Corticosteroid-Sparing Control of Bullous Pemphigoid with Dupilumab and Tripterygium Glycosides: A Real-World Cohort with Longitudinal Transcriptomics

Introduction

Bullous pemphigoid (BP) is a rare autoimmune skin disease that causes blister formation and mainly affects older adults. Standard treatment for BP includes systemic corticosteroids and immunosuppressants, but these drugs can cause notable side effects, especially in patients with multiple comorbidities.^1–3 Dupilumab is a monoclonal antibody that targets the interleukin (IL)-4 receptor α and has become a promising treatment option for BP.^4–9 In a recent multicenter study, dupilumab was shown to be an effective, rapidly acting, and safe treatment for BP, with marked reductions in pruritus and systemic corticosteroids use.⁷ Several studies have shown that dupilumab, when combined with systemic corticosteroids, can effectively control BP symptoms. However, concerns remain regarding the risk of adverse effects from corticosteroids.^4–7,9

In addition to dupilumab, other therapeutic agents have been explored for the treatment of BP. One promising alternative is Tripterygium wilfordii Hook F (TwHF), commonly known as thunder god vine, which has been traditionally used in Chinese medicine for its anti-inflammatory and immunosuppressive properties.^10–13 Tripterygium glycoside (TG), derived from this plant, has been used in the treatment of a variety of autoimmune conditions, including rheumatoid arthritis.^10–12 Similar to methotrexate, which is a first-line treatment for BP but is associated with adverse effects such as bone marrow suppression and liver function impairment, TG functions as a disease-modifying anti-rheumatic drug and may offer a potential adjunctive therapy for BP.^10,12 Our previous study showed that TG tablets alone have good efficacy and safety in treating mild to moderate BP compared with systemic corticosteroids alone, with fewer adverse effects, suggesting that they may be a useful treatment option for BP.¹⁰ The integration of dupilumab with TG tablets could potentially harness the immunomodulatory effects of both agents, providing a synergistic way to treat BP. This regimen may lower the need for systemic corticosteroids and thus reduce their related adverse events, such as a higher risk of infection and metabolic problems.

Here, we carried out an observational clinical cohort study to assess the safety and efficacy of dupilumab combined with TG in patients with moderate to severe BP. Besides this clinical assessment, we also aimed to explore the possible mechanisms of dupilumab-based combination therapy by performing transcriptome sequencing of peripheral blood leukocytes before and after treatment, with a focus on genes involved in key biological processes and signaling pathways that change between the pre- and post-treatment states. The focus will be on identifying genes that may play a critical role in the therapeutic response to dupilumab-based therapy.

Patients and Methods

Clinical Cohort Study

This real-world clinical cohort study retrospectively included consecutive patients with BP who received dupilumab combined with TG between November 2021 and October 2024 at Peking Union Medical College Hospital. The study protocols were reviewed and approved by the ethics committee of PUMCH (I-24PJ2760). Written informed consent was obtained from all participants. All data were anonymized and handled confidentially. This study is reported in accordance with the STROBE guideline.

At our center, this regimen was considered as a corticosteroid-sparing option for patients at high risk of corticosteroid-related adverse events. Systemic corticosteroids were avoided whenever feasible, and rescue initiation or dose escalation was permitted for inadequate control or relapse. The inclusion and exclusion criteria were as follows:

Inclusion Criteria

(1) Age ≥18 years; (2) Diagnosis of BP based on clinical, histological, and immunopathological criteria; (3) Moderate to severe disease (bullous pemphigoid disease area index [BPDAI] ≥20) with active stage;¹⁴ and (4) Willingness to receive dupilumab combined with TG and participate in follow-up.

Exclusion Criteria

(1) Patients with concomitant autoimmune diseases; (2) Known contraindications to dupilumab or TG; (3) Active infections or malignancies; and (4) Inability to comply with regular follow-up.

Of the 13 eligible patients identified in the cohort, one did not initiate treatment and was excluded, leaving 12 patients for analysis. The 12 included patients received an initial dose of 600 mg subcutaneous dupilumab, followed by 300 mg administered every two weeks, combined with oral TG (Zhejiang Deender Pharmaceutical Co., Ltd., China, 40–60 mg daily). Supportive care was limited to topical halometasone/triclosan cream applied to active lesions during blistering episodes. No systemic therapies other than dupilumab, TG, and systemic corticosteroids (continued at baseline dose when applicable or used as rescue) were administered. The primary endpoints were achievement of disease control (no new skin lesions and improvement of existing lesions)¹ and complete remission (absence of new and existing BP lesions, including blisters, eczematous or urticarial plaques, or mucosal lesions, and cessation of pruritus).¹⁵ Secondary endpoints included changes in BPDAI scores, pruritus Numeric Rating Scale (NRS) scores, anti-BP180 antibody levels, eosinophil percentages, and incidence of adverse events. All patients were followed up for a total of six months after treatment initiation. Outcomes (BPDAI and pruritus NRS) were assessed using standardized procedures by trained dermatologists. The same assessor evaluated the same participant whenever feasible. Given the observational design without randomization, confounding by indication cannot be excluded. No missing data were observed for the clinical outcomes (BPDAI, pruritus NRS, BP180, and eosinophil counts) at baseline and week 12 (n = 12).

Transcriptomic Profiling and Bioinformatic Analysis

Transcriptome profiling was performed on longitudinal peripheral blood leukocyte samples collected before (n = 6) and after treatment (n = 5). One post-treatment blood RNA sample was excluded from transcriptomic analysis due to RNA degradation. Total RNA was extracted using standard protocols, and sequencing was conducted on an Illumina platform. Reads were aligned to the GRCh38 reference genome using HISAT2, and quantified with featureCounts. Differential expression was analyzed using the DESeq2. Given the limited sample size, gene-level results were considered exploratory: log2 fold changes and nominal P-values are reported, and Benjamini-Hochberg FDR-adjusted p -values are provided for transparency. Because no genes reached FDR < 0.05, downstream interpretation prioritized pathway-level changes assessed by GO/KEGG enrichment (clusterProfiler) and Gene Set Enrichment Analysis using MSigDB GO Biological Process and ImmuneSigDB gene sets, with FDR-adjusted p < 0.05 (and |NES| > 1 for GSEA) considered significant. For exploratory network analyses, candidate genes were defined using nominal p < 0.05 and |log2FC| > 1 for hypothesis generation. Protein–protein interaction networks were constructed through STRING and visualized in Cytoscape, and hub genes were identified using centrality-based ranking algorithms. Detailed parameters and full analytical outputs are provided in Supplementary Methods.

qRT-PCR Validation

To validate the expression of the identified hub genes, quantitative reverse transcription polymerase chain reaction (qRT-PCR) was performed on peripheral blood leukocytes collected from four pre-treatment and four post-treatment patients. Total RNA was reverse-transcribed, and qRT-PCR was carried out using SYBR Green with glyceraldehyde-3-phosphate dehydrogenase (GAPDH) as the internal reference gene. Each sample was tested in triplicate. Primer specificity was confirmed by agarose gel electrophoresis and single-peak melting curve analysis. The relative expression levels of C-X-C motif chemokine receptor 4 (CXCR4), matrix metalloproteinase 9 (MMP-9), and epidermal growth factor (EGF) were calculated using the 2^(-ΔΔCT) method. Primer sequences:

MMP-9 Forward: 5′-GTCTTCCAGTACCGAGAGAAAGC-3′

MMP-9 Reverse: 5′-TGCAGGATGTCATAGGTCACG-3′

CXCR4 Forward: 5′-CCTCTACAGCAGTGTCCTCATCCT-3′

CXCR4 Reverse: 5′-AGACGCCAACATAGACCACCTT-3′

EGF Forward: 5′-TGCGTGGTGGTGCTTGTCAT-3′

EGF Reverse: 5′-GCCTGCGACTCCTCACATCTC-3′

GAPDH Forward: 5′-CAGGTGGTCTCCTCTGACTTCA-3′

GAPDH Reverse: 5′-CACCCTGTTGCTGTAGCCAAAT-3′

Statistical Analysis and Data Visualization

Statistical analyses were carried out using R version 4.4.1. The Shapiro–Wilk test was used to check the normality of paired differences. Continuous variables were presented as mean (SD) if they were normally distributed, and as median (IQR) if they were not. Changes from baseline to week 12 were tested with paired Student t tests for normally distributed differences, or with Wilcoxon signed-rank tests for non-normal differences. All hypothesis tests were two-sided, and 95% confidence intervals were given for all effect estimates. Categorical variables were summarized descriptively. A p value < 0.05 was considered statistically significant. All figures were created using R software.

Results

Dupilumab Combined with TG Led to Significant Improvement in Clinical Parameters and Showed a Favorable Safety Profile

A total of 12 patients were included in this cohort study. Their baseline characteristics, disease duration, comorbidities, corticosteroid use, and time to disease control were shown in Table 1. Ten (83%) patients reached CR within three months after starting treatment. Nine patients (75%) achieved CR without systemic corticosteroids, using only dupilumab and TG. One patient (Case 11), who had disease relapse while taking 10 mg of prednisone daily at baseline, achieved disease control after dupilumab and TG were added, without increasing the corticosteroid dose. Another patient (Case 10) initially responded to the combination therapy but relapsed at week 10 and needed additional corticosteroids (15 mg daily) to regain control. One patient (Case 12) had an inadequate response to the combination therapy and eventually needed high-dose corticosteroids (50 mg daily) to control the disease. The median time to disease control was eight days (IQR, 7–9.75 days), based on the 11 patients who responded to the combination therapy, including those who did not need corticosteroids or only received low doses (≤10 mg daily). Marked clinical and laboratory improvements were seen after treatment (Figure 1). BPDAI scores decreased significantly (Hodges–Lehmann median difference, −37.50; 95% CI, −62.65 to −30.65; p = 0.003). Pruritus NRS improved (Hodges–Lehmann median difference, −7.58; 95% CI, −9.00 to −6.00; p = 0.002). Anti-BP180 antibody levels declined (Hodges–Lehmann median difference, −33.00 U/mL; 95% CI, −62.00 to −7.00; p = 0.006). EOS% decreased (mean difference, −8.73%; 95% CI, −12.88% to −4.58%; p < 0.001). Although the decrease in Figure 1e appeared less marked than the changes in BPDAI, pruritus NRS, and eosinophil percentage, the paired analysis still showed a significant overall reduction by week 12. This visual impression likely reflects inter-individual heterogeneity, with some patients showing only modest declines during the 12-week follow-up. In terms of safety, during the six-month follow-up, only one patient (Case 6) had an adverse event, with a temporary worsening of dementia symptoms in the first month after starting treatment, on a background of pre-existing Alzheimer’s disease. No adverse events were seen in the other patients during this period. These results suggest that the combination of dupilumab and TG has good clinical efficacy and a favorable safety profile in moderate to severe BP.

Table 1 Clinical Features and Outcomes in BP Patients Treated with Dupilumab and TG

Figure 1 Clinical and Biochemical Improvements in BP Patients Treated with Dupilumab combined with Tripterygium Glycoside. (a and b) Representative clinical photographs showing marked resolution of blisters and erythema after treatment. (c) Pruritus severity assessed by Numeric Rating Scale (NRS). (d) Disease activity assessed by Bullous Pemphigoid Disease Area Index (BPDAI). (e) Serum anti-BP180 antibody levels. (f) Peripheral blood eosinophil percentage (EOS%). Lines connect paired samples from the same patient. Asterisks indicate the significance of baseline vs post-treatment changes, **p < 0.01, ***p < 0.001 (two-sided paired tests).

Differential Transcriptomic Expression and GO/KEGG Enrichment Analyses

This study showed that treatment with dupilumab combined with TG had a marked impact on gene expression in patients with BP. Principal component analysis showed a clear separation between pre- and post-treatment samples, which suggests broad transcriptomic changes (Figure 2a). The volcano plot (Figure 2c) identified 214 nominally upregulated and 593 downregulated candidate genes (nominal p < 0.05 and |log2FC| > 1). No genes reached gene-level FDR < 0.05; therefore, gene-level findings were considered exploratory. The heat map (Figure 2b) showed the distinct expression patterns of these genes. GO enrichment analysis and the gene-concept network indicated that biological processes such as blood coagulation, hemostasis, wound healing, and platelet activation were significantly affected, with key genes like CXCR4 and TLR4 in central positions (Figure 2d and e; Table S1). KEGG pathway enrichment analysis further showed that pathways such as complement and coagulation cascades were significantly enriched, which helps explain the molecular mechanisms of this treatment (Figure 2f).

Figure 2 Transcriptomic Analysis and GO/KEGG Enrichment. (a) Principal Component Analysis (PCA) illustrating the separation of pre- and post-treatment samples. (b) Heatmap showing gene expression patterns. (c) Volcano plot highlighting genes with significant expression changes. (d) Results of GO biological process enrichment analysis. (e) Gene-concept network diagram. (f) Results of KEGG pathway enrichment analysis.

GSEA Analysis

The GSEA results showed that the treatment strongly affected several crucial biological processes and immune pathways. As shown in Figure 3a, processes such as negative regulation of coagulation, hemostasis, platelet activation, and wound healing were all clearly suppressed after treatment. Figure 3b showed immune-related pathways in more detail, including cytokine-mediated signaling, interferon-mediated signaling, responses to type I interferon, IL-8 production and its positive regulation, production of tumor necrosis factor (TNF) superfamily cytokines and their regulation, and the control of inflammatory responses. All of these pathways were significantly suppressed after treatment, indicating broad immune modulation. Figure 3c demonstrated changes in genes that are upregulated or downregulated in neutrophils compared with Th2 responses. Genes that are usually upregulated were clearly suppressed, while genes that are usually downregulated were activated, which demonstrates a two-way regulatory effect of the treatment on neutrophils. Figure 3d identified changes in gene expression in several immune cell types, including basophils, dendritic cells, eosinophils, macrophages, and mast cells, again in relation to Th2 responses. The GSEA results indicated that the activity of all these cell types was generally suppressed after treatment, supporting the hypothesis that the treatment modulates the activity of multiple immune cells.

Figure 3 Gene Set Enrichment Analysis (GSEA) Using Defined Gene Sets. (a) Enrichment plots for biological processes related to hemostasis, negative regulation of coagulation, platelet activation, and wound healing showing significant suppression post-treatment. (b) Detailed plots displaying suppressed cytokine-mediated pathways, highlighting the extensive immune modulation following treatment. (c) Enrichment plots illustrating both suppression and activation of neutrophil genes relative to Th2 responses, indicating the bidirectional effects of the treatment on neutrophils. (d) General suppression in the activity of various immune cells relative to Th2 responses post-treatment, affirming the broad immunomodulatory impact of the therapy.

Network Analysis and Hub Gene Identification

The large PPI network built from the nominally differentially expressed genes was shown in Figure 4a. We used the MCODE plugin to find highly interconnected submodules and identified several clusters (Table S2). Cluster 1 had the highest score (11.391) and contained 24 nodes, including TLR4, MMP-9, and CXCR4, all with high MCODE node scores (Figure 4b). The darker nodes in this subnetwork indicate tighter local connectivity, suggesting that they play a central role in treatment-related processes (Figure 4b). KEGG enrichment analysis of the genes in this cluster (Figure 4c) showed that many of them were involved in inflammatory and immune-related pathways, such as the TNF and IL-17 signaling pathways. Subsequently, the CytoHubba plugin was used to identify central hub genes. The plugin ran seven different centrality algorithms, and each algorithm generated a list of the top 15 genes. The overlap of these lists gave TLR4, JUN, MMP-9, CXCR4, and EGF as core hub genes (Table S3). Expression analysis showed that TLR4, MMP-9, CXCR4, and EGF all decreased after treatment, while JUN increased (Figure 4d). Receiver operating characteristic (ROC) curve analysis suggested that CXCR4 (AUC = 0.800 [0.505–1.000], p = 0.046) and MMP-9 (AUC = 0.867 [0.593–1.000], p = 0.009) could discriminate pre- versus post-treatment samples, warranting further validation in independent cohorts before clinical application (Figure 4e). Taken together, these exploratory results suggest that the hub genes are located in tightly interconnected submodules and undergo marked changes associated with treatment-related immune and inflammatory processes.

Figure 4 Network and Functional Analyses of Hub Genes. (a) The PPI network constructed from all candidate genes (nominal p < 0.05 and |log2FC|>1). (b) The top-scoring subnetwork identified by MCODE. (c) KEGG enrichment analysis of the subnetwork’s constituent genes. (d) Boxplot of hub gene expression profiles pre- and post-treatment, based on normalized TPM matrix from RNA-seq data, showing treatment-associated changes in gene expression. (e) ROC curves for hub genes, assessing their ability to distinguish pre- versus post-treatment samples. (f) qRT-PCR validation of CXCR4, MMP-9, and EGF expression in independent pre- and post-treatment samples (n = 4 per group). A single asterisk (*) in panel (f) indicates p < 0.05.

qRT-PCR Validation of Hub Genes

The qRT-PCR results were consistent with the transcriptomic analysis and confirmed significant changes in gene expression after treatment. In particular, the expression of CXCR4, MMP-9, and EGF decreased after treatment, with p-values of 0.034, 0.018, and 0.038, respectively (Figure 4f). These findings support the downregulation seen in the high-throughput data and confirm the biological effect of the treatment on these key regulatory genes.

Discussion

Dupilumab is an IL-4 receptor α antagonist and is now an effective and safe treatment option for patients with BP. In daily practice, it is usually given together with systemic corticosteroids. However, this combination can cause adverse effects such as infections and metabolic problems. In this study, we assessed a corticosteroid-sparing regimen that dupilumab combined with TG, a traditional Chinese medicine with anti-inflammatory and immunomodulatory effects. In our 12-patient clinical cohort study, this combination led to complete remission within three months in 10 of the 12 patients, and 9 of them reached remission without systemic corticosteroids. Only one patient had a transient adverse event (worsening of pre-existing dementia), and no serious infections or other adverse reactions were seen during the six-month follow-up. These findings suggest that the dupilumab-TG regimen has encouraging clinical efficacy and a favorable safety profile, although causal treatment effects cannot be established due to the observational design. In our combination regimen, dupilumab likely played the main therapeutic role, while TG may have provided additional immunomodulatory and corticosteroid-sparing effects.^6,7,10 At the same time, transcriptomic and bioinformatic analyses pointed to candidate genes and pathways associated with treatment. The PPI network and centrality analyses identified TLR4, JUN, MMP-9, CXCR4, and EGF as central hub genes involved in BP pathogenesis. CXCR4 and MMP-9 were downregulated after treatment, and ROC curve analysis showed that these two genes may help distinguish pre- from post-treatment samples (AUC > 0.8), warranting independent validation before being considered biomarkers.

The CXCL12/CXCR4 axis plays a key role in B cell chemotaxis and differentiation in BP. It promotes the migration and accumulation of CXCR4+ B cells in BP skin lesions, which supports the production of autoantibodies. This is driven by increased levels of CXCL12, the ligand of CXCR4, in the blister fluid and serum of patients with BP.¹⁶ The interaction between CXCR4 and CXCL12 promotes B cell differentiation into antibody-secreting cells by activating the transcription factor c-Myc.^16,17 Recent evidence also shows that IL-13 can stimulate fibroblasts to increase the expression of phospholipase A2 group IIA, which then raises CXCL12 levels and promotes immune cell infiltration through the CXCR4 pathway.¹⁸ In our study, transcriptomic analysis showed a clear decrease in CXCR4 expression after treatment. By blocking IL-13 signaling, dupilumab may interrupt the fibroblast-driven amplification of the CXCL12/CXCR4 loop and reduce the chemotaxis of immune cells into lesion sites. Taken together, the combination of dupilumab and TG may act in part by inhibiting CXCR4, which lowers both immune cell infiltration into lesions and B cell–mediated production of pathogenic autoantibodies.

MMP-9 is an important enzyme involved in the development of several skin diseases, including BP.^19–21 In BP, an abnormal immune response leads to the production of IgG and IgE autoantibodies against the structural proteins BP180 and BP230, which causes neutrophil and eosinophil chemotaxis and then degradation of the basement membrane zone.^19,20 This degradation, which is partly driven by MMP-9, includes the breakdown of collagen and other extracellular matrix components and promotes blister formation and disease progression.^22,23 Anti-BP180 NC16A domain IgE antibodies/BP180 NC16A domain immune complexes can directly induce eosinophils to release MMP-9, and mice that lack MMP-9 are resistant to BP, showing that MMP-9 is crucial for both tissue damage and eosinophil recruitment in this disease.¹⁹ In addition, Th17-related cytokines, such as IL-17 and IL-23, increase MMP-9 production, which then activates downstream neutrophil elastase, leading to BP180 degradation and blister formation.²⁰ Our study found that MMP-9 levels clearly decreased after treatment with dupilumab and TG. This decrease may reduce MMP-9–mediated damage to basement membrane components, such as BP180, and as a result may lessen dermal-epidermal separation and blister formation.^19,20,24 Our findings also suggest that the treatment suppresses Th2 cells and may at the same time affect Th17 cells, which reflects the complex nature of the immune response. These observations indicate that the therapeutic effects are not limited to a single immune cell type but likely involve broader modulation of the immune response. TG may also have contributed through broader immunomodulatory effects. In our study, the reduction in MMP-9 and inflammatory pathways suggests that TG may have helped dampen effector-cell activation and protease-mediated tissue injury.²⁵ However, this interpretation remains speculative because the effects of TG cannot be separated from dupilumab in this combination cohort.

Additionally, our transcriptome analysis showed clear changes in genes related to coagulation pathways. In the treated samples, GO and KEGG analyses indicated enrichment of terms linked to coagulation, hemostasis, wound healing, and platelet activation, and GSEA suggested that these pathways were generally suppressed after treatment. These pathways are important in BP because inflammation can disturb normal coagulation and raise the risk of thrombotic events.^26–28 Previous studies have reported that abnormal coagulation is common in autoimmune diseases and can make the disease more severe.²⁶ By regulating these pathways, the treatment may help stabilize hemostasis and reduce the risk of complications caused by disordered coagulation. However, because whole-blood transcriptomes can be influenced by shifts in leukocyte and platelet composition, coagulation/platelet-related signals should be interpreted cautiously and regarded as hypothesis-generating. Further studies incorporating cell-composition adjustment and clinical endpoints are needed to clarify the clinical relevance of these findings.

GSEA further showed that dupilumab combined with TG importantly affects immune pathways in BP, especially the IL-8 and type I interferon pathways. IL-8 is a chemokine that plays an important role in attracting and activating neutrophils, which are central to BP pathogenesis. Studies have found higher IL-8 levels in the serum and blister fluid of BP patients, which supports its role in the inflammatory processes that lead to blister formation.²⁹ GSEA also showed that the treatment has a marked impact on neutrophil function, along with effects on other immune cells, and this change is closely related to inhibition of the IL-8 pathway.

However, our study has some limitations. This was a small, single-center observational cohort without a comparator group, and confounding by indication cannot be excluded. Therefore, the findings may not be generalizable. Transcriptomic analysis was exploratory (no genes reached gene-level FDR significance), and whole-blood signals may partly reflect shifts in cell composition. Although qRT-PCR supported expression changes, the proposed candidate markers and mechanistic inferences require validation in larger controlled studies.

Conclusions

Dupilumab combined with TG achieved rapid disease control and meaningful clinical improvement in older patients with moderate to severe BP, while limiting systemic corticosteroid exposure. The treatment was generally well tolerated. Transcriptomic changes in CXCR4 and MMP-9 related pathways may help explain the treatment effect and suggest potential biomarkers. Because this was a small, single-center cohort, these findings need to be confirmed before being applied to the broader BP population.