Leiwei Jiang, Email leiweijiang@163.com.Abstract Received 2025 Feb 12; Accepted 2025 May 20; Collection date 2025. © 2025 Fu et al. This work is published and licensed by Dove Medical Press Limited. The full terms of this license are available at [27]https://www.dovepress.com/terms.php and incorporate the Creative Commons Attribution – Non Commercial (unported, v4.0) License ([28]http://creativecommons.org/licenses/by-nc/4.0/). By accessing the work you hereby accept the Terms. Non-commercial uses of the work are permitted without any further permission from Dove Medical Press Limited, provided the work is properly attributed. For permission for commercial use of this work, please see paragraphs 4.2 and 5 of our Terms ([29]https://www.dovepress.com/terms.php). [30]PMC Copyright notice PMCID: PMC12132621 PMID: [31]40463861 Abstract Background Androgenetic alopecia (AGA) is the most common type of androgen-associated hair loss. Emerging evidence highlights inflammation as a critical mediator in follicular miniaturization and disease progression. This investigation systematically explores inflammatory mechanisms in AGA through comprehensive analysis of hair follicles transcriptional profiles combined with cellular heterogeneity. Methods Matched follicular specimens were procured from AGA patients: occipital non-balding units (controls) versus frontal alopecic zones (experimental). Bulk RNA-sequencing was conducted on Norwood-Hamilton grade 3–5 AGA scalp tissues to delineate inflammatory signatures. Subsequent single-nucleus RNA sequencing (snRNA-seq) of grade 5 specimens resolved cellular heterogeneity. Immune subsets (NK/CD8^+ T cells), vascular endothelia (BECs), keratinocytes, and fibroblasts were transcriptionally characterized. Findings were validated through immunofluorescence cytochemistry (IFC) and reverse transcription quantitative PCR (RT-qPCR). Results Bulk RNA-sequencing of AGA hair follicles revealed heightened inflammatory signatures in grade 5 patients compared to grade 3–4 counterparts. To dissect cellular heterogeneity, we systematically investigated the dynamic changes of immune cells in hair follicles of AGA patients using snRNA-seq technology for the first time. The result showed that grade 5 AGA hair follicles, identifying significant enrichment of natural killer (NK) and CD8^+ T cells in balding hair follicles. Concurrently, blood endothelial cells (BECs) in balding follicles exhibited downregulation of angiogenesis-related genes. Notably, IL-15—a cytokine critical for NK/CD8^+ T cell proliferation—was overexpressed in BECs, keratinocytes, and fibroblasts, suggesting a microenvironmental cue for immune cell expansion. Conclusion These findings collectively implicate NK and CD8^+ T cell infiltration as drivers of inflammatory exacerbation in AGA. By blocking IL-15 signaling-mediated immune activation may be an innovative therapeutic approach to promote hair regeneration in AGA patients. Keywords: androgenetic alopecia, natural killer cells, CD8^+ T cells, inflammation and snRNA-seq Introduction Hair loss is a global challenge that significantly impacts patients, leading to issues such as obsessive-compulsive disorder, interpersonal sensitivity, depression, and anxiety.[32]^1^,[33]^2 A substantial number of hair loss cases are attributed to androgenetic alopecia (AGA), which is primarily driven by dysregulation in the microenvironment surrounding hair follicles.[34]^3^,[35]^4 While factors such as androgen, vascular insufficiency and oxidative stress are often mentioned,[36]^5^,[37]^6 the role of the inflammation has emerged as a critical component in the pathophysiology of AGA. Although AGA is commonly categorized as a non-inflammatory and non-scarring form of alopecia, histological evidence indicates that inflammation is indeed present.[38]^7 Jaworsky et al found that the normal hair follicles lacked hair follicular inflammation, whereas hair loss regions of AGA patients consistently exhibited activated T cell infiltrates around the lower portions of the hair follicular infundibula.[39]^7 Mahé et al introduced the term “microinflammation” in AGA, and this term is increasingly recognized in the hair follicles of individuals with alopecia. Inflammatory response can lead to progressive fibrosis of the perifollicular sheath, disrupting normal hair cycling and ultimately contributing to hair loss.[40]^8 The presence of inflammation in the hair follicles is emerging as a significant aspect of understanding the complex etiologies of AGA.[41]^9 It has been established that the immune microenvironment in AGA has significant changes, while the precise patterns of these changes remain unclear. Studies show that the rise in γδ T cells, central memory CD8^+ T cells, activated CD8^+ T cells, and the infiltration of mast cells have contributed to changes in the immune microenvironment in AGA.[42]^10 Hair follicle stem cells (HFSCs) are crucial for the coordinated regeneration of hair follicles. Perifollicular T lymphocytes serve as critical modulators of HFSCs state transitions—encompassing quiescence, proliferation, and differentiation—across homeostatic and pathophysiological conditions.[43]^11 Disruption of this dynamic relationship can lead to significant hair loss. An imbalance in immune cell infiltration may affect the hair follicle cycles, contributing to hair loss.[44]^12 However, research on immune infiltration in male AGA is limited. The hair follicle is a miniature organ containing several cellular components. Due to the intricate cellular composition of the hair follicle, the depth of single-cell RNA sequencing is constrained, making it challenging to accurately detect specific components like the cortex, medulla, and matrix.[45]^13 To further explore the role of different cell in AGA, we utilized bulk RNA-seq and snRNA-seq to analyze the immune differences between the forehead alopecia area (balding group) and the normal occipital region (control group) in AGA patients. Method Selection of Human Samples Human hair follicle samples were obtained from patients with AGA who underwent hair transplant surgery. This study was approved by the Human Research Ethics Committee of Xiang’an Hospital of Xiamen University (XAHLL2023011) and adhered to the principles outlined in the Declaration of Helsinki. Prior to their participation, written informed consent was obtained from all individuals involved. Exclusion criteria included: use of hair growth-promoting agents (eg, finasteride, minoxidil) within the past 18 months; use of immunomodulatory medications (eg, corticosteroids, immunosuppressants) or participation in other drug trials within the past 3 months; and presence of severe visceral/cutaneous diseases affecting hair growth or any systemic disorders. All samples were obtained from AGA patients, with the control group consisting of hair follicles harvested from the non-balding occipital region and the experimental group comprising follicles collected from the frontal balding area. RNA-seq analysis was performed on hair follicles from 9 AGA patients (three patients in each of Hamilton grades 3 to 5), and snRNA-seq was conducted on hair follicles from 3 grade 5 AGA patients, all aged 25–40 years. The detailed information of the study participants is provided in [46]Supplementary Data 1. Bulk RNA Sequencing RNA was extracted from frozen hair follicles using Trizol (Alkali Scientific, TRZ-100). The differences in transcript abundance between different genotypes were calculated for each gene using the R package DESeq2 (version 1.30.0). Genes with adjusted P-value < 0.05 were considered to be differentially expressed. Nucleus Isolation, Sequencing, and Cell Clustering Eligible participants were males aged 25–40 years with grade 5 AGA patients. Hair follicle samples were placed in liquid nitrogen and processed for nucleus isolation, followed by sequencing according to the guidelines of 10X Genomics (USA). The cell-by-gene matrices for each sample were individually imported to Seurat (version 3.1.1)[47]^14 for downstream analysis. To reduce the impact of batch effects and behavioral conditions on clustering, we employed Harmony, an algorithm that maps cells into a common embedding where cells are grouped by cell type rather than dataset-specific factors, allowing us to aggregate all samples. The Harmony algorithm takes a PCA embedding of cells along with their batch assignments as input and produces a batch-corrected embedding as output.[48]^15 For visualization of clusters, uniform manifold approximation and projection (UMAP) and t-distributed Stochastic Neighbor Embedding (t-SNE) was generated.[49]^16 Differentially Expressed Genes Analysis Kyoto Encyclopedia of Genes and Genomes(KEGG) is a meticulously curated database resource that integrates diverse biological entities categorized into systems, genomic information, chemical data, and health-related information.[50]^17 Differentially expressed genes (DEGs) were analyzed using the KEGG database to identify enriched pathways, with significant terms (p < 0.05) highlighted. Additionally, gene expression was assessed through gene set enrichment analysis (GSEA), which revealed hallmark gene sets that were either upregulated or downregulated. Reverse Transcription-Quantitative Real‐time Polymerase Chain Reaction (RT‐qPCR) Assay Total RNA was extracted using NcmZol Reagent (NCM Biotech, China), and cDNA was synthesized using HiScript III Reverse Transcriptase (Vazyme, China). ChamQ Universal SYBR qPCR master mix (Vazyme, China) was used for the RT-qPCR. The primers used are shown in [51]Table 1. Table 1. Primers Used for RT-qPCR of Genes Gene Strand Primer Sequence Homo-Tbp Forward CCTTGTACCCTTCACCAATGAC Reverse ACAGCCAAGATTCACGGTAGA Homo- GZMA Forward TCTCTCTCAGTTGTCGTTTCTCT Reverse GCAGTCAACACCCAGTCTTTTG Homo- GZMB Forward CCCTGGGAAAACACTCACACA Reverse GCACAACTCAATGGTACTGTCG Homo- PRF1 Forward GACTGCCTGACTGTCGAGG Reverse TCCCGGTAGGTTTGGTGGAA Homo-TNFα Forward GAGGCCAAGCCCTGGTATG Reverse CGGGCCGATTGATCTCAGC Homo- IFNG Forward TCGGTAACTGACTTGAATGTCCA Reverse TCGCTTCCCTGTTTTAGCTGC Homo-CCL5 Forward CTGCAGAGGATCAAGACAGCA Reverse GCGGGCAATGTAGGCAAAG [52]Open in a new tab Immunofluorescence Human hair follicle samples were fixed in 10% neutral buffered formalin for 24 h and dehydrated using the LEICA ASP 200S tissue processor. Then processed for paraffin embedding and blocks were sectioned at 10uM. Sections were incubated at 60 °C for 1 hour, then subjected to a series of alcohol dips to clean and rehydrate them. Antigen retrieval was performed using a citrate-based solution at 100 °C for 10 minutes, followed by fixation in 4% paraformaldehyde and washing with PBS. The sections were blocked, incubated with primary antibodies overnight at 4 °C, and then with secondary antibodies before imaging with a Leica Aperio Versa 200. Antibodies used included anti-Ki67 (Abcam, ab15580), anti-CD8 (Abcam, ab316778) and anti-CD56 (CST, 3576). Statistical Analysis The data are expressed as the mean ± SD from a minimum of three independent biological replicates. Differences between groups were compared using a multiple t-test where applicable. A p-value of less than 0.05 was regarded as statistically significant. Statistical analysis was conducted using GraphPad Prism 9.0. Result Inflammation in AGA Patients Gradually Intensify as the Progresses To investigate the cellular and molecular mechanisms about AGA, we conducted bulk RNA sequencing (RNA-seq) and single-nuclei RNA sequencing (snRNA-seq) on samples collected from the forehead alopecia area (balding group) and the normal occipital region (control group) of AGA patients ([53]Figure 1A). Principal component analysis (PCA) demonstrated that the transcriptome data were both reliable and reproducible. In patients with AGA grades 3–5, there was little variation in gene expression patterns among the control group, whereas the balding group exhibited distinct gene expression patterns ([54]Figure 1B). We performed differential gene expression (DGE) analysis for each cluster and found that AGA patients with grade 5 had the highest number of differential genes between control and balding samples ([55]Figure 1C). KEGG pathway enrichment analysis of the differentially expressed genes indicated that, among the top 15 enriched signaling pathways, grades 3 and 4 were associated with fewer pathways linked to immune responses and viral or bacterial infections. In contrast, grade 5 has 14 pathways related to immunity and infection ([56]Figure 1D). Figure 1. [57]Figure 1 [58]Open in a new tab Inflammation in AGA patients gradually intensify as the progresses. (A) Flowchart depicting the overall experimental design of this study. The number of patients with bulk RNA sequencing and snRNA-seq were indicated. (B) PCA illustrates the clustering of samples from various experimental groups. (C) Differential gene expression analysis showing up- and down-regulated genes across all 9 clusters. (D) KEGG pathway enrichment analysis of the differentially expressed genes (DEGs) between balding and control group in AGA patients with grades 3–5 (The red arrows indicate immune-related signaling pathways, while the blue arrows indicate pathways associated with bacterial or viral infections). snRNA-Seq Map of Hair Follicle Comparing Balding and Control in AGA Patients Given the pronounced inflammation in grade 5 AGA patients, we selected 3 such patients for snRNA-seq. Their specific hair loss conditions are illustrated in [59]Figure 2A. Control hair follicles were obtained from the normal occipital region, while balding hair follicles were collected from the frontal hair loss area ([60]Figure 2B). Hair follicles were collected from three donors and subjected to snRNA-seq. Multiple markers were then used to separate different cell clusters, resulting in the identification of 15 distinct cell types ([61]Figure 2C, [62]Supplementary Figure 1A and [63]Supplementary Data 2). As expected, the balding samples contained more SG cells and fewer ORS cells than the control hair follicles. Notably, no difference in bulge cells was observed between the balding and the control samples ([64]Supplementary Figure 1B). Identified bulge (LRIG1^+SOX9^+ ITGA6^+),[65]^18–20 ORS (KRT5^+ KRT17^+),[66]^21 sebaceous glands (PPARA^+ PPARG^+ELOVL5^+),[67]^22–24 immune (PTPRC^+),[68]^25 blood endothelial cells (VWF^+MCTP1^+), fibroblasts (COL1A1^+COL3A1^+DCN^+)[69]^26^,[70]^27 and keratinocytes (KRT10^+ KRT1^+ S100A8^+)[71]^27 marker genes were cell-type specific and matched the established, traditional markers for each cell population ([72]Figure 2D, [73]Supplementary Figure 1C). In the balding samples, there was a significant lack of proliferating cells, particularly in the matrix, bulge, IRS, and ORS cells ([74]Figure 2E and [75]F). Figure 2. [76]Figure 2 [77]Open in a new tab snRNA-seq map of hair follicle comparing balding and control in AGA patients. (A) Clinical pictures of the 3 patients that were included in this study with AGA. (B) Clinical pictures of control hair follicles from the normal occipital region and balding hair follicles from the frontal hair loss area in AGA patients. (C) snRNA-seq of human hair follicle visualized with uniform manifold approximation and projection (UMAP) plot. (D) Marker gene expression in 15 cell clusters. (E) Feature plots of disease groups showing the proliferation marker MKI67. (F) Immunostaining of hair follicles with anti-Ki67 (red) and nuclei were counter-stained with 4′,6-diamidino-2-phenylindole (DAPI) (blue). Zoomed-in images are shown in the right panel. Abbreviations: ORS, outer root sheath; IRS, inner root sheath; BEC, blood endothelial cells; SMC, smooth muscle cells; LEC, lymphatic endothelial cells. CD8^+ T Cells and NK Cells are the Primary Expanded Immune Population in Balding Hair Follicles To obtain a detailed profile of immune cells infiltrating the skin, we analyzed immune cell clusters comprising 7515 cells, identifying five lymphoid clusters, three myeloid clusters, and one lymphatic vessel cluster ([78]Figure 3A, [79]Supplementary Figure 2A and [80]Supplementary Data 2). The CD4 clusters included a CD4^+ native T-cell cluster with a mix of CD8^+ cells (T1: LEF1^+IL7R^+CCR7^+), helper T cells and regulatory T cells (T2: IL32^+IL2RA^+FOXP3^+), and T follicular helper cells (T3: CXCL13+IL21+ICOS+) ([81]Figure 3B). We observed a significant cluster of CD56^+CD16^+ cells that also contained some CD8^+ cells (cluster 0), which we designated as NK_CD8T cells ([82]Figure 3E and [83]Supplementary Figure 2B). Among the transcriptionally distinct subtypes identified, the NK_CD8T cell population was significantly increased in balding compared to control hair follicles ([84]Figure 3C and [85]D, [86]Supplementary Data 3). All NK_CD8T cells expressed the chemokine ligands granzymes A (GZMA), natural killer cell granule protein 7 (NKG7), and perforin 1 (PRF1), which related to cytotoxicity and cell killing[87]^28 ([88]Figure 3E). Figure 3. [89]Figure 3 [90]Open in a new tab CD8+ T cells and NK cells are the primary expanded immune population in balding hair follicles. (A) UMAP plot of the immune cluster. (B) Dot plot with canonical markers indicative of respective immune subsets. (C) Proportion of cells from major cell types identified in balding and control samples. (D) Percentage of each immune cell subset among all immune cells. (E) Feature plots of selected NK_CD8T cell canonical genes. Abbreviations: Mac, Macrophage; Mono, Monocyte; DC, Dendritic cell. The Cytotoxic of CD8^+ T and NK Cells Is Increased in the Hair Follicles of AGA Patients with Grade 5 NK cells and CD8^+ T lymphocytes can kill target cells by secreting cytokines such as interferon gamma (IFN-γ) and tumor necrosis factor-alpha (TNFα), and by producing cytotoxic granules, including PRF1 and granzymes A, B, K, and M (GZMA, GZMB, GZMK, GZMM).[91]^29^,[92]^30 To examine the infiltration condition of these immune cells in different grades of AGA, we performed RT-qPCR on hair follicles from the frontal hair loss area and the occipital control region in AGA grade 3, 4, and 5 patients. Our results showed an increase in the expression of IFNG and TNFα in the frontal hair loss area compared to the occipital control, though this difference was not statistically significant ([93]Figure 4A). Notably, PRF1 expression was significantly higher in the frontal hair loss area of grade 3–5 AGA patients ([94]Figure 4A). Compared to the control group, the expression levels of GZMA and GZMB showed an upward trend in balding hair follicles. Furthermore, in grade 5 AGA patients, the expression of PRF1, GZMA and GZMB were significantly higher in both balding and control hair follicles, compared to grade 3 and grade 4 patients ([95]Figure 4A). Bulk RNA sequencing further revealed significant upregulation of GZMA, GZMB, GZMM, GZMK, PRF1, and NKG7 ([96]Figure 4B). Additionally, CD27, which affects NK cell cytotoxicity and cytokine production,[97]^31 was upregulated in balding hair follicles ([98]Figure 4B). We also observed an increase in CCL5 expression in balding hair follicles; CCL5 is a chemokine primarily produced by NK cells and CD8^+ T lymphocytes[99]^32 ([100]Figure 4A and [101]B). Gene set enrichment analysis (GSEA) revealed significant enrichment of natural killer cell mediated cytotoxicity and cytokine-cytokine receptor interaction pathways in the balding group ([102]Figure 4C, [103]Supplementary Figure 2C). Finally, we confirmed the elevated of CD8^+ T and CD56^+ NK cell at the protein level in balding hair follicles ([104]Figure 4D and [105]E). Figure 4. [106]Figure 4 [107]Open in a new tab The cytotoxic of CD8^+ T and NK cells is increased in the hair follicles of AGA patients with grade 5.(A) Measurement of cytotoxicity-associated genes using qRT-PCR (*p < 0.05). (B) Heat map illustrating the expression of cytotoxicity-associated genes based on previous profiling studies. (C) Gene set enrichment analysis of the natural killer cell-mediated cytotoxicity and cytokine-cytokine receptor interaction pathway in AGA patients with grades 3–5. (D) Representative immunofluorescence images showing CD8 and CD56 expression in hair follicles of control and balding. Red arrows indicate CD8-positive cells, green arrows indicate CD56-positive cells. (E) Quantification of CD8 and CD56 positive cell in hair follicle (**p < 0.01). The Angiogenic Factors are Reduced in the BECs of Balding Hair Follicles Next, we examined the effects of NK and CD8^+ T cells on other cells within the hair follicle. The results showed that genes related to CD8^+ T cell cytotoxicity, such as RORA,[108]^33 were upregulated in the BECs, fibroblasts and keratinocytes of balding hair follicles ([109]Figure 5A). Additionally, genes associated with the activation of NK and CD8^+ T cells—IL1R1,[110]^34 IL33[111]^35 and IL-15[112]^36—were predominantly upregulated in BECs. IL1R1 and IL-15 were also upregulated in fibroblasts and keratinocytes. GZMA was upregulated in both immune cells and BECs ([113]Figure 5A). Figure 5. [114]Figure 5 [115]Open in a new tab The angiogenic factors are reduced in the BECs of balding hair follicles. (A) Dot plot depicting the variations between the control and balding groups for selected genes. (B) UMAP visualization representing the BECs cluster. (C) Dot plot highlighting canonical markers associated with specific BECs subsets. (D) Proportion of each BECs cell subset within the total BECs population. (E) Violin plots of selected genes related to angiogenesis. (F) Violin plots of NK and CD8+T cell associated genes. Abbr: Blood endothelial cell (BEC); proliferating (pro). Within the BECs clusters, we identified four distinct subsets: proliferating BECs (BECpro: TUBA1B^+CENPF^+), venular BECs (VWF^+ACKR1^+SELE^+), arteriolar BECs (TM4SF1^+ICAM2^+STMN1^+IGF2^+), and capillary BECs (SPRY1^+PLVAP^+ENG^+) ([116]Figure 5B and 5C, [117]Supplementary Data 2). The BECs subsets did not show significant differences across the groups ([118]Figure 5D, [119]Supplementary Data 3), but we did observe a decrease in the expression of genes associated with angiogenesis, such as VEGFC, VWF, FGFR1, and FGFR2 ([120]Figure 5E). Interestingly, both RORA and IL15 were elevated in the BECs of balding hair follicles, with IL15 showing a particularly pronounced increase in venular BECs ([121]Figure 5F). Compositional Differences in Fibroblasts and Keratinocytes Among keratinocytes (KCs), subclusters include proliferating KCs (KCpro) marked by MKI67^+TOP2A^+CENPF^+, basal KCs expressing COL17A1^+ KRT15^+ KRT14^+, spinous1 KCs characterized by KRT6^+ KRT6^+ KRT17,^+[122]^37 spinous2 KCs with KRT1^+ KRT10^+, and supraspinous KCs identified by SLURP1^+ CALML5^+ SPRR2E^+([123]Figure 6A and [124]B). Additionally, channel cells were found to express “Channel-ATPase” genes (ATP1A1^+ SAT1^+), which regulate ion channels, while eccrine cells were marked by AQP5^+ KRT19^+([125]Figure 6A and [126]B, [127]Supplementary Data 2). Quantitative analysis revealed no significant differences in the numbers of KCpro, spinous 2, supraspinous, channel, and eccrine cells across the groups, while a significant decrease in spinous1 KCs was observed in balding hair follicles ([128]Figure 6C and [129]Supplementary Data 3). GSEA revealed significant enrichment of natural killer cell mediated cytotoxicity pathways in the KCpro and spinous1 KCs of balding hair follicle ([130]Figure 6D). IL15, IL33, IL1R1, and interferon gamma receptors 1 and 2 (IFNGR1 and IFNGR2), which associated with cytotoxic of NK and CD8+ T cells, were also found to be upregulated in the KCs of balding hair follicles ([131]Figure 6E). Figure 6. [132]Figure 6 [133]Open in a new tab Heterogeneity of fibroblasts and keratinocytes from the hair follicle snRNA-seq dataset. (A, F) UMAP plots depicting the clusters of keratinocytes and fibroblasts identified in hair follicles. (B, G) Dot plots showing the markers characteristic of the respective keratinocyte and fibroblast subsets. (C, H) Percentage distribution of keratinocyte or fibroblast subset within the total cell population (*p < 0.05, **p < 0.01). (D) GSEA plot of the natural killer cell-mediated cytotoxicity pathway in Spinous1 and KCpro subset. (E, I) Violin plots illustrating the expression of genes associated with NK and CD8+ T cells. Abbreviations: KC, Keratinocyte; FB, Fibroblast; pro, proliferating. Fibroblasts (FBs) play a vital role in hair growth by producing and maintaining the connective tissue that supports the hair follicle’s function. In our study, we examined the heterogeneity of FBs in patients with AGA and identified five distinct subgroups through secondary analysis ([134]Figure 6F and [135]G). The largest subgroup was myofibroblasts, characterized by the markers LRRC15^+HOPX^+MMP2^+. Other identified subgroups included proliferating fibroblasts (FBpro), marked by COL15A1^+PI16^+; epithelial FBs (SOX5^+SOX6^+); mesothelial FBs (KRT18^+SFRP1^+); and inflammatory FBs (C3^+CCL19^+PTGDS^+) ([136]Figure 6F and [137]G, [138]Supplementary Data 2). In balding hair follicles, the number of myofibroblasts was significantly reduced. While no significant differences were observed in the inflammatory FBs, there was a noticeable upward trend in their numbers within balding hair follicles ([139]Figure 6H, [140]Supplementary Data 3). Additionally, the expression of CCL19, a cytokine secreted by fibroblasts that enhances the recruitment of CD8^+ T and NK cells,[141]^38–40 was elevated in the inflammatory FBs of balding hair follicles. Similarly, IL-15 expression was also upregulated in inflammatory FBs ([142]Figure 6I). Discussion It is well-established that microinflammation and abnormal immune responses occur in the hair follicles of AGA, yet the exact patterns of immune dysregulation remain unclear. Our bulk RNA-seq analysis showed that inflammation in grade 5 AGA patients’ balding hair follicles is significantly more severe than in those with grades 3 and 4. snRNA-seq further revealed notable differences in the proportions of NK and CD8^+ T cells between balding and non-balding follicles. GSEA showed significant enrichment of NK cell-mediated cytotoxicity and cytokine-receptor interaction pathways in the balding group. IL-15 promotes the expansion and maintenance of CD8^+ T cells and NK cells while enhancing the production of granzymes, IFN-γ, and TNF-α by CD8+ T cells and NK cells, thereby strengthening cytotoxic and inflammatory responses;[143]^41–43 our results further show that IL-15 is highly expressed in BECs, keratinocytes, and fibroblasts, suggesting that CD8^+ T and NK cells may inhibit hair follicle regeneration via IL-15-mediated crosstalk with these structural cells. Although AGA is traditionally considered a non-inflammatory, non-scarring form of alopecia, histological analysis shows evidence of inflammation. One study found that moderate inflammatory infiltrates and fibrosis were present in 36.8% of AGA patients, significantly higher than the 9.1% observed in control subjects.[144]^44 Our research show that early inflammatory responses in AGA are relatively mild, which may the reason of its classification as a non-inflammatory form of hair loss. However, as AGA progresses, inflammation becomes more pronounced, and the impact of immune responses on hair loss becomes increasingly significant. Therefore, we focused on research later-stage AGA patients classified as grade 5 hair loss according to the Hamilton–Norwood system. Research on immune cell regulation in AGA is still in its early stages, and there is currently limited understanding of the specific role immune cells play in AGA. To date, previous study has reported that the increase in γδ T cells, central memory CD8^+ T cells, activated CD8^+ T cells and mast cell infiltration contributes to changes in the immune microenvironment in male AGA.[145]^10 Charoensuksira et al used the scalps of normal patients as controls and found an enrichment of CD4^+ helper T cells in AGA patients.[146]^45 In our study, we selected the occipital area of AGA patients as the control group, based on the observation that this region is less susceptible to androgenetic alopecia-related hair loss. We observed significant differences in the proportions of NK and CD8^+ T cells between balding and non-balding. Comparing the frontal balding area with occipital normal areas can provide difference insights into the molecular mechanisms underlying the condition. CD8^+ T cells have been shown to play a role in immune-mediated alopecia like alopecia areata,[147]^46 primary cicatricial alopecia,[148]^47 and frontal fibrosing alopecia.[149]^48 Moreover, IFN-γ and various other cytokines are pivotal players in alopecia. Our result show that IFN-γ, TNFα, and cytotoxic granules, including PRF1, GZMA and GZMB increased in balding hair follicle of grade 5 AGA patients. These findings suggest that immune-mediated mechanisms, including specific T cell populations and inflammatory cytokines, may play a critical role in the pathogenesis of AGA. Insufficient perifollicular vascularization contributes to premature functional senescence of hair follicles, thereby impeding the transition from telogen to anagen phase in the hair cycle.[150]^49 Our findings demonstrating reduced expression of angiogenesis-associated genes (VEGFC, VWF, FGFR1/2) in balding hair follicle parallel previous reports that DHT-mediated suppression of angiogenic pathways diminishes perifollicular capillary density,[151]^50 a critical pathomechanism in androgen-related alopecia. The therapeutic implications of vascular regeneration strategies are further corroborated by existing evidence: Platelet-rich plasma promotes hair follicle regrowth through Wnt/β-catenin reactivation,[152]^51 while minoxidil’s vasodilatory effects via potassium channel activation in dermal vasculature[153]^52 and transdermal therapeutic platforms utilizing VEGF/Ritlecitinib-encapsulated polyhydroxyalkanoate nanoparticles demonstrate efficacy in AGA treatment. The dual-action mechanism combines Ritlecitinib’s selective JAK3/TEC kinase inhibition - suppressing pathogenic CD8^+ T cell and natural killer (NK) cell activity - with VEGF’s pro-angiogenic signaling, synergistically promoting hair follicular regeneration.[154]^2 These therapeutic outcomes align with our molecular findings of upregulated IL15 and RORA in balding follicle BECs, suggesting compensatory vascular niche remodeling that may functionally interface with the dual immunovascular regulatory axis. Specifically, NK cells and CD8^+ T lymphocytes exhibit context-dependent angiocrine modulation through IL15-mediated cytokine cascades and paracrine VEGF signaling pathways in androgen-driven alopecia pathogenesis. Keratin is a cytoskeletal protein that forms intermediate filaments in epithelial cells, playing a key role in maintaining their structural integrity. It is also a major component of hair, where it contributes significantly to its mechanical strength.[155]^53 In response to skin injury, keratinocytes (KCs) located near the wound site upregulate the expression of KRT16, KRT17, and KRT6, signaling a highly activated and proliferative state.[156]^54 Interestingly, we observed that spinous1 KCs, which exhibit high levels of KRT16, KRT17 and KRT6, show a significant reduction in bald hair follicles. However, the exact relationship between hair growth and spinous1 KCs remains to be fully elucidated and warrants further investigation. Limitation This study has several limitations that should be considered. First, while snRNA-seq is capable of identifying a broader and more comprehensive range of cell types, it has a key limitation in that it excludes RNA molecules from the cytoplasm. This means it may not capture the full spectrum of cellular transcriptional states. Second, the variability observed in findings related to inflammatory infiltration in AGA can be influenced by several factors, including the age of the participants, the severity of their condition, and the different methodologies used to assess and quantify inflammation. Finally, the statistical power of our AGA study is constrained by limited sample size and substantial interindividual heterogeneity, potentially compromising result generalizability. Notably, the observed directional trends in selected gene expression profiles ([157]Figure 4A) lacked statistical significance. These methodological limitations underscore the necessity for expanded human subject recruitment in subsequent validation studies. Conclusion This study combines bulk and single-nucleus RNA sequencing to characterize the immune microenvironment of AGA hair follicles, revealing NK and CD8+ T cells are hyperactivated in disease progression. The overexpression of IL-15 in BECs suggests that IL-15-mediated immune activation and vascular compromise synergistically drive AGA pathogenesis. Future studies should employ single-cell multi-omics to resolve CD8^+ subset dynamics, validate IL-15/JAK inhibition in 3D organoid models, and map spatial crosstalk between semaphorin-guided immune trafficking and pericyte dysfunction. Acknowledgments