DNA methylation profiles for breast cancer subtype classifications: A translational study from microarray to methylation-specific PCR (MSP)

Abstract

Breast cancer subtypes can be categorized based on their gene expression profiles using immunohistochemistry into Luminal A, Luminal B, human epidermal growth factor receptor 2-positive (HER₂₊), and triple-negative breast cancer (TNBC) subtypes. However, immunohistochemistry has certain limitations that can lead to misclassification. DNA methylation is an epigenetic modification, and changes in the promoter region can alter gene expression and the quantity of functional protein synthesized, disrupting gene function. The aim of this study was to identify DNA methylation biomarkers for subtype classification in breast cancer using microarray and methylation-specific polymerase chain reaction (MSP) methods. DNA samples were extracted, subjected to bisulfite conversion and then used for both the microarray and MSP methods. This study successfully identified differentially methylated CpGs (DMCs) as biomarker for each subtype classification of breast cancer: Luminal A (hypermethylation of ADAMTSL₂ gene; cg14397888), Luminal B (hypomethylation of ADAMTSL₂ gene; cg14397888), HER₂₊ (hypermethylation of PTPRN₂ gene; cg25910261), and TNBC (hypomethylation of LCLAT_l gene; cg15652532). The DMC biomarker found for the HER₂₊ subtype, hypermethylation in the PTPRN2 gene (cg25910261), has the potential to be used by healthcare providers to identify HER₂₊ patients and provide the HER2-targeted therapy to improve the patient's survival. In addition, our developed MSP method could produce an effective diagnostic tool for classifying the Luminal A and Luminal B subtypes, with accuracies of 75% and 76%, respectively.

Introduction

Breast cancer is the most prevalent cancer among women, and according to the GLOBOCAN 2022 data, breast cancer had the highest incidence rate in Indonesia, with 66,271 cases, and the number increased compared to previous years [1]. In the same year, the 5-year breast cancer prevalence was 151.3 per 100,000 people in the country, with a 9.3% mortality rate, surpassing the global mortality rate of 6.9% [1]. Currently, breast cancer subtypes can be categorized based on their gene expression profile using immunohistochemistry (IHC) [2]. The standard IHC tests performed for breast cancer include estrogen receptor (ER), progesterone receptor (PR), human epidermal growth factor receptor 2-positive (HER₂), and Ki-67 [2]. The subtypes can be classified into Luminal A, Luminal B, HER₂₊, and triple-negative breast cancer (TNBC) [2]. Distinct subcategories of breast cancer encompass numerous biologically diverse entities with unique pathological characteristics and clinical implications. Current research shows that breast cancer with varying histological and biological characteristics have unique behaviors, resulting in divergent treatment responses that require different therapeutic approaches [3].

IHC is a cost-effective and easily executable technique as it allows for visualization of staining data using a conventional bright field microscope and enables assessment of protein overexpression within the tissue's morphological context. However, the IHC also has certain shortcomings that can result in misclassification [4]. Reproducibility in Ki-67 staining can contribute to the occurrence of misdiagnosis [5]. The interpretation of the IHC relies on semiquantitative evaluations, making it prone to interobserver variability and potentially resulting in significant discrepancies in IHC outcomes [6]. A study revealed a consistent misclassification trend in the alluvial diagram. For example, half of the HER2 + cases were misclassified as basal-like and Luminal B-like and 95% of Luminal B cases were classified as Luminal A [7].

DNA methylation is an epigenetic modification that involves the attachment of a methyl group to the 5^th carbon of the nitrogen base cytosine (C), resulting in the formation of 5- methylcytosine (5-mC). Typically, DNA methylation takes place at C bases next to G bases (known as CpG), and these are commonly located in gene promoter regions [8]. DNA methylation in the promoter region could alter gene expression and the quantity of functional protein synthesized, leading to disruption of gene function [9]. Environmental factors, including exposure to cigarette smoke, food intake, occupational activities, and lifestyle, can affect DNA methylation [10].

There is a lack of research in Indonesia assessing the association between DNA methylation and the classification of breast cancer subtypes, in particular those using peripheral blood as a sample. There are two studies from Indonesia that have discovered the association between methylation of the BRCA1 gene promoter in cancerous tissues and the clinical, as well as pathology of breast cancer [11,12]. Nevertheless, there has been no previous investigation conducted in Indonesia assessing the association between DNA methylation in peripheral blood and the classification of breast cancer subtypes. An examination utilizing peripheral blood offers various advantages, primarily its non-invasive nature, which makes it a relatively easier and more cost-effective procedure. Studies conducted on Polish [13], French [14,15], Arab [16], and Japanese population [17] have demonstrated an association between DNA methylation of the BRCA1/2 promoter region in peripheral blood and susceptibility to breast cancer.

Given the present advancements in technology, the Illumina Infinium Methylation EPIC Beadchip Microarray (EPIC-array) is a widely utilized and cost-effective device for efficiently profiling DNA methylation with more than 850,000 CpG sites.[18] The use of EPIC-array for DNA methylation profiling is expected to yield DNA methylation biomarkers capable of categorizing breast cancer subtypes. Nevertheless, the utilization of this technology requires high costs and complex analysis [19], rendering it challenging to implement in regular clinical practice, particularly in healthcare facilities with more constrained resources such as in Indonesia. A translational approach towards simpler analysis is needed for larger patient benefits.

A translational approach that can be employed is methylation-specific polymerase chain reaction (MSP). Currently, there is no available polymerase chain reaction (PCR) detection method for identifying DNA methylation in breast cancer using population epigenomic data from Indonesia. Therefore, it is important to develop a new detection method specifically designed to identify DNA methylation. The aim of this study was to identify DNA methylations as biomarkers of breast cancer subtype classifications by using microarray and translated into MSP method. The MSP method acts as a substitute for sequencing or microarray technology that is more cost- effective and easier to use, making it suitable for regular clinical use.

Methods

Study design, setting and patients

A cross-sectional study was conducted using blood samples from female breast cancer patients who had already undergone therapy according to their subtypes (based on the IHC results). All patients were outpatients at Cipto Mangunkusumo National Hospital and Dharmais National Cancer Center Hospital, both located in Jakarta, Indonesia. The number of subjects was calculated using the cross-sectional formula, with 46 subjects determined as the minimum required sample size. In this study, 48 breast cancer patients were included, with 12 subjects allocated to each subtype group. The inclusion criteria required female patients with invasive breast carcinoma, aged at least 18 years, with comprehensive medical records and histopathological data. Patients with comorbidities were excluded. Clinical data were collected through participant surveys and computerized medical records.

DNA extractions and DNA isolates quality control

DNA samples were extracted from blood by isolating the buffy coat, which yielded a significantly higher amount of DNA, ranging from five to 10 times more. A high DNA concentration was required, as the DNA-bisulfite conversion process later required 200 ng. Whole blood was first centrifuged at 2500 × g for 10 minutes at room temperature (25°C). The buffy coat, located in the middle layer, was separated for DNA extraction. The DNA was extracted using the Genomic DNA Mini Kit (GeneAid, New Taipei City, Taiwan). DNA purity was assessed using an Implen Nanodrop P300 (Implen, Munich, Germany) by measuring the absorbance ratio at 260 nm and 280 nm. The concentration of double-stranded DNA was determined using the Qubit 3.0 Fluorometer (Thermo Fisher Scientific, Waltham, MA, USA) and the Qubit dsDNA BR Assay Kit (Thermo Fisher Scientific, Waltham, MA, USA).

Microarray preparations

A total of 200 nanograms (ng) of DNA isolates was subjected to bisulfite conversion using the Zymo EZ-96 DNA Methylation Kit (Zymo Research, Orange, CA, USA) prior to the microarray preparation procedure, which was performed using the Infinium MethylationEPIC v2.0 (Illumina, San Diego, CA, USA). The microarray scanning process was conducted on the Illumina iScan platform (Illumina, San Diego, CA, USA). The intensity of each single nucleotide polymorphism (SNP) in the samples was generated during the procedure. The scanning process produced files in the .idat format, which contained two types of intensity colors: red and green.

Microarray data analysis

The imported .idat file in R was examined for data quality using the Minfi program (https://bioconductor.org/packages/release/bioc/html/minfi.html) [20].The p-value parameter was utilized for each CpG probe signal in every sample. The p-value was determined by comparing the combined signal of methylated and unmethylated DNA (M + U) to the background signal. Prior to further analysis, the data were normalized using the “preprocessFunnorm” function from the Minfi package. Upon completion of the normalization process, the data were transformed into Beta values. The Beta value was calculated as the ratio of the intensity of methylated probes to the total intensity (the sum of the intensities of methylated and unmethylated probes), ranging from 0 (fully unmethylated) to 1 (fully methylated) [21]. The Limma tool, developed in R/Bioconductor (https://www.bioconductor.org/packages/release/bioc/html/limma.html) [22], was used to identify differentially methylated CpGs (DMCs), which are locations where methylation differences occur. DMCs were defined as CpG probes with an adjusted p < 0.05, exhibiting hypermethylation with a log2 fold change greater than 1 or hypomethylation with a log2 fold change less than -1 [23]. A CpG probe with a Beta value greater than or equal to 0.3 was classified as methylated, while a Beta value below 0.3 was classified as unmethylated [24].

Comparison of differentially methylated CpGs (DMCs) to other datasets

Given the limited sample size for each subtype in this study, we then compared the identified DMCs with relevant research findings from Gene Expression Omnibus (GEO) dataset. Three distinct datasets were utilized: GSE141338 (consisting of 45 breast cancer subjects), GSE72245 (comprising 118 breast cancer subjects), and GSE72251 (encompassing 119 breast cancer subjects). These datasets contained the DNA methylation profiles of breast cancer patients categorized into four distinct subtypes: Luminal A, Luminal B, HER2+, and TNBC. The Illumina Infinium 450 K Human Methylation BeadChip was employed for the analysis of the datasets. Shared DMCs were identified across the three GEO datasets and the study datasets were extracted using a Venn diagram tool (https://www.biotools.fr/misc/venny).

Survival analysis

Clinical outcomes related to DMCs found in each subtype were then analyzed by using Kaplan- Meier plotter database (https://kmplot.com/analysis/index.php?p = service&cancer = breast) [25] . The Kaplan-Meier curve correlates the gene expression in each subtype with time and the percent of relapse-free survival. The Kaplan-Meier plotter database used gene expression data and relapse-free survival information from GEO and European Genome-Phenome Archive [25]. Molecular subtypes were identified using the StGallen criteria, which are based on estrogen and ERBB2 expression [25]. For patient assignment, the gene expression threshold of 500 for probe set “205225_at” was used to determine estrogen positivity, and the probe set “216836_s_at” with a cutoff of 4800 was determined to assign patients into HER2 positive/negative groups [25]. The total number of breast cancer arrays was 9,423, including 7,830 distinct samples from 55 independent datasets [25]. Relapse-free survival data was available for 5268 patients [25].

In this study, the number of patients used for analysis from Luminal A, Luminal B, HER2+, and TNBC were 2,470, 2,005, 1,080, and 1,671 patients, respectively, provided from the Kaplan- Meier plotter database [25]. All the patients involved in this analysis were given systemic treatments (endocrine therapy and chemotherapy, without targeted therapy) [25]. A significant difference was defined as those with a p < 0.05, using the median as the group cut-off. The hazard ratio (HR) was calculated with a 95% confidence interval (95%CI) to determine the statistical association between gene expression and relapse-free survival.

Primer design of methylation-specific PCR (MSP)

The MSP required the use of two primer sets to detect regions containing CpG sites in bisulfite- treated DNA and to generate distinct PCR results. To achieve this, the main primer pair of MSP was designed to meet two criteria: (1) it had to be specific for bisulfite-converted DNA to prevent false-positive results caused by partial bisulfite conversion; and (2) it needed to clearly differentiate between methylated and unmethylated DNA strands following bisulfite treatment [26] . The identification of the specific genomic region was prioritized based on the results obtained from the EPIC-methylation array. After the desired genomic region was identified, MSP primer pairs were designed using MethPrimer, a web-based program (http://www.urogene.org/cgi-bin/ methprimer/meetprimer.cgi). MethPrimer, a freely accessible tool, provided resources and databases for DNA methylation analysis, including the ability to design MSP primers. The design of the MSP primers adhered to specific criteria: (a) a product length ranging from 100 to 300 base pairs; (b) a primary melting temperature (Tm) between 50°C and 60°C; (c) primer lengths ranging from 20 to 30 base pairs; and (d) a 5°C difference in Tm between the forward and reverse primers [26].

DNA methylation detection methods with methylation-specific PCR (MSP)

The PCR was performed at 95°C for 15 minutes, followed by 38 cycles of 95°C for 50 seconds, 59°C for 50 seconds, and 72°C for 50 seconds, with a final extension at 72°C for 10 minutes. The PCR mixture was prepared by combining 1 μL of bisulfite-converted DNA at a concentration of 15 ng/μL, 10 μL of 2× Vazyme SYBR Green PCR Master Mix (Vazyme, Nanjing, People's Republic of China), 0.8 μL of forward and reverse primers at a concentration of 100 μΜ, and ddH₂O to achieve a total volume of 20 μL. Each sample was processed in duplicate to ensure reproducibility. The PCR products were analyzed by electrophoresis using the BlueGel system (miniPCR bio, Cambridge, MA, USA) with 0.4 g of agarose gel (Invitrogen, Waltham, MA, USA) dissolved in 20 mL of 1× TAE buffer. The PCR product bands were visualized under 254 nm ultraviolet light.

Results

Subject characteristics

A total of 48 breast cancer patients, aged between 32 and 66 years, were included in this study. Among them, eight patients were identified as having familial breast cancer, while the remaining 40 patients were classified as having non-familial breast cancer. The clinical characteristics of the patients are presented in Table 1. Of the total number of patients, 26 individuals (54.2%) were found to have a positive metastatic status.

Table 1. Characteristics of breast cancer patients included in the study (n = 48).

Characteristics	Frequency (percentage)
Age (year), mean ± SD	49 ± 9.1
Breast cancer family history	7 (14–6)
Molecular subtype
Triple-negative breast cancer (TNBC)	12 (25.0)
HER2+	12 (25.0)
Luminal B	12 (25.0)
Luminal A	12 (25.0)
Metastasis status	10 (20.9)
Breast cancer stadium
I	4 (8.3)
II	11 (22.9)
III	23 (47.9)
IV	10 (20.9)

According to the IHC data (Table 1), 12 patients were assigned to each of the four subtype groups: Luminal A, Luminal B, HER2+, and TNBC. The IHC results were used to classify the subtypes as follows: ER/PR positivity (>10%) and K167 negativity (<20%) were categorized as the Luminal A group; ER/PR positivity (>10%) and K167 positivity (>20%) were categorized as the Luminal B group; HER2 + positivity (+3) was classified as the HER2 + group; and negativity for ER/PR/HER2 was classified as the TNBC group following the guidelines [2,3].

Quality control of DNA samples and microarray results

The isolated DNA were subjected to quality control by assessing their purity through the absorbance ratio of 260/280 nm using a Nanodrop and measuring their quantitative concentration with a Qubit 4.0 fluorometer. Based on the quality control results, it was confirmed that all DNA isolate samples exhibited excellent purity (with ratio of absorbance A260/A280 nm ranging from 1.8 to 2.0) and concentrations suitable for bisulfite conversion and microarray analysis (at least 300 ng/μL).

The quality control process of the .idat files, conducted using the Minfi program, revealed that all samples exhibited a p-value of less than 0.01, indicating excellent data quality. CpG probes containing single nucleotide polymorphisms (SNPs) were excluded. Of the 1,105,209 CpG probes initially examined, only 903,232 CpG probes met the quality control criteria. As part of the quality control process, the Beta values of each CpG probe were also assessed. The Beta values were observed to fall within the low range (0–0.2) and the high range (0.8–1.0) (Supplementary 1). A CpG probe with a Beta value of ≥0.3 was classified as methylated, whereas a Beta value below 0.3 was classified as unmethylated [24].

Analysis of differentially methylated CpGs (DMC)

In this step, a comprehensive analysis was conducted to identify probes exhibiting differential methylation in each molecular subtype. This analysis, called paired analysis, involved calculating the differences between data groups within each individual, and then averaging these differences across individuals to determine whether there was an overall significant difference in the mean methylation levels for each CpG site. CpG sites showing significant differences in methylation levels between two different groups were referred to as differentially methylated CpGs (DMCs).

In this study, the DMCs were analyzed in four different groups: Luminal A vs non-Luminal A (Luminal B, HER2+, and TNBC); Luminal B vs non-Luminal B (Luminal A, HER2+, and TNBC); HER2 + vs non-HER2+ (Luminal A, Luminal B, and TNBC); TNBC vs non-TNBC (Luminal A, Luminal B, and HER2+). The Limma software, developed in R/Bioconductor, was used to detect the DMCs. DMCs were defined as CpG probes with an adjusted p < 0.05. Hypermethylation was defined as DMC with log2 fold change (log2FC) > 1 and a Beta value ≥0.3, while hypomethylation was defined as DMC with log2FC < -1 and a Beta value <0.3. The volcano plot of DMCs in each subtype is presented in Figure 1. The DMC list identified from this study is presented in Table 2.

Figure 1.

Volcano plot of differentially methylated CpG (DMC): (A) Luminal A, (B) Luminal B, HER2 + and (D) triple-negative breast cancer (TNBC). Hypomethylations are shown in blue dots (adjusted p < 0.05; log2 fold changes (log2FC) < -1), hypermethylations are shown in red dots (adjusted p < 0.05; log2FC >1).

Table 2. Differentially methylated CpG (DMCs) identified in each subtype group which has the potential to be subtype biomarkers.

Molecular subtype	Methylation status	CpG locus	Methylation positions	CpG regions	Gene	Log2FC
Luminal A (Luminal A vs non-Luminal A)	Hypermethylation	cg11799593	chr12: 68452089	OpenSea/promoter	LINC02384	1.49
		cg14397888	chr9: 133554506	Island/gene body	ADAMTSL2	1.59
Luminal B (Luminal B vs non-Luminal B)	Hypomethylation	cg14397888	chr9: 133554506	Island/gene body	ADAMTSL2	-3.06
HER2+ (HER2 + vs non-HER2+)	Hypermethylation	cg25910261	chr7: 157613273	Island/promoter	PTPRN2	1.06
TNBC (TNBC vs non-TNBC)	Hypomethylation	cg15652532	chr2: 30446893	Shore/promoter	LCLATl	-1.5
		cg26371957	chr12: 630114	OpenSea/promoter	NINJ2	-1.48

Log2FC: log2 fold change

To assess potential biomarkers for the subtypes, a comparison was made between our identified DMCs and those reported in other datasets (using tissue samples) that employed a similar methodology. By examining the Venn diagram (Figure 2), the cg15652532 biomarker was found to overlap in the DMCs identified in our dataset for the TNBC subtype and in the other three datasets as well. This indicated that DNA methylation in TNBC was found not only in germinal tissue but also in somatic tissue. Unfortunately, no overlaps were found in the Luminal A, Luminal B, and HER2 + subtypes. The cg14397888 biomarker, identified as potentially useful for distinguishing between Luminal A and Luminal B subtypes, was not included in the Illumina Infinium 450 K Human Methylation manifest. The absence of cg14397888 in the manifest probes used across the other three datasets may have caused this discrepancy.

Figure 2.

Venn diagram of triple-negative breast cancer (TNBC) subtype's biomarkers from the present study dataset (TNBC) and other three datasets from Gene Expression Omnibus (GEO) databases (GSE141338, GSE72245, and GSE72251).

Based on our analysis, DNA methylation biomarkers were identified for each subtype: cg14397888 (Luminal A and Luminal B), cg25910261 (HER2+), and cg15652532 (TNBC). It was shown that the Beta value scores between the two distinct groups were sufficiently different, indicating methylated and unmethylated DNA (Figure 3). These DMCs have the potential to be further analyzed, as they could serve as biomarkers for classifying the molecular subtypes of breast cancer.

Figure 3.

Violin plot of Beta value for each differentially methylated CpG (DMC): (A) cg14397888 (Luminal A vs non-Luminal A); (B) cg14397888 (Luminal B vs non-Luminal B); (C) cg25910261 (HER2 + vs non-HER2+); and (D) cg15652532 (triple-negative breast cancer (TNBC)) vs non- TNBC).

Survival analysis

We used relapse-free survival analysis with the Kaplan-Meier plotter database to analyze the prognostic outcome due to DMCs in ADAMTSL2, PTPRN2, andLCLAT1 genes (Figure 4). It was informed that relapse-free survival in the HER2 + subtype was significantly different between low and high expression of PTPRN2 expression (p = 0.0096). Hypermethylation in PTPRN2 promoter gene (cg25910261) found in HER2 + subtype could downregulate the PTPRN2 expression. Low expression of PTPRN2 showed lower relapse-free survival probability than high expression of PTPRN2. Therefore, hypermethylation in PTPRN2 (cg25910261) could be proposed as HER2 + identifier and relapse-free survival predictor as well.

Figure 4.

Kaplan-Meier plot: (A) ADAMTSL2 expression in Luminal A patients; (B) ADAMTSL2 expression in Luminal B patients; (C) PTPRN2 expression in HER2 + patients; and (D) LCLAT1 expression in triple-negative breast cancer (TNBC) patients. Significant differences are defined with a p < 0.05; the hazard ratio (HR) with a 95% confidence interval.

DNA methylation analysis of methylation-specific PCR (MSP)

The chosen biomarkers were derived from the findings of the microarray and were subsequently translated using PCR. Epigenomic biomarkers, specifically DNA methylation indicators, were utilized through the application of MSP method. The initial step undertaken was the design of main MSP primers for each group (Table 3). Using MethPrimer software, primers were designed for each DMC biomarker for MSP analysis, consisting of methylated and unmethylated primers.

Table 3. Designed primers for each differentially methylated CpG biomarker.

Group	Gene	cgID	Primer sequences	Product size (bp)
Luminal A /Luminal B	ADAMTSL2	cg14397888	Methylated: F: TTTTTTTTAGGTGTGGAATTAGAAC (25) R: CGAATAACCGAACAACCGAT (20)	238
			Unmethylated: F: TTTTTTTAGGTGTGGAATTAGAATG (25) R: CCAAATAACCAAACAACCAAT (21)	239
HER2+	PTPRN2	cg25910261	Methylated: F: TAGTCGTTTTTGTTAGAATTATCGG (25) R: TAATTCCTTCAACTCCTAAAACGAT (25)	211
			Unmethylated: F: GTAGTTGTTTTTGTTAGAATTATTGG (26) R: TAATTCCTTCAACTCCTAAAACAAT (25)	212
TNBC	LCLAT1	cg15652532	Methylated: F: TTGATTTGTTTTGTGATTTCGC (22) R: TAAAAACTCCTTTAAAAAAACCTCG (25)	239
			Unmethylated: F: TTATTGATTTGTTTTGTGATTTTGT (25) R: AAAACTCCTTTAAAAAAACCTCACT (25)	240

In MSP analysis, it is advisable to initially examine the dissociation curve and determine if there is a distinct peak indicating primary methylation at 81°C. If a dissociation curve has a distinct peak at 81°C for methylated primers, it indicates the presence of methylated DNA. If not, then the sample does not include any methylation DNA. The PCR products showing a peak at 81°C (methylated) and below 81°C (unmethylated) were further analyzed by electrophoresis. The electrophoresis bands also confirmed that the methylated and unmethylated products had the same size as originally designed (Supplementary 2).

The DNA methylation and unmethylation analysis results obtained using MSP were compared with the microarray results using accuracy, sensitivity, and specificity scores. When comparing the MSP analysis to the microarray analysis, the test accuracy for the Luminal A and Luminal B subtypes was determined to be 75% and 76%, respectively (Table 4). The accuracy for the HER2 + and TNBC subtypes was only 49% and 54%, respectively (Table 4). Hence, the utilization of the MSP approach was demonstrated to have promising testing capabilities in the Luminal A and Luminal B subtype groups.

Table 4. Evaluation criteria of the DNA methylation test utilizing the methylation-specific PCR (MSP) technique compared to microarray.

Criteria	Luminal A (cg14397888/ADAMTSL2)	Luminal B (cg14397888/ADAMTSL2)	HER2+ (cg25910261/PTPRN2)	TNBC (cg15652532/LCLAT1)
Accuracy	0.75	0.76	0.49	0.54
Precision	0.76	0.69	0.36	0.46
Sensitivity	0.86	0.56	0.80	1.00
Specificity	0.58	0.86	0.34	0.24

Discussion

This study sought to identify DNA methylation biomarkers that could be used to differentiate the breast cancer subtypes using MSP method. Several DMCs in each subtype group, which had the potential as subtype biomarkers, were identified (Table 2). The cg14397888/ADAMTSL2 was associated with Luminal A and Luminal B subtype but had different methylation status: hypermethylated in Luminal A and hypomethylated in Luminal B. The downregulation or upregulation of ADAMTSL2 may be associated with the transforming growth factor beta (TGFβ) signaling pathway [27]. A study found that ADAMTSL2 has the ability to inhibit the function of TGFβ by attaching itself to fibrillin-i and latent TGFβ-l, hence hindering the release of unbound TGFβ [27]. The expression of ADAMTSL2, TGF-β, and phospho-SMAD2 plays a crucial role in the development of breast cancer, as they are part of the canonical TGF-β signaling pathway [27–29].

We also found a methylation biomarker in the HER2 + subtype, cg25910261/PTPRN2 (Table 2). Hypermethylation of the PTPRN2 promoter gene in the HER2 + subtype can cause gene silencing, that might downregulate the PTPRN2 expression. Previous investigations have found that PTPRN2 promoter hypermethylation is present in both lung glioblastoma and adenocarcinoma, indicating a tumor suppressor function [30,31]. A previous study showed that PTPRN2 expression in breast cancer was higher than in healthy population, but no further data about the PTPRN2 expression in HER2 + subjects [31].

In our study, we found that downregulation of PTPRN2 expression had lower relapse-free survival probability in HER2 + subjects without targeted therapy. A previous study showed that 12 weeks of weekly paclitaxel plus dual HER2 targeted therapy may be an effective de-escalated neoadjuvant regimen in early HER2 + breast cancer patients, with high pathological complete response rates and favorable 5-year survival [32]. Therefore, the DMC biomarker found for HER2 + subtype, hypermethylation in PTPRN2 (cg25910261), can help clinicians identify HER2 + patients early and give HER2-targeted therapy for better patient survival.

Hypomethylation of the LCLAT1 promoter was observed in the TNBC subtype group, specifically in the TNBC patient group (Table 2). This hypomethylation may have led to an increase in the expression of the LCLAT1 gene. A study [28] demonstrated that the acyltransferase LCLAT1/LYCAT was necessary for EGF-induced phosphatidylinositol-3,4,5-trisphosphate and Akt activation in MDA-MB-231 cells. The upregulation of LCLAT1 expression may enhance EGF- induced Akt signaling and play a role in the PI3/Akt signaling pathway.

The present study successfully developed MSP method for classifying Luminal A and Luminal B subtypes (accuracy of 75% and 76%, respectively). Alongside with microarray approach, the MSP is a widely utilized technique that relies on the conversion of modified DNA samples. This method involves subjecting the samples to PCR amplification using specific primers for methylated and unmethylated DNA. MSP has been developed as a minimally invasive method for early diagnosis or prognosis of various types of cancer and is commonly employed for this purpose [33]. The present study successfully produced effective diagnostic tools for classifying Luminal A and Luminal B subtypes by MSP method. Nevertheless, this result is anticipated to have advantages in identifying individuals with low-risk breast cancer (Luminal A), who only require hormone therapy [2] by using less invasive techniques (without biopsy) and at a cost-effective price.

Despite the findings, this study has some limitations. Biomarker testing has not been conducted on other breast cancer patient populations outside the test sample, and cross-validation has not yet been performed. The accuracy, precision, specificity, and selectivity of the MSP method were relatively low for other types of breast cancer and this could be improved by incorporating more data, defining better inclusion or exclusion criteria, and addressing potential sources of bias. Further refinement of the MSP analysis is therefore necessary to achieve more satisfactory accuracy.

Conclusion

This study successfully identified DMCs as biomarkers for each subtype classification of breast cancer: Luminal A (hypermethylation; cg14397888/ADAMTSL2), Luminal B (hypomethylation; cg14397888/ADAMTSL2), HER2+ (hypermethylation; cg25910261/PTPRN2), and TNBC (hypomethylation; cg15652532/LCLAT1). The DMC biomarker found for HER2 + subtype, hypermethylation in PTPRN2 (cg25910261), had the potential to help clinicians identify HER2 + patients early and prescribe HER2-targeted therapy for a better survival rate. For the translational approach, MSP method produced effective diagnostic tools for classifying the Luminal A and Luminal B subtypes, with accuracies of 75% and 76%, respectively.

Ethics approval

The study received clearance from the Ethical Committee of the Faculty of Medicine, Universitas Indonesia, on September 4, 2023 (approval number: KET-1140/UN2.F1/ETIK/PPM.00.02/ 2023). The procedures were conducted in accordance with the relevant norms and regulations.

Prior to the collection of the patients’ samples, all individuals provided their informed consent for participation in this study.

Acknowledgments

None.

Competing interests

All the authors declare that there are no conflicts of interest.

Funding

This research was funded by RIIM-BRIN/LPDP Research Grant (Grant number: 34/IV/KS/05/2023; 22/PKS/WRIII-DISTP/UI/2023; Grant receiver name: Sonar Soni Panigoro)

Underlying data

The full data of Beta values of each CpG probe (Supplementary 1) are available from: https://doi.org/10.6084/m9.figshare.27634503.v1 and the results of electrophoresis bands during the methylation-specific PCR (MSP) (Supplementary 2) are available from: https://doi.org/10.6084/m9.figshare.27634506

Declaration of artificial intelligence use

We hereby confirm that no artificial intelligence (AI) tools or methodologies were utilized at any stage of this study, including during data collection, analysis, visualization, or manuscript preparation. All work presented in this study was conducted manually by the authors without the assistance of AI-based tools or systems.

How to cite

Panigoro SS, Paramita RI, Wanandi SI, et al.DNA methylation profiles for breast cancer subtype classifications: A translational study from microarray to methylation-specific PCR (MSP). Narra J 2025; 5 (1): e1364 - http://doi.org/10.52225/narra.v5i1.1364.