Soluble N-terminal EphA2 fragment as a diagnostic biomarker and functional mediator of tumor progression in triple-negative breast cancer

Highlights

• •Soluble EphA2 fragments are elevated in serum from patients with TNBC.
• •A dual-epitope ELISA detects soluble EphA2 fragments in biological samples.
• •Serum EphA2 shows exploratory associations with grade, nodal status, and survival.
• •Recombinant EphA2-P328 promotes migration and alters EMT markers in vitro.

Abstract

Triple-negative breast cancer (TNBC) lacks reliable biomarkers for early detection and remains a highly aggressive disease. Ephrin type-A receptor 2 (EphA2), a receptor tyrosine kinase frequently overexpressed in advanced cancers, undergoes MMP14-mediated cleavage that releases soluble N-terminal fragments. Here, we investigated the diagnostic and functional significance of this fragment in TNBC. Immunohistochemical analysis of clinical specimens revealed elevated EphA2 expression in TNBC tissues, with a trend toward higher levels in tumors of advanced stage and grade. To enable non-invasive detection, we established a dual-epitope sandwich ELISA targeting spatially distinct regions of the EphA2 extracellular domain. The assay showed concentration-dependent detection of recombinant EphA2 (EC₅₀, 48 ng/mL) and detected soluble EphA2 in TNBC cell-conditioned media. Serum analysis showed significantly higher EphA2 levels in TNBC patients compared with healthy controls (median 93.3 vs 25.6 ng/mL). In this exploratory cohort, ROC analysis yielded an AUC of 0.912 (95% CI, 0.862–0.962), with a sensitivity of 95.0% (95% CI, 86.3–98.6%) and a specificity of 78.3% (95% CI, 66.4–86.9%), and higher levels were associated with lymph node metastasis and higher histological grade. Functionally, recombinant EphA2-P328 promoted TNBC-cell migration and induced cell context-dependent changes in EMT-associated markers in vitro. Collectively, these findings support the soluble N-terminal EphA2 fragment as a candidate serum biomarker and an in vitro modulator of TNBC cell migration, warranting validation in larger independent cohorts and further mechanistic investigation.

Introduction

Triple-negative breast cancer (TNBC) is a subtype of breast cancer characterized by the absence of estrogen receptor (ER), progesterone receptor (PR), and human epidermal growth factor receptor 2 (HER2), accounting for approximately 12–17% of all breast cancer cases [1,2]. TNBC disproportionately affects younger women and is known for its pronounced molecular and clinical heterogeneity [1,2]. Clinically, TNBC is associated with aggressive behavior, poor prognosis, high recurrence rates, and shorter overall survival compared to hormone receptor-positive or HER2-positive breast cancer subtypes [3,4]. Due to the lack of specific therapeutic targets, TNBC does not respond to endocrine therapies (e.g., tamoxifen) or anti-HER2 agents (e.g., trastuzumab), rendering conventional chemotherapy the mainstay of treatment. While TNBC may initially respond to chemotherapy, long-term outcomes remain unsatisfactory, with a high risk of relapse. These challenges underscore the urgent need for improved strategies in TNBC management, where earlier detection may offer a critical window for curative intervention [3,4]. Although circulating tumor DNA (ctDNA) and circulating tumor cells (CTCs) have been investigated in breast cancer, their sensitivity and reproducibility remain limited, especially in TNBC. Protein-based biomarkers detectable in serum may therefore provide a more robust and complementary strategy for early diagnosis and dynamic monitoring.

The clinical management of TNBC remains challenging due to its aggressive progression and lack of actionable molecular targets [4]. More critically, TNBC lacks reliable biomarkers for early diagnosis or prognostic stratification. Diagnostic decisions still largely depend on imaging or invasive biopsies, both of which may have limited sensitivity and specificity, especially during early-stage disease. To overcome these limitations, liquid biopsy platforms based on circulating tumor proteins, extracellular vesicles, or cleaved surface molecules have emerged as promising tools for non-invasive cancer detection. These platforms allow dynamic disease monitoring and provide real-time insights into tumor biology [5]. Importantly, receptor tyrosine kinases such as Ephrin type-A receptor 2 (EphA2) are frequently overexpressed and proteolytically cleaved in aggressive tumors, releasing soluble N-terminal fragments into the bloodstream. These fragments can serve as potential circulating biomarkers, linking molecular alterations within the tumor to detectable changes in peripheral blood [6]. Thus, developing robust assays capable of sensitively and specifically detecting such cleaved EphA2 fragments in serum may offer a valuable strategy for early TNBC screening and risk assessment.

EphA2, a member of the Eph receptor tyrosine kinase family, is frequently overexpressed in various cancers, including breast, lung, brain, pancreatic, and prostate [[7], [8], [9], [10], [11]]. Eph receptors can signal through canonical (ligand-dependent) and non-canonical (ligand-independent) pathways. Canonical signaling involves bidirectional communication initiated upon interaction with ephrin ligands, regulating normal cell-cell communication. In contrast, non-canonical signaling is typically triggered through phosphorylation of EphA2 at serine 897 (S897), mediated by kinases such as AKT, RSK, or PKA, and often promotes tumor progression. In tumor cells, EphA2 is subject to proteolytic cleavage by membrane-type matrix metalloproteinases (MT-MMPs), particularly MT1-MMP (MMP14), which removes its extracellular ligand-binding domain [12]. This cleavage abrogates canonical signaling and facilitates ligand-independent activation of pro-oncogenic pathways, in particular the AKT cascade [13,14]. Notably, both EphA2 and MMP14 are often co-upregulated in aggressive cancers and are under the regulation of oncogenic Ras-MAPK signaling [15,16].

In TNBC, EphA2 overexpression is associated with poor prognosis, including reduced overall survival (OS) and disease-free survival (DFS) [17]. Recent studies have shown that the shed N-terminal fragment of EphA2, generated by MMP14 cleavage, can be detected in the tumor microenvironment and potentially released into the circulation [13,14]. This fragment retains functional activity, promoting cancer cell proliferation and invasion via activation of downstream signaling pathways. Despite these findings, the diagnostic and functional significance of circulating EphA2 N-terminal fragments has been reported only in selected cancer types, such as pancreatic or ovarian cancer. Still, it has not been systematically evaluated in breast cancer, particularly in TNBC. Moreover, whether soluble EphA2 fragments exert direct functional effects on TNBC progression has not been demonstrated, leaving an important gap in understanding their biological relevance. In this study, we investigate the potential of this fragment as a biomarker for TNBC through the development of a custom ELISA platform and assess its clinical significance as a serum-based indicator of tumor progression and invasiveness.

Materials and methods

Clinical serum samples and ethical approval

Human serum samples were obtained from the Taipei Medical University Joint Biobank (TMU-JBB; Biobank application number B202402001). The study protocol was approved by the Taipei Medical University Joint Institutional Review Board (IRB number N202312057). A total of 60 TNBC patient serum samples and 60 healthy control samples were acquired and used in this study. Serum samples were stored at −80 °C according to the standardized operating procedures of the TMU-JBB. Freeze-thaw cycles were minimized and limited to two before analysis. Individual haemolysis information was not available for retrospective evaluation.

Cell culture

TNBC cell lines MDA-MB-231, BT549, Hs578T, and MDA-MB-468, along with other breast cancer cell lines and the non-tumorigenic mammary epithelial cell line 184B5, were obtained from the American Type Culture Collection (ATCC, Manassas, VA, USA). All cell lines were maintained under standard culture conditions recommended by ATCC and incubated at 37 °C in a humidified incubator with 5% CO₂. Conditioned media were collected at specified time points to assess the release of soluble EphA2.

Bioinformatics analysis

To evaluate the RNA expression of EphA2 between TNBC and non-TNBC cell lines, normalized transcript expression (nTPM) data were retrieved and analyzed using the Protein Atlas database (https://www.proteinatlas.org/). The comparative analysis assessed EphA2 expression levels across different breast cancer cell line subtypes. In addition, because no complete experimental structure of the EphA2 extracellular domain (ECD) is available in the Protein Data Bank (PDB), the predicted EphA2 ECD structure was obtained from the AlphaFold Protein Structure Database (https://alphafold.ebi.ac.uk/) with the identifier AF-P29317-F1-v4.

Immunohistochemical staining

A TNBC tissue microarray slide (TissueArray.com, catalog no BR1301a) was used to evaluate EphA2 expression in clinical tissue samples. The immunohistochemical (IHC) staining was performed using a commercially available anti-EphA2 antibody (Cell Signaling Technology), and staining patterns were observed under a Zeiss Axioskop-2 microscope (Carl Zeiss). The staining intensity was semi-quantitatively scored from 0 to 3 by a board-certified pathologist. Because scoring was performed by a single pathologist, inter-observer variability was not assessed.

Western blotting

For Western blot analysis, cells were seeded in 6-well plates and treated with the recombinant EphA2-P328 fragment at the indicated concentrations. After treatment, cells were lysed in 200 μL of lysis buffer containing 50 mM Tris–HCl (pH 7.5), 50 mM NaCl, 5 mM EDTA, 1% Triton X-100, and a protease inhibitor cocktail (Roche Applied Science). Protein concentrations were determined using the Coomassie Plus (Bradford) Protein Assay Kit (Thermo Fisher Scientific). Protein lysates were separated by reducing SDS-PAGE and transferred to membranes for immunoblotting. The following primary antibodies were used: EphA2, p-EphA2, MMP14, p-AKT, AKT, p-ERK, ERK, p-FAK, FAK, p-Src2, and Src2 (Cell Signaling Technology), and β-actin and GAPDH (GeneTex) as loading controls. HRP-conjugated secondary antibodies (anti-mouse IgG and anti-rabbit IgG) were obtained from Jackson ImmunoResearch. Band intensities were quantified using ImageJ. Phosphorylated protein levels were normalized to the corresponding total protein levels, whereas E-cadherin, N-cadherin, and vimentin levels were normalized to GAPDH. All values were expressed relative to the untreated control for each cell line, which was set to 1.00.

Generation of recombinant Epha2 fragments

Peptide EA2pepI was prepared and expressed according to previously published methods [18]. Peptide EA2pepII, designed to avoid the ligand-binding domain and located within the cysteine-rich region (aa 182–328), was cloned into the pET-32a vector, expressed in Escherichia coli BL21(DE3), and purified by standard procedures. To generate recombinant soluble EphA2 proteins, two N-terminal truncated fragments of human EphA2-P328 (aa 27–328) and -N534 (aa 27–534) were designed by removing the native signal peptide sequence and cloned into the pSecTag2A expression vector (Thermo Fisher Scientific). The following primer sequences were used for PCR amplification; they are forward primer (EphA2-pSec_F): 5′-GACGC GGCC CAGCC GGCC GCG CAG GGC AAG GAA GTG-3′, reverse primer for P328 (EphA2-pSec_P328_R): 5′-CCTCGAGCGGCCGCGGGTCGTGTGCAAGGCAT-3′ and reverse primer for N534 (EphA2-pSec_N534_R): 5′-CCTCGAGCGGCCGCgttgccagatccctccgg-3′. The PCR-amplified fragments were cloned into the pSecTag2A vector and sequence-verified before transfection into Expi293 cells using the Expi293 expression system (Thermo Fisher Scientific), following the manufacturer’s instructions. After 5 days of expression, culture supernatants containing secreted His-tagged recombinant proteins were harvested and purified via Ni²⁺-NTA resin column (GE Healthcare).

Chicken immunization and phage display selection

Female white leghorn chickens (Gallus domesticus) were purchased from the National Laboratory Animal Center, Taiwan, and maintained at the Taipei Medical University animal facility under standard husbandry conditions. Chickens were immunized with recombinant EphA2-N534 protein (aa 27–534), and spleens were collected after the final immunization for library construction. Antibody library generation and phage display procedures were carried out as previously described, with minor modifications [18,19]. For biopanning, recombinant EA2pepII (aa 201–328) was used as the target antigen to enrich EphA2-specific scFv clones, which led to the isolation of the scFv S10 employed in this study.

ScFv and IgG expression and purification

Recombinant scFv S10 clone was cultured in 2 × YT medium at 37 °C until an optical density at 600 nm (OD₆₀₀) of 0.6 was reached. Protein expression was induced with 0.5 mM isopropyl β-_d-1-thiogalactopyranoside (IPTG) for 6 hours. Bacterial cells were harvested by centrifugation and resuspended in binding buffer (500 mM NaCl, 20 mM Tris–HCl, 20 mM imidazole, pH 7.4). Cells were lysed by three freeze-thaw cycles (−80 °C/37 °C) followed by sonication (QSONICA LLC.) on ice. The clarified lysate was subjected to His-tag purification using a Ni²⁺-charged resin column (GE Healthcare) according to the manufacturer’s instructions.

The full-length IgG hSD5 antibody was produced in Expi293 cells (Thermo Fisher Scientific) according to the manufacturer’s protocol. Following the expression, the antibody-containing culture supernatant was harvested, and purification was carried out using protein A affinity chromatography. The binding specificity of IgG hSD5 had been previously validated by flow cytometry in EphA2-overexpression cell lines [18].

Competitive ELISA

To assess the epitope-specific binding of scFv antibodies, a competitive ELISA was performed using recombinant EphA2-P328 protein as the immobilized antigen. Microtiter plates were coated with 0.5 μg/well of recombinant EphA2 and blocked with 5% skim milk in PBS to prevent non-specific binding. Purified scFv antibodies (either hSD5 or S10) were individually pre-incubated with various concentrations (1.5–100 μg/mL) of recombinant peptide competitors (EA2pepI or EA2pepII) in PBS for 1 hour at room temperature. These antibody–peptide mixtures were then transferred to the wells and incubated for an additional hour under the same conditions to allow competitive binding. After washing with PBST, bound scFv was detected using an HRP-conjugated anti-HA antibody (Cell Signaling Technology Inc.), targeting the HA tag fused to the scFv constructs. The colorimetric signal was developed using a TMB substrate and stopped with 1 N HCl. Absorbance at 450 nm was measured to evaluate antibody binding in the presence of either epitope-specific or non-specific peptide competitors.

Sandwich ELISA platform

A custom sandwich ELISA was developed to quantify soluble N-terminal EphA2 fragments in serum and conditioned media. Flat-bottom 96-well plates (Nunc MaxiSorp) were coated with 1 μg/well of the anti-EphA2 capture antibody scFv S10 in PBS and incubated overnight at 4 °C. The plates were then blocked with 3% BSA in PBS for 1 hour at room temperature. Serial dilutions of recombinant human EphA2 fragment P328 were prepared starting at 800 ng/mL in PBS to generate a standard curve. Clinical serum samples were diluted 1:10 in PBS before application. After loading the samples, plates were incubated for 1 hour at room temperature, followed by three washes with PBST (PBS with 0.05% Tween-20). The biotin-conjugated detection antibody hSD5 was applied at 0.1 μg/mL and incubated for 1 hour at room temperature. After washing, HRP-conjugated streptavidin was added and incubated for 1 hour at room temperature. Following a final wash, TMB substrate solution was added for signal development, and the reaction was terminated by the addition of 1 N HCl. Absorbance was measured at 450 nm using a microplate reader. All samples were analyzed in duplicate, and the mean values were used for quantification. The EphA2 concentration in each sample was determined by interpolating the standard curve.

Generation of epha2-knockdown (KD) and epha2-overexpressing cells

Stable repression of EphA2 expression in TNBC cell lines was achieved using a CRISPR interference (CRISPRi) approach based on a catalytically inactive Cas9 fused to a KRAB transcriptional repressor (dCas9-KRAB). A lentiviral dCas9-KRAB expression construct (Addgene plasmid #71236) was used to establish the CRISPRi system. A single-guide RNA (sgRNA) targeting the promoter region of the EPHA2 gene (5’-GCGCAACTTTGGAACCTACC-3’) was designed and cloned into the corresponding lentiviral sgRNA vector. Lentiviral particles were generated in HEK293T cells via co-transfection of the dCas9-KRAB or sgRNA constructs together with packaging and envelope plasmids provided by the National RNAi Core Facility (Academia Sinica, Taiwan). TNBC cells were transduced with the resulting lentiviruses and subsequently selected using puromycin (2 μg/mL; Sigma) for 7 days to establish stable EphA2-KD cell lines. The efficiency of EphA2 repression was confirmed by immunoblot analysis.

EphA2-overexpressing 293T cells were generated by stable transfection with an EphA2 expression construct using PolyJet in vitro DNA transfection reagent (SignaGen Laboratories), followed by selection with hygromycin B (InvivoGen). EphA2 overexpression was confirmed by immunoblotting and flow cytometry.

Cell migration

TNBC cells were seeded at a density of 5 × 10⁴ cells per well in 24-well transwell migration inserts (Corning). Cells were exposed to recombinant EphA2-P328 protein at concentrations of 0.1 or 1 μM for either 6 or 24 h. Following treatment, cells were fixed in pre-chilled 100% methanol for 10 min and stained with DAPI at room temperature. After rinsing with distilled water, non-migrated cells on the upper surface of the membrane were gently removed using cotton swabs. Migrated nuclei on the underside of the membrane were visualized by fluorescence microscopy and quantified using ImageJ software. All migration assays were performed in at least three independent biological replicates to ensure reproducibility.

Statistical analysis

Statistical analyses were performed using GraphPad Prism software (GraphPad Software, San Diego, CA, USA), whereas multivariable Cox regression and proportional hazards diagnostics were performed using Python lifelines version 0.30.0. Normality was assessed using the Shapiro-Wilk test. Because serum EphA2 concentrations were non-normally distributed and the subgroup sizes were small and unequal, non-parametric tests were used. Two-group comparisons were performed using the two-sided Mann-Whitney U test, whereas T-stage and N-stage comparisons were performed using the Kruskal-Wallis test followed by Dunn’s multiple-comparisons test, with multiplicity-adjusted p values reported. Receiver operating characteristic (ROC) curve analysis was conducted to evaluate diagnostic performance. The area under the curve (AUC) and its 95% confidence interval were calculated using the nonparametric method implemented in GraphPad Prism. The optimal cutoff value was determined by maximizing Youden’s J index (sensitivity + specificity − 1). The 95% confidence intervals for sensitivity and specificity were calculated using the hybrid Wilson/Brown method implemented in Prism [20].

Overall survival was estimated using the Kaplan-Meier method, and patients were stratified into high and low serum EphA2 groups according to the median serum EphA2 level. Survival curves were compared using the log-rank (Mantel-Cox) test. The unadjusted hazard ratio and its 95% confidence interval were estimated using the Mantel-Haenszel method. Median follow-up duration was estimated using the reverse Kaplan-Meier method by reversing the event and censoring indicators.

To evaluate whether serum EphA2 provided prognostic information after adjustment for clinicopathological variables, multivariable Cox proportional hazards regression was performed with serum EphA2 group (high vs low), age (per 10-year increase), tumor T category (T2-T3 vs T1), nodal status (N1-N3 vs N0), and histologic grade (grade 3 vs grades 1–2). Patients with unknown nodal status were excluded from the complete-case analysis. Hazard ratios and 95% confidence intervals were estimated using the Efron approximation for tied event times. The proportional hazards assumption was assessed using scaled Schoenfeld residuals. All assays were performed in duplicate or in at least three independent biological replicates to ensure reproducibility. A p-value < 0.05 was considered statistically significant (*p < 0.05, **p < 0.01, ***p < 0.001).

Result

EphA2 is overexpressed in TNBC and tends to be elevated in advanced tumors

We investigated EphA2 expression in TNBC by performing IHC staining on tissue microarrays containing 119 TNBC samples and 8 normal breast tissues. Representative images from grade 3 tumors of stage II and stage III, along with normal controls, are shown in Fig. 1A. In high-magnification views, intense membrane-localized EphA2 staining was observed in tumor cells (indicated by red arrows). Overall, 68 of 119 TNBC samples (57.1%) were positive for EphA2, whereas all normal tissues were negative (Fig. 1B). When stratified by clinical stage, samples at stages II and III exhibited higher staining intensity. Among positive cases, 91.5% (43/47) of stage II tumors were scored as (+) and 8.5% (4/47) were scored as (++), while in stage III, 75% (9/12) were scored as (+) and 25% (3/12) were scored as (++). In contrast, stage I tumors showed only weak staining (Fig. 1C). Semi-quantitative analysis of EphA2 IHC staining revealed a significant difference in staining intensity distributions among normal tissues and TNBC stages (Kruskal-Wallis test, p = 0.0379). A similar pattern was observed when stratified by histological grade: positive staining was detected only in tumors of grades II and III. Among positive cases, score (++) staining was observed in 6% (1/17) of grade II and 14% (6/44) of grade III tumors (Fig. 1D). When stratified by histological grade, semi-quantitative analysis of EphA2 IHC staining showed a trend toward higher intensity in higher-grade tumors; however, the difference did not reach statistical significance (Kruskal-Wallis test, p = 0.0688).

Fig. 1.

EphA2 expression in TNBC tissues and cell lines. (A) Representative IHC images of EphA2 staining in TNBC tissues, including stage II and III tumors with grade 3 histology, and in normal breast tissue. Higher-magnification views (red boxes) show membrane-localized staining indicated by red arrows. (B) Summary of IHC staining results from 119 TNBC and 8 normal breast tissue samples. (C) Distribution of EphA2 IHC staining scores stratified by clinical stage. Staining intensity was scored on an ordinal scale (0–3) and compared among normal tissues and TNBC stages using the Kruskal-Wallis test (p = 0.0379). Sample sizes: Normal (n = 8), Stage I (n = 6), Stage II (n = 90), Stage III (n = 23). The right panel shows the staining intensity percentage among EphA2-positive cases within each stage. (D) Distribution of EphA2 IHC staining scores stratified by histological grade. Staining intensity was analyzed using the Kruskal-Wallis test (p = 0.0688). Sample sizes: Grade I (n = 2), Grade II (n = 41), Grade III (n = 76). The right panel shows the staining intensity percentage among EphA2-positive cases within each grade. (E) Western blot analysis of EphA2 and MMP14 protein expression in a panel of nine breast cell lines, including four TNBC, four non-TNBC, and one normal breast epithelial cell line (184B5). (F) EphA2 mRNA expression levels in TNBC (n = 21) and non-TNBC (n = 41) breast cancer cell lines based on the Protein Atlas database (CCLE cell line transcriptomics). Statistical comparisons were performed using the Mann-Whitney U test, with significance defined as *p < 0.05, **p < 0.01, ***p < 0.001.

Western blot analysis of nine breast cell lines, including four TNBC, four non-TNBC, and one normal mammary epithelial cell line (184B5), revealed that EphA2 was strongly expressed in three TNBC cell lines (MDA-MB-231, BT549, and Hs578T) but absent in MDA-MB-468 and in all non-TNBC and normal cell lines (Fig. 1E). MMP14, a metalloproteinase implicated in EphA2 cleavage, was co-expressed in the three EphA2-positive TNBC cell lines, supporting the possibility that MMP14-mediated cleavage generates a soluble EphA2 fragment. Finally, the RNA expression analysis of normalized transcript expression from the Protein Atlas database (derived from Cancer Cell Line Encyclopedia, CCLE) confirmed that TNBC cell lines (n = 21) expressed significantly higher levels of EphA2 mRNA than non-TNBC cell lines (n = 41), with statistical significance (Fig. 1F).

Design and structural modeling of a dual-epitope sandwich ELISA for detecting soluble Epha2

To evaluate the clinical applicability of detecting cleaved, soluble EphA2 in liquid biopsy specimens such as serum, we developed a custom sandwich ELISA system. As shown in Fig. 2A, the epitope selection strategy was guided by the predicted ECD structure of EphA2 obtained from the AlphaFold Protein Structure Database (AF-P29317-F1-v4). The model exhibited a high average pLDDT score of 81.83 across the full-length protein, indicating overall reliable confidence in the structural prediction. Two spatially distinct and non-overlapping epitopes were selected, with EA2pepI (red) located within the ligand-binding domain (LBD; aa 27–200) [18], and EA2pepII (yellow) positioned in the cysteine-rich domain (CRD; aa 201–328). Both epitopes were purposefully selected away from the membrane-proximal MMP14 cleavage site to ensure compatibility with the detection of cleaved EphA2. The antibody pair consisted of hSD5, a previously characterized neutralizing antibody that binds EA2pepI [18], and scFv S10, which was isolated from a chicken immune phage display library using recombinant EA2pepII as bait. Structural prediction of both hSD5 Fab and scFv S10 using AlphaFold3 revealed high-confidence models with pTM scores exceeding 0.9, except for expected variability in the CDR loops, supporting confidence in the overall predicted folds (Fig. 2B). To assess the feasibility of dual engagement on the same cleaved EphA2 molecule, we performed ternary complex modeling using AlphaFold3, incorporating both antibody fragments and the EphA2-P328 fragment. The predicted model suggested that hSD5 Fab and scFv S10 may bind simultaneously without substantial steric interference. This computational prediction was considered hypothesis-generating, whereas simultaneous antibody binding and epitope specificity were evaluated experimentally using BLI and competitive ELISA, respectively (Fig. 2C).

Fig. 2.

Structural modeling and epitope validation for dual-antibody detection of EphA2. (A) Schematic representation of epitope mapping based on the predicted ECD structure of EphA2 obtained from the AlphaFold Protein Structure Database. Two spatially distinct epitopes were selected: EA2pepI (red) within the ligand-binding domain (LBD, aa 27–200) and EA2pepII (yellow) within the cysteine-rich domain (CRD, aa 201–328), both located away from the C-terminal MT1-MMP cleavage site. (B) Structural models of the antibodies hSD5 Fab and scFv S10 generated using AlphaFold3 prediction. The confidence metrics of AlphaFold 3 include pLDDT (predicted Local Distance Difference Test) and predicted Template Modeling (pTM) score. (C) Predicted ternary complex of cleaved EphA2-P328 with hSD5 Fab and scFv S10 modeled by AlphaFold3. The predicted model suggests simultaneous binding without substantial steric interference, supporting the feasibility of a sandwich assay configuration. (D) BLI analysis of scFv S10 binding to EphA2-P328, showing dose-dependent binding kinetics. (E) Competitive ELISA validating the non-overlapping nature of the two antibody epitopes. Binding of hSD5 and S10 to recombinant EphA2 was specifically blocked by their corresponding peptides (EA2pepI and EA2pepII, respectively), but not by irrelevant peptides.

To validate the performance of our sandwich ELISA platform, we expressed a truncated recombinant EphA2 fragment (EphA2-P328; aa 27–328), which mimics the extracellular domain shed by MMP14 under pathological conditions (Supplementary Figure 1). The binding kinetics of scFv S10 to EphA2-P328 were further characterized by BLI, showing dose-dependent binding with a calculated affinity (K_D) of 48.9 nM (Fig. 2D). Together with the previously reported affinity of hSD5 (K_D = 2.06 nM), this antibody pair provides an effective tool for capturing and detecting soluble EphA2. Moreover, to confirm the distinct binding specificities of the two antibodies, we performed a peptide competition ELISA. The binding of hSD5 and S10 to EphA2 was specifically blocked by their corresponding peptides (EA2pepI and EA2pepII, respectively) but not by the non-matching peptide, thereby validating the independent and non-competing binding specificities of the two antibodies (Fig. 2E).

Characterization of the sandwich ELISA for detecting cleaved Epha2

The antibodies hSD5 and S10 have been shown to recognize separate, non-competing epitopes on the EphA2 ECD. To confirm their ability to recognize the same EphA2 molecule simultaneously, we performed a sequential binding assay using BLI system. As shown in Fig. 3A, the immobilized hSD5 IgG effectively captured EphA2-P328, and the subsequent addition of increasing concentrations of scFv S10 led to a dose-dependent signal enhancement, indicating that both antibodies can co-engage the same antigen. We next examined whether the C-terminal truncation in EphA2-P328 affected epitope presentation by comparing its detection with that of a full-length ECD construct, EphA2-N534 (aa 27–534). ELISA results showed that both recombinant proteins were efficiently recognized with comparable signal intensities, confirming that the shorter P328 fragment retains full antibody accessibility. Accordingly, EphA2-P328 was used as the standard for assay calibration and as a representative model of circulating fragments. Serial dilutions of EphA2-P328 generated a concentration-dependent response with a calculated EC₅₀ of 48 ng/mL (Fig. 3B) and showed linearity over the tested range (R² = 0.9918) (Fig. 3C). To test the assay under more physiologically relevant conditions, we collected conditioned media from three EphA2-high TNBC cell lines at 0, 24, 48, and 72 h of culture. As shown in Fig. 3D, a time-dependent accumulation of soluble EphA2 was observed, reaching approximately 1.2–1.5 μg/mL at 72 h (approximately 0.04 μM based on EphA2-P328), which is comparable in magnitude to the lower concentrations used in the subsequent functional experiments. To further assess assay selectivity, soluble EphA2 was measured in conditioned media from parental MDA-MB-231 cells, CRISPRi-mediated EphA2-KD MDA-MB-231 cells, and EphA2-overexpressing 293T cells. At 48 and 72 h, the detected EphA2 levels were lower in samples from EphA2-KD cells than in those from parental MDA-MB-231 cells and were highest in samples from EphA2-overexpressing 293T cells at 72 h, supporting EphA2-dependent detection by the sandwich ELISA (Supplementary Figure 2). These findings indicate that TNBC cells actively release cleaved EphA2 fragments under standard culture conditions, and support the ability of our sandwich ELISA to detect such fragments in biologically complex samples.

Fig. 3.

Characteristics of the sandwich ELISA for detecting recombinant and soluble EphA2 fragments. (A) BLI assay showing sequential binding of recombinant EphA2-P328 to immobilized hSD5 IgG followed by scFv S10 at increasing concentrations. (B) Dose-response curve of the sandwich ELISA using recombinant EphA2 fragments EphA2-N534 and EphA2-P328 to assess detection range and assay performance. (C) Standard curve generated from serial dilutions of the recombinant EphA2 fragment EphA2-P328 to evaluate assay linearity via regression analysis. (D) Measurement of soluble EphA2 levels in conditioned media collected from three EphA2-high TNBC cell lines at 0, 24, 48, and 72 h of culture.

Serum epha2 levels are elevated in TNBC patients and are associated with clinical parameters

To explore the discriminatory performance of the sandwich ELISA platform within the current cohort, we measured serum EphA2 levels in 60 TNBC patients and 60 healthy individuals. As shown in Fig. 4A, TNBC patients exhibited significantly elevated serum EphA2 levels compared to healthy controls, with median concentrations of 93.3 ng/mL and 25.6 ng/mL, respectively. We performed the receiver operating characteristic (ROC) analysis to evaluate the ability of serum EphA2 to distinguish patients with TNBC from healthy controls. Exploratory ROC analysis yielded an AUC of 0.912 (95% CI, 0.862–0.962). At the optimal cutoff of 51.9 ng/mL determined by maximizing Youden’s J index, the assay showed a sensitivity of 95% (95% CI, 86.3–98.6%) and a specificity of 78.3% (95% CI, 66.4–86.9%) (Fig. 4B). We further analyzed serum EphA2 levels in relation to clinicopathological features. Stratification by lymphatic metastasis status (Fig. 4C) and histological grade (Fig. 4D) revealed higher EphA2 levels in patients with lymph node metastasis and tumors of higher histological grade, suggesting a positive association with tumor aggressiveness. Serum EphA2 levels differed across T categories (Kruskal-Wallis p = 0.0011), with higher levels in T2 and T3 tumors than in T1 tumors (Dunn-adjusted p = 0.0084 and 0.0082, respectively). Serum EphA2 levels also differed across N categories (Kruskal-Wallis p < 0.0001), with higher levels in N1, N2, and N3 cases than in N0 cases (all Dunn-adjusted p < 0.0001) (Fig. 4E-F).

Fig. 4.

Serum EphA2 levels in TNBC patients and correlation with clinical parameters. (A) Serum EphA2 levels measured by sandwich ELISA in healthy controls (n = 60) and TNBC patients (n = 60). (B) ROC curve analysis of serum EphA2 for distinguishing patients with TNBC from healthy controls. The AUC was 0.912 (95% CI, 0.862–0.962). At the optimal cutoff of 51.9 ng/mL determined using Youden’s J index, sensitivity was 95% (95% CI, 86.3–98.6%) and specificity was 78.3% (95% CI, 66.4–86.9%). (C) Serum EphA2 levels in TNBC patients grouped by lymphatic metastasis status. (D) Serum EphA2 levels grouped by histological grade. (E) Serum EphA2 levels stratified by tumor size (T stage). (F) Serum EphA2 levels stratified by lymph node involvement (N stage). Two-group comparisons in panels A, C, and D were performed using the two-sided Mann-Whitney U test, whereas panels E and F were analyzed using the Kruskal-Wallis test followed by Dunn’s multiple-comparisons test. The p values for panels A, C, and D were < 0.0001, < 0.0001, and 0.0152, respectively. For panel E, the adjusted p values were 0.0084 for T1 versus T2 and 0.0082 for T1 versus T3; all indicated comparisons in panel F had adjusted p < 0.0001. (G) Overall survival was analyzed using the Kaplan-Meier method in patients with available follow-up data (n = 58). Patients were stratified into high and low serum EphA2 groups based on the median serum EphA2 level (n = 29 per group). The median follow-up duration was 88 months, as estimated using the reverse Kaplan-Meier method. Survival curves were compared using the log-rank (Mantel-Cox) test (p = 0.0334). The unadjusted hazard ratio for high versus low serum EphA2 was 2.09 (95% CI, 1.06–4.12), as estimated using the Mantel-Haenszel method. Numbers at risk at the indicated time points are shown below the survival curves.

Among the 58 patients with available survival data, the median follow-up duration was 88 months, as estimated using the reverse Kaplan-Meier method. In the unadjusted survival analysis, patients with high serum EphA2 levels had poorer overall survival than those with low serum EphA2 levels (log-rank p = 0.0334; HR for high versus low EphA2 = 2.09, 95% CI, 1.06–4.12) (Fig. 4G). In the multivariable Cox model, high serum EphA2 was associated with poorer overall survival after adjustment for age, T category, nodal status, and histologic grade (adjusted HR, 5.89; 95% CI, 1.25–27.76; p = 0.025; n = 56, with 40 deaths) (Supplementary Table 1). No evidence of violation of the proportional hazards assumption was detected for the EphA2 group (p = 0.231) or the included covariates (all p > 0.17). Given the wide confidence interval and the substantial overlap between EphA2 group and nodal status, this exploratory estimate should be interpreted cautiously.

Cleaved Epha2 fragment promotes migration and activates downstream signaling in TNBC cells

To investigate the biological effects of cleaved EphA2, we evaluated the response of TNBC cells to EphA2-P328 fragment, designed to mimic the ECD released upon MMP14-mediated cleavage. In a transwell migration assay, treatment with 0.1 or 1 μM EphA2-P328 significantly enhanced the migration capacity of MDA-MB-231, BT549, and Hs578T cells in a dose-dependent manner. Notably, BT549 and Hs578T exhibited a 7- to 10-fold increase at 1 μM, whereas MDA-MB-231 cells showed a more modest increase of around 4-fold under the same condition (Fig. 5A, 5B). In contrast, MDA-MB-468 cells, characterized by low endogenous EphA2 expression (Fig. 1E), exhibited no migration response. Similarly, non-TNBC cell lines (T47D and AU565) and normal mammary epithelial cells (184B5) did not respond to EphA2-P328, consistent with their minimal EphA2 expression. To further investigate the receptor dependence of soluble EphA2-mediated signaling, we generated CRISPRi-mediated EphA2-KD TNBC cell lines, including MDA-MB-231 EphA2-KD and Hs578T EphA2-KD, and examined their migration responses to serial dilutions of recombinant EphA2-P328. In contrast to vehicle controls, which exhibited a dose-dependent increase in migration upon EphA2-P328 stimulation, EphA2-KD cells failed to show enhanced migration, even at higher concentrations of EphA2-P328 (Fig. 5C). These findings indicate that the pro-migratory effect of soluble EphA2-P328 requires cellular EphA2 expression; however, whether this dependence involves direct homotypic interaction, receptor clustering, or co-receptor-mediated signaling remains unresolved.

Fig. 5.

Recombinant EphA2-P328 promotes migration and activates intracellular signaling in TNBC cells. (A) Transwell migration assay of MDA-MB-231, BT549, and Hs578T cells treated with 0, 0.1, or 1 μM EphA2-P328. MDA-MB-231 and BT549 cells were treated for 24 h, while Hs578T cells were treated for 6 h. Migrated cells were fixed, stained with DAPI, and imaged. (B) Quantification of migrated cells shown as fold change relative to untreated controls. Values represent the mean ± SD of three independent experiments. Statistical significance is indicated as ***p < 0.001. (C) Transwell migration assay of vehicle and EphA2-KD MDA-MB-231 and Hs578T cells treated with serial dilutions of EphA2-P328, as indicated. Quantification of migrated cells is shown as fold change relative to untreated controls. Values represent the mean ± SD of three independent experiments. Statistical significance is indicated as ***p < 0.001. The right panels show immunoblot analysis confirming EphA2 knockdown in vehicle and EphA2-KD cells, with β-actin as a loading control. (D) Western blot analysis of intracellular signaling pathways, including phosphorylation of EphA2, Src, FAK, Erk, and AKT, in MDA-MB-231, BT549, and Hs578T cells after 24 h stimulation with varying concentrations of EphA2-P328. (E) Western blot analysis of EMT markers (E-cadherin, N-cadherin, and Vimentin) under the same treatment conditions. GAPDH was used as a loading control. Numbers below the immunoblot bands indicate relative densitometric ratios normalized to the corresponding total protein in panel D or to GAPDH in panel E and expressed relative to the corresponding untreated control, which was set to 1.00.

Western blot analysis revealed that EphA2-P328 stimulation induced the phosphorylation of EphA2, Src, FAK, and AKT in BT549 and Hs578T cells, consistent with their enhanced migratory responses. In contrast, MDA-MB-231 cells did not exhibit a clear induction of EphA2 phosphorylation and showed inconsistent changes in downstream signaling components (Fig. 5D). Treatment with EphA2-P328 produced cell line- and dose-dependent changes in EMT-associated proteins. In MDA-MB-231 cells, E-cadherin and vimentin increased, whereas N-cadherin changed only modestly. In BT-549 cells, all three markers increased. In Hs578T cells, E-cadherin decreased, N-cadherin increased, and vimentin remained largely unchanged. These noncanonical patterns are consistent with context-dependent hybrid or partial EMT-associated states rather than a uniform canonical EMT transition (Fig. 5E). These findings suggest that soluble EphA2 fragments can promote migration and EMT-associated signaling in TNBC; however, the phenotypic outcomes vary depending on the intrinsic signaling context of each cell line.

Discussion

In this study, we identified the soluble N-terminal EphA2 fragment as a candidate diagnostic biomarker and demonstrated its pro-migratory activity in TNBC cells in vitro. By combining tissue-level IHC analysis, serum-based ELISA quantification, and functional assays, our findings show that EphA2 fragments are clinically detectable in TNBC and can enhance TNBC cell migration under in vitro conditions. This dual role highlights the translational relevance of soluble EphA2, positioning it as a promising candidate for liquid biopsy applications. Serum EphA2 fragments were significantly elevated in patients with TNBC compared with healthy controls, showing strong apparent discriminatory performance within the current exploratory cohort (AUC = 0.912; 95% CI, 0.862–0.962) (Fig. 4A-B). These results are concordant with our IHC analysis, which revealed higher EphA2 expression in advanced-stage and higher-grade TNBC tumors, supporting a positive association with aggressiveness (Fig. 1A-D). Consistent with its association with aggressive clinicopathological features, elevated soluble EphA2 levels were associated with poorer overall survival in both the unadjusted survival analysis and the multivariable categorical model. However, the adjusted estimate was imprecise, and the substantial overlap between EphA2 group and nodal status precludes definitive conclusions regarding its independent prognostic value (Fig. 4G). Our findings are consistent with reports in pancreatic cancer, where soluble EphA2 fragments were detected in patient blood and correlated with disease status [6,13], and in ovarian tumors where activated EphA2 processing by MT1-MMP was linked to malignant transformation [14]. Our results are also consistent with findings in bone sarcomas, where EphA2 expression has been linked to metastasis, chemoresistance, and stemness, and is currently under exploration as both a diagnostic and therapeutic target [21]. Our data extend these findings to breast cancer and highlight the clinical potential of serum EphA2 as a non-invasive biomarker for TNBC detection and risk stratification.

We developed a dual-epitope sandwich ELISA that showed concentration-dependent detection and linearity over the tested range and applied it for exploratory measurement of soluble EphA2 in conditioned media and patient sera. Conventional serum breast cancer markers have demonstrated relatively modest diagnostic performance. In a recent large-scale multicenter study, CA15-3 and CEA yielded AUCs of 0.641 and 0.733, respectively, in the validation cohort [22]. Although serum EphA2 yielded an AUC of 0.912 in our TNBC cohort, differences in study populations and assay designs preclude direct comparison, and independent head-to-head validation is required. The current study integrates an exploratory clinical cohort assessment with in vitro evidence of functional activity of an EphA2 fragment, supporting further translational evaluation. The use of two spatially distinct EphA2 epitopes, together with the EphA2-dependent signal changes observed in conditioned media from genetically manipulated cells, supports assay selectivity in a complex biological matrix. Compared with liquid biopsy modalities such as ctDNA or CTCs, which often suffer from technical variability and limited sensitivity in TNBC [5], protein-based assays offer a complementary approach that is practical for longitudinal monitoring. Recent studies also demonstrated that N-terminal fragment of EphA2 (EphA2-NF) serves as an early marker in pancreatic ductal carcinoma, further reinforcing the feasibility of extending such assays into clinical diagnostics beyond breast cancer [23].

Although recombinant EphA2-P328 promoted migration and induced phosphorylation of downstream signaling molecules in all three tested TNBC cell lines, the magnitude of response varied, with MDA-MB-231 cells showing relatively modest migration enhancement and inconsistent activation of downstream effectors (Fig. 5). Taken together, the heterogeneous migratory and EMT-associated responses observed across TNBC cell lines point to subtype-dependent baseline signaling states rather than a uniform response to soluble EphA2 fragments. This observation underscores the intrinsic functional heterogeneity of TNBC and indicates that the biological effects of soluble EphA2 fragments are highly context-dependent. Accordingly, soluble EphA2 may be particularly relevant in EphA2-high, mesenchymal-like TNBC contexts, whereas its utility across basal-like subtypes remains to be established. The variability may reflect the intrinsic heterogeneity of TNBC subtypes, as MDA-MB-231 cells exhibit a distinct mesenchymal-like phenotype and constitutive activation of several signaling pathways, including high basal Src and FAK activity, which may dampen the additive effect of exogenous EphA2 fragments [2,24]. Additionally, analysis of EMT-associated markers revealed heterogeneous responses among the TNBC cell lines. MDA-MB-231 and BT-549 cells showed increased E-cadherin and vimentin expression, whereas N-cadherin changed only modestly in MDA-MB-231 cells but increased in BT-549 cells. In Hs578T cells, E-cadherin decreased, N-cadherin increased, and vimentin remained largely unchanged. These observations are consistent with a hybrid or partial EMT-associated state rather than a complete EMT transition, reflecting the phenotypic plasticity increasingly recognized in aggressive TNBC [25]. However, changes in these protein markers alone are insufficient to establish a complete EMT program or increased invasive capacity, which would require additional transcriptional and functional validation. Accordingly, differences in baseline EMT status and signaling plasticity are likely to contribute to the variable functional responses to soluble EphA2 fragments observed among TNBC cell lines. For instance, constitutive Raf-MAPK activation can negatively feedback to suppress Ras-PI3K-AKT via EphA2 modulation [26]. Moreover, EphA2 has been shown to drive EMT in TNBC under conditions of high extracellular matrix stiffness, indicating that EphA2 is mechanosensitive in regulating metastatic behavior [27]. In line with this concept, previous studies have shown that intrinsic cellular heterogeneity, including distinct biophysical properties at the single-cell level, can contribute to metastatic potential [28].

Prior studies have highlighted the complex roles of EphA2 signaling, where its downstream effects vary depending on the phosphorylation status, interaction with co-receptors, or cleavage-derived fragments, potentially explaining the divergent responses observed across different TNBC subtypes [29,30]. These findings support the hypothesis that cleaved EphA2 fragments may engage in autocrine signaling and promote tumor progression, but the biological outcome is shaped by the cellular background, pre-existing signaling activity, and EMT status. Importantly, genetic knockdown of EphA2 abolished EphA2-P328-induced migration, establishing EphA2 as a necessary mediator of the functional activity of soluble EphA2 fragments (Fig. 5C). The response was observed without exogenous ephrin stimulation and therefore did not reflect experimentally imposed canonical ligand-dependent signaling; however, involvement of endogenous ephrin and direct fragment-receptor binding cannot be excluded. These findings highlight the importance of considering subtype-specific cellular contexts when evaluating the biological effects of soluble EphA2 fragments, which nevertheless demonstrate significant pro-migratory activity and diagnostic relevance in multiple TNBC models.

Despite these promising findings, several limitations should be acknowledged. The cohort was modest in size and derived from a single center; therefore, the observed ROC performance may be optimistic and requires validation in larger, independent cohorts. Although the conditioned-media experiments support assay selectivity, cross-reactivity with a complete panel of soluble Eph-family proteins was not directly evaluated in human serum and therefore cannot be fully excluded. Formal analytical validation of the assay, including determination of the LOD and LLOQ, intra- and inter-assay precision, and spike-recovery in serum, was not performed. Individual haemolysis information was also unavailable for retrospective evaluation. Therefore, the present ELISA should be regarded as a research-use platform for exploratory biomarker assessment rather than a clinically validated diagnostic assay. Comprehensive analytical validation will be required before clinical application. The survival cohort was also limited in size, and serum EphA2 group substantially overlapped with nodal status, resulting in a wide confidence interval for the adjusted hazard ratio. Therefore, the multivariable prognostic finding should be regarded as exploratory and requires validation in a larger independent cohort with a prespecified biomarker threshold.

In addition, while our in vitro functional assays and EphA2 knockdown experiments support EphA2-dependent pro-migratory activity in vitro for soluble EphA2 fragments in promoting TNBC cell migration, in vivo validation will be essential to confirm their contribution to tumor progression within the physiological tumor microenvironment. The mechanistic basis of how soluble EphA2 fragments engage membrane-bound EphA2 or potential co-receptors also remains to be fully elucidated. Importantly, serum represents a diluted systemic compartment; local pericellular EphA2 concentrations generated by tumor-cell shedding are therefore the more relevant comparator for the functional assays, although these local concentrations were not directly measured. Previous studies have shown that EphA2 cleavage can reprogram signaling from tumor-suppressive to oncogenic outputs depending on cellular context [12,13], and recent preclinical data demonstrate that aberrant EphA2 expression mediates EMT and metastatic progression in response to elevated extracellular matrix stiffness in breast cancer models, reinforcing the context-dependent nature of EphA2 signaling [27]. Collectively, these findings suggest that EphA2 is a highly context-dependent regulator, with functional effects likely influenced by locally elevated concentrations within the tumor microenvironment, and future studies integrating in vivo models, receptor interaction analyses, and dose-response analyses under more physiologically relevant conditions may further enhance the precision and translational value of soluble EphA2-based diagnostics in TNBC.

CRediT authorship contribution statement

Tsai-Yu Lin: Validation, Investigation, Formal analysis, Data curation. Fu-Ling Chang: Methodology, Investigation, Formal analysis, Data curation. Shang-Ju Hsieh: Methodology, Formal analysis, Data curation. Jia-Huei Zheng: Investigation, Formal analysis. Yen-Chen Lu: Validation, Investigation. Ming-Hui Cheng: Methodology, Formal analysis. Keng-Chang Tsai: Visualization, Methodology, Formal analysis. Wei-Chun HuangFu: Project administration, Conceptualization. Ching-Sheng Hung: Conceptualization. Yu-Ching Lee: Writing – review & editing, Writing – original draft, Supervision, Project administration, Funding acquisition, Conceptualization.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.