Immunotherapy may be more appropriate for ERBB2 low-expressing extramammary paget’s disease patients: a prognosis analysis and exploration of targeted therapy and immunotherapy of extramammary paget’s disease patients

Abstract

Extramammary Paget’s disease (EMPD) is a rare cutaneous malignancy characterized by its uncertain etiology and metastatic potential. Surgery remains the first-line clinical treatment for EMPD, but the efficacy of radiotherapy and chemotherapy remains to be fully evaluated, and new therapies for EMPD are urgently needed. In this study, we initially screened 815 EMPD patients in the Surveillance, Epidemiology, and End Results (SEER) database and analyzed their clinical features and prognostic factors. Using the dataset from the Genome Sequence Archive (GSA) database, we subsequently conducted weighted gene coexpression network analysis (WGCNA), gene set enrichment analysis (GSEA), gene set variation analysis (GSVA), and immune infiltration analyses, grouping the samples based on EMPD disease status and the levels of ERBB2 expression. The prognostic analysis based on the SEER database identified increased age at diagnosis, distant metastasis, and receipt of radiotherapy as independent risk factors for EMPD. Moreover, our results indicated that patients who received chemotherapy had worse prognoses than those who did not, highlighting the urgent need for novel treatment approaches for EMPD. Functional analysis of the GSA-derived dataset revealed that EMPD tissues were significantly enriched in immune-related pathways compared with normal skin tissues. Compared with those with high ERBB2 expression, tissues with low ERBB2 expression displayed greater immunogenicity and enrichment of immune pathways, particularly those related to B cells. These findings suggest that patients with low ERBB2 expression are likely to benefit from immunotherapy, especially B-cell-related immunotherapy.

Introduction

Mammary and extramammary Paget’s disease are rare epithelial-derived intraepidermal adenocarcinomas that are classified as cutaneous malignancies. The lesions of extramammary Paget’s disease (EMPD) predominantly manifest in areas such as the perianal region, scrotum, penis, and vulva [1]. The diagnosis of EMPD relies on Paget cells and other specific molecular markers identified via histopathological examination [2]. In addition to traditional markers, such as cytokeratin (CK) and carcinoembryonic antigen (CEA), recent studies have introduced several other immunohistochemical (IHC) markers, which can be utilized to predict prognosis, classify the disease, and guide clinical treatments. Among these markers, overexpression of the oncogene HER2/ERBB2 is observed in 30–50% of EMPD patients and is associated with a poorer prognosis and a more aggressive, recurrent disease status [3, 4]. The overexpression of HER2/ERBB2 also provides EMPD patients with an alternative choice for clinical treatment. In recent years, trastuzumab was successfully used to treat patients with EMPD and high ERBB2 expression [5, 6].

In addition to the standard treatment options for EMPD, which include surgical excision, local radiotherapy, photodynamic therapy, and conservative chemotherapy for unresectable lesions [6], the advent of immunotherapy has led to revolutionary advancements in the field of cancer treatment. Immunotherapy is closely related to the genetic instability of tumors, and genetic diversity contributes to the emergence of tumor antigens, allowing tumors to be recognized and eliminated by the human immune system [7, 8]. In recent years, numerous clinical trials have confirmed the efficacy of immunotherapy in the treatment of various types of tumors [9–11]. The results of the large KEYNOTE-355 clinical trial also suggested that pembrolizumab significantly prolonged progression-free survival (PFS) of patients with metastatic triple-negative breast cancer (TNBC) and a baseline combined positive score (CPS) of PD-L1 ≥ 10 [12]. However, recent studies have revealed that tumor cells can also evade immune surveillance through multiple mechanisms, such as the induction of an immunosuppressive tumor microenvironment (TME) and impaired antigen presentation [13], which weaken the efficacy of immunotherapy. Thus, understanding the composition of the TME is crucial to improve the efficacy of immunotherapy and select appropriate patients for immunotherapy treatment. In mammary Paget’s disease (MPD) and EMPD, immune-related treatments are also beginning to gain attention. Several studies have indicated the efficacy of the topical application of imiquimod, an immune response modifier, for treating EMPD [14, 15]. Studies have also shown that PD-L1 is expressed in the tumor cells of EMPD patients and in the tumor-associated lymphocytic infiltrates of MPD patients. In MPD, patients who are HER2/neu-positive exhibit lower PD-L1 lymphocyte expression than HER2/neu-negative patients [16]. In some EMPD patients with low HER2/neu positivity, the targets of immune checkpoint inhibitors can be positive [17]. Additionally, one patient in a comparative study who did not respond to pembrolizumab treatment was found to have HER2 amplification and subsequently started trastuzumab therapy [18]. However, research on the role of the TME in EMPD and its relationship with HER2 expression is lacking.

This study reports an analysis of clinical risk factors in EMPD patients based on the Surveillance, Epidemiology, and End Results (SEER) database. In addition, this study used a genome sequence archive (GSA)-sourced dataset to investigate the functional differences between the EMPD and normal skin groups, as well as the correlations between HER2 expression and immune cell infiltration in EMPD tissues. The findings suggest the potential for using immunotherapy in EMPD patients with low HER2 expression, providing a theoretical foundation for the precise clinical treatment of EMPD.

Materials and methods

Datasets and data preprocessing

First, we collected data from the SEER database, the largest publicly available data source on cancer incidence and survival in the United States [19], to investigate the prognostic factors for survival in EMPD patients. The following demographic and clinical characteristics were collected for analysis: age at diagnosis, sex, race, year of diagnosis, marital status, primary site, tumor size, SEER historic stage A, surgical and chemoradiotherapy information, survival time, and survival status. SEER*Stat Version 8.4.2 (Surveillance Research Program, NCI, Bethesda, MD, USA) was used to screen and collect information on representative patients according to the process shown in Fig. 1. We initially included patients who were diagnosed with primary EMPD [International Classification of Diseases for Oncology, Third Edition (ICD-O-3): 8542/3] with a primary site in the vulva, truncal skin, perianal area, penis, scrotum, or other area between 2000 and 2020. Patients with the following criteria were excluded: (1) patients with unknown age at diagnosis, marital status, race, diagnosis confirmation, histopathology, stage (SEER historic stage A), treatment information, and survival months; (2) patients with nonfirst primary malignant tumors; (3) patients who died from causes other than cancer; and (4) patients who survived < 1 month (to exclude patients lost to follow-up).

Fig. 1.

Screening process for the data from the SEER program

Each index of the SEER database data was presented in the form of counting data and expressed in the form of the number of cases. The “compareGroups” package of R was used for statistical analysis of the clinical characteristic data. The “survminer” and “survival” packages of R were used for univariate and multivariate Cox regression analysis, and the “ggsurvplot” package was used to draw Kaplan‒Meier survival curves.

Subsequently, bulk RNA-seq fastq files were obtained from the HRA001914 dataset in the GSA database (https://ngdc.cncb.ac.cn/gsa-human/browse/HRA001914), following approval from the dataset owner and submitter. Subsequent data processing was performed on a Linux system. Data quality was assessed via “FastQC”, and data were preprocessed via “Trim-Galore”. The sequencing data were aligned to the GRCh38.p14 reference genome using “HISAT2”, and the final expression matrix was obtained using “featureCounts”.

Weighted gene coexpression network analysis (WGCNA)

WGCNA can be used to identify modules of highly correlated genes and to relate these modules to each other and external sample traits via eigengene network methodology [20]. After normalizing the count form expression data from the HRA001914 dataset to the TPM format, we first performed log₂(x + 1) transformation and then conducted the analysis via the “WGCNA” R package. We identified the modules that were most strongly associated with the phenotypes of interest and subsequently performed GO functional enrichment analysis on the genes within the selected modules using the “clusterProfiler” package [21].

Gene set enrichment analysis (GSEA) and gene set variation analysis (GSVA)

GSEA employs differentially expressed genes between sample groups for functional enrichment analysis [22], whereas GSVA estimates the variations in gene set enrichment for each sample based on the expression profiles to elucidate functional enrichment differences between sample groups [23]. GSEA and GSVA in this study were conducted using the OmicStudio tools at https://www.omicstudio.cn/tool [24].

Estimation of immune microenvironment scores, immune cell infiltration, and signature set sco res

The normalized expression profile data were further analyzed using the “IOBR” R package to assess immune microenvironment scores, immune cell infiltration, and tumor-related signature set scores. The “IOBR” package incorporates various existing immune microenvironment deconvolution methods, including CIBERSORT, ESTIMATE, IPS, EPIC, xCell and quanTIseq, and also integrates feature construction tools from published studies [25]. Additionally, immune scoring of the samples was performed via The Cancer Immunome Atlas (TCIA). TCIA develops immune scoring methods based on data from The Cancer Genome Atlas (TCGA) and multiple immunotherapy trials. The calculated immunophenoscore (IPS) is effective in predicting the outcomes of immunotherapy [26].

Results

Analysis of the clinical features and survival prognostic factors of EMPD patients

In total, 815 patients in the SEER database who were diagnosed with EMPD from 2000 to 2020 according to our criteria were included. The clinical characteristics of the 815 patients are summarized in Table 1. The median age at diagnosis was 68 years (range, 61–76 years). of the cohort was predominantly female (68.1% vs. 31.9%). In terms of racial distribution, 76.3% of patients were White; 22.7% were American Indian, Alaskan Native or Asian/Pacific Islander; and only 1% were Black. A total of 543 (66.6%) patients were married, 181 (22.2%) patients were widowed, divorced or separated, and 91 (11.2%) patients were never married. We divided the primary sites of EMPD into the following six categories: perianal region, trunk skin, vulva, penis, scrotum and others. The top three locations of primary lesions were the vulva (58.2%), trunk skin (20.2%) and scrotum (13.7%), whereas lesions on the penis, perianal and other sites accounted for less than 3% of all lesions. Among the 38.8% of patients with known tumor sizes, the median size of the lesion was 38 millimeters (mm), with 15.4% of patients having tumor lesions larger than 42 mm. The majority of EMPD patients presented with localized lesions (85.8%), with 12.3% exhibiting regional metastasis and only 1.9% showing distant metastasis. In terms of treatment, surgery was the main treatment method (86.1%), and only a few patients received radiotherapy (4.9%) or chemotherapy (4.4%). The 5-year survival rate was 87.7%, and the median overall survival (OS) was 76 months (range 31–140 months).

Table 1.

Clinical characteristics of EMPD patients in the SEER database

Clinical characteristics	No. of patients	Clinical characteristics	No. of patients
Age at diagnosis (years, IQR, 68.0 [61.0; 76.0])		Tumor size (mm, IQR, 38.0 [20.0; 53.2])
< 65	298 (36.6%)	≤ 20	89 (10.9%)
65–74	275 (33.7%)	21–41	102 (12.5%)
≥ 75	242 (29.7%)	≥ 42	125 (15.4%)
Sex		NA	499 (61.2%)
Female	555 (68.1%)	SEER historic stage A
Male	260 (31.9%)	Localized	699 (85.8%)
Race		Regional	100 (12.3%)
Black	8 (1.0%)	Distant	16 (1.9%)
Other*	185 (22.7%)	Surgery
White	622 (76.3%)	No	113 (13.9%)
Marital status		Yes	702 (86.1%)
Married	543 (66.6%)	Radiation
Other#	181 (22.2%)	No	755 (95.1%)
Never married	91 (11.2%)	Yes	40 (4.9%)
Primary site		Chemotherapy
Other	21 (2.6%)	No	779 (95.6%)
Penis	19 (2.3%)	Yes	36 (4.4%)
Perianal area	24 (2.9%)	Survival state
Scrotum	112 (13.8%)	Alive	715 (87.7%)
Truncal skin	165 (20.2%)	Death	100 (12.3%)
Vulva	474 (58.2%)	Overall survival(month, IQR, 76.0 [31.0; 140.0])

^*Other: American Indians, AK Natives, or Asian/Pacific Islanders.

^#Other: Widowed/Divorced/Separated, W/D/S

To explore the risk factors for EMPD, we performed univariate and multivariate Cox regression analyses on the enrolled population (Table 2) and used the Kaplan‒Meier method for survival analysis. Univariate Cox regression analysis revealed that OS-related risk factors included age at diagnosis, sex, marital status, tumor stage, surgery, radiotherapy and chemotherapy (p < 0.05), whereas race, primary site and tumor size were not significantly correlated with OS (p > 0.05). Multivariate Cox regression analysis was used to screen out independent risk factors associated with OS, including age at diagnosis, tumor stage, and radiotherapy. Combining the above results with survival curve analysis, we found that patients with a diagnosis age of 65–74 years and over 75 years had a significantly lower survival rate than patients with a diagnosis age of less than 65 years. Furthermore, a higher age at diagnosis corresponded to a worse prognosis (p < 0.0001) (Fig. 2A). Moreover, female patients had slightly better survival than male patients (p = 0.015) (Fig. 2B). Marital status was also associated with the survival prognosis of EMPD patients, with widowed, divorced or separated patients exhibiting significantly lower survival rates (p = 0.00027) (Fig. 2D). Patients with lesions in the perianal region had a worse prognosis than those with lesions on the penis, vulva, trunk skin, or scrotum (p = 0.0032) (Fig. 2E). Compared with EMPD patients with distant metastasis, patients with localized lesions and regional metastasis had a significantly lower risk of death, and those with distant metastasis had the lowest survival rate (p < 0.0001) (Fig. 2G). Tumor stage was also significantly correlated with disease prognosis, suggesting that earlier stages correlate with better clinical outcomes. Patients who underwent surgery had a better prognosis than those who did not undergo surgery (p < 0.0001) (Fig. 2H). Our analysis indicated that patients who did not receive radiotherapy had a greater survival rate than those who received radiotherapy (p < 0.0001) (Fig. 2I). Furthermore, patients who did not receive chemotherapy also had a better prognosis than those who received chemotherapy (p < 0.0001) (Fig. 2J). However, race (Fig. 2C) and tumor size (Fig. 2F) were not significantly associated with survival. Our results revealed that conventional nonsurgical treatments for EMPD have shown limited efficacy, emphasizing the need to explore more effective therapeutic approaches.

Table 2.

Univariate and multivariate Cox analyses of EMPD patients in the SEER database

Clinical characteristics	No.of patients	Univariate analysis		Multivariate analysis
		HR (95%CI)	p-value	HR (95%CI)	p-value
Age at diagnosis
<65	298	1.0		1.0
65–74	275	1.83 (1.03–3.24)	0.038	2.29 (1.26–4.16)	0.007
≥ 75	242	4.77 (2.83–8.05)	<0.001	4.64 (2.63–8.19)	<0.001
Sex
Female	555	1.0		1.0
Male	260	1.63 (1.09–2.42)	0.016	1.50 (0.99–2.26)	0.055
Race
Black	8	1.0
White	622	3,330,000 (0-Inf)	0.993
Other*	185	3,340,000 (0-Inf)	0.993
Marital status
Married	543	1.0		1.0
Other#	181	2.36 (1.53–3.64)	<0.001	1.50 (0.92–2.44)	0.100
Never married	91	1.67 (0.91–3.07)	0.096	1.64 (0.88–3.07)	0.119
Primary site
Other	21	1.0
Penis	19	0 (0-Inf)	0.993
Perianal area	24	1.37 (0.05–4.18)	0.585
Scrotum	112	0.61 (0.22–1.67)	0.339
Truncal skin	165	0.60 (0.23–1.57)	0.297
Vulva	474	0.40 (0.16–1.01)	0.053
Tumor size
≤ 20 mm	89	1.0
21–41 mm	102	1.50 (0.72–3.11)	0.279
≥ 42 mm	125	1.68 (0.73–3.86)	0.224
NA	499	–	–
SEER historic stage A
Distant	16	1.0		1.0
Localized	699	0.04 (0.02–0.07)	<0.001	0.04 (0.02–0.09)	<0.001
Regional	100	0.15 (0.08–0.28)	<0.001	0.16 (0.07–0.35)	<0.001
Surgery
No	113	1.0		1.0
Yes	702	0.37 (0.24–0.59)	<0.001	0.71 (0.39–1.28)	0.251
Radiation
No	775	1.0		1.0
Yes	40	3.65 (2.04–6.55)	<0.001	2.00 (1.03–3.89)	0.042
Chemotherapy
No	779	1.0		1.0
Yes	36	4.41 (2.50–7.76)	<0.001	1.40 (0.66–2.89)	0.381

^*Other: American Indians, AK Natives, or Asian/Pacific Islanders.

^#Other: Widowed/Divorced/Separated, W/D/S

Fig. 2.

Kaplan‒Meier survival analysis of OS in EMPD patients based on age at diagnosis A, sex B, race C, marital status D, primary site E, tumor size F, SEER historic stage A G, surgery H, radiotherapy I, and chemotherapy J

Enrichment of Immune-related pathways in EMPD tissues

Understanding the disease characteristics of EMPD is crucial to explore other effective therapeutic strategies for EMPD. Thus, to preliminarily investigate the functional differences between normal skin tissues and EMPD tissue samples, WGCNA was used to analyze the GSA database-sourced expression profiles of normal skin tissues and EMPD tissues, with a soft threshold calibrated to 9 (scale free R² = 0.80) (Fig. 3A). Subsequently, 14 gene modules (each containing at least 30 genes) were identified and associated with phenotypes (normal skin versus EMPD) (Fig. 3B–C). After evaluating the correlations of each gene module with the phenotypes, we performed GO functional enrichment analysis using the genes from the modules that were most highly correlated with the normal skin group (brown and green modules) and the EMPD group (greenyellow module) (Fig. 3D–E). The results indicated that the genes most correlated with the normal skin group were enriched primarily in the regulation of normal skin functions, such as epidermis development, the cornified envelope and the hair cycle, whereas the genes most correlated with the EMPD group were enriched in immune-related pathways, particularly those related to B-cell function and antibody-mediated immune responses (Fig. 3F). Additionally, GSEA and GSVA revealed consistent results. Both the HALLMARK pathway and the GO enrichment analysis revealed enrichment of immune-related pathways in EMPD tissues, including the inflammatory response, adaptive immunity, chemokine receptor binding, and B-cell receptor signaling pathways (Fig. 3G–I).

Fig. 3.

Construction of weighted gene coexpression networks and functional analyses based on the expression profiles of the HRA001914 dataset. A Calibration of the best soft-threshold power. B Construction of a weighted coexpression network and identification of gene modules. C Correlation analysis and heatmap of module-trait relationships. D–E Eigengene adjacency heatmap illustrating the correlations among gene modules and between gene modules and phenotypes (normal skin and EMPD). F GO functional enrichment analysis of genes from the brown, green, and green yellow modules. G–H GSEA results based on the HRA001914 dataset. I GSVA analysis results for the HRA001914 dataset

EMPD patients with low ERBB2 gene expression exhibit increased immunogenicity

We further stratified EMPD tissue samples into high- and low-expression groups based on ERBB2 gene expression levels and conducted functional enrichment analysis for both groups. WGCNA (with a soft threshold calibrated to 7, scale-free R² = 0.80) identified 14 gene modules (each containing ≥ 30 genes) (Fig. 4A–C). GO functional enrichment of the genes in the purple module, which was most correlated with the ERBB2-low-expression group, revealed enrichment in pathways such as B-cell receptor signaling, B-cell-mediated immunity, B-cell proliferation, and immunoglobulin binding. In contrast, the green and red modules, which were most strongly correlated with the high ERBB2 expression group, were enriched in pathways related to DNA and chromosomal functions, metabolism, microtubule binding, and microtubule motility (Fig. 4D–F). Moreover, the GSEA results indicated that the ERBB2-high-expression group was enriched in pathways such as the adherens junction, base excision repair, the cell cycle, and several common tumor-related regulatory signaling pathways, including the mTOR signaling pathway, the Notch signaling pathway, and the Wnt signaling pathway. In contrast to the negative regulation of innate immune response pathways observed in the ERBB2-high-expression group, the ERBB2-low-expression group was significantly enriched in immune-related pathways, including pathways related to the circulation and binding of the immunoglobulin complex secreted by plasma cells, the B-cell signaling pathway and antigen binding, which was consistent with the WGCNA findings (Fig. 4G–I).

Fig. 4.

Construction of weighted gene coexpression networks and functional analyses based on different ERBB2 gene expression levels in the HRA001914 dataset. A Calibration of the best soft-threshold power. B Construction of a weighted coexpression network and identification of gene modules. C Correlation analysis and heatmap of module-trait relationships. D–E Eigengene adjacency heatmap illustrating the correlations among gene modules and between gene modules and phenotypes (ERBB2-Low and ERBB2-High). F GO functional enrichment analysis of genes from the green, purple, and red modules. G–I GSEA results based on the HRA001914 dataset

The functional enrichment results, along with previous research findings, prompted us to further investigate the correlation between ERBB2 gene expression and the immune microenvironment of EMPD tissues. Using algorithms such as ESTIMATE, quanTIseq, and xCell, we analyzed immune microenvironment scores, immune cell infiltration, and signature gene set scores based on the gene expression profiles of EMPD tissues. The results revealed significant positive correlations between ERBB2 expression and the infiltration of cells, such as sweat gland cells, keratinocytes, and epithelial cells (Fig. 5A), whereas immune-related microenvironment scores were significantly negatively correlated with ERBB2 expression. The immunophenoscore calculated via the TCIA platform was also negatively correlated with ERBB2 expression (Fig. 5B). The TME plays a crucial role in clinical therapeutic responses, with tumor-infiltrating immune cells significantly influencing tumor progression and treatment efficacy by exerting pro- or antitumor effects. A previous multiomics study identified cytotoxic cells, NK cells, and B cells as components of the antitumor TME, whereas Treg cells, neutrophils, MDSCs, macrophages, and Th2 cells contributed to the protumor TME [27]. Our study revealed consistent results across multiple algorithms, indicating significant negative correlations between ERBB2 expression and B-cell infiltration (including B-cell lineage, pro-B cell-cell, common lymphoid progenitor (CLP), and memory B-cell infiltration) (Fig. 5C). With respect to the correlations between signature gene set scores and ERBB2 expression, the gene set scores of immune cells associated with the protumor TME, such as MDSCs, Th2 cells, and macrophages, exhibited significant positive correlations with ERBB2 expression. Signatures related to tumor progression and malignancy, such as cancer-associated fibroblasts (CAFs) and epithelial‒mesenchymal transition (EMT), were also positively correlated with ERBB2 expression (Fig. 5D–H). These results corroborated the results of the previous functional enrichment analysis, suggesting that tissues with low ERBB2 expression exhibited increased immunogenicity, particularly significantly increased B-cell immune infiltration. Additionally, ERBB2 expression was associated with the malignancy and progression of EMPD, likely because of its presence in the protumor immune microenvironment.

Fig. 5.

Comparison of tumor microenvironment features across different ERBB2 gene expression levels. A‒C Pearson analysis and scatter plots showing the correlations between ERBB2 expression and different levels of immune cell infiltration and immune microenvironment scores evaluated via the IOBR package and the TCIA platform. D–H Pearson analysis and coexpression heatmaps illustrating the correlation between ERBB2 expression and various tumor-related signature scores evaluated by the IOBR package

Discussion

Our study based on the SEER database indicated that patients who underwent surgery had significantly better survival outcomes than those who did not. However, patients who received radiotherapy and chemotherapy had markedly poorer prognoses than those who did not receive these treatments. These findings are consistent with those of previous studies suggesting that surgery is still the first-line treatment for EMPD patients. Whole lesion excision (WLE) surgery can achieve an overall survival rate of 77.2%. Furthermore, additional management of lesion margins, including comprehensive assessment and control of circumferential peripheral and deep margins, such as margin control and Mohs micrographic surgery (MMS), can further improve the survival outcomes of EMPD patients to 92.5% and 90.2%, respectively, and can also reduce the recurrence rate from 37% with WLE to less than half [28]. Because EMPD is a rare form of intraepidermal adenocarcinoma, previous studies investigating the efficacy of radiotherapy and chemotherapy for EMPD have involved only limited cases, resulting in inconsistent conclusions. A review of radiotherapy in EMPD patients summarized previous related studies, which involved 1 to 14 patients, with disease-free survival (DFS) ranging from 10 to 96 months and overall survival (OS) ranging from 10 to 70.8 months [29]. Another study involving 41 EMPD patients who received radiotherapy reported that 16 patients experienced disease progression, including local progression within the radiotherapy field, as well as lymph node and distant metastases [30]. Additionally, in a multicenter study of 76 EMPD patients, 12 patients with distant metastases received systemic chemotherapy. Despite some tumor regression, all patients developed new metastatic lesions during chemotherapy [31]. These findings suggest that, in addition to surgery as the first-line treatment, the effects of nonsurgical therapeutic approaches for EMPD are very limited, particularly for patients with metastases. This finding underscores the urgent need for more effective treatment strategies.

Although the pathogenesis of EMPD remains controversial, its IHC characteristics are similar to those of breast cancer, particularly in terms of HER2 expression [32]. HER2 amplification in EMPD appears to be a significant risk factor for this disease [29]. In a previous study, 37.1% (13 out of 35 EMPD patients) of metastatic lesions exhibited HER2 gene amplification. Recent results from a Phase II clinical study of HER2-positive advanced EMPD indicated that combined therapy with trastuzumab and docetaxel effectively prolonged the prognosis of patients with HER2-positive advanced EMPD, with no reported adverse events related to trastuzumab, such as cardiac dysfunction [32]. HER2 amplification not only serves as a risk factor for disease progression in EMPD but also provides a new therapeutic target. Because increasing evidence shows that HER2-high patients may benefit from HER2-targeted therapy, we are also considering immunotherapy as a new treatment option for patients with low HER2 expression.

Our results revealed enrichment of immune-related pathways in EMPD tissues compared with normal skin tissues. In EMPD, patients with low ERBB2 gene expression exhibited increased immunogenicity, particularly as evidenced by increased B-cell infiltration and functional enrichment related to immunoglobulin secretion by B cells and plasma cells. These findings indicate that patients with low ERBB2 expression might benefit from immunotherapy. Although recent immunotherapy studies have focused predominantly on T cells, increasing evidence highlights the role of B cells in immunotherapy. Studies have indicated that tumor-infiltrating B cells (TIL-Bs) and plasma cells may play crucial roles in tumor control and have important predictive and prognostic implications [33]. A pathology study of the lung tissues of patients with metastatic lung adenocarcinoma treated with immune checkpoint inhibitors (ICBs) revealed that B cells and tumor-infiltrating lymphocytes (TILs) are complementary predictive factors for the efficacy of ICB therapy in non-small cell lung cancer patients [34]. These therapeutic effects may be related to plasma cells recognizing and producing specific antibodies against tumor-associated antigens, secreting immunoglobulins, and mediating tumor lysis through antibody-dependent cellular cytotoxicity (ADCC), which activates immune cells such as natural killer cells or complement-dependent cytotoxicity (CDC). Our findings suggest that in EMPD tissues, B-cell signaling pathways and antibody-related pathways, as well as T-cell receptor complexes, lymphocyte-mediated immunity and chemokine receptor binding pathways are significantly enriched. We also observed enrichment of pathways related to B-cell receptor signaling, chemokine binding, phagocytosis, and antigen binding in EMPD patients with low ERBB2 expression. Previous studies have shown that B cells can capture and internalize antigens through B-cell surface receptors and present antigens to CD8 + T cells via MHC-I and CD4 + T cells via MHC-II [35]. Moreover, with the rapid development of immunotherapies in recent years, such as immune checkpoint blockade therapy, an increasing number of studies have highlighted the important role of tertiary lymphoid structures (TLSs), which are ectopic aggregates of immune cells, in cancer immunity. T cells and B cells are key components of TLSs [35]. Research has shown that dysfunctional CD8 + T cells can secrete the chemokine CXCL13, inducing B-cell migration into tumor tissue. TLSs can coordinate the joint response of T cells and B cells to enhance antitumor immunity, suggesting that B cells are potential targets for immunotherapy [33, 36]. Based on our results and previous research, we speculate that B cells in the immune microenvironment of EMPD tissue may promote and regulate T-cell activation, proliferation, and formation, which contributes to an antitumor response and affects the efficacy of immunotherapy [37]. The antibodies secreted by plasma cells have been widely used in scientific research and clinical translation because of their ability to bind to specific antigens. Many antibodies (targeting PD-1, CD20, etc.) have demonstrated utility in cancer therapy [38]. Heemskerk et al. introduced an IgG1 bispecific antibody based on the structural characteristics of plasma cell-secreted antibodies. These results suggest that this antibody can increase ADCC and antibody-dependent cell phagocytosis (ADCP) in melanoma mice by recruiting neutrophils, natural killer cells, and macrophages to kill tumor cells [39]. A study based on the analysis of the immune microenvironment of head and neck squamous cell carcinoma (HNSCC) patients revealed differences in B-cell types among different subtypes of HNSCC. These findings suggest that promoting the expression of SEMA4A in B cells may help TIL-Bs develop into germinal center (GC) TIL-Bs and TLSs. The presence of GC TIL-Bs in TLSs is associated with a better clinical prognosis, and this finding may also be a follow-up direction for B-cell-related immunotherapy in patients with tumors such as lung cancer and melanoma [40]. In addition, several studies have shown that immune checkpoint receptors, such as PD-1 and PD-L1, are expressed on the surface of TIL-Bs in various tumors, such as hepatocellular carcinoma and thyroid cancer, suggesting that immune checkpoint inhibitors can also affect TIL-Bs [41, 42]. A large-scale study involving 608 cases of soft tissue sarcoma revealed that high infiltration of B cells is associated with a better survival prognosis and a better response to PD-1 therapy in patients [43]. Among melanoma patients receiving ICB neoadjuvant chemotherapy, B-cell-related genes are among the genes associated with the greatest differences between ICB responders and nonresponders [44]. Moreover, a study based on a mouse model of breast cancer also demonstrated that adoptive transfer of B effector cells can inhibit the development of spontaneous metastatic tumors, including the activation of host T-cell antitumor immunity [45]. Together with the results of our study, these findings suggest that immunotherapy may be useful in EMPD patients with low ERBB2 expression because it may leverage the tumoricidal functions of B cells and their synergistic interaction with T cells. In recent years, studies have indicated that the sequential order of targeted therapy and immunotherapy in tumor treatment may affect the survival prognosis of certain cancer patients. As noted in the DREAMseq trial and SECOMBIT trial, sequential treatment starting with immune checkpoint inhibitors rather than BRAF-targeted drugs significantly improved survival outcomes for patients with BRAF-mutated metastatic melanoma [46, 47]. This finding prompted us to consider the timing of immunotherapy initiation in EMPD patients with low ERBB2 expression. Although the number of existing cases is not sufficient to conclusively determine differences in the efficacy of sequential treatment and traditional therapy, such as surgery, chemotherapy or radiotherapy, and immunotherapy, the findings of this study can guide the design and clinical practice of EMPD immunotherapy-related clinical trials in the future.

However, our study has many limitations. First, our research data were drawn from the large-scale cancer-related epidemiological database SEER and the public GSA omics database, and research validation of clinical cases at our center is lacking. The understanding of the enrichment of B-cell infiltration in EMPD tissues with low ERBB2 expression is limited, and experimental or clinical sample verification of the relationship between B-cell infiltration levels and ERBB2 expression in EMPD tissues is lacking. Moreover, specific molecular biology research on the biological and immunological roles of B-cell infiltration in the immune microenvironment of EMPD tissues, as well as its interaction mechanisms with other immune cells, is lacking. These findings provide a theoretical basis for future research on the immune microenvironment of EMPD tissues and suggest directions for future experimental and clinical studies. Second, our study indicates that EMPD tissues with low ERBB2 expression exhibit stronger immunogenicity and higher levels of B-cell infiltration, suggesting that patients with low ERBB2 expression may benefit from immunotherapy. However, the dataset we utilized in this study lacked relevant clinical data on the treatment and prognosis of patients, limiting our ability to conduct subgroup analysis to evaluate the differences in efficacy between different types of immunotherapy and the ideal sequence of traditional treatment methods, such as surgery, chemotherapy, and immunotherapy, in sequential clinical treatment. Moreover, we still lack relevant experimental validation or clinical case support, which requires further EMPD samples to be collected in the future to verify the expression levels of ERBB2, B-cell markers, and some immune checkpoints. This finding also suggests that more attention should be given to the possibility of applying immunotherapy in EMPD patients with low ERBB2 expression in future clinical practice, and relevant cases should be collected to further improve the conclusions of this study.

In summary, this study identified advanced age at diagnosis, distant metastases, and radiotherapy treatment as independent risk factors for EMPD patients based on the analysis of epidemiological and clinical factors from 815 EMPD patients extracted from the SEER database. Notably, patients who received radiotherapy and chemotherapy had a worse prognosis than those who did not, suggesting the urgent need for more effective clinical treatment strategies. Further bioinformatic analysis revealed significant enrichment of immune-related pathways in EMPD tissues compared with normal skin tissues. Surprisingly, we observed increased immunogenicity in EMPD tissues with low ERBB2 expression, and notable enrichment of B cells, plasma cells, and related antibody pathways was detected. These findings provide a theoretical basis for the use of immunotherapy, particularly B-cell-related immunotherapy, in patients with EMPD with low ERBB2 expression.

Author contributions

All authors contributed to the study conception and design. Ning Zhu and Ying Yuan contributed to the study conception and design. Jiawen Yang and Yurong Chen analyzed the data. Xiuyuan Zhang, Ziyan Tong and Shanshan Weng helped data interpretation and discussion. The first draft of the manuscript was written by Jiawen Yang and Yurong Chen. Ying Yuan and Ning Zhu approved the final draft. All authors read and approved the submitted version.

Funding

This work was supported by the National Natural Science Foundation of China (Grant No. 82373415), Zhejiang Provincial Clinical Research Center for CANCER (2022E50008, 2024ZY01056) and Provincial Key R&D Program of Zhejiang Province (2021C03125).

Data availability

No datasets were generated or analysed during the current study.

Declarations

Competing interests

The authors declare no competing interests.

Ethical approval

No ethical approval is required.