Clinical Significance of Tumor Invasion Gene Profiling in Early‐Stage Hormone Receptor‐Positive Breast Cancer: A Cross‐Sectional Study

ABSTRACT

Background and Aims

Advanced breast cancer is a leading cause of cancer‐associated death in women. It is critical to find biomarkers to identify patients who are more prone to progress to advanced disease and would be better served with more aggressive treatments. Hormone receptor‐positive breast cancers make up about 70% of all breast cancers. The analysis of gene amplification and expression has been utilized to classify breast tumors into medically useful risk groups. In the present study, we selected five genes (MMP11, CTSV, CD68, SCUBE2, and MYBL2) that are important in tumor invasion and metastasis to identify a pattern associated with well‐established pathology markers indicative of poor prognosis in early‐stage hormone receptor‐positive breast cancers.

Methods

Two hundred and thirty‐nine early‐stage hormone receptor‐positive breast cancer patients were analyzed. Patients were required to have FFPE (formalin‐fixed paraffin‐embedded) samples from the primary tumor. Expression of the selected genes was determined using TaqMan expression assays. MYBL2 amplification was examined using fluorescent in situ hybridization (FISH).

Results

Expression of CTSV was significantly associated with increasing tumor size (p = 0.001) and lymph node involvement (p = 0.016). In multivariate analysis, CTSV expression remained significantly associated with lymph node involvement (p = 0.049). Expression of MYBL2 was associated with higher tumor grade (p = 0.001). We identified MYBL2 amplification in one sample. Expression of SCUBE2 was inversely associated with tumor grade (p = 0.04). Receiver‐operator characteristic (ROC) analysis was used to predict lymph node involvement using clinicopathologic features (age at diagnosis, tumor size, and histologic grade), and the five gene expression markers, which resulted in AUC = 0.83.

Conclusion

We found a significant association between CTSV, MYBL2, and SCUBE2 expression and pathological variables of poor prognosis. Our study reveals that analysis of the expression of these genes may help identify early‐stage hormone receptor‐positive breast tumors that are more likely to metastasize and need more intense treatment modalities.

1. Introduction

Globally, breast cancer is the leading malignancy diagnosed among women. Breast cancer is a heterogeneous disease, and multiple subtypes have been identified. Hormone receptor‐positive breast cancers, which express either estrogen receptor (ER) or progesterone receptor (PR), make up about 70% of all breast cancers [1]. Advances in the understanding of the biological heterogeneity of breast cancer have driven the development of novel therapeutic strategies, resulting in substantial improvements in patient survival. Despite all these developments, advanced breast cancer continues to be a major cause of cancer‐related death among women worldwide [2, 3]. Therefore, finding biomarkers to identify patients who are more prone to progress to advanced disease and would be better served with more aggressive treatments is critical.

The analysis of genes important in the invasion and metastasis pathways in hormone receptor‐positive breast tumors can help identify patients more likely to progress to advanced disease and develop prognostic biomarkers. In routine clinical practice, ER, PR (which are markers of ER signaling pathway), HER2 (which is a marker of HER2 signaling pathway), and ki67 (which is a marker of cancer cell proliferation pathway) are analyzed for every breast cancer patient by immunohistochemistry (IHC) [4]. Still, no marker of cellular invasion and metastasis is routinely checked [5]. The analysis of gene expression has been used to classify breast tumors into clinically useful risk groups [6, 7]. Several commercial breast cancer gene expression signatures have been developed, including the 21‐gene signature (Oncotype DX) [8], the 70‐gene signature (MammaPrint) [9], and the 50‐gene signature (PAM50) [10].

Because of the high costs, these tests are unavailable in most low‐ to middle‐income countries. In order to develop an in‐house test, we selected five genes (MMP11, MYBL2, SCUBE2, CTSV, and CD68) that are involved in tumor invasion and metastasis. These genes are incorporated in multiple breast cancer gene expression profiles: MMP11 (Oncotype DX, PAM50), MYBL2 (Oncotype DX, PAM50), SCUBE2 (Oncotype DX, MammaPrint), CTSV, and CD68 (Oncotype DX). All of these genes are included in the Oncotype DX test, which examines the expression of 16 genes. This test is validated to predict the likelihood of chemotherapy benefit in early‐stage, hormone receptor‐positive breast cancer [8, 11, 12]. The five‐gene assay, compared to commercial gene signatures, has reduced costs, a more simplified design, and potential for local adaptation.

Matrix metalloproteinases (MMPs) are zinc‐dependent endoproteases that can degrade extracellular matrix (ECM) proteins. MMPs are the most significant protease family involved in tumorigenesis, invasion, and metastasis [13, 14]. MMP11 is anchored to the cell membrane, and its expression in invasive breast tumors is higher than in normal tissue [15]. MMP11 expression is restricted to the stromal cells surrounding invasive breast cancer cells, but not the in situ component of the breast carcinomas [16]. Cysteine proteinases (cathepsins) are proteolytic enzymes within the lysosomes that are important in extracellular matrix degradation during tumor invasion and metastasis [17]. CTSV (that encodes cathepsin V, also known as cathepsin L2), is overexpressed in breast carcinomas [18], and its elevated expression is associated with adverse prognosis. CTSV is expressed in both breast cancer cells and the surrounding stroma [19].

Tumor‐associated macrophages (TAMs) regulate tumor cells and their microenvironment, and are important in controlling tumor growth and response to therapy [20, 21, 22, 23]. CD68 is considered a marker of TAM, and its expression is associated with reduced survival in hormone receptor‐positive breast cancers [24, 25]. Signal peptide, CUB domain, and EGF‐like domain‐containing 2 (SCUBE2) protein is a member of the evolutionary conserved SCUBE gene family. When overexpressed, it is a secreted glycoprotein that remains stably associated with the cell surface. SCUBE2 overexpression in invasive breast carcinomas is associated with better prognosis through inhibition of breast cancer cell migration and invasion [26, 27]. MYBL2 amplification is observed in 9%–13% of breast cancers and drives abnormal cell proliferation and invasion [28]. MYBL2 overexpression disrupts cell adhesion and promotes epithelial‐mesenchymal transition (EMT), increasing the metastatic potential of breast tumors [29].

In the present study, we analyzed the expression of five genes to identify a gene expression pattern associated with well‐established pathology markers indicative of poor prognosis in early‐stage hormone receptor‐positive breast cancers. We also analyzed the amplification of MYBL2 gene by Fluorescence In Situ Hybridization (FISH) method.

2. Methods

2.1. Study Cohort

We screened 276 hormone receptor‐positive breast cancer patients who received treatment in several oncology clinics and hospitals in Tehran (including Tajrish Shohada Hospital, Mehrad Hospital, and Erfan Hospital), Ahvaz, Kermanshah, and Karaj from February 2022 to October 2025. The study was a case series. Inclusion criteria were female breast cancer patients over the age of 18 with a diagnosis of invasive hormone receptor‐positive early‐stage breast cancer. Early‐stage breast cancer was defined as non‐metastatic cancer with 0–3 axillary lymph node involvement. Patients were required to have FFPE (formalin‐fixed paraffin‐embedded) samples from the primary tumor. Exclusion criteria were ER‐negative, ER‐low (< 10% stained cells), HER2‐positive tumors, prior neoadjuvant chemotherapy and the absence of invasive breast carcinoma in the available FFPE tissue blocks. The institutional review board of cancer research center, Shahid Beheshti University of Medical Sciences, approved the study (IR.SBMU.CRC.REC.1403.007), and informed consent was obtained from all patients in accordance with the recommendations of our Institutional Review Board (IRB). The study specimens had previously undergone immunohistochemical (IHC) evaluation for ER, PR, HER2, and Ki67. Clinical and pathological variables analyzed included age of cancer diagnosis, tumor histology, grade, focality, size, axillary nodal status, lympho‐vascular invasion, tumor stage, ER, PR, HER2, and Ki67 IHC status. Fluorescence In Situ Hybridization (FISH) was conducted on samples that yielded a 2+ score in HER2 IHC testing.

2.2. Fluorescence In Situ Hybridization (FISH)

FISH for HER2 (ERBB2) gene was performed using HER2 dual probe kit (CytoCell, Oxford Gene Technology Inc, UK). HER2 and centromere 17 signals were analyzed using ASI HiFISH software (Applied Spectral Imaging, CA, USA). Signal quantification was performed across 40 nuclei sampled from the invasive breast carcinoma. HER2 amplification status was determined in accordance with the 2023 American Society of Clinical Oncology/College of American Pathologists (ASCO/CAP) guidelines [30].

FISH for MYBL2 gene was performed using MYBL2 FISH probe (Empire Genomics, New York, USA). MYBL2 signals were analyzed using ASI HiFISH software. A total of 40 nuclei from invasive breast tumor cells were evaluated for signal enumeration. Samples with an average of ≥ 6 MYBL2 signals per cell were considered MYBL2 amplified, while having 3‐5 copies was defined as copy number gain.

2.3. RNA Extraction and Purification

A representative section from each specimen was stained with hematoxylin and eosin (H&E) to identify the invasive tumor component and assess the proportion of tumor cells within the tissue. Samples containing < 50% invasive tumor content underwent macrodissection to increase purity. Total RNA was subsequently isolated from three 10‐µm formalin‐fixed paraffin‐embedded (FFPE) sections using the Qiagen RNeasy FFPE kit (QIAGEN, Hilden, Germany), following the manufacturer's protocols. Briefly, tumor sections underwent dewaxing in xylene (Merck Millipore, Germany). RNA release was initiated via incubation at 56°C in a lysis buffer supplemented with proteinase K, followed by an 80°C incubation step to reverse formalin‐induced crosslinking. Genomic DNA was removed through DNase treatment. The resulting lysate was processed using an RNeasy MinElute spin column, where RNA was captured on the membrane, washed of contaminants, and finally eluted in RNase‐free water. RNA concentration and purity were assessed using a Qubit RNA High Sensitivity kit (Thermo Fisher Scientific, CA, USA) and a NanoDrop spectrophotometer (Thermo Fisher Scientific, OR, USA), respectively. To ensure the absence of residual DNA, the TaqMan ACTB assay was utilized, incorporating genomic DNA from breast tissue as a positive control and a no‐template sample as a negative control. Any samples exhibiting DNA contamination underwent secondary treatment with DNase I (DNA‐free kit, Thermo Fisher Scientific, OR, USA) before re‐assessment.

2.4. Reverse Transcription and Real‐Time PCR

Complementary DNA (cDNA) was synthesized from the extracted RNA using the Omniscript RT kit and random hexamers (Qiagen, Hilden, Germany), with reaction mixtures incubated at 37°C for 60 min. TaqMan gene expression assays (Thermo Fisher Scientific, Pleasanton, CA) were used for expression analysis of MMP11, CTSV, CD68, SCUBE2, and MYBL2, with normalization performed using five established reference genes for breast tissue: ACTB, GAPDH, GUSB, RPLP0, and TFRC [31]. (Supporting Table 1).

Triplicate quantitative real‐time PCR (qPCR) reactions were carried out using the QuantStudio 3 system (Thermo Fisher Scientific, CA, USA). Each 20 µL amplification reaction utilized the qPCRBIO Probe Mix Lo‐ROX (PCR Biosystems Ltd., London, UK). The thermal cycling profile consisted of an initial denaturation at 95°C for 2 min, followed by 40 cycles of 95°C for 5 s and 60°C for 25 s. The ΔCt (cycle threshold) method was used to calculate gene expression, with ΔCt defined as the difference between the Ct values of the reference genes and the target gene (Ct_reference − Ct_target). Normalization was based on the mean Ct of the five reference genes using QuantStudio 3 software.

2.5. Statistical Methods

Descriptive statistical analyses were performed to characterize the study population, and summarize baseline demographic and clinical variables. The normality of continuous variables was assessed using the Kolmogorov–Smirnov test to guide the selection of appropriate parametric or non‐parametric statistical methods. All statistical tests were conducted using two‐sided hypotheses. A p‐value of < 0.05 was considered statistically significant, indicating a meaningful association between mRNA expression and clinicopathological factors.

The following clinicopathological characteristics were dichotomized: age at cancer diagnosis ≤ 50 versus > 50 (to assess its impact on mRNA expression), tumor size (to evaluate its relationship with expression levels), histologic grade (to determine if expression varies by tumor differentiation), lymph node status (to examine its prognostic significance), and histologic subtype (to compare between different breast cancer types).

The relationship between mRNA expression levels and clinicopathological variables was evaluated using the independent‐samples t‐test for normally distributed data to compare mean expression levels between groups and the Wilcoxon rank‐sum test for non‐normally distributed data to evaluate median differences between groups. Bonferroni correction was used to compensate for multiple comparisons.

Statistical analysis was performed using SPSS version 27.0.1 (IBM Corp., Armonk, NY) and R Statistical Software version 4.3.1 (Vienna, Austria).

2.6. Machine Learning Analysis

To identify the optimal predictive model for nodal involvement, 12 supervised machine learning algorithms were developed and compared, including Random Forest, XGBoost, Logistic Regression, SGD Classifier, Decision Tree, Extra Trees, Gradient Boosting, AdaBoost, LightGBM, Multi‐Layer Perceptron (MLP), k‐Nearest Neighbors (kNN), and Naive Bayes classifiers. All models were implemented using Python (scikit‐learn and imbalanced‐learn). The dataset was randomly split into training (80%) and test (20%) sets, maintaining class balance through stratified sampling. Because the number of node‐positive cases was smaller than the number of node‐negative cases, the SMOTEENN hybrid resampling technique was applied to the training data. SMOTEENN combines Synthetic Minority Oversampling (SMOTE) with Edited Nearest Neighbor (ENN) undersampling to both oversample the minority class and remove ambiguous samples near decision boundaries, enhancing sensitivity and generalization. Each model was trained using a pipeline integrating the resampling step and classifier. Hyperparameters were optimized using GridSearchCV with five‐fold stratified cross‐validation, with the F₂‐score as the optimization metric (a metric that weights recall (sensitivity) higher than precision to prioritize identification of node‐positive patients). The F₂‐score was defined as:

$$F_{\beta }=\frac{(1+{\beta }^2)\times (\text{Precision}\times \text{Recall})}{({\beta }^2\times \text{Precision})+\text{Recall}},\beta =2$$

This metric was used as the primary optimization criterion. For each model, the best‐performing hyperparameter combination was identified, and the final model was evaluated on the held‐out test set. Performance metrics included sensitivity (recall), specificity, F₂‐score, F₁‐score, and ROC‐AUC. A confusion matrix was also generated for each model to visualize classification outcomes (true positives, false positives, true negatives, and false negatives).

3. Results

3.1. Patient Characteristics

Out of 276 initially identified patients, 242 met the study selection criteria, with 239 ultimately included in the final analysis. Two patients were excluded after initial selection: one due to the presence of ductal carcinoma in situ (DCIS) only without invasive breast tumor in the FFPE block, and one due to insufficient RNA yield (Supporting Figure 1). One sample was ER‐low‐positive; therefore, it was excluded from the study. All tumor samples were obtained from surgical resections, and the corresponding FFPE blocks had been stored for 1 to 6 months prior to analysis. Patient characteristics are summarized in Table 1. Patients had a median age of 50 years at breast cancer diagnosis (IQR, 43–60 years). Breast cancer was unilateral and unifocal in 219 patients (91.6%). Two hundred and fourteen (89.5%) patients presented with invasive ductal carcinoma, and seventeen (7.1%) presented with invasive lobular carcinoma. 230 tumors (96.2%) were PR‐positive, 8 tumors (3.3%) were PR‐low‐positive, and 9 tumors (3.8%) were PR negative.

Table 1.

Clinical and Pathological characteristics of the patients (N = 239).

Variable	No.	%
Age at breast cancer diagnosis
Median (IQR)	50 (43–60)
≤ 50	126	52.7
> 50	113	47.3
Tumor histology
IDC	214	89.5
ILC	17	7.1
Mucinous	5	2.1
Other*	3	1.3
Tumor focality
Unifocal	219	91.6
Multifocal	17	7.1
Bilateral	3	1.3
Tumor grade
1	53	22.2
2	169	70.7
3	14	5.9
Missing	3	1.3
Tumor size (cm)
≤ 1	39	16.3
> 1–2	91	38.1
> 2–4	102	42.7
> 4	4	1.7
Missing	3	1.3
Nodal status
0	216	90.4
1–3	23	9.6
Lympho‐vascular Invasion
Present	39	16.3
Absent	197	82.4
Missing	3	1.3
Tumor infiltrating lymphocytes (TIL)
< 20%	104	43.5
20–50%	25	10.5
> 50%	1	0.4
Missing	109	45.6
Stage
I	121	50.6
II	113	47.3
III	2	0.8
Missing	3	1.3
ER IHC Nuclei Staining
10–50%	8	3.3
> 50%	219	91.6
ER staining percentage missing (reported as ER positive)	12	5
PR IHC Nuclei Staining
< 1%	9	3.8
1–10%	8	3.3
11–50%	58	24.3
> 50%	153	64
PR staining percentage Missing (reported as PR positive)	11	4.6
HER2 IHC Score
0	137	57.3
1	61	25.5
2	39	16.3
3	0	0
HER2 score missing (reported as HER2 Negative)	2	0.8
Ki67 IHC Staining
≤ 5%	20	8.4
6–29%	170	71.1
≥ 30%	42	17.6
Missing	7	2.9

Abbreviations: ER, Estrogen receptor; FISH, Fluorescence In Situ Hybridization; HER2, Human epidermal growth factor receptor 2; IDC, Invasive ductal carcinoma; IHC, Immunohistochemistry; ILC, Invasive lobular carcinoma; IQR, Interquartile range; PR, Progesterone receptor.

*Invasive apocrine cancer, Invasive micropapillary cancer, Invasive tubular cancer.

3.2. FISH Analysis

HER2 FISH was performed for 40 samples with a 2 + HER2 IHC. One specimen demonstrated HER2 amplification by FISH and was therefore excluded from the analysis. A representative case with a normal HER2 copy number is shown in Figure 1A.

Figure 1.

(A) Illustration of two‐color FISH (fluorescent in situ hybridization) analysis of HER2, representing centromere 17 (green) and HER2 (ERBB2) gene (red) of a sample without amplification, and (B) FISH analysis of MYBL2 gene of a tumor exhibiting MYBL2 copy number gain.

MYBL2 FISH was performed for all samples. We identified one sample with MYBL2 amplification, and four cases with MYBL2 copy number gain. A representative case with MYBL2 copy number gain is depicted in Figure 1B.

3.3. Differential Expression of the Selected Genes in Tumor Samples

The association of expression levels of the selected genes and pathological characteristics is presented in Figure 2. Expression levels of CTSV gene were significantly associated with increasing tumor size (p = 0.001) and lymph node involvement (p = 0.016). Expression levels of MYBL2 gene were associated with higher tumor grade (p = 0.001). Expression levels of SCUBE2 gene were inversely associated with tumor grade (p = 0.04). Higher expression of CTSV and MMP11 was associated with a trend toward a greater frequency of lympho‐vascular invasion, although the association was not statistically significant. Multivariable logistic regression analysis of all five genes and age demonstrated that lymph node involvement was significantly associated with CTSV gene expression (p = 0.049) and age (p < 0.001) (Figure 3). The percentage of Ki67‐positive tumor cells was moderately correlated with MYBL2 mRNA expression (r = 0.43, p < 0.001). (Figure 4, Supporting Figure 2). Table 2 summarizes the gene functions and their clinicopathological associations.

Figure 2.

Boxplot of the association between gene expression and tumor pathologic characteristics. (A) mRNA ΔCt values/tumor size > 2 cm or ≤ 2 cm; (B) mRNA ΔCt values/tumor grade; (C) mRNA ΔCt values/Lymph node involvement; (D) mRNA ΔCt values/lympho‐vascular invasion. Ct, Cycle threshold; LVI, Lympho‐vascular invasion.

Figure 3.

Multivariable analysis of the association of the expression of all five genes (CTSV, MMP11, MYBL2, SCUBE2, and CD68) and (A) Lymph node involvement, (B) lympho‐vascular invasion, (C) tumor grade, and (D) tumor size. LVI, Lympho‐vascular invasion

Figure 4.

Scatter plots of MYBL2 mRNA ΔCt values and ki67 IHC. Ct, Cycle threshold; IHC, Immunohistochemistry.

Table 2.

The function and clinical association of the selected genes (Univariate analysis).

Gene	Function	Association with pathological markers
CTSV	Extracellular matrix degradation during tumor invasion and metastasis	Increased tumor size, More lymph node involvement
MMP11	Extracellular matrix degradation during tumor invasion and metastasis	None
MYBL2	Abnormal cell proliferation and invasion, cell adhesion disruption and epithelial‐mesenchymal transition promotion	Increased tumor grade
SCUBE2	Inhibition of tumor cell migration and invasion	Decreased tumor grade
CD68	CD68 is considered a marker of TAMs, which regulate tumor cells and their microenvironment	None

Abbreviation: TAMs, Tumor‐associated macrophages.

A post‐hoc power analysis was performed to determine the statistical power for detecting the observed difference in CTSV gene expression between lymph node‐negative and lymph‐node‐positive patients. Based on the observed mean ∆Ct difference of 0.847, the respective standard deviations (1.503 and 0.966), and a significance level (α) of 0.05, the achieved power of the study was 96.3%.

3.4. Receiver‑Operator Characteristic Analyses to Predict Lymph Node Involvement

Receiver‐operator characteristic (ROC) analysis was used to predict lymph node involvement using standard clinicopathologic features (age at cancer diagnosis, tumor size, and histologic grade), and five gene expression markers (CD68, CTSV, MMP11, MYBL2, and SCUBE2). ROC analysis to predict lymph node involvement using the five gene expression model resulted in AUC = 0.743. ROC analysis to predict lymph node involvement using the clinical model resulted in AUC = 0.777. Integrating the clinicopathological model with the five‐gene model increased the AUC to 0.83 (Figure 5).

Figure 5.

Receiver‐operator characteristic (ROC) analysis. Red Line. ROC analysis for predicting lymph node involvement using standard clinicopathologic features (age at cancer diagnosis, tumor size, and histologic grade), AUC = 0.777. Green Line. ROC analysis for predicting lymph node involvement using the five gene model, AUC = 0.743. Blue Line. ROC analysis for predicting lymph node involvement, clinicopathological model with the five‐gene model, AUC = 0.83.

3.5. Evaluation of Machine Learning Models in Predicting Nodal Status

Among the 12 machine learning models trained on 190 samples, the k‐nearest neighbors (kNN) model achieved the best performance, with an area under the ROC curve (AUC) of 0.90. This model was developed to predict lymph node involvement in hormone receptor–positive early breast cancer using three clinicopathologic features (age at cancer diagnosis, tumor size, and histologic grade) and five gene expression markers (CD68, CTSV, MMP11, MYBL2, and SCUBE2). The kNN model demonstrated excellent sensitivity (100%), but moderate specificity (45.5%) on the test set with 49 patients. The ROC‐AUC score was 0.898, indicating strong overall discriminative ability. The confusion matrix and ROC curve are presented in Figure 6.

Figure 6.

Machine learning analysis for predicting lymph node involvement. (A) Receiver‐operator characteristic (ROC) analysis, (B) Confusion matrix.

4. Discussion

Univariate analysis of the present study samples demonstrated a significant association between expression levels of CTSV (encoding cathepsin V), MYBL2 (encoding myb‐related protein B), and SCUBE2 (encoding signal peptide, CUB and EGF‐like domain‐containing protein 2) mRNAs and the clinicopathological characteristics of early‐stage hormone receptor‐positive breast tumors. CTSV exhibited significantly higher expression levels in breast tumors with larger size and lymph node involvement, and there was a trend toward more lympho‐vascular invasion. Toss et al. analyzed breast ductal carcinoma in situ (DCIS) specimens with immunohistochemistry for cathepsin V protein. Cathepsin V expression was associated with larger DCIS size. DCIS associated with invasive breast cancer showed higher cathepsin V expression compared to DCIS alone. They showed that higher expression of cathepsin V is associated with poor prognosis, which is similar to our findings regarding CTSV mRNA expression associated with poor prognostic pathology markers [22]. The interaction between cancer cells and their surrounding stroma is among the hallmarks of cancer [32]. Cathepsin V exhibits extensive elastase activity and is expressed in human activated macrophages [33], although its expression in cancerous tissue macrophages has not been documented in breast cancer. Toss et al. showed that cathepsin V is expressed in the cytoplasm of breast cancer cells and cancer‐associated fibroblasts (CAFs) [19]. Sun et al. analyzed the expression of different cathepsins in breast tumors and reported that cathepsin V expression is associated with distant metastasis [34]. Sereesongsaeng et al. showed that cathepsin V silencing by shRNA significantly reduced tumor cell invasion in hormone receptor‐positive breast cancer cell lines [35]. Cathepsin V directly affects the basement membrane and extracellular matrix degradation, facilitating tumor cell invasion into the surrounding stroma. Our findings highlight the interplay between tumor cells and the surrounding stromal microenvironment during tumor progression, in which CTSV gene (which encodes cathepsin V) may have a significant role. Multivariable analysis of the five gene expression model showed that CTSV expression analysis can predict axillary lymph node involvement.

Matrix metalloproteinases (MMPs) are a family of zinc‐dependent endopeptidases capable of degrading components of the extracellular matrix to promote invasion and metastasis of tumor cells [36]. This study did not identify a significant association between lymph node involvement and MMP11 expression. Previous studies have shown that overexpression of MMP11 is associated with poor survival [37]. Matrix metalloproteinase 11 is expressed in stromal cells of invasive breast tumors. While most MMPs are secreted as inactive enzymes, MMP11 is processed in the cytoplasm and secreted as active, making it important in tumor invasion and metastasis [38]. Cheng et al. showed that overexpression of MMP11 is associated with poorly differentiated breast cancers and lymph node metastasis [15]. We did not find a significant association between MMP11 gene expression, tumor grade, and lymph node involvement. This may be due to the small number of lymph node‐positive tumors in our cohort. In a more recent study, Cheng et al. demonstrated that MMP11 expression is associated with CD8 + T cell and macrophage infiltration within the breast tumor [39]. In contrast, Kim et al. reported low tumor‐infiltrating lymphocytes and worse survival in breast cancers with high MMP11 expression [36]. We did not find a significant association between tumor‐infiltrating lymphocytes and MMP11 expression.

In the present study, expression of MYBL2 was concordant with ki67 IHC percentage and was associated with higher tumor grade. MYBL2 is a transcription factor important in tumor proliferation and progression [40]. MYBL2 is frequently amplified in breast tumors, and previous studies have shown that its overexpression is associated with poor prognosis [29, 41]. Although there is no clear definition for MYBL2 gene amplification in breast cancer, based on previously published studies [42, 43, 44], we defined MYBL2 amplification as the presence of ≥ 6 copies per cell and MYBL2 copy number gain as the presence of 3‐5 copies per cell. In this patient cohort, we identified MYBL2 amplification in only one sample, and three samples showed MYBL2 copy number gain. Ginestier et al. analyzed the chromosomal locus 20q13 (which includes MYBL2 gene) in 466 breast cancer samples and detected MYBL2 amplification in 4% of samples [45]. The detection of MYBL2 amplification in only a single sample in our cohort is likely attributable to tumor subtype distribution. As reported by Li et al., MYBL2 is predominantly overexpressed in non‐luminal breast cancers, particularly the basal‐like subtype, whereas its expression is typically low in luminal tumors. Given that our cohort is mainly composed of luminal cases, the low frequency of MYBL2 amplification is therefore expected [46]. The tumors included in this study cohort are all luminal breast cancers, and MYBL2 amplification may not be a common mechanism in luminal tumors. In addition, there is no standard definition for MYBL2 amplification. All four samples reported in this study, as MYBL2 copy number gain showed high copy number gain. Other studies may have different definitions of MYBL2 amplification.

We found an association between lower expression of SCUBE2 gene and higher tumor grade. Cheng et al. analyzed the expression of SCUBE2 protein in breast cancer patients and found an association with a favorable outcome [26]. Overexpression of SCUBE2 protein inhibits breast tumor cell migration and invasion by increasing epithelial E‐cadherin‐containing adherence junction and reversal of epithelial‐mesenchymal transition (EMT) [27]. Association of decreased expression of SCUBE2 gene and poor prognosis has been previously shown in other cancers, including gastric, colorectal, and brain cancers [47, 48, 49].

In the present study, we used a five‐gene model to develop a lightweight, efficient diagnostic approach using minimal materials to predict lymph node involvement in early‐stage hormone receptor‐positive breast cancer. We acknowledge the value of highly multiplexed assays in identifying prognostic/predictive biomarkers, especially when combined with machine learning and AI [50, 51]. This five‐gene panel includes genes that are incorporated in the established breast cancer gene expression assays. One gene is included in MammaPrint, two genes in PAM50, and all five genes are included in the Oncotype DX test. All of these assays have prognostic value in early‐stage hormone receptor‐positive breast cancer. In addition, the Oncotype DX test can predict the benefit of adding chemotherapy to hormonal therapy in this group of patients.

Our study has limitations. The study is cross‐sectional, assessing gene expression and pathological markers at a single time point. It does not provide information on long‐term patient outcomes (e.g., recurrence, metastasis, survival), which limits its predictive clinical usefulness. All analyses are correlative; the study does not include functional experiments (e.g., knockdown/overexpression studies, pathway analysis) to elucidate causal mechanisms behind gene expression changes and tumor aggressiveness. In addition, our study does not delve deep into the full spectrum of the tumor microenvironment (such as stromal or immune cell phenotyping). Other limitations of this study include the small number of genes compared to validated signatures and the exploratory nature of the correlations with pathology markers. In addition, independent external validation has not yet been performed.

In conclusion, we found a significant association between CTSV, MYBL2, and SCUBE2 gene expression and clinicopathological features. Our study reveals that analysis of the expression of CTSV gene can help identify early‐stage hormone receptor‐positive breast cancers that are more likely to metastasize; therefore, they need more intense treatment modalities. In addition, we found that higher MYBL2 and lower SCUBE2 gene expression are associated with more aggressive breast tumors. Integrating the clinicopathological model with the five‐gene model increased the ability of each model to predict lymph node involvement. Further studies are needed to analyze the association of this five‐gene model with survival outcomes, such as overall survival and progression‐free survival. If validated in independent cohorts, it can be clinically implemented as an in‐house low‐cost alternative to commercial expression profiling tests.

Author Contributions

Shayan Forghani: methodology, data curation, investigation, formal analysis, writing – original draft. Hamid Rezvani: conceptualization, supervision, project administration, writing – review and editing. Arman Forghani: methodology, data curation, investigation, formal analysis. Fatemeh Mahdavi Sabet: methodology, data curation, investigation, formal analysis. Atieh Akbari: conceptualization, supervision, writing – review and editing. Sanaz Tabarestani: conceptualization, methodology, investigation, supervision. Hamid Reza Mirzaee: conceptualization, writing – review and editing, supervision.

Funding

The authors have nothing to report.

Ethics Statement

The study was approved by the ethics committee of cancer research center (IR.SBMU.CRC.REC.1403.007).

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Figure S1: Patient selection diagram. This diagram provides an outline of the patients selected and those excluded due to the lack of consent, tumor FFPE block unavailable, no invasive tumor in the available FFPE block, ER‐low positive tumors, or RNA extraction not successful. FFPE: Formalin‐Fixed Paraffin‐Embedded; DCIS: Ductal carcinoma in situ.

Figure S2: Scatter plots of mRNA ΔCt values and ki67 IHC. A) CTSV mRNA; B) CD68 mRNA. C) SCUBE2 mRNA. D) MMP11 mRNA. Ct: Cycle threshold; IHC: Immunohistochemistry.

Supporting File: hsr272780‐sup‐0003‐Supplementary_Table_1.docx.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request. The datasets generated and/or analyzed during the current study are available from the corresponding author upon reasonable request.