Multi-omics analysis unveiled fibroblast-mediated pathogenesis in male genital lichen sclerosus

Abstract

Male genital lichen sclerosus (LS), a chronic inflammatory dermatological condition, has been recognized for its profound implications on the quality of life among males. The exact etiological factors behind this prevalent condition remained largely enigmatic. In this research, we employed a multi-omics strategy to identify and elucidate the underlying histological biomarkers and the fundamental pathogenesis associated with male genital LS. A comprehensive cell atlas of male genital LS disease was constructed, highlighting a pronounced increase in T cells and a remarkable reduction in keratinocytes within the male genital LS samples. Further insights elucidated the enhanced crosstalk between fibroblasts and T cells via the collagen-CD44 axis, and between fibroblasts and keratinocytes through the APP-CD74 signaling pathway. This molecular dialogue was implicated in the immune infiltration and hyperkeratosis observed in the dermal-epidermal layer of male genital LS. Subsequently, we integrated single-cell RNA sequencing data with genome-wide association study findings to explore the cell-type-specific genes predisposing to the development of male genital LS. The analysis underscored the pivotal role of GAS1, which was enriched in fibroblasts and implicated in the pathogenesis of male genital LS progression. Collectively, we highlighted the critical role of fibroblasts in initiating male genital LS onset, generating interactions with T cells and keratinocytes, and eliciting the classical histological features of male genital LS.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13578-025-01453-3.

Introduction

Male genital lichen sclerosus (LS) is considered a chronic inflammatory dermatological disorder that has been observed to have a significant impact on male sexual health, leading to issues such as dyspareunia and complications like urethral stricture disease (up to 20-30%, [1–7]). Additionally, a concerning link exists between male genital LS and the development of penile squamous cell carcinoma (PSCC, [1–3]). This underscores the importance of early diagnosis and pathogenesis exploration for affected individuals. The identification of male genital LS is predominantly based on clinical assessment, with symptoms such as sexual discomfort, itching, and skin irritation being the primary indicators. The presentation of the condition can vary from mild to pronounced, often appearing as patches that are either depigmented and atrophic or lilac in color, accompanied by slight scales, visible blood vessels, and occasional bruising [8, 9]. Furthermore, there may be manifestations of lichenoid balanitis, along with structural changes, hardening, and tightness of the foreskin [1, 9]. Characteristic histological features of male genital LS encompass thinning of the epidermis, hyperkeratosis, the presence of edema, and hyaline degeneration of collagen in the upper dermis, as well as telangiectasia and purpura [8, 10, 11]. However, the precise etiology of this relatively common disorder remains elusive [10, 12].

Research has revealed the connections of male genital LS to genetic, infectious, and autoimmune factors; however, the most plausible pathological mechanism is the chronic exposure of susceptible epithelial tissue to the urine, which is obstructed by the foreskin [2, 9, 13, 14]. Individuals who underwent circumcision in early childhood demonstrated the lowest likelihood of male genital LS, with those circumcised at a later age showing a slightly higher probability [2, 4]. In vulvar LS, a disrupted immune response is observed, which is believed to play a pivotal role in the etiology of the disease [15–17]. The gene expression landscape substantiates vulvar LS as an inflammatory disorder, evidenced by the heightened expression of T-helper type I cytokines and the impaired function of T-regulatory cells [15]. On the other hand, LS is comparatively less understood in males, who exhibit a significantly lower incidence rate, estimated to be roughly one in ten relative to their female counterparts [18–21]. Nonetheless, emerging evidence suggests that a distinct pathophysiological mechanism may be at play within this population [21, 22].

Therefore, a deeper comprehension of the histological biomarkers and the underlying pathogenesis at play in male genital LS is imperative for advancing precision-targeted therapies in the future. In this study, we conducted single-cell RNA sequencing (scRNA-seq) analysis of 9 foreskin samples [3 normal, 3 tissues adjacent to lichen sclerosus (paraLS), 3 LS] to delineate the cell atlas of normal and LS tissues. We observed a notably increased immune infiltration, especially T cells, and a drastically reduced number of keratinocytes in LS samples. Further evidence pointed to enhanced communications from fibroblasts to T cells and keratinocytes, resulting in immune infiltration and excessive keratinization of the epidermis. All these pathophysiological processes formed the typical histological characteristics of LS. Then we integrated the scRNA-seq data with the genome-wide association study (GWAS) to explore the specific mechanism during male genital LS development. Altogether, we relied on a multi-omics approach to discern and elucidate the underlying histological biomarkers and the fundamental etiologies associated with male genital LS.

Methods and materials

This research was conducted following the principles outlined in the Declaration of Helsinki and received approval from the Ethics Committee of West China Hospital of Sichuan University (approval number: 20241358). The consents of all participants were duly obtained in an informed manner.

Patients

All patients were recruited in both the Department of Urology, West China Hospital of Sichuan University, and the Department of Daytime Service Center, West China Tianfu Hospital of Sichuan University. A total of 27 patients were included for analysis, encompassing 19 patients with normal-appearing foreskins undergoing circumcision, 5 patients with LS-appearing foreskins undergoing circumcision, and 3 patients diagnosed with LS-induced anterior urethral stricture undergoing urethroplasty. Demographic and clinical data were meticulously gathered from the participants, complemented by the collection of foreskin samples for further analysis. Detailed information was presented in Table S1 and Table S2.

Preparation of single-cell suspension from foreskin samples and scRNA-seq library

In this research, the 10x Genomics platform was strategically implemented. The process commenced with the meticulous preparation of a single-cell suspension, followed by a stringent evaluation of its viability. It was confirmed that the cells demonstrated an exceptional level of activity, exceeding the threshold of 85%, characterized by outstanding uniformity and negligible impurities. To enhance the efficiency of capturing individual cells, the cell concentration was meticulously adjusted to a precise range between 700 and 1200 cells per microliter.

In the phase of constructing the scRNA-seq library, each cell, accompanied by its cellular enzymes, was affixed to a gel bead. Subsequently, this was encapsulated within an oil droplet, resulting in the formation of a single-cell gel bead-in-emulsion (GEM). Each gel bead was distinctively labeled with a unique barcode, equipped with a unique molecular identifier (UMI) sequence, and furnished with a poly-dT primer sequence intended to initiate the reverse transcription process. The GEM droplets were then pooled into Eppendorf tubes, where the cell membranes were lysed using tailored recovery reagents, thereby releasing the mRNA. This liberated mRNA interacted with the enzymes, dNTP substrates, and the poly-dT primer sequence encapsulated within the droplets. Under meticulously set conditions within a polymerase chain reaction (PCR) apparatus, the process of reverse transcription was initiated, leading to the synthesis of complementary DNA (cDNA). In the final stage, the cDNA underwent PCR amplification, thereby completing the assembly of the scRNA-seq library.

The process of scRNA-seq library

Upon the finalization of library assembly, a preliminary assessment of quantity was conducted employing the Qubit 3.0 fluorometer (Life Technologies, Thermo Fisher Scientific). This process involved diluting the library to a standard concentration of 1ng/µL. Following this, the Agilent 2100 Bioanalyzer (Agilent Technologies) was used to scrutinize the insert size of the DNA segments integrated into the library. Upon confirmation that the insert size aligned with the desired specifications, the StepOnePlus Real-Time PCR System (Thermo Fisher Scientific) was employed for quantitative PCR (qPCR) to ascertain the precise concentration of the library, ensuring that it possessed an effective concentration of no less than 10nM. This step was crucial for verifying the library’s quality. Libraries that surpassed this benchmark were advanced to the sequencing phase, where they were analyzed on the Illumina sequencing platform, using a paired-end 150-base pair approach for comprehensive sequencing.

The scRNA-seq data have been deposited in the NCBI Gene Expression Omnibus (GEO) database, with the accession number GSE290798.

The quality control (QC) of scRNA-seq data

Employing the Cell Ranger v7.1.0 toolkit from 10X Genomics, the FASTQ files were processed for alignment with the human GRCh38 reference genome. Subsequently, the scRNA-seq data were subjected to QC using the Seurat v5.0.3 package [23]. Typically, cells that expressed fewer than 200 genes or more than 90% of the top gene count were filtered out to prevent the inclusion of doublets or cellular debris. Additionally, cells with a mitochondrial gene content exceeding 20% and those with more than 5% erythrocyte-related genes were removed. To mitigate batch effects, data from distinct sequencing batches were consolidated through the harmony algorithm offered by the Seurat R package (v5.0.3). The scRNA-seq data underwent a series of standard procedures, including normalization, identification of variable features, scaling, principal component analysis (PCA), and harmonization integration.

Cell clustering and cluster annotation

Post batch correction, the scRNA-seq data underwent further analysis using the k-nearest neighbors algorithm facilitated by the FindNeighbours function. Employing the shared nearest neighbor (SNN) modularity optimization, distinct cell clusters were delineated using the FindClusters function, set at a resolution parameter of 0.4. A parallel methodology was applied for the subclustering process.

The sum of expression counts for each cluster was determined utilizing the AggregateExpression function. Markers with significant differential expression—defined by a log2 fold change (log2FC) greater than 0.25 and an adjusted P value less than 0.05—were extracted through the FindAllMarkers function. Cluster annotation and visualization were achieved by integrating these findings with the CellMarker v2.0 database [24] and the recognized signature genes for the pertinent cell types. The following signature genes for the initial clustering were included: B cells (MS4A1, CD19, CD79A); T cells (CD3D, TRBC2, TRBC1, CD3E, CD2, CD3G, TIGIT, CTLA4, PDCD1); Macrophage (CD68, ITGAM, CD14); Mast cells (TPSAB1, IL1RL1, CPA3, TPSB2); Endothelial cells (PECAM1, CDH5, VWF, CLDN5, EGFL7, CLEC14A, ICAM1, SELE, SELP, ENG, VCAM1, MCAM, TEK, KDR, NOS3, ANGPT2); Smooth muscle cells (DES, ACTA2, TAGLN, SMTN, TPM2, CNN1); Fibroblasts (PDPN, COL6A1, APCDD1, PTGDS, MFAP5, ELN, FGF7, MME); Keratinocytes (SPRR2A, KRT19, KRT1, KRT4, GJB2, UBE2C, DEFB1, CDH1); Melanocytes (PMEL, MITF, MLANA, DCT, TYR, TYRP1); and Schwann cells (MPZ, CDH19, SCN7A, PLP1, S100B, SOX10). Subsequent subclustering was analyzed by extending a similar approach used for the initial clustering.

Differentially expressed genes analysis

Utilizing the FindMarkers function, we employed a Wilcoxon Rank Sum test to discern differential gene expression between the normal and LS groups, both across all the samples and within specific cell types.

Inference and analysis of cell-cell communication

The interactions among different cell types were inferred by the CellChat v1.1.3 package [25]. The package serves as a comprehensive tool for inferring, analyzing, and visualizing cell-cell interactions derived from single-cell datasets. It is designed to empower users with the capability to discern and elucidate cell-cell interactions within a framework that prioritizes clarity, visual appeal, and interpretability. The communication analysis was mainly anchored in the initial cell type classifications.

Single-cell gene set enrichment analysis (scGSEA)

Based on the FindAllMarkers and FindMarkers functions, differentially expressed genes of subsequent subclusters were in descending order according to the log2FC value. The clusterProfiler package was employed to conduct the scGSEA and to facilitate the comprehensive visualization of the results [26].

Trajectory analysis of keratinocytes

The keratinocyte subcluster underwent further transformation to be compatible with the monocle3 package, enabling the execution of trajectory analysis [27–33]. Monocle3 is designed to decipher the progression of cells through a genetic expression sequence. In this context, every cell is conceptualized as a coordinate within a multi-dimensional space, with each axis representing the expression level of a distinct gene. The task of uncovering the sequence of gene expression alterations is akin to mapping out a path that cells traverse through this space. By utilizing the uniform manifold approximation and projection (UMAP) technique for dimensionality reduction and by pinpointing the initial positions of the cells along the trajectory, it becomes feasible to chart the cells and their developmental paths. This approach not only offers a visual representation of the cell’s journey but also aids in the elucidation of the underlying biological processes. In this study, the keratinocyte subcluster was selected for the trajectory analysis to observe the keratinization process, and possible differentially expressed genes along the trajectory were explored with the help of the graph_test function.

GWAS information on the gene expression and LS disease

The GWAS profile of gene expression was derived from the eQTLGen consortium [34]. The eQTLGen consortium, established to uncover the effects of genetic variants linked to specific traits, encompasses 37 datasets from a collective of 31,684 individuals’ blood samples. The consortium identified a total of 16,987 genes for the cis-expression quantitative trait locus (cis-eQTL), 6,298 genes for the trans-eQTL, and 2,568 genes for the expression quantitative trait score (eQTS). To enhance biological interpretability, the analysis was narrowed down to include solely cis-eQTLs. The summary statistics are readily accessible to the public on their website at https://www.eqtlgen.org/.

The LS-related GWAS data were obtained from the FinnGen consortium R10 version [35]. The LS disease was defined according to the International Classification of Diseases (ICD) codes. Following a meticulous selection process involving various inclusion and exclusion criteria, the GWAS incorporated a total of 2,755 cases and 385,509 controls. The summary statistics are readily accessible to the public on their website at https://www.finngen.fi/en.

The cis-eQTLs and LS GWAS datasets specifically targeted the European demographic. Further details regarding these datasets are outlined in Table S3.

Summary-based mendelian randomization (SMR)

Mendelian randomization (MR) leverages genetic variation as a natural experiment to explore causal links between factors and diseases, offering advantages over traditional epidemiological methods. Since genetic alleles are randomly allocated at conception, this approach can mitigate the impact of confounding factors, thereby facilitating the identification of genuine causal relationships [36–41]. In this context, a novel software tool, SMR, has been crafted to facilitate the application of both the SMR and heterogeneity in dependent instruments (HEIDI) methodologies [42]. These methods are designed to ascertain whether the influence of a single-nucleotide polymorphism (SNP) on a phenotype is mediated through gene expression levels. Consequently, this tool is instrumental in prioritizing genes associated with GWAS findings for subsequent validation through scRNA-seq analysis. Here, we applied the SMR software to delve into the causal influence exerted by 16,987 cis-eQTLs on the LS disease phenotype. Genes with a p-value of < 0.05 for the SMR test and > 0.05 for the HEIDI test were regarded as causally related to the LS disease phenotype.

Hematoxylin-Eosin (HE), Picric acid-Sirius red, immunohistochemistry (IHC), and multiplex immunohistochemistry (mIHC) staining

The foreskin specimens were preserved in formalin for three days, after which they underwent the standard histological procedures for all the staining. Subsequently, the specimens were embedded in paraffin and sectioned into 5-micrometer slices. Following this, a series of steps such as deparaffinization, rehydration, and antigen retrieval were meticulously executed. For HE and Picric acid-Sirius red staining, the tissue sections were meticulously treated with their respective staining agents to ensure accurate coloration and visualization of cellular structures. For IHC and mIHC staining, the slices were blocked before incubating with the primary antibodies. The primary antibodies included: ELN (1:50 dilution, Cat# sc-58756, Santa Cruz Biotechnology, Inc.); COL1A1 (1:200 dilution, Cat# ab34710, Abcam plc.); COL6A1 (1:100 dilution, Cat# A9738, ABclonal Technology Co., Ltd.); CD44 (1:100 dilution, Cat# 37259, Cell Signaling Technology, Inc.); CD3E (1:150 dilution, Cat# ab16669, Abcam plc.); LAG3 (1:200 dilution, Cat# 15372, Cell Signaling Technology, Inc.); APP (1:100 dilution, Cat# 193895, Cell Signaling Technology, Inc.); CD74 (1:250 dilution, Cat# 772745, Cell Signaling Technology, Inc.); and KRT15 (1:200 dilution, Cat# 60247-1-Ig, Proteintech Group, Inc.). Post-staining, the samples were visualized using the fully automated inverted fluorescence microscope (Olympus, Japan) or the Akoya Vectra Polaris (Akoya Biosciences, Inc., the U.S.A.).

Cell culture

Human foreskin fibroblasts (HFF-1) were purchased from the National Collection of Authenticated Cell Cultures in Shanghai, China. These cells were maintained in Dulbecco’s Modified Eagle Medium (DMEM) supplemented with 15% fetal bovine serum (FBS, Nanjing BioChannel Biotechnology Co., Ltd., China) and 1% penicillin-streptomycin (PS, Hyclone, Cytiva, USA). The culture medium was refreshed every other day, or more frequently if the cells exhibited rapid growth. When the HFF-1 cells reached 70–80% confluence in a 10-cm culture dish (LABSELECT, China), the medium was aspirated and the cells were gently washed with an adequate volume of phosphate-buffered saline (PBS, Solarbio, Beijing, China). Subsequently, the cells were trypsinized for subculturing.

Plasmid construction and cell transfection

The knockdown short-hairpin RNA (shRNA) plasmids of human GAS1 were bought from the MiaoLing Plasmid Platform (Wuhan, China). Three sequences were designed, and detailed information was provided in Table S4. The plasmid vector was: pLV3-U6-shRNA-CopGFP-Puro. Transfection procedures adhered to the manufacturer’s instructions provided with the transfection reagent (Lipofectamine 3000, Cat# L3000015, Invitrogen, USA). Following a 12- to 24-hour incubation period, the conditioned medium was replaced with fresh complete medium to complete the transfection process. Subsequently, after an additional 48- to 72-hour period, whole-cell proteins were extracted for further analysis.

Urea intervention

Before the HFF-1 cells were subjected to subsequent experiments, the Minimum Essential Medium (MEM) was utilized. HFF-1 was cultured in FBS-free MEM with 10% PS for 24 h, followed by different concentrations of urea intervention (Macklin Inc., Shanghai, China).

Western blots (WB)

The WB experiments were carried out following a standardized procedure. In summary, SurePAGE™ high-performance precast mini gels (GenScript Biotech Corporation) were procured for sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE). To ascertain the molecular weight spectrum, a multicolor prestained protein ladder (Cat# WJ103, Shanghai Epizyme Biomedical Technology Co., Ltd., China) was employed. The transmembrane procedure was facilitated by the WB transfer buffer (Cat# D1060, Solarbio, Beijing, China) and the polyvinylidene fluoride (PVDF) membrane featuring a 0.2 μm pore size (Immobilon^®, Millipore, Merck KGaA, Darmstadt, Germany, and/or its affiliates). Subsequently, the PVDF membrane was incubated in the protein-free rapid blocking buffer (Cat# DB307L, Shanghai Jumpingfrog Biotechnology Co., Ltd., China) at room temperature for a minimum duration of 10 min to ensure proper blocking. The primary antibodies included: GAPDH (1:5000 dilution, Cat# ab181602, Abcam plc.); ST13 (1:1000 dilution, Cat# BF8687, Affinity Biosciences); COL1A1 (1:1000 dilution, Cat# R10022, ZEN-BIOSCIENCE); COL6A1 (1:1000 dilution, Cat# A9738, ABclonal Technology Co.,Ltd.); and GAS1 (1:1000 dilution, Cat# DF4098, Affinity Biosciences). The membrane, following the blocking step, was subjected to an overnight incubation at 4 °C with the primary antibodies. The secondary antibodies utilized were goat-anti-rabbit HRP-conjugated IgG (diluted to 1:5000, Cat# S0001, Affinity Biosciences) and goat-anti-mouse HRP-conjugated IgG (diluted to 1:5000, Cat# 511103, ZEN-BIOSCIENCE). After the primary antibody incubation and subsequent washing with 1x tris buffered saline with tween-20 (TBST, Cat# T1082, Solarbio, Beijing, China), the membrane was further incubated with the secondary antibodies for one hour at room temperature. The membrane was then prepared for detection using the Oriscience Ultrasensitive ECL Kit (Cat# PD203, Oriscience Biotechnology (Chengdu) Co., Ltd) in conjunction with the chemiluminescence imaging system (Bio-Rad Laboratories, Inc.).

Statistical analysis

The main analysis was completed on the R (v4.3.2), GraphPad Prism 9 (GraphPad Software Inc., San Diego, CA, USA), and ImageJ (version 1.54f, National Institutes of Health, the U.S.A.) software. The Seurat v5.0.3 [23], CellChat v1.1.3 [25], clusterProfiler v4.10.1 [26], monocle3 v1.4.18 [27–33], CMplot v4.5.1 (https://github.com/YinLiLin/CMplot), and ggsci (https://github.com/nanxstats/ggsci) packages were implemented. Partial plots were generated on the Hiplot Pro platform (https://hiplot.com.cn/), a comprehensive web service for biomedical data analysis and visualization. The WB images were processed using the Image Lab (Bio-Rad Laboratories, Inc.) software. For the mIHC images, the QuPath software (v0.4.3) was employed for the enhancement and processing of the illustrations [43]. The data were presented as the mean with the standard error of the mean (SEM). For the comparison among three groups with similar variances, the one-way ANOVA test followed by Dunnett’s multiple comparisons test was applied. In contrast, when the variances were not comparable, the Welch ANOVA test coupled with Dunnett’s multiple comparisons test was utilized. The threshold for statistical significance was established at 0.05, with the following notations to indicate the levels of significance: * for P < 0.05, ** for P < 0.005, and *** for P < 0.001.

Results

The morphological appearance and histological traits of male genital LS were distinctly divergent

A total of 27 patients were included in our study, encompassing 19 with normal foreskins and 8 with LS-appearing foreskins. Their demographic and clinical information was provided in Table S1. The foreskin samples of LS exhibited a distinctly white and hardened appearance compared to those of normal individuals (Fig. 1A). In some cases, a white and hardened appearance of balanus could be observed (Fig. 1A). The Hematoxylin-Eosin (HE) staining images of foreskin samples from normal, paraLS, and LS groups also showed distinct features (Fig. 1B). At the dermal-epidermal junction of the foreskin tissue affected by LS, a pronounced homogenization of collagen is observed (the black triangles in Fig. 1B). To go a step further, we quantified the thickness of the epidermis and dermis. A significantly thicker dermis was observed in the LS group (P < 0.001 for both LS vs. paraLS group and LS vs. normal group, Fig. 1C). Interestingly, the age of patients diagnosed with LS-induced urethral stricture (US) was older than that of patients with normal-appearing foreskins (P < 0.005, Fig. 1D).

Fig. 1.

The morphological presentation and histological characteristics of male genital LS were markedly different. A. The morphological presentation of the foreskin and balanus from the normal and LS group. The black triangles indicated the white and hardened lesions. B. The HE staining images of the foreskin samples. The black triangles indicated a pronounced homogenization of collagen at the dermal-epidermal junction. C. The quantitative analysis of the epidermal and dermal thickness among the three groups. The data were presented as the mean ± SEM (n = 5). D. The age differences among the four groups. The data were presented as the mean ± SEM. E. The Picric acid-Sirius red staining images of the foreskin samples. The black triangles indicated a modest quantity of collagen at the dermal-epidermal junction of the paraLS group, and an extensive collagen deposition in the LS group. F. The heatmap of the pathological features in LS patients. LS: lichen sclerosus; US: urethral stricture; SEM: standard error of the mean

The application of Picric acid-Sirius red staining facilitated the observation of collagen distribution. In contrast to the normal group, a modest quantity of collagen was noted at the dermal-epidermal junction within the paraLS group, whereas an extensive collagen deposition was characteristic of the LS group (Fig. 1E). Based on the pathological features, a heatmap was plotted to figure out the most common histological characteristics of LS disease. We discovered that both collagen homogenization and inflammatory infiltration were observed in all cases (Fig. 1F and Table S2). On the other hand, the LS-appearing foreskins exhibited either atrophic epidermis or hyperkeratosis (Fig. 1F and Table S2).

The scRNA-seq data disclosed a distinct variation in the proportion of T cells and keratinocytes

To map the single-cell landscape of LS disease, we generated single-cell suspensions from nine samples (3 normal, 3 paraLS, 3 LS). Following the scRNA-seq library construction, preprocessing, and QC (Annoroad Gene Technology Co., Ltd., Beijing, China), a comprehensive set of expression profiles was extracted from 81,700 living cells (Fig. 2A). Further processing was based on the Seurat v5.0.3 package [23], eventually identifying 18 clusters (Figure S1A and Figure S1B). The top differentially expressed genes (DEGs) in each cluster and a series of classical cell markers provided by the CellMarker v2.0 database [24] were determined (Figure S1C, Figure S1D, and Figure S1E). A total of ten cell types (fibroblasts, keratinocytes, endothelial cells, smooth muscle cells, T cells, B cells, macrophages, mast cells, melanocytes, and Schwann cells) were annotated (Fig. 2B and C, and Figure S1F). In the separate cell atlas of each group, we noted a pronounced shift in the representation of T cells and keratinocytes, particularly in the LS group (Fig. 2B). The quantitative analysis of each cell type also revealed a slightly increased proportion of macrophages and B cells in the LS group (Fig. 2D and Table S5). Due to the drastic alterations in T cells and keratinocytes, we focused on the HE-staining images of the normal, paraLS, and LS groups (Fig. 2E). Indeed, we found a progressive immune infiltration from the normal group to the LS group (the orange triangles in Fig. 2E), along with a decreasing number of nucleated keratinocytes (the green regions in Fig. 2E).

Fig. 2.

The scRNA-seq data revealed a distinct distribution of T cells and keratinocytes, highlighting a notable disparity in their proportions. A. The flowchart of the scRNA-seq analysis. B. The cell landscapes in the normal, paraLS, and LS groups. C. The top DEGs in each cell type. The grey dashed lines indicated the log2-fold change of 0.25 and − 0.25. D. The proportion analysis of the annotated cell types and sample groups. E. The alterations of the immune cells and nucleated keratinocytes among the three groups. The orange triangles indicated a progressive immune infiltration from the normal group to the LS group. The green regions indicated a decreasing number of nucleated keratinocytes from the normal group to the LS group. DEGs: differentially expressed genes; B: B cells; Endo: endothelial cells; Fibro: fibroblasts; Kera: keratinocytes; Macro: macrophages; Mast: mast cells; Me_Sc: melanocytes and Schwann cells; SMC: smooth muscle cells; T: T cells

A panel of molecular markers was discerned for the LS samples

Given that the uniform deposition of collagen at the dermal-epidermal junction was a hallmark of LS, we systematically assessed the expression patterns of a spectrum of collagen genes across the cellular landscape. A pronounced enrichment of collagen was specifically detected in the fibroblasts (Fig. 3A and Figure S2A), which implied the pathological markers of the disease and the non-negligible role of fibroblasts. We further investigated the DEGs between the fibroblasts in the normal group and those in the LS group. A variety of significant extracellular matrix (ECM) markers were found, including COL3A1, COL1A1, and ELN (Fig. 3B). The expression patterns of collagen and elastin exhibited an inverse relationship (Fig. 3A and Figure S2A). A subsequent scatter plot confirmed the observed trend in ELN expression, demonstrating a specific accumulation in the fibroblast population (Fig. 3C). However, another ECM marker, fibronectin (FN1), did not show an evident alteration (log2-fold change: -0.05, adjusted P value < 0.001, Figure S2A and Figure S2B).

Fig. 3.

A panel of distinctive markers was identified for the LS samples. A.The violin plot of collagen gene expression among all the cell types. B.The plot of DEGs between the fibroblasts in the normal group and those in the LS group. The grey dashed lines indicated the log2-fold change of 0.25 and − 0.25. C.The scatter plot of the ELN expression among the three groups. D.The images of Picric acid-Sirius red staining samples photographed under a polarized microscope. E.The IHC images of ELN expression between the paraLS and LS groups. The dashed arrow pointed from the epidermis to the hypodermis. F.The quantitative analysis of ELN expression along the dashed arrow in Fig. 3E. DEGs: differentially expressed genes; ELN: elastin

To validate our findings, the Picric acid-Sirius red staining samples were photographed under a polarized microscope to distinguish the different types of collagen. In the normal group, collagen type I (Col-1) and collagen type III (Col-3) were in a sparse arrangement, aligned with the rete ridges (Fig. 3D). Nevertheless, in the LS group, the distribution of Col-1 and Col-3 was dense and disordered, without a clear structure of the rete ridges (Fig. 3D). On the other hand, the expression of elastin in foreskin samples was also detected in both the paraLS group and the LS group. As illustrated by the dashed arrow (Fig. 3E), which delineated the transition from the epidermis to the hypodermis, the distribution of elastin in the paraLS group was observed to be uniformly dispersed across the epidermal, dermal, and hypodermal layers (Fig. 3F). However, the expression pattern of elastin in the LS group was completely different, with elastin fibers being confined to the hypodermis instead of the affected epidermis and dermis (Fig. 3F).

Cell-cell interactions highlighted the communications from fibroblasts to T cells and keratinocytes

We have stressed the role of fibroblasts in producing collagen and noted marked alterations in the number of T cells and keratinocytes. The potential interactions among these cells piqued our interest, as elucidating such cellular crosstalks could provide critical insights into the complex biological processes at play. Therefore, a cell-cell interaction analysis was conducted based on the CellChat v1.1.3 package [25]. Our analysis revealed only minor fluctuations in the number of cellular communications across the three distinct groups (Fig. 4A). Nevertheless, the interaction strength was reduced drastically in both paraLS and LS groups (Fig. 4A). To be more specific, we delineated the cellular crosstalk among all the cell types. The cellular crosstalks among the three groups were complicated (Figure S3A and Figure S3B). Hence, we emphasized the differential interactions between the normal and LS groups (Fig. 4B and Figure S3C). Among the myriad cellular interactions observed, the most striking form of communication was identified as the signaling from fibroblasts to T cells (about 1.5-fold change in LS samples compared with normal samples, Fig. 4B). In addition to the prominent fibroblast-to-T cell signaling, another notable cellular dialogue from fibroblasts to keratinocytes came into focus (about 1.2-fold change in LS samples compared with normal samples, Fig. 4B). The intricate interplay we observed could potentially drive significant quantitative shifts in the T cell and keratinocyte populations, implicating a complex modulation of the immune and epidermal compartments in the LS disease.

Fig. 4.

The cell-cell crosstalks highlighted the signaling from fibroblasts to T cells and keratinocytes. A.The number and strength of interactions among the three groups. B.The heatmap of differential interactions between the normal and LS groups. C.The plot of the general incoming and outgoing interaction strength among the three groups. D.The specific differential interaction strength of fibroblasts, T cells, and keratinocytes between the normal and LS groups. E.The increased signaling from fibroblasts to T cells. A detailed plot was provided in Figure S4A. F.The increased signaling from fibroblasts to keratinocytes. G.The mIHC images demonstrated a strong interaction (COL6A1-CD44) between the fibroblasts (COL6A1+) and the T cells (CD3E+), as shown by the orange arrows. Scale bar: 100 μm for the images of the upper row, and 20 μm for the images of the lower row. mIHC: multiplex immunohistochemistry

Moreover, the two-dimensional plot depicted the main source and target cell types among the three groups (Fig. 4C). Our study revealed that both the afferent and efferent signaling pathways within fibroblasts exhibited a gradient descent from the normal group to the LS group (Fig. 4C). On the contrary, the incoming interaction strength of T cells was abnormally elevated (Fig. 4C). We further concentrated on the signaling of fibroblasts, T cells, and keratinocytes between the normal and LS groups (Fig. 4D and Figure S3D). Interestingly, we observed a raised incoming signaling of collagen pathways in the T cells of the LS group compared to those of the normal group (Fig. 4D), which was substantially produced by the fibroblasts (Fig. 3A and B, and Figure S2A). Based on these findings, we calculated the communication probability from the fibroblasts to the T cells and keratinocytes. For fibroblasts-to-T cells crosstalks, we focused on the up-regulated pathways (Fig. 4E and Figure S4A). For fibroblasts-to-keratinocytes interactions, both the up-regulated and down-regulated signalings were illustrated (Fig. 4F, Figure S4B, and Figure S4C). Expectedly, the most notable signaling from fibroblasts to T cells in the LS group was the collagen pathway (about 1.3-fold change of both COL1A1-CD44 and COL6A1-CD44 signaling in LS samples compared with normal samples, with P < 0.01, Fig. 4E). Conversely, the collagen pathway between fibroblasts and keratinocytes was reduced (Figure S4B and Figure S4C), which was consistent with the results in Fig. 4D and Figure S3D. Yet, the APP-CD74 pathway was drastically elevated in the fibroblasts-to-keratinocytes interplay of the LS group (only observed in LS samples compared with both normal and paraLS samples with P < 0.01, Fig. 4F). To substantiate the crosstalk between fibroblasts and T cells, the multiplex immunohistochemistry (mIHC) staining technique was applied (Fig. 4G). COL6A1 was stained for the marker of fibroblasts (Fig. 2C and Figure S1D) and the ligand of the collagen pathway (Fig. 4E and Figure S4A). CD3E was stained for the marker of T cells (Fig. 2C and Figure S1D). CD44 was stained for the receptor of the collagen pathway (Fig. 4E and Figure S4A). We did observe a strong interaction between fibroblasts and T cells in the mIHC images of the LS group (the orange arrows in Fig. 4G).

Fibroblast sub-clustering unveiled distinctive communication patterns with T cells and keratinocytes

Previous results inferred an enhanced collagen-CD44 communication between fibroblasts and T cells and APP-CD74 crosstalk between fibroblasts and keratinocytes (Fig. 4E and F). Hence, we plotted the gene expression of COL1A1, COL1A2, COL6A1, COL6A2, CD44, APP, and CD74 in the corresponding cell types. We found that collagen expression was up-regulated in the fibroblasts of the LS group compared to those of the normal group (log2-fold change: 1.03 for COL1A1, 0.72 for COL1A2, 0.36 for COL6A1, and 0.21 for COL6A2; adjusted P value < 0.001 for all four collagen genes, Fig. 5A). On the other hand, CD44 and CD74 were increased in T cells and keratinocytes, respectively (log2-fold change: 0.56 for CD44, and 3.13 for CD74; adjusted P value < 0.001 for both, Fig. 5A and B).

Fig. 5.

The sub-clustering of fibroblasts revealed unique patterns of interaction with T cells and keratinocytes. A.The gene expression of COL1A1, COL1A2, COL6A1, COL6A2, and CD44 in fibroblasts and T cells. B.The gene expression of APP and CD74 in fibroblasts and keratinocytes. C.The sub-clustering of fibroblasts. D.The plot of DEGs between the fibroblasts in the normal group and those in the LS group. The grey dashed lines indicated the log2-fold change of 0.25 and − 0.25. E.The mIHC images demonstrated a strong interaction (COL1A1-CD44) between the fibroblasts (COL1A1+) and the T cells (CD3E+). Scale bar: 100 μm. F.The volcano plot of the top DEGs in each sub-cluster of fibroblasts. G.The sub-clustering criteria of fibroblasts according to the four DEGs (MDK, COL1A1, CXCL12, and ANGPTL1) provided in Fig. 5D. +/-: significant with log2-fold change > 0.25; ++/--: significant with log2-fold change > 0.5; +++/---: significant with log2-fold change > 0.75; ++++/----: significant with log2-fold change > 1; +++++/-----: significant with log2-fold change > 2; /: in significant. H. The communication pattern annotation of fibroblasts. DEGs: differentially expressed genes; Log2FC: log2-fold change; mIHC: multiplex immunohistochemistry

Taking an additional step, we conducted a classification of fibroblasts into sub-clusters to uncover unique interaction patterns with T cells and keratinocytes. Generally, the fibroblasts were divided into 8 clusters (Fig. 5C). There were no significant differences among the three groups (Figure S5A). Based on the interaction plots (Fig. 4E and F, and Figure S4), we identified four DEGs (MDK, COL1A1, CXCL12, and ANGPTL1) between fibroblasts of the LS group and those of the normal group (Fig. 5D), which were putative ligands in the fibroblast crosstalks. To further validate the dialogue between fibroblasts and T cells, we have re-applied the mIHC staining technique. This reiteration reinforced the understanding of cellular collagen interactions (Fig. 5E). Similarly, we observed that COL1A1, CD44, and CD3E were co-expressed and localized together in the majority of areas within the LS group (Fig. 5E).

Furthermore, we tried to uncover distinctive communication patterns within fibroblasts (Fig. 5G), according to the top DEGs of each fibroblast sub-cluster (Fig. 5F and Figure S5B) and the four DEGs (MDK, COL1A1, CXCL12, and ANGPTL1) provided in Fig. 5D. Ultimately, a total of 6 communication patterns were identified (Fig. 5H and Figure S5C). Single-cell gene set enrichment analysis (scGSEA) was employed to explore the unique function of each pattern, based on the clusterProfiler package [26]. We discovered that the six patterns had a specific enriched function (Figure S5D). Although pattern 1 and pattern 6 were enriched in the ribosome pathway, the enrichment profiles of pattern 1 and pattern 6 diverged within the apoptosis pathway (Figure S5E). Particularly, pattern 3 emerged as a notable category within our analysis, where the collagen genes were enriched (Figure S5F and Figure S5G). Therefore, pattern 3 was deduced as a potential cluster to interact with the T cells through the collagen pathway. Other patterns might communicate with T cells or keratinocytes mainly through different pathways.

T cell sub-clustering exhibited an elevated proportion of exhausted phenotype in LS disease

The T cells were subjected to the sub-clustering analysis to identify a specific subtype that interacted with fibroblasts. A total of 8 clusters were found (Fig. 6A and Figure S6A). The top DEGs of each sub-cluster were illustrated in Figure S6B and Figure S6C. In Fig. 6B, we delineated the phenotypic markers characteristic of several prevalent T-cell subtypes, encompassing cytotoxic, exhausted, anergic, regulatory, and helper T cells. According to the DEGs and phenotypic markers, we annotated the T-cell clusters (Fig. 6C and Figure S6D). Based on the annotation, we discovered that a large proportion of exhausted T cells were located in the LS group, although the absolute number of all the T-cell subtypes increased simultaneously (Fig. 6D). We further performed the mIHC experiment to investigate the findings. Anticipatively, an increase in the population of CD3E+/LAG3 + cells was observed in the paraLS and LS groups (Fig. 6E).

Fig. 6.

The T cell sub-clustering exhibited an elevated proportion of exhausted phenotype in LS disease. A.The sub-clustering of T cells among the three groups. B.The dot plot of the expression of the phenotypic markers characteristic of several prevalent T-cell subtypes in the T-cell sub-clusters. C.The subtype annotation of T cells among the three groups. D.The proportion analysis of the annotated T-cell subtypes and sample groups. E.The mIHC images demonstrated an increase in the population of CD3E+/LAG3 + cells in the paraLS and LS groups. Scale bar: 20 μm. F.The plot of DEGs between the T cells in the normal group and those in the LS group. The grey dashed lines indicated the log2-fold change of 0.25 and − 0.25. DEGs: differentially expressed genes; mIHC: multiplex immunohistochemistry

To go a step further, based on the interaction plots (Fig. 4E, and Figure S4A), we identified six DEGs (TIGIT, CD74, ITGB1, CD44, PTPRC, and CXCR4) between T cells of the LS group and those of the normal group (Fig. 6F), which were putative receptors in the fibroblast-to-T cell crosstalks. It was intriguing that half of the up-regulated DEGs depicted in Fig. 6F were enriched in the exhausted T cells (Figure S6E). This observation implied that the exhausted T cells might constitute the predominant subtype engaging with fibroblasts. The expression profiles of the six DEGs among the exhausted T cells of the three groups substantiated these findings (Figure S6F). The scGSEA indicated that, compared with the exhausted T cells of the normal group, those of the LS group demonstrated a strengthened ability of antigen processing and presentation of exogenous peptide antigen, along with a weakened competence of protein folding (Figure S6G).

Keratinocyte sub-clustering pinpointed a specific subtype engaging with fibroblasts through the APP-CD74 interaction

To discern the pattern of keratinocytes interacting with fibroblasts, the keratinocytes were sub-clustered into 10 populations (Figure S7A and Figure S7B). Our observations indicated a notable reduction in the number of keratinocyte sub-clusters, with the most pronounced decrease observed in the LS group (Fig. 7A). The most prominent DEGs within each sub-cluster were depicted in Figure S7C and Figure S7D. Based on the DEGs (Figure S7C and Figure S7D) and a series of epithelial markers (Fig. 7B), four subtypes were identified, including type I basal keratinocytes, type II basal keratinocytes, spinous keratinocytes, and granule keratinocytes (Fig. 7C and Figure S7E). The proportion analysis implicated that it was type I basal keratinocytes, rather than type II basal keratinocytes, reduced progressively from the normal group to the LS group (Fig. 7D).

Fig. 7.

Sub-cluster analysis of keratinocytes delineated a unique subtype in interaction with fibroblasts via the APP-CD74 pathway. A.The sub-clustering of keratinocytes among the three groups. B.The dot plot of the expression of the epithelial markers in the keratinocyte sub-clusters. C.The subtype annotation of keratinocytes among the three groups. D.The proportion analysis of the annotated keratinocyte subtypes and sample groups. E.The expression profile of CD74 among the keratinocyte subtypes of the three groups. F.The mIHC images demonstrated a robust interaction (APP-CD74) between the fibroblasts (COL6A1+) and the keratinocytes (KRT15+). Scale bar: 40 μm. mIHC: multiplex immunohistochemistry

Expanding our analysis, we scrutinized the interaction plots (Fig. 4F, Figure S4B, and Figure S4C) to pinpoint six key genes (CD74, NRP1, ITGAV, PTPRZ1, ACVR1, and CD44). These genes, identified as potential receptors mediating fibroblast-to-keratinocyte crosstalk, were found to be differentially expressed between keratinocytes of the LS group and those of the normal group (Figure S7F). Among the identified DEGs, CD74 stood out as the most prominent, mainly expressed in the keratinocytes of the LS group (Fig. 7E). Taken together, the discovery highlighted the interaction of the APP-CD74 pathway between the fibroblasts and keratinocytes (Figs. 4F, 5B and 7E, and Figure S7F). To verify the inference, the mIHC staining technique was applied. COL6A1 was stained for the marker of fibroblasts. KRT15 was stained for the marker of basal keratinocytes (Fig. 7A and B, and Fig. 7C). APP and CD74 were stained for the ligand and receptor of the pathway. We observed a robust interaction between fibroblasts and basal keratinocytes in the mIHC images of the LS group (Fig. 7F).

Trajectory analysis revealed type I keratinocytes as a key population in initiating epidermal hyperkeratosis of LS disease

The integumentary system, comprising the foreskin, is in a perpetual state of renewal, reflecting the dynamic nature of this protective barrier. The basal keratinocytes of normal skin continuously migrate toward the stratum corneum, where keratinocytes lose their nucleus and die, which is known as epidermal keratinization. In the LS disease, hyperkeratosis has been identified as a classical characteristic [8, 10, 11]. Hence, it was within our expectations that the number of keratinocytes was drastically reduced in the LS group (Fig. 7A and C). More specifically, we discovered a unique subtype of keratinocytes, type I basal keratinocytes, which decreased notably (Fig. 7C and D). We aimed to identify the specific subgroup that played a pivotal role in the hyperkeratosis process, a critical aspect of our ongoing investigation into the molecular mechanisms underlying LS pathology. Thus, we conducted the pseudotime analysis of keratinocytes to delineate their developmental trajectory, utilizing the monocle3 package [27–33].

Interestingly, the trajectory analysis suggested a developmental pathway from the type I basal keratinocytes, through the spinous keratinocytes, to the granule keratinocytes, excluding the type II basal keratinocytes (Fig. 8A). The analysis implied an important role of type I basal keratinocytes in the occurrence and progression of LS disease. Further, scGSEA indicated an enriched basal cell carcinoma signaling in the type I basal keratinocytes of the LS group compared to those of the normal group (Fig. 8B). By contrast, the basal cell carcinoma signaling was not enriched in the type II basal keratinocytes of the LS group (Fig. 8C). Moreover, we performed the scGSEA to explore the function of type I basal keratinocytes in the LS group compared with those in the normal group (Figure S8). A significant inflammation response pathway was enriched in the type I basal keratinocytes in the LS group (adjusted P value: 1.43e-04, Figure S8).

Fig. 8.

Trajectory analysis delineated type I keratinocytes as a pivotal sub-cluster in the initiation of epidermal hyperkeratosis of LS disease. A.The pseudotime analysis of keratinocytes. B.The scGSEA to explore the differential functions of type I keratinocytes in the LS group compared with those in the normal group. C.The scGSEA to explore the differential functions of type II keratinocytes in the LS group compared with those in the normal group. D.The DEGs along the developmental trajectory of keratinocyte sub-clusters. DEGs: differentially expressed genes; scGSEA: single-cell gene set enrichment analysis; Adj. P-val: adjusted P value

Following the pseudotime analysis, we endeavored to elucidate the DEGs that manifested along the developmental trajectory, aiming to uncover the molecular underpinnings of the keratinization process. A total of eight DEGs were found (Fig. 8D and Figure S9). They were: KRT15, COL17A1, KRT1, KRTDAP, CALML5, DMKN, SBSN, and SPRR1B. Among these genes, KRT15 and COL17A1 were markers for the basal keratinocytes, and KRT1 was the marker for spinous and granule keratinocytes in our study (Fig. 7B). The three DEGs (KRT15, COL17A1, and KRT1) along the developmental trajectory confirmed the accuracy of keratinocyte sub-cluster annotation. The other five DEGs (KRTDAP, CALML5, DMKN, SBSN, and SPRR1B) could serve as novel markers in future dermatology research (Fig. 8D and Figure S9).

Implementing the multi-omics approach to filter out additional susceptible genes in LS disease

To investigate additional susceptible genes of LS disease within a large population, we made efforts in the GWAS of gene expression and LS disease. We obtained the GWAS expression profile from the eQTLGen consortium [34]. The GWAS data pertinent to LS disease were sourced from the FinnGen consortium [35]. Additional particulars concerning these datasets were articulated in Fig. 9A and Table S3. Post-GWAS analysis, the summary statistics were utilized for subsequent Mendelian randomization (MR) analysis. The MR analysis served as a powerful instrument for deducing causal relationships between two traits. This approach circumvented some of the confounding factors inherent in traditional observational studies, thereby enhancing the credibility of causal inferences drawn from genetic association data [36–41]. In this study, we implemented both the summary-based MR (SMR) and heterogeneity in dependent instruments (HEIDI) methodologies [42], to delve into the causal influence exerted by 16,987 cis-eQTLs on the LS disease phenotype (Fig. 9A).

Fig. 9.

Employing the multi-omics approach to identify additional susceptible genes in LS disease. A.The flowchart of the multi-omics analysis. B.The integration of the SMR data with the scRNA-seq data ultimately yielded two susceptible genes in LS disease. C.The expression profiles of the two filtered-out genes among all the cell types. D.The protein expression level of the two filtered-out genes among the three groups. E.The quantitative analysis of the protein expression level in Fig. 9D. The data were presented as the mean ± SEM (n = 3). F.The protein expression level of the GAS1, COL1A1, and COL6A1 genes in shRNA plasmid-transfected HFF-1. G.The expression profiles of the TGFB1 gene between GAS1⁺ and GAS1^- fibroblasts. H.The protein expression level of the GAS1, COL1A1, and COL6A1 genes in urea-stimulated HFF-1 at different concentrations. cis-eQTL: cis-expression quantitative trait locus; GWAS: genome-wide association study; SMR: summary-based Mendelian randomization; SEM: standard error of the mean; shRNA: short-hairpin RNA; HFF-1: human foreskin fibroblasts

Following the SMR analysis, a total of 1,042 genes were considered significant with p_SMR value < 0.05 (Fig. 9B and Table S6). Based on the HEIDI test, 841 genes with p_HEIDI value > 0.05 were further included (Fig. 9B and Table S6). Then we integrated the genetic data with the scRNA-seq data (Fig. 9A and B). Ultimately, we filtered out 19 genes that exhibited causal directions consistent with the differential expression patterns observed in the scRNA-seq data from the LS group, as compared to the normal group (Fig. 9B). Six genes (RPS26, ETS1, APBB1IP, RILPL2, CTSL, and BRAF) were up-regulated in the LS group and thought of as positive causality. The other thirteen genes (Sect. 63, CHMP3, ST13, NCOA7, AFAP1, RPS10, LATS2, SPSB1, EMP1, IRF2BPL, RPS20, GPX3, and GAS1) were down-regulated in the LS group and thought of as negative causality.

We plotted the expression profile of the 19 genes among all the cell types (Figure S10). Seven genes (CHMP3, ST13, RILPL2, AFAP1, GAS1, RPS10, and LATS2) were found to be enriched in a certain cell type (Fig. 9B). We concentrated on the fibroblast-T cell-keratinocyte interaction to explore the mechanism of LS disease development. Finally, two genes (ST13 enriched in the keratinocytes and GAS1 enriched in the fibroblasts) were identified (Fig. 9B and C). Both genes were hypothesized to be down-regulated in the LS group (Table S6). To certify our discoveries, the protein expression level of ST13 and GAS1 was detected (Fig. 9D). We just observed a significant reduction of GAS1 in both the paraLS and LS groups (P < 0.005 between the paraLS and normal groups, and between the LS and normal groups, Fig. 9E).

In addition, we examined the role of GAS1 in the collagen secretion of HFF-1. We knocked down the GAS1 expression in HFF-1 and found that the protein expression of COL1A1 and COL6A1 was elevated (Fig. 9F). To go a step further, we analyzed the expression of the TGFB1 gene, which was considered a classical factor promoting fibrosis, between GAS1⁺ and GAS1⁻ fibroblasts. We discovered that the expression level of TGFB1 in GAS1⁻ fibroblasts was significantly higher than that in GAS1⁺ fibroblasts (P < 0.001, Fig. 9G). Moreover, we stimulated the HFF-1 with different concentrations of urea. We found that the GAS1 expression showed a concentration-dependent decreasing trend, while the COL1A1 and COL6A1 expression showed an increasing trend (Fig. 9H).

Discussion

In this research, we employed a multi-omics strategy to identify and elucidate the underlying histological biomarkers and the fundamental pathogenesis associated with male genital LS. Generally, a comprehensive cell atlas of LS disease was constructed, highlighting a pronounced increase in T cells coupled with a remarkable reduction in keratinocytes within the LS samples. A panel of ECM markers was found, including COL3A1, COL1A1, and ELN. Further insights indicated a heightened interaction from fibroblasts to T cells through the collagen-CD44 pathway and to keratinocytes through the APP-CD74 signaling, which contributed to the immune infiltration and hyperkeratosis observed in the epidermis and dermis of LS disease. These pathophysiological processes collectively led to the characteristic histological features of LS (Fig. 10). Subsequently, we integrated our scRNA-seq findings with data from the GWAS to investigate the susceptible genes underlying the development of LS, emphasizing the role of GAS1 enriched in the fibroblasts in inducing LS progression. In summary, the multi-omics methodology facilitated our understanding of the potential mechanism associated with male genital LS.

Fig. 10.

The mechanism diagram of LS disease

So far, the diagnosis of LS frequently hinges on the recognition of clinical manifestations, including persistent pale, atrophic lesions around the anogenital region accompanied by pruritus and discomfort [44]. For male genital LS, the diagnosis can also rely on a circumcised foreskin sample. However, the histological identification of LS can be intrinsically difficult, especially during the initial or less severe phases, as the characteristics may be invisible and not distinctive, easily confused with other disorders exhibiting an interface reaction pattern [10, 12]. Consequently, we meticulously compiled the histological characteristics observed in LS cases, thereby establishing a comprehensive set of diagnostic criteria. Our findings indicated that in all instances, both the presence of collagen homogenization and inflammatory infiltration were noted. Conversely, the foreskin samples presenting LS characteristics showed either atrophic epidermis or hyperkeratosis (Fig. 1F). That was why the thickness of the epidermis in the LS group was not distinguished from that in the normal group (Fig. 1C). From our point of view, a suspicious case with all the following pathological characteristics would be considered LS: collagen homogenization, inflammatory infiltration, and atrophic epidermis or hyperkeratosis.

Another clinical dilemma was the current absence of a diagnostic panel for LS disease. This research introduced novel diagnostic markers for the disease, highlighting the value of COL3A1, COL1A1, and ELN. In particular, the COL3A1 and COL1A1 were massively sedimented at the dermal-epidermal junction, where the expression of ELN was notably absent in this region. This discovery could potentially enhance the diagnostic accuracy for LS disease by providing specific markers to target.

The cell number proportion and cell-cell interaction analysis implicated enhanced crosstalk among the fibroblasts, T cells, and keratinocytes. Currently, few studies have investigated the underlying mechanism of male genital LS disease [45–48]. They predominantly relied on retrospective observations, which were deficient in a thorough exploration of the underlying mechanisms. Therefore, we for the first time delineated the landscape of male genital LS, indicating the fibroblast-to-T cell interaction through the collagen-CD44 pathway, which was a novel insight. The CD44 protein is involved in a diverse array of cellular processes, encompassing the activation and homing of lymphocytes [49–51]. We considered the fibroblast-produced collagen recruited and induced the local infiltration of T cells. Additionally, the collagen-CD44 signaling axis has been documented as playing a role in the migrative and invasive behavior of head and neck squamous cell carcinoma [52]. Altogether, the extensive deposition of collagen in LS disease contributed to the migration and activation of T cells to the dermal-epidermal junction.

On the other hand, pieces of evidence pointed to a strengthened communication between the fibroblasts and keratinocytes through the APP-CD74 pathway. The APP (amyloid precursor protein) was first reported in Alzheimer’s disease [53, 54]. However, in recent years, the protein has been associated with other diseases [55, 56]. While those researchers have established that elevated expression levels were intimately linked to disease progression and invasiveness [57–61], the precise mechanisms at play were not yet fully understood, and few prior dermatological studies reported the association between APP and dermatological diseases. We speculated that the APP-CD74 interaction from fibroblasts to keratinocytes facilitated the proliferation of basal keratinocytes, resulting in hyperkeratosis characteristic of the LS disease. Our ongoing research will delve deeper into this subject matter.

We observed an elevated proportion of exhausted T cells in the LS disease. Exhausted T cells, emerging in chronic stimulation and cancer, signify a compromised immune cell function state [62, 63]. As we know, LS could be acknowledged as a disease typical of chronic inflammation and irritation. After infiltrating the collagen deposition area, T cells gradually exhibited an exhausted phenotype due to constant collagen stimulation. During the process, other signalings, like APP-CD74, CXCL12-CXCR4, and NECTIN-TIGIT pathways, might exert their roles.

We identified a specific keratinocyte subtype that impacted the keratinization process. It was the type I basal keratinocytes rather than the type II basal keratinocytes that functioned in the LS pathogenesis. The type I basal keratinocytes were over-activated and progressively migrated toward the stratum corneum in the LS disease. That was why only an enriched basal cell carcinoma signaling was observed in the type I basal keratinocytes of the LS group compared to those of the normal group, indicating a state of over-proliferation. We assumed that an inflammatory state of keratinocytes was able to accelerate the keratinization process. Moreover, a series of markers along the keratinization trajectory were discerned (KRT15, COL17A1, KRT1, KRTDAP, CALML5, DMKN, SBSN, and SPRR1B). In our research, KRT15 and COL17A1 were markers for the basal keratinocytes, while KRT1 was for spinous and granule keratinocytes, which validated the precision of the keratinocyte sub-cluster annotation. Some of the other markers were also consistent with the results of high-quality papers, including KRTDAP, DMKN, and SBSN [64–66]. The CALML5 and SPRR1B were novel markers for spinous and granule keratinocytes, with few reports, which could benefit future studies on keratinocyte sub-categorization.

Utilizing a multi-omics strategy coupled with experimental verification, GAS1 was pinpointed as the most dependable candidate gene associated with the progression of LS. The expression profile of GAS1 indicated an enrichment in the fibroblasts, which was thought of as a protective factor. The preliminary findings of our study implied that fibroblasts were instrumental in initiating immune cell penetration and hyperkeratosis, both of which were hallmark features of LS disease. Till now, the predominant etiological theory for male genital LS implicates the persistent contact of the foreskin with urine that is impeded in its evacuation, leading to a chronic state of irritation that may precipitate the disorder [2, 9, 13, 14]. We inferred that the continuous irritation of urine down-regulated the expression of GAS1 in the fibroblasts, thereby instigating an overproduction of collagen and prompting the subsequent cell-cell interactions. GAS1 has been documented to function as a suppressor in the oncogenic processes associated with various cancers [67, 68]. Future research endeavors would be directed towards dissecting the role of GAS1 in fibroblasts, with a specific emphasis on its potential to initiate the overabundance of collagen and the cascade of mechanisms it set into motion.

Interestingly, a recent multi-modal study on vulvar LS has incorporated both scRNA-seq and spatial transcriptomics to characterize fibroblast heterogeneity, keratinocyte differentiation, and immune cell states, which was posted on a pre-printed platform [69]. They highlighted the collaborative roles of fibroblasts, keratinocytes, and immune cells in vulvar LS pathogenesis, which was consistent with our study findings. Moreover, the study also discovered the immune dysregulation featuring T cell infiltration and keratinocyte stress responses marked by premature differentiation. However, the study demonstrated that knockdown of ATF3 and APOE in keratinocytes recapitulated vulvar LS-like stress and differentiation defects, which were different from our results and seemed to be sex-specific. Future research should focus on establishing correlations between the findings observed in male genital LS and those documented in vulvar LS.

To our knowledge, this investigation represented the inaugural study to delve into the intrinsic etiology of LS disease, employing a comprehensive multi-omics methodology. However, several limitations should be acknowledged. We faced difficulty in procuring lesion tissues from individuals who had undergone circumcision and subsequent treatment, including the administration of glucocorticoids. The aforementioned circumstances precluded us from substantiating our deductions. On the other hand, our scRNA-seq dataset was primarily derived from an Asian demographic, whereas the GWAS dataset was predominantly European in origin. The GWAS dataset of LS also included lesions from the extra-genital region. Nevertheless, our efforts were directed toward uncovering the conservative susceptible gene that influenced the advancement of LS. Consequently, our research carried substantial implications, and further inquiries were essential to validate our findings in various animal models including primates and mice [70, 71, 72].

In conclusion, our research pioneered a holistic view of LS disease by leveraging a multi-omics approach. We underscored the pivotal role of fibroblasts in initiating LS onset, generating interactions with T cells and keratinocytes, and eliciting the classical histological features of LS. Additionally, we determined a set of diagnostic markers for LS that could inform future clinical assessments.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Author contributions

Conception and design of the study: Lede Lin, Yu Liu, Shiqian Qi, and Liang Zhou. Acquisition of data: Lede Lin, Yu Liu, Xiaocheng Wang, Kun Liu, Wei Wang, Linhu Liu, Yaohui Jiang, and Jiawei Chen. Data analysis and/or interpretation: Lede Lin, Yu Liu, Xiaocheng Wang, Kun Liu, and Wei Wang. Drafting of manuscript and/or critical revision: Lede Lin, Yu Liu, Xiaocheng Wang, Kun Liu, Wei Wang, Dan Tang, Di Jiang, Xiang Li, Banghua Liao, Shiqian Qi, and Liang Zhou. Approval of final version of manuscript: Lede Lin, Yu Liu, Xiaocheng Wang, Kun Liu, Wei Wang, Linhu Liu, Yaohui Jiang, Jiawei Chen, Dan Tang, Di Jiang, Xiang Li, Banghua Liao, Shiqian Qi, and Liang Zhou.

Funding

The study was supported by the National Natural Science Foundation of China (Grant No. 32171301, 82200851, 32201025, 32071214), the Project of Science and Technology Department of Sichuan Province (Grant No. 2023NSFSC1533), the Project of Technology Transfer of West China Hospital of Sichuan University (Grant No. CGZH21012), and the Project of West China Hospital of Sichuan University (Grant No. ZYYC23012).

Data availability

All data generated or analyzed during this study will be made available upon reasonable request.

Declarations

Ethics approval and consent to participate

This research was conducted following the principles outlined in the Declaration of Helsinki and received approval from the Ethics Committee of West China Hospital of Sichuan University (approval number: 20241358). The consents of all participants were duly obtained in an informed manner.

Consent for publication

Not applicable.

Competing interests

All authors have no conflicts of interest or financial ties to disclose.