DNA 5-Hydroxymethylcytosine Landscape and Transcriptional Profile Highlight the TUBB4B-Mediated Th17/Th1/Treg Imbalance in Behçet's Uveitis

Abstract

Purpose

Behçet's uveitis (BU) is an auto-inflammatory disease frequently with a poor prognosis. Here, we performed an integrated analysis of DNA 5-hydroxymethylcytosine (5-hmC) landscape and transcriptomic profiling in CD4⁺ T cells from active BU patients and healthy controls.

Methods

We conducted an integrated analysis of DNA 5-hmC modifications and transcriptomic data from CD4⁺ T cells of active BU patients and healthy individuals. Publicly available single-cell RNA-sequencing data were analyzed to validate findings across cell clusters. Functional experiments were performed to assess the effects of TUBB4B on cytokine production and T-cell frequencies.

Results

Bioinformatics analyses identify 801 downregulated and 2489 upregulated differential hydroxymethylated genes, predominantly enriched in pathways related to pathogen defense and various immune responses. RNA sequencing reveals 958 downregulated and 572 upregulated genes, with significant enrichment in pathways related to bacterial infection. Integration of epigenetic and transcriptional data highlights 74 candidate genes and three genes including TUBB4B, SKI and ZFPM1 are validated to be downregulated in active BU patients. Further analysis of publicly available single-cell RNA-sequencing data reveals TUBB4B downregulation across multiple cell clusters, particularly in naïve CD4⁺ T cells. Functional experiments indicate that overexpressing TUBB4B could decrease the frequencies of IL-17⁺ and IFN-γ⁺ CD4⁺ T cells and reduce the production of IL-17 and IFN-γ, while upregulating the frequency of CD4⁺CD25⁺FOXP3⁺ T cells and enhancing secretion of IL-10.

Conclusions

This study underscores the pivotal role of pathogen-induced immune dysregulation in BU development and identifies TUBB4B as a potential therapeutic target for the study on prevention and treatment of this disease.

Behçet's disease (BD) is a recurrent and chronic auto-inflammatory disease that typically manifests as recurrent oral ulcers, uveitis, genital ulceration and multiform skin lesions.^1–4 The prevalence of this disease varies geographically, with a relatively high incidence observed in the ancient “Silk Road” countries extending from the Mediterranean countries to China, Japan, and Korea.^5–7 Ocular involvement occurs in 50% to 75% of BD patients and is characterized by recurrent bilateral pan-uveitis or retinal vasculitis.⁸ Behçet's uveitis (BU), with an aggressive and relapsing nature, poses a significant risk for severe visual impairment and blindness.⁹

Recent advances in multi-omics sequencing technologies have shed light on the pathogenesis of BU, highlighting genetic susceptibility, immune dysregulation and infectious triggers as pivotal factors.^9–12 Multiple genetic studies have identified associations between BU susceptibility and specific human leukocyte antigen (HLA) alleles, such as HLA-B51, HLA-A26, and HLA-C0704.^10,13 Bulk RNA sequencing of iris specimens from patients with BU revealed a crucial role of T-cell-mediated immune responses in this disease.¹⁴ Numerous studies have shown that over-activation of T-helper (Th) 17 and Th1 cells, alongside the dysregulation of regulatory T (Treg) cells, is crucial in auto-inflammatory and autoimmune uveitis, especially in BU development.^15–17 A previous study suggested that epigenetic modification plays a vital role in the cytokine expression of Th17 and Th1 cells.¹⁸ However, the epigenetic mechanisms underlying this disease remain poorly understood.

Epigenetic modifications, a bridge between genetic predispositions and environmental factors, could be triggered by internal or external stimulation and directly linked to transcriptional regulation.^19,20 Among these modifications, DNA 5-hydroxymethylcytosine (5-hmC), an epigenetic modification associated with active demethylation processes, plays a critical role in regulating gene expression and has been implicated in the dysregulation of T-cell differentiation and activation.¹⁹ This dysregulation contributes to the pathogenesis of primary inflammatory responses and has gained much attention in various autoimmune diseases, such as systemic lupus erythematosus, allergic rhinitis, and Vogt-Koyanagi-Harada (VKH) disease.^21–23 The reversible nature of hydroxymethylation modifications suggests a promising avenue for therapeutic intervention in these autoimmune diseases.²⁰ Despite these advancements, the specific DNA 5-hmC profile in BU remains largely unexplored, presenting a critical gap in our knowledge that warrants further investigation to elucidate its potential role in BU pathogenesis.

In this study, we conducted a comprehensive analysis of the genome-wide DNA 5-hmC pattern and transcription profile in active BU CD4⁺ T cells. By integrating our sequencing data with publicly available single-cell RNA sequencing (scRNA-seq) data from healthy individuals and BD patients, we determined that TUBB4B, a key gene linked to pathogen infection and antigen presentation, was downregulated in active BD naïve CD4⁺ T cells. Functional analysis revealed that TUBB4B inhibited Th17 and Th1 responses and promoted Treg response. These results underscore a potential target of TUBB4B for the study on prevention and treatment of BU.

Methods

Subject Information

Twelve active BU patients and 16 healthy controls (HCs) from the uveitis center of the First Affiliated Hospital of Chongqing Medical University from June 2022 to September 2024 were included in this study (12 BU and 12 HCs samples for sequencing and in vitro validation experiments, other 4 HCs samples for differentiation of naïve CD4⁺ T cells). BU diagnosis followed two criteria established by the International Revision Team and the International Study Group.^24,25 Patients with active acute anterior uveitis (AAU, n = 6), active pediatric uveitis (PU, n = 6), and active VKH disease (n = 6) according to relevant criteria^26–28 were chosen as disease controls for validation. All participants did not systemically use medicines for at least two weeks before enrollment. The exclusion criteria encompassed cardiovascular disease, malignant tumors, diabetes, hypertension or other autoimmune diseases. Active intraocular inflammation was assessed using the criteria of the Standardization of Uveitis Nomenclature working group.²⁹ Informed consent was obtained from all participants, and the study was conducted in adherence to the principles of the Declaration of Helsinki and approved by the Ethics Committee of the First Affiliated Hospital of Chongqing Medical University. The detailed demographic and clinical information of all the participants is detailed in Supplementary Table S1.

Isolation and Culture of CD4⁺ T Cells and Naïve CD4⁺ T Cells

To isolate peripheral blood mononuclear cells (PBMCs), a lymphocyte separation medium (TBD Science, Tianjin, China) was prepared. Following PBMC extraction, CD4⁺ T cells were subsequently purified using CD4⁺ microbeads (Miltenyi Biotec, Bergisch Gladbach, Germany). The freshly isolated CD4⁺ T cells were cultured in a complete RPMI 1640 medium supplemented with 10% fetal bovine serum (Gibco, Thermo Fisher Scientific, Waltham, MA, USA). For activation, the cells were stimulated with an anti-CD28/CD3 antibody (1 µg/mL; Miltenyi Biotec) and seeded in 24-well or 48-well plates at a density of 5 × 10⁵ cells/mL for a period of three days.

In parallel, naïve CD4⁺ T cells were isolated from PBMCs using a naïve CD4⁺ T cell isolation kit II (Miltenyi Biotec). These cells were re-suspended at a concentration of 5 × 10⁵ cells/mL in RPMI 1640 complete medium and were co-cultured with an anti-CD28/CD3 antibody (1 µg/mL, Miltenyi Biotec) in 24-well plates and differentiated for six days in the presence of cytokine combinations as reported previously.^30–31 Th1 differentiation was induced with recombinant human interleukin-2 (IL-2, 10 µg/mL; TargetMol, Wellesley Hills, MA, USA) and IL-12 (5 ng/mL; MedChemExpress, Monmouth Junction, NJ, USA). Th17 differentiation was induced with human IL-1β (50 ng/mL, MedChemExpress), IL-6 (5 ng/mL, MedChemExpress) and IL-23 (50 ng/mL, MedChemExpress). Treg differentiation was induced with recombinant human IL-2 (10 ng/mL, TargetMol) and TGF-β I (50 ng/mL, MedChemExpress).

5hmC-Seal-seq

The genomic DNA of each sample was extracted from CD4⁺ T cells via Quick-gDNA MicroPrep (Zymo Research, Irvine, CA, USA). DNA was prepared into libraries and then sequenced as methods reported previously.³² Briefly, after fragmentation and purification, glycosylation reactions were used to transfer azido-glucose to 5-hmC using T4-beta-glucosyltransferase. The resulting DNA fragments were subsequently captured by binding to C1 streptavidin beads and subjected to amplification through PCR. This approach enabled the selective enrichment and analysis of 5-hmC-containing DNA regions. The quantification of the DNA library was conducted using the Qubit fluorimeter (Life Technologies, Carlsbad, CA, USA), and sequencing was carried out on the HiSeq platform. Bioinformatics analysis was performed using R packages (version 3.5.0), with differential hydroxymethylated genes (DhmGs) identified on the basis of a P value < 0.05 and a log2-fold change greater than 1.5.

RNA-Seq

Total RNA was obtained from CD4⁺ T cells using TRIzol reagent (Invitrogen, Carlsbad, CA, USA). The extracted RNA was subsequently used to construct a complementary DNA library, which was prepared via a strand-specific library preparation kit to ensure an accurate representation of transcript orientation. High-throughput sequencing was then conducted on an Illumina NovaSeq platform, followed by PCR amplification and cluster generation to achieve sufficient sequencing depth. For bioinformatics analysis, R software (version 3.5.0) was used to identify differentially expressed genes (DEGs) with thresholds of a P value < 0.05 and a log2fold change P ≥ 1.0.

Enrichment Analysis

Enrichment analysis of DhmGs and DEGs was conducted via the ClusterProfiler R package (version 3.8.1) to evaluate their associations with Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways and Gene Ontology (GO) terms. A P value threshold of ≤0.05 was considered statistically significant. Gene set enrichment analysis (GSEA) was carried out using the KEGG and GO datasets, and the GSEA results were considered significant if the normalized enrichment score was greater than 1, the false discovery rate q value was less than 0.25, or the nominal P value was below 0.05.

Analysis of the Protein-Protein Interaction (PPI) Network

PPI networks of the overlapping genes were constructed via the STRING database and visualized with Cytoscape software (version 3.9.1). Key nodes or Hub genes within these networks were identified using the Cytoscape plugin cytoHubba (version 0.1). To enhance the accuracy of hub gene identification, five topological algorithms (degree, MNC, EPC, closeness, and radiality) were employed, and the top 10 hub genes were identified according to their respective metrics.

Drug-Gene Interactions

Drug-gene interactions were analyzed via the Drug-Gene Interaction Database (DGIdb 4.0) to determine whether the genes of interest were targeted by existing pharmaceuticals or classified as potentially druggable genes. The DGIdb integrates data from various sources, including the Therapeutic Target Database, ChEMBL, Drug Target Commons, DrugBank, and PharmGKB, facilitating the identification of potential therapeutic connections.

Real-Time Quantitative PCR (RT-qPCR)

RNA was extracted from CD4⁺ T cells using the SteadyPure Universal RNA Extraction Kit (Accurate Biology, Guangzhou, China). The Evo M-MLVRT Mix Kit with gDNA Clean for PCR (Accurate Biology) was used to synthesize complementary DNA from the extracted RNA, ensuring the removal of genomic DNA contamination. Subsequently, qPCR was performed on an ABI 7500 RT-qPCR thermocycler with 2 × Universal Blue SYBR Green qPCR Master Mix (Servicebio, Wuhan, China) to amplify target genes. To ensure accurate quantification, melt curve analysis was conducted to verify the specificity of the amplified products. The data were analyzed employing the 2^−ΔΔCT method with normalization to GAPDH. The primer sequences, synthesized by Accurate Biology Company or Sangon Biotech Company (Shanghai, China), are listed in Supplementary Table S2.

ScRNA-seq Analysis

The scRNA-seq data of PBMCs from 4 active BD patients and 4 HCs were downloaded from the GEO database (GSE198616). The data were processed using Seurat (v4.3.0) with the Read10X function for quality control, normalization, and variable feature identification. Batch effects were corrected using cross-dataset anchors. Mitochondrial percentages were calculated with the percentage feature set function. Cells with 200 to 3,000 expressed genes and a mitochondrial content below 15% were included. Following the removal of inferred doublets using DoubletFinder (v2.0.2), a total of 39,561 qualified PBMCs were deemed suitable for further analyses. The data were normalized and conducted with the NormalizeData function, and the top 2,000 variable genes were selected through the FindVariableFeatures function. Scaling of the data was performed with the ScaleData function, followed by principal component analysis (PCA) of these highly variable genes. Cell clustering was executed using the top 50 principal components with a resolution of 0.6 in the FindClusters function, and uniform manifold approximation and projection (UMAP) was utilized for cluster visualization. Clusters were assigned based on established cellular markers from the literature, as detailed in Supplementary Table S3. The ClusterProfiler R package (version 3.14.0) was used to perform enrichment analysis. Cell interactions were inferred using CellChat (v2.1.0), with significance determined by a permutation test at a threshold of P < 0.05.

CD4⁺ T Cell Stimulation and Transfection

To explore the effect of interferon α2a (IFNα2a; MedChemExpress) in vitro, freshly isolated CD4⁺ T cells were treated with or without IFNα2a (2000 U/ml) for 24 hours. For transfection experiments, CD4⁺ T cells were transfected with siRNAs (Wuhan GeneCreate Biological Engineering Co., Ltd., Wuhan, China; Supplementary Table S2) using the INVI DNA RNA Transfection Reagent (Invigentech, San Diego, CA, USA) for 72 hours. Two specific siRNAs targeting TUBB4B were confirmed to effectively knock down TUBB4B at both mRNA and protein levels, as compared to a nonsense control (Supplementary Figs. S3A, S3B). TUBB4B overexpression was achieved by infecting CD4⁺ T cells with adenovirus constructs (multiplicity of infection = 40) purchased from the GenePharma Company (Shanghai, China). After 72 hours of co-culture, the adenoviral constructs were confirmed to be effective by comparison with the empty constructs (Supplementary Figs. S3A, S3B).

Western Blot

Proteins extracted from CD4⁺ T cells were separated by using FuturePAGE 4%–20% gels (ACE Biotechnology, Hunan, China). After transferring onto PVDF membranes (Merck Millipore, Burlington, MA, USA) and blocked with MinuteBlock Blocking Buffer (Affinibody, Wuhan, China) for 20 minutes, the membranes were then incubated overnight at 4°C with primary antibodies against TUBB4B (1:1000 dilution; Abnova, Amyjet Scientific, Wuhan, China) and GAPDH (1:3000 dilution; Proteintech, Rosemont, IL, USA). After washing, membranes were incubated with HRP AffiniPure Goat anti-rabbit IgG (1:5000 dilution; EarthOx, Millbrae, CA, USA) for 1.5 hours. Protein bands were visualized and quantified using a gel imaging system (Bio-Rad Laboratories, Ann Arbor, MI, USA).

Flow Cytometry

For intracellular cytokine analysis, CD4⁺ T cells were stimulated with the Cell Activation Cocktail (BioLegend, San Diego, CA, USA) for five hours, followed by fixation and permeabilization using either the Foxp3/Transcription Factor Staining Buffer Set or the eBioscience Intracellular Fixation/Permeabilization Buffer Set (Invitrogen). The cells were then incubated with anti-human CD25/FOXP3 antibodies or anti-human IL-17/IFN-γ (BioLegend) at 4°C for one hour. Flow cytometry analysis was conducted on a CytoFLEX flow cytometer (Beckman Coulter, Inc, Southfield, MI, USA) to assess cell populations and their specific markers. The resulting data were analyzed using FlowJo software (version 10.0.7).

ELISA

Levels of IFN-γ, IL-17, and IL-10 in the supernatants were quantified via Duoset ELISA Development Kits (R&D Systems, Minneapolis, MN, USA), following the manufacturer's guidelines. Signal values were quantified using the Varioskan LUX multimode microplate reader (Thermo Fisher Scientific).

Statistical Analyses

Statistical analyses were conducted using GraphPad Prism software (version 8.0.2). For data with a non-normal distribution, the Mann-Whitney test was employed, while normally distributed data were assessed using Student's t-test. Age-adjusted comparisons between the PU and HC groups were conducted using generalized estimating equations with robust standard errors. The data are presented as the means ± SD, with statistical significance defined by a P value < 0.05.

Results

DNA 5-hmC Landscape in Active BU CD4⁺ T Cells

To characterize the epigenetic pattern of BU, we performed DNA 5hmC-Seal-seq on CD4⁺ T cells derived from six active BU patients and six sex- and age-matched HCs. Our results revealed distinct differences in the distribution of 5-hmC peaks across various genomic regions, exhibiting an increased proportion of 5-hmC within promoter regions (5.66%) and intergenic areas (30.23%) in BU patients compared to HCs (4.31% and 25.97%, respectively) (Fig. 1A). PCA demonstrated a pronounced separation in 5-hmC patterns between BU patients and HCs, underscoring the presence of aberrant hydroxymethylation in active BU patients (Fig. 1B). A total of 801 downregulated DhmGs and 2489 upregulated DhmGs were identified in active BU patients (Fig. 1C, Supplementary Table S4). GO and KEGG analyses revealed that 3290 DhmGs were significantly enriched in 416 biological processes (BPs), 110 molecular functions (MFs), 81 cell components (CCs), and 50 KEGG pathways (Figs. 1D, 1E). Notably, DhmGs were prominently associated with signaling pathways related to bacterial defense responses and major histocompatibility complex (MHC) protein complexes. Additionally, immune-related pathways such as T-cell co-stimulation, regulation of interleukin and TNF production, the IFN-γ signaling pathway, chemokine production and cytokine secretion were significantly implicated in the pathogenesis of BU. Oxidative stress response-related pathways and apoptotic-related signaling pathways have also emerged as important contributors to BU pathogenesis (Fig. 1F). These findings suggest a substantial role of microbial infection in modulating epigenetic modifications and driving the hypersensitive immune reactions of BU.

Figure 1.

DNA 5-hmC pattern in CD4⁺T cells from active BU patients and HCs. (A) Distribution of 5-hmC peak proportions across seven genomic regions in CD4⁺ T cells from 6 active BU patients and 6 HCs. (B) PCA results of 5-hmC sequencing data from 6 active BU patients and 6 HCs. (C) Volcano plots illustrating differential 5-hmC modifications: peaks that are significantly upregulated (red) or downregulated (blue) in active BU patients versus HCs. (D) GO enrichment analysis of DhmGs, with a bar graph depicting the top ten most significantly enriched GO terms across the BP, CC, and MF categories. (E) KEGG pathway enrichment analysis of DhmGs, represented by a bubble plot showing significantly enriched pathways. (F) Classification of GO enrichment results for all DhmGs, with a bar graph depicting the significantly enriched terms across different categories.

Transcriptional Profiling in Active BU CD4⁺ T Cells

To delineate the transcriptional alterations associated with active BU, we further conducted RNA sequencing on CD4⁺ T cells from these participants, as outlined above. PCA revealed a clear separation between the BU and control groups, reflecting substantial differences in gene expression (Supplementary Fig. S1A). In CD4⁺ T cells from active BU patients, 958 genes were downregulated, while 572 genes were upregulated by comparing with HCs (Supplementary Fig. S1B, Supplementary Table S5). Bioinformatics analyses indicated that 1530 DEGs were significantly involved in 717 BPs, 27 CCs, and 86 MFs respectively (Fig. 2A). The functional enrichment analysis revealed that cellular response to biotic stimulus was the most significantly enriched biological process among all DEGs, followed by T-cell differentiation, response to molecules of bacterial origin and lipopolysaccharide, and response to IFN-gamma. Consistent with the GO results of the 5-hmC pattern, the pathogenesis of BU was also notably associated with responses to the bacterium, MHC interactions, oxidative stress response and apoptosis. Specifically, pathways related to T cell differentiation, interleukins and TNF, chemotaxis and cytokines, as well as IFN-related signaling pathways, were also prominently implicated in BU (Supplementary Fig. S1C). Besides, GSEA of the GO datasets further demonstrated prominent associations of IFN-related signaling pathways, signaling pattern recognition receptor activity and the myd88-dependent Toll-like receptor signaling pathway with BU (Fig. 2B). KEGG pathway analysis identified 48 enriched pathways, including the Toll-like receptor, NF-Kappa B, MAPK, IL-17, Nod-like receptor, chemokine and TGF-beta signaling pathway, cytokine-cytokine receptor interaction and viral protein interaction with cytokine and cytokine receptor (Fig. 2C). Similarly, GSEA of KEGG datasets further corroborated the critical roles of the Toll-like and Nod-like receptor signaling pathway, as well as Staphylococcus aureus infection in BU development (Fig. 2D). These results indicate important roles of inflammation-related responses and pathogen infection prominently in the pathogenesis of BU.

Figure 2.

Transcription profile of CD4⁺T cells from active BU patients and HCs. (A) GO enrichment analysis of DEGs, with a bar graph showing the top 10 most significantly enriched GO terms across the BP, CC, and MF categories. (B) GSEA plot for the significantly enriched GO terms. (C) KEGG pathway enrichment analysis of DEGs, visualized through a bubble plot of significantly enriched pathways. (D) GSEA plots for the significantly enriched KEGG pathways.

Identification of Candidate Target Genes Based on DNA 5-HmC Profile and Transcription Pattern

To elucidate candidate target genes implicated in the pathogenesis of active BU disease, we performed a comprehensive analysis via integrating the DNA 5-hmC pattern with the transcriptional profile. Through overlapping DhmGs and DEGs, 74 genes were identified as potential candidates for subsequent experiments (Fig. 3A, Supplementary Table S6). Functional enrichment analyses revealed that IFN-related signaling pathways, enriched by RARA, ZFPM1, FADD, and IFITM3, were significantly implicated in BU pathogenesis. Besides, pathogen defense responses and MHC-related signaling pathways, enriched by TUBB4B, FADD, and IFITM3, were also prominently associated with BU (Figs. 3B, 3C). To further elucidate the interactions among the 74 genes, we examined the PPI network of these genes using the STRING database. Overall, a network comprising 18 edges (protein-protein interactions) and 53 nodes (protein-coding genes) was identified with a PPI enrichment P value of 0.000858. The top 5 protein-protein interactions, involving 10 protein-coding genes including S1PR4, CXCR5, RARA, SKI, SLC16A3, SLC2A1, TUBB4B, PLK3, FOXA1 and ZFPM1, were selected as candidates according to their interaction score (Fig. 3D). Additionally, based on overlapping genes obtained via the use of five algorithms in Cytoscape software, two other hub genes (FLNA and PLEC) were screened within the PPI network (Fig. 3E). Finally, a total of 11 downregulated candidate genes were identified.

Figure 3.

Integrated analysis of the DNA 5-hmC pattern and transcription profile in CD4⁺T cells from active BU patients and HCs. (A) Venn diagram representing the 74 candidate genes identified by the integrated analysis of DNA 5-hmC and transcription data. (B) GO enrichment analysis of the 74 candidate genes, displayed as a bar graph showing the top ten most significantly enriched GO terms in the BP, CC, and MF categories. (C) KEGG pathway enrichment analysis of the 74 candidate genes presented as a bubble plot depicting all significantly enriched KEGG pathways. (D) PPI network analysis shows the top five interactions among the 74 candidate genes, as determined using the STRING database. (E) Venn diagram showing the overlapping hub genes identified by five distinct algorithms in the cytoHubba tool.

Validation of Candidate Target Genes

To validate the selected candidate genes, we assessed the mRNA expression levels of the aforementioned genes in CD4⁺ T cells. The results showed significant downregulation of TUBB4B, SKI, and ZFPM1 in active BU patients (Figs. 4A–C). However, no significant differences were observed in the mRNA levels of the remaining eight genes between active BU patients and HCs (Fig. 4D). Additionally, we measured the mRNA expression of TUBB4B, SKI, and ZFPM1 in patients with active VKH, active AAU, and active PU. The results showed decreased expression of TUBB4B and SKI across these uveitis entities, whereas the mRNA expression of ZFPM1 did not differ between HCs and any of the other uveitis entities (Figs. 4A–C). In an effort to explore the therapeutic implications, we evaluated the impact of IFNα2a treatment on these genes in active BU patients. The result showed that it significantly upregulated TUBB4B and SKI but did not have any effect on ZFPM1 expression (Fig. 4E). Furthermore, a search of the DGIdb identified TUBB4B as a potentially druggable gene, with 83 drugs predicted to interact with it possibly. Notably, Combretastatin A4, Nocodazole, and Mebendazole had the highest interaction scores with TUBB4B (Fig. 4F). Both SKI and ZFPM1 did not appear to be druggable genes in DGIdb. These results suggest the potential of TUBB4B as a promising candidate for novel therapeutic strategies in BU.

Figure 4.

Validation of candidate genes in CD4⁺ T cells from active BU patients, disease controls and HCs. (A–C) Relative mRNA expression of TUBB4B (A), SKI (B), and ZFPM1 (C) in CD4⁺ T cells from active BU patients, active VKH patients, active AAU patients and active PU patients compared with HCs (n = 6:6:6:6:6), as measured by q-PCR and normalized to GAPDH. (D) Relative mRNA expression of additional candidate genes in CD4⁺ T cells from active BU patients compared with HCs (n = 6:6), as assessed by q-PCR and normalized to GAPDH. (E) Relative mRNA expression of TUBB4B, SKI and ZFPM1 in CD4⁺ T cells from six active BU patients after treatment with IFNα2a. (F) Overview of the top eight drugs exhibiting the highest interaction scores with TUBB4B, as identified through the DGIdb. *P < 0.05, **P < 0.01. ***P < 0.001, ****P < 0.0001, no statistical significance (ns) means P > 0.05.

Analysis of TUBB4B Relevance to Immune Cells Based on ScRNA-seq Data

We further explored the expression patterns of TUBB4B within the scRNA-seq data derived from PBMCs of four BD patients and four HCs. Using canonical markers and cell type-specific gene sets for identification (Supplementary Fig. S2A), the UMAP plots delineated 17 distinct cell subtypes within PBMCs (Fig. 5A). Our analysis revealed a marked reduction in expression of TUBB4B across several cell clusters, including memory CD4⁺ T cells (CD4 Tmemory), naïve CD4⁺ T cells (CD4 Tnaïve), activate natural killer cells (NKs), naïve CD8⁺ T cells (CD8 Tnaïve), memory CD8⁺ T cells (CD8 Tmemory), effector CD8⁺ T cells (CD8 Teffector), innate-like T cells, memory B cells and naïve B cells (Fig. 5B). To elucidate the regulatory interactions among these clusters, we performed cellChat analysis. The data indicated an increased number and strength of interactions from naïve CD4⁺ T cells to innate-like T cells in BD patients compared with HCs, in association with a reduced number of interactions from effector CD8⁺ T cells and activate natural NKs to naïve CD4⁺ T cells (Fig. 5C). Pro-inflammatory characteristics were also shown in BU-associated signaling pathways, including the IL-1 signaling pathway and the IFN-II signaling pathway (Fig. 5D). Further examination of re-clustering TUBB4B⁺CD4⁺ T cells, revealing 60 upregulated DEGs, as well as 15 downregulated DEGs compared with TUBB4B⁻CD4⁺ T cells (Supplementary Table S7). GO analysis suggested that TUBB4B⁺CD4⁺ T cells were associated with Toll-like receptor binding, antigen binding and MHC class II protein complex binding, implicating a significant role of TUBB4B in pathogen infection and antigen presentation (Supplementary Fig. S2B). Besides, KEGG analysis showed enrichment of inflammatory signaling pathways in TUBB4B⁺CD4⁺ T cells, including the TNF, IL-17, and NOD-like receptor signaling pathways (Supplementary Fig. S2C). To further delineate the heterogeneity and increase cell-level resolution, we re-clustered CD4⁺ T cells and identified four distinct subtypes, including effector CD4⁺ T cells, memory CD4⁺ T cells, naïve CD4⁺ T cells and regulatory T cells (Supplementary Fig. S2D, Fig. 5E). TUBB4B was broadly distributed across these CD4⁺ T cell clusters, with the scRNA-seq results confirming a significant downregulation of TUBB4B in all CD4⁺ T cells (adjusted P ≤ 0.0001), most notably in naïve CD4⁺ T cells (adjusted P < 0.003) from BD patients compared with HCs (Figs. 5F, 5G). The distributions of SKI and ZFPM1 are presented in Supplementary Figures S2E and S2F. Additionally, cellChat analysis of the four clusters revealed increased interaction strengths from memory CD4⁺ T cells to both effector CD4⁺ T cells and naïve CD4⁺ T cells, with increased activity in pro-inflammatory pathways such as the macrophage migration inhibitory factor and IL-7 signaling pathways (Figs. 5H, 5I). These findings suggest a role of TUBB4B in the correlations with increased pro-inflammatory signaling and the functional dynamics of naïve CD4⁺ T cells in active BD patients.

Figure 5.

Analysis of TUBB4B relevance to immune cells in publicly available scRNA-seq data. (A) UMAP visualization shows 17 distinct cell clusters in PBMCs from four BD patients and four HCs. (B) Identification of cell clusters with downregulated TUBB4B expression in active BD patients compared with HCs. (C) CellChat analysis depicting the number and strength of interactions significantly upregulated (red) or downregulated (blue) among TUBB4B-downregulated cell clusters in active BD patients versus HCs. (D) Enrichment of significant signaling pathways in BD patients (blue) versus HCs (red) based on the ligand-receptor interactions. (E) UMAP visualization depicting four distinct cell clusters in CD4⁺ T cells from four BD patients and four HCs. (F) Expression levels of TUBB4B across the four CD4⁺ T cells clusters in BD patients compared with HCs. (G) UMAP visualization exhibiting TUBB4B distribution within CD4⁺ T cell clusters, with red intensity indicating the normalized gene expression level: dark red for high expression and gray for absence of expression. (H) CellChat analysis shows the number and strength of interactions significantly upregulated (red) or downregulated (blue) within CD4⁺ T cell clusters in active BD patients versus HCs. (I) Dot plot of significant ligand-receptor interactions (Y-axis) among different CD4⁺ T cell subtypes. Incoming interactions are defined by receptor-expressing cell subtypes, while outgoing interactions are defined by ligand-expressing subtypes. The circle size and color represent the variation in interaction scores across interacting populations.

Role of TUBB4B in T-Cell Differentiation

To explore the role of TUBB4B in T cell differentiation, siRNAs were used to downregulate TUBB4B and adenoviral constructs were employed to overexpress TUBB4B. Silencing TUBB4B increased the frequencies of IL-17⁺ and IFN-γ⁺ CD4⁺ T cells and the secretions of IL-17 and IFN-γ while reducing IL-10 production and CD4⁺CD25⁺FOXP3⁺ Treg cells frequency in CD4⁺ T cells from HCs. Conversely, TUBB4B overexpression reduced the frequencies of IL-17⁺ and IFN-γ⁺ CD4⁺ T cells and IL-17/IFN-γ production but increased IL-10 level and CD4⁺CD25⁺FOXP3⁺ Treg cell frequency (Fig. 6A, Supplementary Figs. S3C, S3D). A similar pattern was observed in the differentiation of naïve CD4⁺ T cells, where TUBB4B silencing promoted the frequencies of IL-17⁺ and IFN-γ⁺ CD4⁺ T cells but inhibited Treg cell differentiation, whereas TUBB4B overexpression exerted an opposite effect (Fig. 6B). These findings suggest that TUBB4B is as an anti-inflammatory gene involved in BU pathogenesis.

Figure 6.

Effects of TUBB4B on Th1, Th17 and Treg responses in CD4⁺ T cells. (A) Representative images and quantification of frequencies of IFN-γ⁺ CD4⁺ T cells, IL-17⁺ CD4⁺ T cells and CD4⁺CD25⁺FOXP3⁺ Treg cells by flow cytometry in CD4⁺ T cells from six HCs and six active BU patients after treatment with TUBB4B siRNAs or overexpression adenovirus constructs. (B) Representative images and quantification of frequencies of IFN-γ⁺ CD4⁺ T cells, IL-17⁺ CD4⁺ T cells and CD4⁺CD25⁺FOXP3⁺ Treg cells during the differentiation of naïve CD4⁺ T cells from four HCs after incubation with TUBB4B siRNAs or overexpression adenovirus constructs. EC, empty constructs; NC, nonsense control; OE, overexpression. *P < 0.05, **P < 0.01.

Discussion

Our study, for the first time, provides a comprehensive epigenetic and transcriptional pattern of CD4⁺ T cells from active BU. Enrichment analyses of these two omics highlighted the involvement of pathogen infection, the oxidative stress response, apoptosis and various immune and inflammatory signaling pathways in BU pathogenesis. Integrated analyses identified TUBB4B as an important target and showed its downregulation in active BU patients and its upregulation in response to IFNα2a treatment. Additionally, scRNA-seq analyses revealed downregulation of TUBB4B across multiple cell clusters, particularly in naïve CD4⁺ T cells. Functional experiments confirmed that TUBB4B could promote Treg response and inhibit Th17 and Th1 responses. Collectively, our findings underscore the contribution of pathogen infection to BU development through integrating DNA hydroxymethylation patterns and transcriptional alterations and highlight TUBB4B as a promising therapeutic target for BU (Fig. 7).

Figure 7.

Integration of DNA 5-hmC and transcriptomic profiling in CD4⁺ T cells from active BU patients highlight the critical role of pathways related to pathogen defense in BU pathogenesis. Functional experiments demonstrate that TUBB4B overexpression promotes Treg response and inhibits Th17 and Th1 responses. (This figure was created using BioRender.com and has been granted a license for publication.)

BU, one of the most severe uveitis entities, is characterized by intricate auto-inflammatory and autoimmune responses. The pathogenesis of BU is multifactorial, involving interactions between genetic and environmental factors.^9,12 Specifically, the MHC gene region, particularly the HLA-B51 allele, has been identified as a critical genetic risk factor for BU.^13,33 Additionally, epigenome-wide DNA methylation analysis has highlighted differential loci between BU patients and HCs.³⁴ However, the role of DNA hydroxymethylation in BU remains largely unexplored. In this study, we provide the first analysis of DNA 5-hmC patterns in CD4⁺ T cells from active BU patients, thereby expanding the understanding of BU pathogenesis at the hydroxymethylation level. We observed a significant increase in 5-hmC within promoter and intergenic regions in BU patients, suggesting that altered DNA hydroxymethylation is involved in BU development. This aberrant hydroxymethylation pattern is associated with a plethora of biological processes and pathways, including oxidative stress signaling pathways, apoptotic-related signaling pathways and various immune-related signaling pathways involving interleukin, TNF, and IFN-γ. In parallel with our epigenetic findings, RNA-seq analysis corroborated these findings. Notably, significant enrichment in pathogen-induced signaling pathways was found in both 5-hmC sequencing and RNA sequencing data, highlighting the essential role of pathogen infection in the development of BU. Pathogens are recognized by pattern recognition receptors (PRRs) or pathogen-associated molecular patterns, such as NOD-like and Toll-like receptors, which activate innate immune responses and steer subsequent adaptive immunity.³⁵ These PRRs also stimulate IFN gene expression, playing a crucial role in defending against microbial infections.³⁶ In this study, inflammatory signaling pathways, including those related to IFN, Toll-like receptors, NOD-like receptors, and IL-17, were markedly enriched in BU patients compared with HCs. The GSEA further revealed that IFN-related and PRR-related signaling pathways were closely associated with intraocular inflammation in BU patients. The convergence of these pathways suggests that an abnormal sensitivity to microbial exposure may cause excessive activation of PRR signaling in BU, thereby promoting downstream IFN-related signaling pathways, enhancing the expression of cytokines and chemokines, and eventually driving pathogenic Th responses.

Identifying novel therapeutic targets is crucial for the effective management of BU. The integration of epigenetic and transcriptional data allowed us to pinpoint candidate target genes that may be implicated in the pathogenesis of active BU. Our validation studies highlighted the downregulation of TUBB4B, SKI, and ZFPM1 in patients with active BU, supporting their potential role in the disease process. Specifically, reduced expression of TUBB4B has been linked to increased metastasis in colon and breast cancers,^37,38 and this gene has also been associated with inflammatory and pro-metastatic environments in tumor-associated macrophages.³⁹ SKI has been recognized as a potential therapeutic target in autoimmune diseases due to its involvement in IL-21-induced Th17 differentiation.⁴⁰ Additionally, ZFPM1 has been identified as a potential mediator of increased susceptibility to Th2-driven allergic asthma.⁴¹ In our study, we further observed diminished levels of TUBB4B and SKI in active VKH, AAU, and PU patients compared with HCs, indicating the presence of a common pathway or mechanism underlying the pathogenesis of these ocular inflammatory conditions. Although TUBB4B may not be exclusively specific to BU, the identification of a shared molecular signature across different uveitis subtypes raises the possibility of developing therapies that target common inflammatory pathways. A previous clinical study has demonstrated that IFNα2a combined with corticosteroids is superior to cyclosporine plus corticosteroids in reducing uveitis relapse rates in patients with severe BU.^42,43 In our study, we observed that IFNα2a treatment modulated TUBB4B and SKI expression, suggesting a potential mechanism through which IFNα2a exerts its therapeutic effects in BU. DGIdb analysis further supported the druggability of TUBB4B, while SKI and ZFPM1 were not identified as druggable targets. Based on these findings, we have emphasized TUBB4B as a promising candidate for novel therapeutic strategies in BU. The particular therapeutic promise of TUBB4B may stem from its involvement in pathogen defense and MHC-related signaling pathways. Our scRNA-seq analysis further confirmed these findings, showing that TUBB4B⁺CD4⁺ T cells were involved in Toll-like receptor binding, MHC class II protein complex binding, and antigen binding. The widespread downregulation of TUBB4B across a range of immune cell clusters may lead to dysfunction of these cells in BU. Notably, the most significant downregulation of TUBB4B was observed in naïve CD4⁺ T cells, which was coupled with increased interaction strength from memory CD4⁺ T cells to naïve CD4⁺ T cells. This finding was associated with the increased pro-inflammatory pathways, such as the migration inhibitory factor and IL-7 signaling pathways in BU. Functionally, our findings demonstrated that TUBB4B played a role in the pathogenesis of BU by inhibiting Th17 and Th1 responses and promoting Treg response, suggesting that TUBB4B may serve as a potential anti-inflammatory target in BU. Overall, our findings elucidate a critical role for TUBB4B in pathogen defense and the modulation of immune cell interactions, underscoring its potential as a novel therapeutic target for the prevention and treatment of BU and other uveitis entities.

In summary, our study highlights the complex interplay between epigenetic modifications and transcriptional changes in BU, providing a deeper understanding of the disease's molecular mechanisms of this disease. The integration of 5-hmC and gene expression data identified that TUBB4B, which is related to immune dysregulation and pathogen defense, is a promising candidate for targeted therapeutic interventions. Future research should focus on validating these findings in larger cohorts and exploring the therapeutic potential of TUBB4B to pave the way for personalized medicine approaches in BU.

Although our study offers novel insights into the molecular mechanisms underlying BU, several limitations warrant consideration. An important limitation of our study is that the scRNA-seq dataset used for external validation was derived from patients with systemic BD. Further single-cell studies using ocular samples from BU patients are warranted to more precisely capture local immune dysregulation. The sample size, although adequate for the discovery phase, may limit the broader applicability of our findings. Further studies involving larger cohorts are needed to validate the identified candidate genes and their potential as therapeutic targets. Additionally, the cross-sectional design of our study precludes the ability to establish causality or explore the temporal dynamics of molecular changes. Longitudinal studies are needed to further elucidate the role of TUBB4B in both the progression and resolution of BU. Moreover, the integration of multi-omics data with clinical parameters may offer a more comprehensive perspective on disease heterogeneity and facilitate the development of precision medicine strategies for BU patients.