) Circle plot demonstrates the connections via IL‐13‐IL4R/IL13RA1 complex and IL‐13–IL13RA1 and heatmap shows the relative importance of each cell group for IL‐4 signaling network specific for AD. (C) Circle plot demonstrates the connections via IL‐17‐IL17RA/IL17RC complex and heatmap shows the relative importance of each cell group for IL‐17 signaling network specific for psoriasis. (D) Boxplots represent the proportion of POSTN ⁺ IL4R ⁺ IL13RA1 ⁺ fibroblasts (calculated from Table S2) in healthy, AD, and psoriasis skin. Kruskal‐Wallis one‐way ANOVA with Dunn's post‐hoc test. (E) Scatter plots and correlations for the expression of POSTN with IL4R and IL13RA1 in POSTN ⁺ IL4R ⁺ IL13RA1 ⁺ fibroblasts. Spearman correlation coefficient (R) and p‐value are shown. Different colors indicated conditions as shown in the legend.

To provide the connection for IL‐4/IL‐13 signaling and periostin in fibroblasts, we examined the proportion of cells that co‐expressed POSTN, IL4R, and IL13RA1. The proportion of POSTN ⁺ IL4R ⁺ IL13RA1 ⁺ fibroblasts in AD was significantly higher than in healthy and quasi‐significantly higher than psoriasis (p = .067) (Figure 2D and Table S2). Moreover, the expression level of POSTN was positively correlated with the expression levels of IL4R and IL13RA1 in AD and psoriasis, but not in healthy skin (Figure 2E, Figure S3). These results supported the connection between IL‐4/IL‐13 signaling input and periostin output, which is prominent in AD fibroblast.

Our study has some limitations. Firstly, due to technical limitations, the causal‐consequence in gene expression in scRNA‐seq cannot be explored. Secondly, four datasets included in our analysis were generated using different RNA sequencing platforms. Although integrated analysis with batch effect correction (Appendix S1) was taken, possible impacts on downstream analysis cannot fully be excluded. Thirdly, the analyses in this study were exclusively based on the transcriptome. Thus, interactions involving post‐translational modifications and dynamic cellular infiltration cannot be explored.

Collectively, we present transcriptomic evidence for cross‐talk among T cells, fibroblasts, and keratinocytes via IL‐13‐POSTN‐IGTAV/ITGB5 in AD. Our observations highlight periostin as a key molecule linking Th2 response to keratinocyte activation in AD, suggesting modulation of periostin signaling may be beneficial for AD.

AUTHOR CONTRIBUTIONS

The study was conceptualized and designed by AN and KI². NQVT, YK, YN, and KI¹ screened and selected datasets and performed the analysis. All authors discussed and interpreted data analysis. NQVT prepared the figures and tables. NQVT and AN wrote the first draft of the manuscript. AN and KI² finalized the manuscript. All authors approved the final version of the manuscript.

FUNDING INFORMATION

This research was funded by a grant‐in‐aid for scientific research to AN from the Ministry of Education, Culture, Sports, Science and Technology, Japan (grant number 22 K19427).

CONFLICT OF INTEREST STATEMENT

The authors declare no conflicts of interest.

Supporting information

Appendix S1.

DATA AVAILABILITY STATEMENT

Data sharing not applicable to this article as no datasets were generated or analysed during the current study.