Highly CD81-expressing stromal cells characterize a tumor-suppressive immune microenvironment in invasive breast cancer

Abstract

Background

CD81, a member of the tetraspanin family, is implicated in tumor progression, and its elevated expression in tumor cells has been associated with poor prognosis across multiple cancers. However, the functional significance of CD81 within the stromal compartments of the tumor microenvironment (TME) remains unclear. This study investigated the clinical relevance and mechanistic role of stromal CD81 (sCD81) in invasive breast cancer.

Methods

A total of 462 human primary invasive breast cancer samples, along with 13 normal breast tissues and 10 ductal carcinoma in situ samples collected between 2000 and 2013, were analyzed. Immunohistochemistry and spatial analyses were used to evaluate sCD81 expression, and its association with cancer-associated fibroblast (CAF) subsets and immune components in the TME, as well as correlation with clinicopathological correlations and survival outcome.

Results

sCD81 expression was significantly higher in invasive breast cancer than in normal breast stromal tissue but did not differ across clinical subtypes. High sCD81 expression was associated with significantly longer disease-free survival, and multivariate analysis identified sCD81 as an independent prognostic factor alongside nuclear grade, estrogen receptor status, and lymph node metastasis. Co-expression analyses demonstrated that sCD81-high stromal cells exhibited a CD34-positive, FAP-low, and PDGFRα-low phenotype, indicating that they are distinct from established CAF subtypes. Both sCD81 and podoplanin (PDPN) expression correlated with increased immune cell infiltration. However, PDPN-positive CAFs promoted an immunosuppressive TME characterized by enrichment of regulatory T cells and M2-like macrophages, whereas CD81-high stromal cells did not exert such effects. Spatial analyses revealed that CD81-high stromal cells were preferentially localized near tumor vasculature and CCL19-expressing cells, suggesting a potential association with T-cell recruitment into the TME through CCL19-mediated mechanisms.

Conclusions

These findings identify stromal CD81 expression as a novel prognostic marker in invasive breast cancer and highlight its association with a tumor-suppressive immune microenvironment, distinct from the immunosuppressive phenotype typically associated with PDPN-positive CAFs.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13058-026-02326-4.

Background

Breast cancer is the most common malignancy among women worldwide. Although mortality rates have gradually declined due to advances in early detection and systemic therapy, incidence continues to rise and breast cancer remains responsible for more than 600,000 deaths annually in the world [1]. To further improve patient outcomes, the development of combination strategies incorporating standardized immune checkpoint blockade or novel therapeutics based on newly identified mechanisms of action is urgently needed.

The tumor microenvironment (TME) is a highly complex and dynamic system composed of malignant cells, extracellular matrix (ECM), cancer-associated fibroblasts (CAFs), vascular structures, and a variety of immune cell populations, all of which collectively shape tumor progression and therapeutic responses [2]. Among these components, CAFs represent a particularly heterogeneous population of stromal cells that can be broadly categorized into tumor-promoting and tumor-suppressive subsets, based on their transcriptional states, secretory profiles, and interactions with tumor, immune and vascular compartments [3–6]. Tumor-promoting CAFs facilitate immune evasion, angiogenesis, and ECM remodeling, whereas tumor-suppressive CAFs can constrain tumor proliferation and enhance antitumor immune activity.

CD81, a member of the tetraspanin family, is a ubiquitously expressed membrane protein that regulates cell adhesion, motility, proliferation, and signal transduction [7, 8]. Elevated CD81 expression in tumor cells has been associated with poor prognosis across multiple cancer types, including breast cancer, melanoma, lymphoma, and prostate cancer [9–13]. Despite this evidence, the functional significance of CD81 in stromal compartments of the TME remains poorly defined. Of note, stromal expression of CD9, another tetraspanin family member, has been linked to favorable prognosis in melanoma [14], suggesting that stromal CD81 may also exert context-dependent effects on tumor–stroma interactions and clinical outcomes.

Cytotoxic T lymphocytes (CTLs) are central effectors of antitumor immunity, and their efficient recruitment into tumor tissues is a prerequisite for effective immune surveillance. Chemokines such as C-C motif chemokine ligand 19 (CCL19) and CCL21 have been reported as key regulators of lymphocyte homing, and their expression within TME promotes effector T-cell infiltration and antitumor immune responses [15]. In studies of non-small cell lung carcinoma (NSCLC), CCL19-producing cells arising within the TME have been shown to originate from two distinct cellular lineages, perivascular reticular cells and T-cell zone reticular cells [16, 17]. Furthermore, high endothelial venules (HEVs), which are specialized postcapillary venules, mediate lymphocyte extravasation and trafficking [18–20]. Ectopic HEVs have been identified within tumors which facilitate organized aggregates of immune cells that enable local antigen presentation and immune activation [21–24].

Given these observations, the present study investigated the clinical relevance of CD81 expression in stromal compartments of invasive breast cancer and its mechanistic contribution to shaping the tumor immune microenvironment. Specifically, we evaluated the spatial relationships among CD81-high stromal cells and CCL19-expressing stromal cells, and examined their association with patient prognosis. Through these analyses, we aimed to determine whether stromal CD81 represents a novel determinant of intratumoral immune contexture and a potential stratification marker for patient-specific therapeutic strategies.

Methods

Patients and samples

A total of 462 primary invasive breast cancer tissue samples were collected along with13 normal breast tissues and 10 ductal carcinoma in situ (DCIS) samples at Nagoya City University Hospital between 2000 and 2013. A single 2.0-mm core of tumor tissue from each formalin-fixed, paraffin-embedded specimen was incorporated into multitissue blocks containing up to 48 cores. Disease-free survival (DFS) was defined as the interval from the date of primary surgery to the earliest occurrence of one of the following events: locoregional recurrence, distant organ metastasis, or death from any cause. The median follow-up duration was 143 months (maximum, 243 months).

Immunohistochemistry

The antibodies used in this study are listed in Table S1. Immunohistochemistry was performed using a Bond-Max automated stainer (Leica Biosystems, Nussloch, Germany). Immunoreactivity was visualized with 3,3′-diaminobenzidine. Tumor CD81 expression (tCD81) was evaluated using the membrane intensity algorithm, with modifications, implemented in the Aperio CS2 digital pathology system (Leica Microsystems, Wetzlar, Germany), and scored as 0, 1+, 2+, or 3+. The H-score for each tissue sample was calculated as follows:

Stromal CD81 (sCD81) expression for each sample was quantified using the positive area measurement algorithm in the Aperio CS2. The positive areas for podoplanin (PDPN), alpha-smooth muscle actin (ACTA2), CD34, vimentin (VIM), CD31, CD146, MECA-79, and CCL19 were measured using ImageJ/FIJI (version 2.16.0/1.54p; National Institutes of Health, Bethesda, MD, USA). The densities of sC81-, PDPN-, ACTA2-, CD3-, CD8-, FoxP3-, CD68-, CD163-, CD11b-, and PD-1-positive cells, as well as the Ki-67 labeling index and cyclin A (CCNA) positivity rate, were determined using the positive cell count algorithm in the Aperio CS2. Phosphohistone H3 (PHH3)-positive cells were counted in four high-power fields (HPFs), and the mean value was used for statistical analysis.

Fluorescent immunohistochemistry

Antigen retrieval was performed using Histofine deparaffinization and antigen retrieval buffer pH 9 (Nichirei Biosciences, Tokyo, Japan) according to the manufacturer’s instructions. After blocking, primary antibodies were applied at room temperature (RT) for 1 h. The dilutions of the primary antibodies are summarized in Table S1. Signals were visualized using secondary antibodies labelled with fluorescein, tetramethylrhodamine, or Alexa Fluor™ 647 applied at a dilution of 1:100 (Molecular Probes^®, Thermo Fisher Scientific K. K., Tokyo, Japan). Fluorescent images were acquired using an FV3000 confocal laser scanning microscope (Evident corporation, Tokyo, Japan).

Hierarchical clustering analysis

Hierarchical clustering analysis was conducted using the ClustVis web tool [25]. Dissimilarities were calculated based on Euclidean distances of the log-transformed cell density values, and clustering was conducted using Ward’s method (unsquared distances).

Statistical analyses

Statistical analyses were performed using EZR software version 1.66 [26]. Cutoff values for immunohistochemical and immune cell parameters were determined as follows. The cutoff for sCD81 expression was defined as the point maximizing the Youden index on the receiver operating characteristic curve for 10-year patient survival (area under the curve = 0.571). The same method was applied to determine cutoff values for the densities of CD3-positive cells (458 cells/mm²), CD8-positive cells (197 cells/mm²), and the regulatory T cell (Treg)/CD8-positive cell ratio (0.063). Cutoff values for tCD81, PDPN, cellular density of pan-macrophages (563 cells/mm²), and the M2/pan-macrophage ratio (0.252) were set at their respective medians. Associations between categorical variables were assessed using the chi-squared test. For continuous variables, an unpaired two-tailed Welch’s t-test or Welch’s one-way ANOVA followed by Bonferroni’s post hoc test was applied. A simple Bonferroni correction for multiple hypothesis testing was used, with a two-sided alpha level of 0.005 (= 0.05/10) (Table 2). For survival analyses, Kaplan–Meier survival curves were generated and compared using the log-rank test for DFS. A Multivariate analysis using Cox proportional hazards model was performed to examine the association between DFS and clinicopathological factors, including nuclear grade (NG1/2 versus NG3), estrogen receptor status (negative versus positive), lymph node metastasis (negative versus positive), sCD81 expression (low versus high), and tCD81 expression (low versus high). Co-expression analyses of CD81 or PDPN with various markers in stromal cells were performed as follows: Pearson’s correlation coefficients were calculated for individual cells using ImageJ/FIJI. For each patient, 20 cells were analyzed from three independent tissue specimens (total n = 60 cells).

Table 2.

Clinicopathological characteristics of patients with invasive breast cancer according to stromal CD81 expression

	All patients		sCD81 expression				p value
			Low		High
Patient No.	370	(100%)	222	(60.0%)	148	(40.0%)
Age	57.1	±13.4	57.1	±12.9	57.2	±12.5	0.96a
Ki-67 labeling index	18.1	±15.9	14.0	±12.9	23.3	±16.4	<0.0005 a
Nuclear grade							0.37b
NG1,2	201	(54.3%)	123	(55.4%)	78	(52.7%)
NG3	154	(41.6%)	86	(38.7%)	68	(45.9%)
Unknown	15	(4.1%)	13	(5.9%)	2	(1.4%)
Histology							0.18b
IDC	313	(84.6%)	180	(81.1%)	133	(89.9%)
ILC	19	(5.1%)	12	(5.4%)	7	(4.7%)
Others	28	(7.6%)	21	(9.5%)	7	(4.7%)
Unknown	10	(2.7%)	9	(4.1%)	1	(0.7%)
Tumor size							<0.0005 b
≦2cm	184	(49.7%)	90	(40.5%)	94	(63.5%)
＞2cm	171	(46.2%)	119	(53.6%)	52	(35.1%)
Unknown	15	(4.1%)	13	(5.9%)	2	(1.4%)
ER expression							0.69b
Negative	49	(13.2%)	27	(12.2%)	22	(14.9%)
Positive	305	(82.4%)	181	(81.5%)	124	(83.8%)
Unknown	16	(4.3%)	14	(6.3%)	2	(1.4%)
PgR expression							0.059b
Negative	94	(25.4%)	47	(21.2%)	47	(31.8%)
Positive	260	(70.3%)	161	(72.5%)	99	(66.9%)
Unknown	16	(4.3%)	14	(6.3%)	2	(1.4%)
HER2 expression							0.95b
Negative	318	(85.9%)	187	(84.2%)	131	(88.5%)
Positive	37	(10.0%)	21	(9.5%)	16	(10.8%)
Unknown	15	(4.1%)	14	(6.3%)	1	(0.7%)
Lymph node metastasis							>0.99b
Negative	220	(59.5%)	131	(59.0%)	89	(60.1%)
Positive	143	(38.6%)	85	(38.3%)	58	(39.2%)
Unknown	7	(1.9%)	6	(2.7%)	1	(0.7%)
Distant organ metastasis							0.075b
Negative	302	(81.6%)	171	(77.0%)	131	(88.5%)
Positive	46	(12.4%)	33	(14.9%)	13	(8.8%)
Unknown	22	(5.9%)	18	(8.1%)	4	(2.7%)

^aWelch’s t-test. ^bChi-squared test. The Bonferroni-corrected threshold for statistical significance was 0.005 (0.05/10). sCD81: stromal CD81; NG: Nuclear grade; IDC: Invasive ductal carcinoma; ILC: Invasive lobular carcinoma; ER: Estrogen receptor; PgR: Progesterone receptor; HER2: Human epidermal growth factor receptor 2

P values with statistically significance are presented in bold

Results

High stromal CD81 expression is associated with favorable prognosis

Representative CD81 expression patterns observed in normal breast tissue and invasive breast cancer are shown in Fig. 1A. In normal breast tissue, CD81 expression was weakly detected in the stromal compartment, with no detectable expression in terminal ductal lobular units. In case 1 of invasive breast cancer, both tumor cells and stromal cells showed low CD81 expression. In case 2, stromal cells showed strong CD81 expression, whereas tumor cells showed minimal expression. In case 3, the opposite pattern was noted, with strong expression in tumor cells and weak expression in stromal cells. In case 4, high expression of CD81 was observed in both tumor cells and stromal cells.

Fig. 1.

CD81 expression in stromal and tumor cells in invasive breast cancer tissues. A Representative tissue images of normal and invasive breast cancer stained with H&E (upper panels) or anti-CD81 antibody. Scale bar = 100 μm. B Average expression of sCD81 in normal, noninvasive and invasive breast cancer tissues. C, D Average sCD81 or tCD81 expression in HR⁺/HER2⁻ (hormone receptor-positive and HER2-positive), HER2⁺ and triple-negative subtypes. E Correlation of CD81 expression between tumor and stromal cells. F, G Kaplan-Meier survival curves stratified by CD81 expression in stromal or tumor cells. H Survival curves classified according to sCD81 and tCD81 expressions. s: stromal cells; t: tumor cells; ns: not significant. *p < 0.05

Quantitative analysis revealed that stromal CD81 expression was significantly higher in invasive breast cancer than in normal breast tissue (Fig. 1B), suggesting an increased abundance of CD81-expressing stromal cells in invasive disease. CD81 expression in both stromal and tumor compartments did not significantly differ among clinical subtypes of invasive breast cancer (Fig. 1C, D).

A strong positive correlation was noted between tumor and stromal CD81 expression levels (r = 0.72, p < 0.0005; Fig. 1E). Kaplan–Meier analysis of DFS based on stromal CD81 (sCD81) expression revealed significantly longer DFS in patients with high stromal expression compared with those with low stromal expression (Fig. 1F). In contrast, tumor CD81 (tCD81) expression showed no significant association with DFS (Fig. 1G). These findings suggest that stromal CD81 expression may be linked to tumor CD81 expression, but that their biological roles in tumor and stromal cells in disease progression differ.

Further stratification into four groups based on combination of stromal and tumor CD81 expression revealed that the stroma-low/tumor-high (Low/High) group tended to have poorer survival compared with the stroma-low/tumor-low (Low/Low) group, whereas the stroma-high/tumor-high (High/High) group showed significantly better prognosis than the Low/High group (Fig. 1H). Collectively, these findings suggest that high CD81 expression in tumor cells may be associated with adverse prognosis, while high stromal CD81 expression is linked to improved clinical outcomes.

Univariate Cox proportional hazard regression analyses identified nuclear grade (NG) and lymph node metastasis as significant predictors of shorter DFS, whereas estrogen receptor (ER) positivity and high sCD81 expression were associated with prolonged DFS (Table 1). In multivariate analysis including NG, ER status, lymph node metastasis, sCD81 expression, and tCD81 expression, high sCD81 expression emerged as an independent favorable prognostic factor. In contrast, high tCD81 expression revealed a nonsignificant trend toward worse prognosis (HR = 1.26, p = 0.43).

Table 1.

Univariate and multivariate analyses using Cox proportional hazards model for patients with invasive breast cancer

	Univariate				Multivariate
	Hazard	95%CI		p value	Hazard ratio	95%CI		p value
	ratio	min	max			min	max
Nuclear grade
NG1,2 (n=223)	1 (ref.)				1 (ref.)
NG3 (n=189)	1.76	1.18	2.62	0.0060*	1.94	1.21	3.10	0.0057*
Tumor size
≦2cm (n=218)	1 (ref.)
>2cm (n=194)	1.45	0.96	2.20	0.078
ER expression
Negative (n=69)	1 (ref.)				1 (ref.)
Positive (n=343)	0.60	0.37	0.97	0.035*	0.53	0.29	0.96	0.035*
HER2 expression
Negative (n=371)	1 (ref.)
Positive (n=42)	1.25	0.66	2.34	0.50
Lymph node metastasis
Negative (n=261)	1 (ref.)				1 (ref.)
Positive (n=160)	2.58	1.73	3.83	<0.0005*	2.50	1.56	3.98	<0.0005*
sCD81 expression
Low (n=222)	1 (ref.)				1 (ref.)
High (n=148)	0.57	0.36	0.92	0.021*	0.40	0.22	0.75	0.0038*
tCD81 expression
Low (n=205)	1 (ref.)				1 (ref.)
High (n=205)	0.89	0.60	1.33	0.58	1.26	0.71	2.22	0.43

ER: Estrogen receptor; HER2: Human epidermal growth factor receptor 2; sCD81: Stromal CD81; tCD81: Tumor CD81. *p < 0.05 were considered statistically significant

P values with statistically significance are presented in bold

Clinicopathological comparisons between high and low sCD81 expression groups revealed that tumors in the high sCD81 group were significantly smaller in size, despite exhibiting higher Ki-67 proliferation indices (Table 2). Other cellular proliferation markers also showed similar trends (Fig. S1). This apparent discrepancy will be further addressed in the Discussion section.

CD81-high stromal cells exhibit a CD34-positive, PDGFRα-low, and FAP-low phenotype

The potential associations of sCD81 with PDPN and ACTA2, which are representative markers of CAFs, were evaluated (Fig. 2A). A weak positive correlation was noted between sCD81 and PDPN (r = 0.31, p < 0.0005), whereas no significant correlation was found between sCD81 and ACTA2 (r = 0.096, p = 0.071) (Fig. 2B). The expression levels of PDPN and ACTA2 observed across breast cancer clinical subtypes are shown in Fig. S2A. Three representative cases illustrating the proportions of CD81-high, PDPN-positive, and ACTA2-positive stromal cells were presented in Fig. S2B. These observations indicate that sCD81-high stromal cells may represent a distinct stromal population within the TME in a subset of cases.

Fig. 2.

Co-expression of CD81 with various markers. A Representative immunohistochemical staining for CD81, PDPN, and ACTA2 in invasive breast cancer tissue. Scale bar = 150 μm. B Scatterplot matrix showing the relationships among sCD81, PDPN, and ACTA2 expression levels. C Fluorescent immunostaining of CD81 and various markers. Scale bars: 100 μm (left panels), 20 μm (right panels). D Average Pearson’s correlation coefficients between CD81 and each marker. *p < 0.05; **p < 0.005; ***p < 0.0005

To investigate the cellular origin of CD81-high cells, co-expression analyses were performed between CD81 and multiple lineage- and stromal-related markers (Fig. 2C). VIM exhibited a strong positive correlation with CD81 (R = 0.69) (Fig. 2D), suggesting that CD81-high stromal cells are predominantly associated with the mesenchymal stromal compartment. In contrast, ACTA2 showed a negligible correlation with CD81 (R = 0.03). CD146, immunoglobulin superfamily containing leucine-rich repeat (ISLR/Meflin), platelet-derived growth factor receptor alpha (PDGFRα), CD31, and fibroblast activation protein (FAP) demonstrated only weak positive correlations with CD81 (R = 0.11, 0.14, 0.18, 0.20 and 0.21, respectively). PDPN and CD34 exhibited relatively strong positive correlations with CD81 (R = 0.51 and 0.57, respectively), whereas Decorin (DCN) showed a relatively weaker positive correlation (R = 0.47).

Additional co-expression analyses between PDPN and a selected subset of markers were conducted to further compare CD81-high stromal cells with PDPN-positive CAFs (Fig. S2C). FAP and PDGFRα exhibited relatively strong positive correlations with PDPN (R = 0.51 and 0.43, respectively), whereas CD34 showed only a very weak positive correlation with PDPN (R = 0.10) (Fig. S2D). Moderate positive correlations were observed between PDPN and DCN and between PDPN and VIM (R = 0.38 and 0.34, respectively). ACTA2 showed relatively weak positive correlation with PDPN (R = 0.24).

Collectively, these results suggest that CD81-high stromal cells are distinct from PDPN-positive CAFs associated with PDGFRα- and FAP-expressing activated fibroblasts. Instead, CD81-high stromal cells appear to represent a VIM- and CD34-positive stromal population with low PDGFRα and FAP expression. Consistently, sCD81 expression showed weak positive correlations with both CD34 and VIM (Fig. S2E).

Stromal CD81 and podoplanin expression are associated with immune cell infiltration into the TME

To elucidate the potential role of CD81-high stromal cells in the TME, we investigated their association with immune cell infiltration. The infiltration of T cells, including CD8-posittive cells and Tregs, as well as macrophages, including M2-like macrophages, and tumor-associated myeloid cells (TAMCs), into the TME was quantitatively assessed (Fig. 3A). Based on these immune cell density profiles, TMEs were stratified into three distinct clusters using hierarchical clustering analysis (Fig. 3B). Cluster 1 was characterized by a marked paucity of immune cells within the TME, corresponding to a so-called “cold” tumor phenotype [27]. In contrast, Cluster 3 showed abundant immune cell infiltration, whereas Cluster 2 displayed intermediate immune cell density (Fig. 3C). Notably, Cluster 3 contained a significantly higher density of CD8-positive cells compared with the other clusters, whereas no statistically significant differences were noted in Treg densities (Fig. 3D, left and middle). The ratio of M2-like macrophages increased progressively with overall immune cell density across clusters (Fig. 3D, right). In addition, the frequency of PD-1–expressing cells was higher in Cluster 3 compared with the other clusters (Fig. 3E), suggesting that patients within this subgroup may be more responsive to immune checkpoint blockade.

Fig. 3.

Immune cell infiltration and stromal comopartments in the TME. A Representative immunohistochemical staining of CD3, CD8, FoxP3, CD68, CD163, and CD11b. Scale bar = 100 μm. B Hierarchical clustering and heatmap based on immune cell densities. C Comparison of average immune cell densities among the three identified clusters. D Ratios of CD8-positive cells to CD3-positive cells, Tregs to CD8-positive cells, and M2- to Pan-macrophages across the three clusters. E Comparison of PD-1 expression among the three clusters. F Representative immunohistochemical images of CD8, CD81, PDPN, and ACTA2 for each cluster. Scale bar = 2 mm. G Average expression levels of sCD81, PDPN, and ACTA2 across the three clusters. *p < 0.05; **p < 0.005; ***p < 0.0005

Representative images of CD81, PDPN, and ACTA2 expression patterns across clusters, together with CD8-positive cells, are shown in Fig. 3F. Both sCD81 and PDPN expression levels increased in parallel with immune cell infiltration, whereas ACTA2 showed no such association (Fig. 3G). These findings suggest that both sCD81 and PDPN expression are positively associated with immune cell infiltration into the TME. Comparison of immune cell densities among clinical subtypes of breast cancer revealed that all immune cell types were significantly more abundant in the triple-negative subtype compared with the HR⁺/HER2⁻ subtype (Fig. S3A). Indeed, the proportion of triple-negative cases was significantly higher in Cluster 3 (Fig. S3B). To exclude potential bias introduced by subtype distribution, the same analyses were repeated in HR⁺/HER2⁻ subtype cases alone, yielding largely consistent results (Fig. S3C–F).

Differential effects of CD81-high stromal cells and PDPN-positive CAFs on immune cell infiltration in the TME

Hierarchical clustering analysis suggested that both CD81-high stromal cells and PDPN-positive CAFs are potentially involved in immune cell infiltration within the TME. To further clarify their distinct roles, we evaluated the effects of CD81-high and PDPN-positive cells on specific immune cell populations—particularly T cells and macrophages—and their associations with patient prognosis. Kaplan–Meier survival analysis stratified by PDPN expression revealed no significant difference in DFS between subgroups (Fig. 4A). In both PDPN-low and PDPN-high subgroups, patients with high sCD81 expression tended to exhibit a more favorable prognosis, although this trend did not reach statistical significance in the PDPN-high subgroup (Fig. 4B, C).

Fig. 4.

Comparison of immune cell recruitment profiles between sCD81 and PDPN expression levels. A Kaplan-Meier survival curves stratified by PDPN expression levels. B, C Kaplan-Meier survival curves stratified by sCD81 expression levels within the PDPN-low (B) and PDPN-high (C) subgroups. D, E Comparison of immune cell densities between groups defined by sCD81 and PDPN expression levels. *p < 0.05; **p < 0.005; ***p < 0.0005

Both CD3-positive cells and CD8-positive cells exhibited significantly higher densities in the PDPN-high group, accompanied by an increased Treg/CD8-posive cell ratio (Fig. 4D). Furthermore, in both PDPN-low and PDPN-high subgroups, high sCD81 expression was associated with increased densities of CD3-positive cells and CD8-positive cells, whereas the Treg/CD8-positive cell ratio remained unchanged. With respect to macrophages, both high sCD81 and PDPN expression were associated with higher densities of pan-macrophages and M2-like macrophages; however, only high PDPN expression contributed to an increased M2-like/pan-macrophage ratio (Fig. 4E).

In our cohort, higher densities of CD3-positive cells and CD8-positive cells were each associated with favorable prognosis (Fig. S4A, B), whereas a higher Treg/CD8-positive cell ratio correlated with poorer patient outcomes (Fig. S4C). Although no significant prognostic differences were observed with respect to macrophage densities (Fig. S4D, E), these findings collectively suggest that both CD81-high stromal cells and PDPN-positive CAFs facilitate immune cell infiltration into the TME. However, only PDPN-positive CAFs appear to promote the development of an immunosuppressive microenvironment enriched in Tregs and M2-like macrophages, whereas CD81-high stromal cells do not exert such immunosuppressive effects.

CD81-high stromal cells are associated with tumor vasculature and CCL19-expressing cells

To investigate the mechanisms by which CD81-high stromal cells may promote immune cell infiltration into the TME, we analyzed tumor-associated vasculature, including CCL19-expressing cells and HEVs. Endothelial cells, pericytes, and CCL19-expressing cells were quantified by immunostaining for CD31, CD146, and CCL19, respectively, along with HEVs (Fig. 5A and Fig. S5B). A positive correlation was observed between endothelial cells (CD31) and pericytes (CD146), whereas neither HEVs (MECA-79) nor CCL19-expressing cells showed significant associations with CD31 or CD146 (Fig. S5A).

Fig. 5.

Association of sCD81 and PDPN expression levels with tumor-associated vasculature. A Representative images of H&E staining and immunohistochemical staining for CD31, CD146, and CCL19. Scale bar = 200 μm. B–D Scatter plots showing the correlations between CD8-positive cell density and CD31 (B), CD146 (C) or CCL19 (D). E–G Comparison of average expression levels of each vessel-related marker between groups defined by sCD81 and PDPN expression levels

Among these stromal components, both CCL19-expressing cells and HEVs exhibited significant positive correlations with CD8-positive cell density (Fig. 5D and Fig. S5C), while endothelial cells and pericytes did not (Fig. 5B, C). Regarding two cell types, only CD81-high stromal cells were specifically associated with tumor vasculature (Fig. 5E, F), whereas both CD81-high and PDPN-positive cells showed significant associations with CCL19-expressing cells (Fig. 5G). In contrast, only PDPN-positive CAFs correlated with HEVs (Fig. S5D). Collectively, these findings indicate that CD81-high stromal cells are linked to tumor vasculature and CCL19-expressing cells, whereas PDPN-positive CAFs are associated with both HEVs and CCL19-expressing cells.

CD81-high stromal cells surround perivascular cells producing CCL19

We next examined the spatial distribution of CD81-high stromal cells, PDPN-positive CAFs, HEVs, and CCL19-expressing cells within the TME. Immunofluorescence analyses revealed vessel-like structures in which CD81-high stromal cells were preferentially localized in close proximity to, and often surrounding, CCL19-expressing cells; however, direct spatial overlap between CD81 and CCL19 signals was not observed (Fig. 6A). Within these vessel-like structures, CD31 and CD146 were co-expressed with CCL19 and CD81, supporting the presence of vascular structures encircled by both CCL19-expressing cells and CD81-high stromal cells (Fig. 6B). Notably, these vascular structures lacked PDPN expression.

Fig. 6.

The spatial distribution of CD81-high and PDPN-positive CAFs as well as CCL19-expressing cells. A CD81-high CAFs surround CCL19-expressing cells that construct vessel-like structures. Scale bars: 100 μm (left panel), and 20 μm (right panel). B CCL19-expressing cells surrounded by CD81-high stromal cells are associated with the tumor vasculature. Scale bar = 50 μm. C Non-vascular CCL19-expressing cells that also express PDPN. Scale bars: 200 μm (low power field) and 30 μm (high power field)

In addition, we identified CCL19-positive, pavement-like structures that were not associated with the vascular marker CD31, although weak CD146 expression was detected (Fig. 6C). These non-vascular CCL19-positive cells also expressed PDPN and exhibited minimal spatial colocalization with CD81-high stromal cells.

Collectively, these findings indicate that CCL19-expressing cells can be divided into two distinct subsets: (i) a vascular-associated, PDPN-negative population surrounded by CD81-high stromal cells, and (ii) a non-vascular, PDPN-positive population lacking CD81 association. Consistently, HEVs expressing MECA-79 also expressed PDPN but not CD81, further indicating that HEVs are not closely associated with CD81-high stromal cells (Fig. S6).

Discussion

This study identifies a previously unrecognized subset of stromal cells characterized by high CD81 expression within the TME of invasive breast cancer. Importantly, the presence of CD81-high stromal cells was associated with favorable prognosis, potentially through enhanced immune cell infiltration, particularly of CD8-positive effector cells including CTLs. Spatial analyses demonstrated that CD81-high stromal cells were preferentially localized in proximity to tumor vasculature and CCL19-expressing cells, implicating this stromal compartment in the regulation of T-cell recruitment into the TME, possibly via CCL19-mediated mechanisms.

The classification of CD81-high stromal cells as a mesenchymal compartment has been supported by their spindle-shaped morphology and expression of VIM [28–33]. Although VIM is widely expressed in CAF populations, it is not CAF-specific and does not distinguish among established CAF subtypes, including myofibroblastic CAFs (myCAFs), inflammatory CAFs (iCAFs), matrix CAFs (mCAFs), antigen-presenting CAFs (apCAFs), vascular CAFs (vCAFs) and other CAF subtypes.

vCAFs are typically localized adjacent to blood vessels and are functionally associated with angiogenesis [4, 28, 29, 33, 34]. Recent single-cell transcriptomic and proteomic analyses in breast cancer have identified CD146 as a representative marker of vCAFs [35, 36]. In our study, co-expression of CD81 and CD146 was uncommon, suggesting that CD81-high stromal cells are unlikely to correspond to vCAFs.

The negligible expression of ACTA2 further argues against a myCAF phenotype, as myCAFs are characterized by high ACTA2 expression [37]. iCAFs are generally characterized by enrichment of PDGFRα and FAP expression together with inflammatory gene signatures. While CD34 has been reported among genes enriched in iCAF clusters in single-cell analyses, the low expression of PDGFRα and FAP in CD81-high stromal cells suggests that they are phenotypically distinct from typical iCAFs [36]. mCAFs are commonly associated with extracellular matrix production and may express matrix-related genes such as DCN [38]. The relatively weak co-expression of DCN in CD81-high stromal cells may indicate partial overlap with mCAF-like populations. However, the low expression of FAP suggests that these cells are phenotypically distinct from conventional activated CAF subsets, including mCAFs, apCAFs, and other FAP-enriched CAF populations [36]. Nevertheless, the possibility that these cells correspond to mCAFs or apCAFs cannot be excluded without evaluation of canonical markers, such as collagen-related markers for mCAFs and CD74/MHC class II molecules for apCAFs. The very low expression of ISLR/Meflin in CD81-high stromal cells suggests that these cells are unlikely to correspond to ISLR/Meflin-positive CAFs, which are known as a tumor-restraining CAF subtype [39].

The relatively high expression of PDPN in CD81-high stromal cells raises the possibility that this population may belong to the CAF compartment, as PDPN has been described as a broadly expressed CAF marker [37]. However, the distinct expression patterns observed between CD81-high stromal cells and PDPN-positive CAFs (Fig. 2D and Fig. S2D) suggest that these populations are phenotypically distinct and unlikely to represent a single homogeneous lineage. Collectively, these findings suggest that CD81-high stromal cells represent a previously unrecognized stromal subset distinct from established CAF subtypes, with potential biological and clinical significance.

Telocytes have recently been described in the stromal compartments of various organs, including breast tissue, and are characterized as CD34-positive and vimentin-positive spindle-shaped interstitial cells with long cytoplasmic processes [40, 41]. Telocytes are thought to contribute to stromal structural organization and intercellular communication through direct cellular contacts and extracellular vesicles including CD81-enriched exosomes. The phenotypic similarities between telocytes and CD81-high stromal cells raise the possibility that this population may correspond to telocyte-like stromal cells or telocyte-derived CAFs rather than conventional activated CAF subsets.

Indeed, the co-expression analyses performed in this study have important limitations due to the restricted set of markers examined. Comprehensive genetic and spatial omics analyses, flow cytometry or single-cell analysis will be required in future studies to further clarify the origin and identity of CD81-high stromal cells. Such approaches may also help determine whether these cells represent a distinct stromal population or a transitional state of stromal components.

Tumor cells in the high sCD81 group showed elevated proliferation markers, including Ki-67, despite their smaller tumor size. In breast cancer, Ki-67 labeling index is known to correlate strongly with nuclear grade and histological grade [42, 43], and previous studies have also reported a positive correlation between Ki-67 and tumor size [44]. In contrast, our study found no statistically significant differences in nuclear grade and an inverse relationship between tumor size and Ki-67 status between high and low CD81 groups. These findings suggest that CD81-high stromal cells may not directly drive tumor growth. Instead, the strong association of CD81-high stromal cells with tumor vasculature raises the possibility that these stromal cells promote tumor cell proliferation indirectly. The increased vascularization may provide a richer oxygen supply, thereby supporting tumor cell viability without necessarily contributing to overall tumor mass expansion. This mechanism could help explain the noted discrepancy between higher proliferative activity and smaller tumor size in the CD81-high group.

The present study confirmed that PDPN-positive CAFs preferentially recruit Tregs and M2-like macrophages into the TME [45], and PDPN is also expressed in HEVs, consistent with previous reports [18–20]. By contrast, CD81-high stromal cells showed distinct characteristics from PDPN-positive CAFs. These findings suggest that CD81-high stromal cells represent a functionally distinct stromal compartment, differing from the immunosuppressive phenotype typically associated with PDPN-positive CAFs. apCAFs have been proposed to promote the recruitment of Tregs in the TME [37]. These shared immunosuppressive features of apCAFs and PDPN-positive CAFs further support the interpretation that CD81-high stromal cells are unlikely to correspond to apCAFs.

In a study of NSCLC, CCL19-expressing stromal cells were classified into two distinct subsets: perivascular reticular cells and T-cell zone reticular cells, which originate from different precursors [17]. Similarly, in the present study, we identified two types of CCL19-expressing cells: a tumor vasculature–associated population and a non-vascular, PDPN-positive population. These populations likely correspond to perivascular reticular cells and T-cell zone reticular cells, respectively. Our findings further suggest that CD81-high stromal cells are preferentially localized in close proximity to perivascular reticular cells, highlighting a potential role for these stromal elements in regulating T-cell trafficking within the TME.

In this study, no significant differences in the abundance of CD81-high stromal cells were observed across clinical subtypes within the TME. However, immune cell densities were significantly higher in the triple-negative subtype than in the so-called luminal subtype, as shown in Fig. S3A. Consistent with this observation, an increased tendency for HEVs, as indicated by MECA-79 expression, was observed in the triple-negative subtype. In contrast, a weaker trend was observed for CCL19-positive cells (Fig. S5E, F). Taken together, these findings suggest that HEV-associated immune cell recruitment is more active in the triple-negative subtype than in luminal subtypes [46], potentially through mechanisms that are at least partly independent of CCL19 and CD81-high stromal cells.

A strong positive correlation in CD81 expression between tumor cells (tCD81) and stromal cells (sCD81) suggests the presence of cell–cell interactions mediated by yet unidentified mechanisms. CD81 is known to be enriched in extracellular vesicles, including exosomes, raising the possibility that high CD81 expression within cells reflects enhanced utilization of such intercellular communication pathways. Consistent with previous reports [12], high tCD81 expression was associated with poor prognosis, as also demonstrated in this study (Fig. 1H); however, the mechanisms underlying this adverse clinical outcome remain unclear. Within the tumor immune microenvironment, high tCD81 expression was associated with a significant reduction in CCL19-expressing cells in tumors with high sCD81 expression (Fig. S7A). In contrast, sCD81 expression was associated with increased immune cell infiltration in this study. Collectively, these findings suggest that CD81 may play context-dependent roles in regulating CCL19-associated immune cell recruitment within the TME, potentially exerting divergent effects depending on its cellular source. However, the precise role of CD81 in modulating CCL19-associated immune cell recruitment in the TME, whether mediated by direct or indirect tumor–stroma crosstalk, cannot be determined based on the current data. Further studies will be required to elucidate the precise mechanisms underlying this potential mode of tumor–stroma crosstalk.

Conclusions

Stromal CD81 expression in invasive breast cancer defines a previously unrecognized stromal compartment distinct from established CAF subsets. These cells are preferentially localized to perivascular regions, where they appear to be closely associated with CCL19-producing reticular cells and may be involved in the recruitment of effector T cells into the TME. The abundance of CD81-high stromal cells in the TME of invasive breast cancer may not only stratify patients into a favorable prognostic category but also contribute to the selection of appropriate therapeutic decision-making, including the use of immune checkpoint therapies.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

ACTA2

alpha-smooth muscle actin

CAFs

Cancer-associated fibroblasts

CCL19

C-C motif chemokine ligand 19

CD

Cluster of differentiation

CTLs

Cytotoxic T lymphocytes

DCIS

Ductal carcinoma in situ

DCN

Decorin

DFS

Disease-free survival

ECM

Extracellular matrix

FAP

Fibroblast activation protein

HER2

Human epidermal growth factor receptor 2

HEVs

High endothelial venules

HR

Hormone receptor

ISLR

Immunoglobulin superfamily containing leucine rich repeat protein

Macrophages

NSCLC

Non-small cell lung carcinoma

PDGFRα

Platelet-derived growth factor receptor alpha

PDPN

Podoplanin

PHH3

Phosphorylated histone H3

TAMCs

Tumor-associated myeloid cells

TME

Tumor microenvironment

Tregs

Regulatory T cells

VIM

Vimentin

Author contributions

MK conceived and designed the study and drafted the manuscript. KN and MK performed all experiments. HK, AN, AN-I, and ST contributed to the interpretation of the data. SI provided guidance on statistical analyses. TT provided patient data and tissue specimens. All authors reviewed and approved the final manuscript.

Funding

This work was supported by the Japan Society for the Promotion of Science (JSPS) through a Grant-in-Aid for Scientific Research (C) (Grant Number 23K06427 to M.K.).

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethics approval and consent to participate

The study protocol was approved by the Institutional Review Board of the Nagoya City University Graduate School of Medical Sciences (approval number: 60-25-0050) and was conducted in accordance with the principles of the Declaration of Helsinki.

Consent for publication

All authors have consented to the publication of this article.

Competing interests

The authors declare no competing interests.