Skip to main content

DNA methylation profile of inflammatory breast cancer and its impact on prognosis and outcome



Inflammatory breast cancer (IBC) is a rare disease characterized by rapid progression, early metastasis, and a high mortality rate.


Genome-wide DNA methylation analysis (EPIC BeadChip platform, Illumina) and somatic gene variants (105 cancer-related genes) were performed in 24 IBCs selected from a cohort of 140 cases.


We identified 46,908 DMPs (differentially methylated positions) (66% hypomethylated); CpG islands were predominantly hypermethylated (39.9%). Unsupervised clustering analysis revealed three clusters of DMPs characterized by an enrichment of specific gene mutations and hormone receptor status. The comparison among DNA methylation findings and external datasets (TCGA-BRCA stages III-IV) resulted in 385 shared DMPs mapped in 333 genes (264 hypermethylated). 151 DMPs were associated with 110 genes previously detected as differentially expressed in IBC (GSE45581), and 68 DMPs were negatively correlated with gene expression. We also identified 4369 DMRs (differentially methylated regions) mapped on known genes (2392 hypomethylated). BCAT1, CXCL12, and TBX15 loci were selected and evaluated by bisulfite pyrosequencing in 31 IBC samples. BCAT1 and TBX15 had higher methylation levels in triple-negative compared to non-triple-negative, while CXCL12 had lower methylation levels in triple-negative than non-triple-negative IBC cases. TBX15 methylation level was associated with obesity.


Our findings revealed a heterogeneous DNA methylation profile with potentially functional DMPs and DMRs. The DNA methylation data provided valuable insights for prognostic stratification and therapy selection to improve patient outcomes.

Graphical Abstract


Inflammatory breast cancer (IBC) is the most lethal breast cancer type (BC). IBC is characterized by rapid progression and aggressive behavior from the early stages of the disease [1]. High body mass index (BMI), insulin resistance, hormone-related factors (such as early age at menarche and pre-menopause), ethnicity (independent predictor of increased risk of breast cancer mortality), and socioeconomic status (independent predictor of advanced stages at diagnosis) have been described as risk factors associated with IBC [2, 3].

Initial efforts were undertaken to identify specific molecular signatures based on transcriptome analysis and prognostic stratification of IBC patients. Two studies showed similar gene signatures between non-IBC and IBC cases [4, 5]. Hyperactivation of interferon (alpha and gamma), hypoactivation of epidermal growth factor receptor (EGFR), TGFβ (attenuated transforming growth factor β), and p53 pathways were associated with complete pathological response [6]. Recently, Zare et al. (2021) [7] used an approach based on machine learning and reported a 50 genes signature independent of ER/HER2 status with 100% accuracy in classifying IBC cases. Frequent mutations of TP53, BRCA2, and homologous recombination (HR) genes [8, 9] and copy number alterations (gains and upregulation of MYC and MDM4, and losses of TP53 and RB1) were described in IBC [9].

Although, to date, no specific molecular differences between IBC and other advanced BC types have been found, some intrinsic IBC features contribute to its aggressiveness, including stemness phenotype, high rates of motility, invasion and metastasis, tumor emboli formation leading to inflammation signals, tumor microenvironment (TME) promoting tumor epithelial cells, and immune evasion [10, 11].

DNA methylation is the most widely studied epigenetic mark. In the human genome, the enzymatic addition of the methyl radical to the carbon 5′ position of the cytosine (5-methylcytosine) occurs mainly in cytosine-phosphate-guanine (CpG) dinucleotides. DNA methylation alterations are particularly relevant in the context of global methylome changes during the development and progression of complex diseases. It is well established that weight, diet, and exercise status can influence BC risk and outcomes and that DNA methylation has a preeminent role in this process [11]. The methylation of genes involved in the inflammatory process may be associated with lifestyle and increased risk of BC development [12]. Another aspect is that DNA methylation has been associated with drug resistance, predicts response to treatment, and potentially controls the stemness phenotype [13]. Overall, these features pointed out the relevance of DNA methylation as a critical mechanism involved in IBC development and progression.

Up-to-date information is restricted to rare studies investigating DNA methylation of CpG sites associated with candidate genes showing that promoter hypermethylation is a common feature in IBC [11]. Using a 27-k microarray-based approach, Van der Auwera et al. identified two clusters characterized by low and high methylation levels; the last was enriched with metastatic and poor prognosis cases [14].

In this study, we conducted a genome-wide DNA methylation screening using a robust microarray with extensive coverage of 850,000 methylation sites across the genome to identify biologically significant differentially methylated positions (DMPs) and regions (DMRs) in IBC and their association with clinical and molecular features.

Materials and methods


A total of 140 IBC patients were admitted at Barretos Cancer Hospital, São Paulo, Brazil, between 2002 and 2019. We selected only biopsies collected from primary tumors prior to surgery and before chemo-radiotherapy treatment. The eligibility criteria include patients with no synchronous or metachronous tumors at diagnosis, biopsies available in the biobank, complete clinical and histopathological information in medical records, and revisions by the oncologist (C.P.S.) and pathologist (I.V.V.S.) specialists. Tumors with lymph vascular invasion and having no clinical hallmarks of IBC were excluded. The cases were classified according to the American Joint Committee on Cancer 8th Edition (AJCC 2018) [15].

According to the adopted criteria, 116 available biopsies were analyzed by an expert pathologist (I.V.V.S.) and macrodissected (to scrape off regions of tumors labeled on hematoxylin and eosin-stained tissue sections). Thirty-two cases resulted in DNA of high quality and sufficient quantity for the experiments. The study was conducted following the ethical guidelines and regulations of the Declaration of Helsinki and approved by the institutional Ethics Committee (CEP# 37779220.5.1001.5437). Written informed consent was obtained from all patients before sample collection. Clinical and epidemiological data are summarized in Table 1.

Table 1 Clinical and pathological characteristics of 32 female patients with inflammatory breast carcinomas evaluated in this study

DNA isolation

Genomic DNA was isolated from 32 fresh frozen macrodissected IBC tissue biopsies and six surrounding normal breast frozen tissues using the QIAsymphony kit (Qiagen, Valencia, CA, USA) at Barretos Cancer Hospital Biobank. The selected tumor area had at least 80% of tumor cells. The DNA was quantified using Qubit® dsDNA BR Assay (Life Technologies, USA).

Targeted Next-Generation Sequencing (t-NGS)

A targeted NGS panel of 105 cancer-related genes (all exons, 3′ UTR, and 5′UTR) was performed, as previously described [10]. Twenty-one IBC samples had their mutational profile described in our previous study [9], and eight additional cases were presented herein. The libraries were prepared using SureSelectQXT Library Prep Kit (Agilent) and sequenced on the NextSeq 550 system (Illumina, San Diego, CA, USA) according to the manufacturer’s instructions. The pipeline used to select the variants was previously described [16]. In summary, we selected variants classified as pathogenic or likely pathogenic in ACMG (American College of Medical Genetics and Genomics), ClinVar (, or Clinvitae (, and variants of uncertain significance (VUS) in ACMG presenting loss of function (e.g., frameshift, stop gain, canonical splice site). Variants classified as synonymous, intergenic or intronic regions, or mapped in the homopolymer regions were excluded. The variants were manually curated using the Genome Browse software (Golden Helix, Inc., Bozeman, MT, USA) and visualized using the ComplexHeatmap package (v.2.10.0) [17].

DNA methylation profiling

Bisulfite conversion was performed with the EZ DNA Methylation-Gold™ Kit (Zymo Research, USA), according to the manufacturer’s recommendations. The genome-wide DNA methylation profiling of 24 IBC and six normal breast samples was evaluated using the Infinium Human Methylation EPIC BeadChip platform (Illumina, USA), according to the manufacturer’s recommendations. The Illumina HiScan system was used to scan the beadchip platform, and the data processing was performed with R software (v.4.1.0). Briefly, probes with low quality (detection p > 0.05), mapped in X/Y chromosomes or to SNPs, and cross-reactive were filtered out using the minfi package v.1.42.0 [18]. Differences in Type-I and Type-II probes were adjusted, and β values were normalized using the Beta MIxture Quantile dilation (BMIQ) method [19]. The batch effect was corrected with the ComBat function from the sva package (v.3.44.0) [20]. Annotation was performed with Illumina’s manifest file.

Screening of differentially methylated sites (DMPs) and regions (DMRs)

Differentially methylated probes (DMP) comparing tumor and normal samples were identified using the limma package (v.3.50.3) [21] considering FDR-adjusted p < 0.05 and |∆β|≥ 0.2 annotated with the Illumina manifest file. An unsupervised K-means clustering algorithm was performed considering DMPs, while the ComplexHeatmap package was used to plot the results [17]. Differentially methylated regions (DMR) analysis was carried out using the DMRcate package (v.2.10.0) [22]. Genomic regions with at least two consecutive DMPs (distance less than 1000 nucleotides to the next CpG site) and smoothed FDR-adjusted p < 0.05 were considered statistically different.

Site-specific DNA methylation evaluated by bisulfite pyrosequencing (BS-P)

Thirty-one IBC specimens (including 23 evaluated by DNA methylation array above described) and six normal breast tissues were investigated using BS-P to confirm the DNA methylation levels of BCAT1 (cg04543413), CXCL12 (cg06702993 and cg11267527), and TBX15 (cg14565725, cg24884142, cg07892597, and cg10703826). The CpGs were selected based on the comparison performed in the experiments, gene function, and CpG position. The CpGs selected for BCAT1 and TBX15 genes were mapped in differentially methylated regions. CXCL12 was selected based on the literature pointing out its role in triple-negative BC [23]. Primer sequences and PCR conditions used for each CpG are described in Table S1. Ten ng of gDNA bisulfite converted was amplified using the PyroMark PCR kit (Qiagen, Redwood, CA, USA), according to the manufacturer’s recommendations. The PCR products were sequenced on the PyroMark Q48 system (Qiagen, USA).

Methylation data comparison with IBC and non-IBC external datasets

The DNA methylation data of our IBC cases were compared with those published by Van der Auwera et al. (2010) [14]. In this study, the authors analyzed 62 breast tissues (19 IBC and 43 non-IBC) in comparison with the corresponding mean β-value of 10 normal tissues generated by the Infinium Human Methylation 27 K, Illumina (data not available for download and re-analysis), considering p < 0.00001 and |∆β|≥ 0.17 to identify DMPs. Additionally, we explored the DNA methylation profiles (TCGA-BRCA dataset) of 178 advanced non-IBC (stages III and IV) and 93 adjacent normal samples evaluated using the Infinium Human Methylation 450 K platform (Illumina). The β values matrix and clinical information were downloaded with the TCGA biolinks package (v.2.25.2) [24] and analyzed with minfi (v.1.42.0) [17] and limma (v.3.50.3) packages [20], as previously described for our internal dataset. The intersection of the three DMP lists was acquired by drawing Venn diagrams using the ggplot2 package (v.3.4.0).

Pathway enrichment analysis

The three final gene lists were obtained based on DMPs annotated in gene regions from the internal and external datasets comparison. Pathway enrichment analysis was performed based on the Kyoto Encyclopedia of Genes and Genomes (KEGG) database using the cluster Profiler package (v.4.4.4). Significantly enriched pathways were acquired considering FDR < 0.05, and the top results were plotted.

Gene expression data acquisition

Public transcriptomic data of IBC samples were manually curated from the Gene Expression Omnibus repository (NCBI GEO, GSE45581 was selected as the onliest dataset containing microdissected fresh frozen IBC samples from untreated patients and normal tissues [5]. To this extent, we retrieved raw probe intensity values of 20 IBC and five healthy tissue samples profiled using the Agilent Whole Human Genome 4 × 44 K array platform. The limma package was used to perform background correction and quantile normalization, and the expression data were annotated with the Agilent microarray manifest file [20]. Differentially expressed genes (DEG) were determined considering FDR < 0.05 and |FC|≥ 1.5. To our knowledge, GSE207248 was the unique public RNA-seq dataset available for in silico analysis. Herewith, count data of 22 IBC samples from untreated HER2 + patients of this dataset were downloaded and integrated with normal breast tissue samples (N = 22) obtained from The Cancer Genome Atlas (TCGA) and Genotype-Tissue Expression (GTEx) databases by randomly selecting 11 samples from each platform. For restricted validation purposes, we performed a variance stabilizing transformation (VST) following DESeq2 package workflow (v.1.34.0) [25], annotated gene expression matrix using biomaRt package (v.2.50.3) [26], and screened DEGs with cutoff values set to FDR < 0.05 and |FC|≥ 2.

The transcriptomic data of 196 advanced breast cancer (stages III and IV), 101 triple-negative breast cancer (TNBC), and 106 adjacent normal samples from TCGA-BRCA dataset were used to evaluate the differences in the gene expression levels of the selected targets (BCAT1, CXCL12, and TBX15). Clinical and RNA-seq count data (STAR method) were downloaded with the TCGAbiolinks package (v.2.25.2) [24], and DESeq2 package pipeline was performed for data normalization (v.1.34.0) [25]. Wilcoxon test was applied for data comparison considering p < 0.05 as significant.

Statistical analysis

The β values of the DMPs identified in the internal dataset and the normalized values of the corresponding DEGs obtained from IBC samples (GSE45581) were used to determine the correspondence between methylation and expression pattern of altered genes. The statistical significance of the mean DNA methylation levels of normal versus IBC samples and IBC-TNBC versus IBC-non-TNBC samples were examined using two-way ANOVA (p ≤ 0.05). Data from BS-P experiments of distinct CpG dinucleotides were analyzed with the Mann–Whitney test (p ≤ 0.05). Logistic regression was used to estimate the association between DNA methylation levels and clinicopathological variables. Survival curves were generated using the Kaplan–Meier method and the log-rank test. Overall survival was defined as the period (in months) between the biopsy date and death. Statistical analysis was performed in Graph-Pad Prism v.9.5.1 and R software v.4.2.1 ( Prognostic factors underwent assessment through univariate and multivariate Cox regression analyses. Significant variables (univariate regression p < 0.5) and potential pairwise risk factors were incorporated into the multivariate regression model. All values related to these analyses were presented in the log2 scale.


Clinical and pathological characteristics of IBC patients

The clinicopathological information of 32 women diagnosed with IBC is summarized in Table 1. The age at diagnosis ranged from 29 to 82 years, with 14 patients aged 50 years or younger and 18 patients showing a family history of cancer. According to the Body Mass Index (BMI), 22 patients were overweight or obese (BMI ≥ 25). Twelve patients presented histological grade III, 19 had clinical stage III at diagnosis, and 13 were triple-negative (TNBC: ER/PR/HER2 negatives). Thirteen patients presented distant metastases at diagnosis and eleven during follow-up. Twenty-four women died at a medium time of 18.3 months (3.06 to 105.2 months), six were alive, and two died by other causes.

Patients at clinical stage III were initially treated with neoadjuvant chemotherapy with four cycles of doxorubicin and cyclophosphamide, followed by paclitaxel for 12 weeks. In HER2-positive patients, trastuzumab was administered, and some hormone receptor-positive IBC patients also received tamoxifen or anastrozole. Patients alive after the neoadjuvant treatment were referred to surgery and adjuvant radiotherapy. Five patients presented a pathological complete response (pCR) after the neoadjuvant treatment. Stage IV patients were treated with palliative therapy, including fluorouracil, doxorubicin, and cyclophosphamide.

Mutational profile

The mutational profile was performed in 28 IBC samples (21 previously reported) [9] using t-NGS (105 cancer driver genes). The most common variants were detected in TP53 (18 cases), BRCA2 (9 cases), and PIK3CA (6 cases) (Fig. 1E). Interestingly, 20 out of 28 cases showed variants in genes involved in the homologous recombination pathway (Table S2).

Fig. 1
figure 1

Methylation profile of the IBC internal dataset. A Distribution of the differentially methylated CpG probes between tumor and normal tissues (NT). The proportion of CpGs in relation to its gene location (TSS200/TSS1500, body/exons boundaries, intergenic regions, 5′UTR/first exon and 3′UTR); and to the CpG islands context (island, open sea, shelf, and shore) was based on the Illumina EPIC annotation. B Unsupervised K-means clustering analysis based on 46,908 identified DMPs revealed four clusters: cluster 1 is composed of adjacent normal samples and three clusters with IBC samples. Rows indicate the CpG sites, while columns represent samples. Clinical features of each case are represented below the heatmap along with targeted next-generation sequencing data for specific genes. The estrogen receptor (ER), progesterone receptor (PR), HER2 status, and the mutational pattern of TP53, BRCA1, BRCA2, and homologous recombination genes are indicated below the heatmap. C The survival curves (Kaplan–Meier and log-rank test) showed no significant statistical differences among clusters. D Univariate Cox regression analysis based on clinical variables and overall survival of 24 IBC patients evaluated by DNA methylation profiling. The forest plot shows the hazard ratios (squares), and the horizontal bars represent the range between the lower and upper limits of the 95% confidence intervals (CI) in the log2 scale. IBC patients positive for estrogen (ER), progesterone (PR), and or human epidermal growth factor receptor 2 (HER2) were taken as a reference against triple-negative tumors. In the analysis of the BMI variable, we excluded patients within the normal weight range due to the small sample size (N = 4); overweight patients were used as a reference. For the remaining variables, we considered the absence of the corresponding predictive factor as a reference for analysis (*p < 0.05). E A panel of 105 cancer-related genes was investigated in 28 IBC patients using t-NGS. Highlights include clinical, molecular, and vital status information. Genes are organized in descending order of alterations. The top bar plots illustrate the number of altered genes detected in each sample, and the percentages on the right indicate the number of samples with genetic alterations for a given gene among all analyzed samples. Genes without selected variants were excluded

Identification of DMPs and DMRs in IBC

The DNA methylation performed in 24 IBC and six adjacent normal breast samples (NT) revealed 46,908 DMPs (FDR < 0.05 and |∆β|≥ 0.2), in which 30,919 (66%) were hypomethylated and 15,989 (34%) hypermethylated (Fig. 1A). Hypermethylation was predominant in the CpG islands (39.9%), followed by open sea regions (32.7%), while most hypomethylation was found in the open sea region (80.8%). Hypermethylated sites (32.4%) were frequently found in CpGs mapped in gene bodies and hypomethylated DMPs in intergenic regions (42.1%). Hypermethylated CpGs mapped in promoter regions (21.2%) were more frequently observed than in hypomethylated DMPs (11.7%). Annotations related to shore, shelf, and other regions are illustrated in Fig. 1A. Supplementary Fig. 1 summarizes the study design and main findings.

The unsupervised clustering analysis of all DMPs revealed four clusters: One grouped all normal samples (cluster 1), and three had IBC samples (Fig. 1B, Table S3). Clinical and molecular features distribution, such as estrogen (ER) and progesterone receptors (PR), HER2 status, metastasis at diagnosis, TP53, BRCA1, BRCA2, and HR genes mutations were distributed among the three clusters (Fig. 1B). Cluster 3 was enriched with triple-negative cases, TP53 mutation, and cluster 4 with ER/PR positive, HER2-negative. Both clusters 3 and 4 were enriched with HR mutated genes. Although no significant difference was observed in the survival curves and BMI among these groups of patients, cluster 3 presented shorter survival (Fig. 1C; Table S4).

The Cox regression model was applied to evaluate the prognostic role of molecular and clinical information gathered from our cohort of IBC cases screened for DNA methylation. We found that TNBC (p = 0.036; HRatio = 1.131; CI 0.095–2.175), metastasis at diagnosis (p = 0.00048; HRatio = 2.23; CI 0.993–3.466), and HR mutated genes (p = 0.011; HRatio = 2.64; CI 0.588–4.700) were significantly correlated with overall survival as independent factors for increased risk (Fig. 1D). Multivariate regression analysis revealed that only metastasis at diagnosis (p = 0.0032; HRatio = 2.10; CI 0.704–3.487) remained a prognostic indicator for overall survival. A pairwise comparison of potential variables showed a robust association between triple-negative tumors (p = 0.015; HRatio = 1.45; CI 0.288–2.622) and metastasis at diagnosis (p = 0.0003; HRatio = 2.57; CI 1.178–3.969), indicating a poorer overall survival. Similar findings were observed in patients presenting metastasis at diagnosis (p = 0.0073; HRatio = 1.71; CI 0.460–2.953) and HR genes mutation (p = 0.0305; HRatio = 2.36; CI 0.222–4.499). Cox regression analysis highlighted the significant role of metastasis at diagnosis in predicting poorer overall survival. Furthermore, our model underscores that the separate interaction of this variable with TNBC and tumors with HR mutations is significantly associated with an increased risk of death.

Among the 46,908 significant DMPs, 17,868 CpGs were mapped in intergenic regions and 29,040 (11,135 hypomethylated and 17,905 hypermethylated) in known genes. The DNA methylation data is available in the Gene Expression Omnibus database (GSE238092). The enrichment analyses showed pathways involved in signal transduction (MAPK signaling pathway, PI3K-Akt signaling pathway, Ras signaling pathway), cellular adhesion and movement (focal adhesion, ECM-receptor interaction, regulation of actin cytoskeleton, cell adhesion molecules), and neuronal system (axon guidance, neuroactive ligand-receptor interaction, glutamatergic synapse, synapse vesicle cycle, morphine addiction, and nicotine addiction), among others. The top 20 KEGG pathways are shown in Fig. 2A.

Fig. 2
figure 2

Differentially methylated probes (DMPs) comparison performed among the internal dataset, 27 k Van der Auwera et al. (2010), and TCGA-BRCA advanced tumors. A KEGG pathway analysis shows five shared enriched pathways among three datasets. B Venn diagram obtained from dataset comparisons (internal dataset, TCGA-BRCA, and Van der Auwera et al., 2010) shows 385 shared DMPs among the datasets comparison

A total of 4369 DMRs mapped on known genes was found (2392 DMR hypomethylated and 1977 hypermethylated) (Table S4). Two genes presented more than 30 CpGs in the same DMR, TBX15 (36 CpGs) and OR2I1P (31 CpGs).

Comparison of DNA methylation profiles of IBC and non-IBC using external datasets

The comparison of our findings with the Van der Auwera et al. (2010) [27] data (1353 DMPs; p < 0.00001, |∆β |≥ 0.17) revealed 417 shared DMPs with the same methylation status, 92 hypomethylated and 325 hypermethylated (Figure S1). Using the same criteria adopted in the internal dataset (FDR < 0.05 and |∆β|≥ 0.2) to analyze advanced breast tumors from TCGA, we found 38,662 DMP, in which 14,611 DMP were found in both datasets (6298 hypomethylated and 8313 hypermethylated). The comparison among these three datasets resulted in 385 shared DMPs (Fig. 2B) mapped in 333 genes, 69 hypomethylated and 264 hypermethylated (Table S5). Enrichment analysis revealed five main pathways: Focal Adhesion, Calcium Signaling Pathway, cAMP Signaling Pathway, Neuroactive Ligand-Receptor Interaction, and Protein Digestion and Absorption (Fig. 2A, Table S6).

The gene expression data of IBC samples retrieved from the GSE45581 (Whole Human Genome Microarray 4 × 44 K Agilent platform) was compared with our DNA methylation results (Pearson correlation, p < 0.05), revealing 151 DMPs mapped in 110 genes; 68 probes mapped in 50 genes presented a negative correlation (Table 2). The gene annotation using CancerMine ( and MSigDB ( databases showed 50 genes, of which 31 are cancer-related (Table 2). We also accessed the expression levels of these genes using the GSE207248. Ten genes were hypomethylated and overexpressed (such as the oncogenes CHST11, KIF26B, LAIR1, NTM, RUNX2, VOPP1, and WDR5), and six hypermethylated and down expressed (including the tumor suppressor genes CDO1, CTDSPL, EBF1, ROBO3, and TGFBR3) (Table 2).

Table 2 List of probes and genes with a negative correlation between the internal dataset and expression data from the GSE45584. Significant values of correlation are presented (p < 0.05)

Locus-specific DNA methylation analysis

Bisulfite pyrosequencing was used to quantitatively determine the methylation of CGs mapped on BCAT1, CXCL12, and TBX15 genes. Figure 3 shows the mean DNA methylation levels of consecutive CpG dinucleotides flanking those interrogated by microarrays as well as the methylation levels of individualized DMPs. The analysis of gene body DMR encompassing 16 CpGs of BCAT1 gene confirmed increased methylation levels in IBC compared with normal samples. We also detected a trend to increase methylation levels in IBC versus normal tissues (p = 0.0725), but no differences were observed between triple-negative IBC compared with non-triple-negative IBC. The comparison with the expression levels of BCAT1 of TCGA-BRCA stages III-IV revealed no difference between non-TNBC and normal samples. However, a significant increase in expression was observed between normal versus TNBC and non-TNBC versus TNBC.

Fig. 3
figure 3

Loci-specific DMRs analyzed by bisulfite pyrosequencing of CpGs associated with selected genes (BCAT1, CXCL12, and TBX15) using the internal cohort and cross-validation with gene expression levels using external cohorts (advanced III-IV stages from TCGA-BRCA and IBC from GSE45581). The first column shows the comparison of the means of DNA methylation levels of each associated DMRs containing the interrogated CpG probes and the flanking CpG dinucleotides between normal and IBC samples. Tumor samples were also dichotomized in non-TNBC and TNBC, and the methylation levels were compared. The DNA methylation levels of distinct DMPs confirmed higher methylation levels of BCAT1 and TBX15 genes and lower levels of CXCL12 in IBC. The methylation levels of TBX15 gene were associated with the subtype TNBC and obesity (*p ≤ 0.05, **p ≤ 0.01; ***p ≤ 0.001, and ****p ≤ 0.0001). ns: not significant (p > 0.05)

Based on the unsupervised K-means clustering analysis of DMPs and in the literature data, two differentially methylated CpGs (cg06702993 and cg11267527) of CXCL12 were evaluated. Triple-negative IBC showed a significant decrease in DNA methylation levels of CXCL12 compared to non-triple-negative samples (p = 0.0481). These findings are in accordance with the expression levels of CXCL12 of TCGA-BRCA stages III-IV, showing a decreased expression in advanced tumors compared to normal samples and TNBC.

We selected four differentially methylated CpGs extracted from the DMR of TBX15 (10 analyzed), which were also altered in TCGA-BRCA dataset. We confirmed increased methylation levels in IBC compared to normal samples. Triple-negative tumors presented increased methylation levels of TBX15 compared with non-triple-negative IBC cases. The expression level of TBX15 was confirmed as significantly decreased in advanced BC tumors (TCGA-BRCA stages III-IV) and IBC (GSE45581) compared to normal samples (Fig. 3G). In addition, we found a significant negative association between methylation levels of TBX15 and obesity (cg14565725, OR 0.9182, CI 0.8332 to 0.9864, p = 0.0004; cg07892597, OR 0.9391, CI 0.8661 to 0.9996, p = 0.0486; cg10703826 OR 0.9336, CI0.8690 to 0.9892, p = 0.0184; and cg24884142, OR 0.8864, CI 0.7909 to 0.9659, p = 0.0042).


Cancer-specific alterations in DNA methylation are closely related to a variety of malignancies. Although the main risk factors associated with IBC (such as obesity, exposition to endocrine disruptors, socioeconomic status, and racial disparities) have been linked with epigenetic alterations, limited DNA methylation data are available in the literature [11].

Using a robust microarray platform with extensive coverage of CpG islands, genes, and enhancers (850 k Infinium Methylation EPIC BeadChip platform, Illumina), we verified that 66% of the DMP were hypomethylated mainly in open sea regions, while 34% were hypermethylated frequently in CpG islands. CpG islands hypermethylation was previously reported in IBC [14]. Genome-wide DNA hypomethylation have been described in cancer cells and associated with chromosomal instability, de-repression of imprinted genes and retrotransposons, and aberrant gene expression [27]. The high proportion of hypomethylated CpG sites in the gene body is a feature of many tumors, but its biological effect is still under investigation. Gene body methylation suppresses gene expression through chromatin and its interaction with regulatory elements (alternative promoters, enhancers, transcription factors, and repetitive elements) [28]. Gene body methylation can also activate transcription by suppressing spurious gene transcription, regulating, and allowing the correct transcription process [29, 30]. Hypermethylation in gene bodies has been associated with increased oncogenes expression and related to tumor progression [30].

Unsupervised hierarchical clustering analysis of DNA methylation showed a perfect separation between normal and IBC samples. Three clusters composed exclusively of IBC samples were found, and their clinical features were compared. Body mass index (BMI) is recognized as an independent risk factor for IBC development. However, no significant difference was observed among IBC clusters, probably due to the small sample size. Although no difference was observed in the overall survival among IBC clusters, cluster 3 (TNBC enriched) showed a trend toward shorter survival than clusters 2 and 4. DNA methylation profiles in IBC could provide new molecular stratification with potential implications for individual patient prognosis and therapy selection.

Cluster 3 was enriched by TNBC subtype (7 of 9 cases), TP53 mutations, and homologous recombination mutated genes. Triple-negative BC is more frequently detected in IBC than the other subtypes, and it is associated with worse overall survival, distant metastasis, and locoregional relapse [31]. Gene expression datasets available in GEO showed no differences in the comparison between triple-negative IBC and triple-negative non-IBC, which suggested a molecular similarity [32]. A recent report showed that TNBC subtype luminal androgen receptor subtype presented the most hypermethylated CpGs, while the mesenchymal subtype the most hypomethylated CpGs [33]. Differences among the triple-negative IBC subtypes should also be explored and contribute to developing new treatment strategies for this devastating disease.

The relationship between clinical variables and outcomes of our IBC cohort (univariate and multivariate Cox proportional-hazards regression models) revealed that the TNBC subtype, metastasis at diagnosis, HR mutations were significantly correlated with overall survival. However, only metastasis at diagnosis remained a significant prognostic factor in the multivariate analysis. Like our findings, Dano et al. (2021) utilized multivariate analysis and demonstrated that overall survival was significantly shorter in metastatic IBC cases compared with metastatic non-IBC [34]. Deficiency in homologous recombination pathway, such as BRCA2 mutations frequently altered in our cases, has been described in IBC cases [8, 9]. The limited sample size may have contributed to the lack of statistical significance of our parameters since the TNBC phenotype is a well-known risk factor for unfavorable prognosis in breast cancer and IBC [35, 36].

Fifteen IBC patients presented somatic TP53 mutations distributed into the three clusters. TP53 and HR mutations are frequently associated with triple-negative BC [37], a finding also detected in our IBC samples. The impact of somatic TP53 mutations on the prognosis was reported as independent of hormone receptor status [38], a feature also observed in our dataset. Somatic TP53 mutations alter the level of transcripts like thymine-DNA glycosylase that are implicated in the physiological control of promoter demethylation of several genes involved in embryogenesis and development [39]. Promisor cancer treatment strategies have been described based on targeting dysfunctional p53, reactivating the mutant to its wild-type form, and inhibiting the interaction between wild-type p53 and MDM2/MDM4 [40]. Antagonists of p53-MDM2/MDM4 are undergoing clinical trials. Moreover, two mutant p53-reactivating compounds have progressed to clinical trials, COTI-2 and APR-246 [40]. These strategies could be tested in IBC in conjunction with other therapies to improve outcomes.

We found an enrichment of homologous recombination genes mutation in clusters 3 (6 of 7 IBC cases) and 4 (8 out of 9 IBC cases). Although not statistically significant, both showed shorter survival compared to cluster 2. Homologous recombination is the most effective mechanism to repair double-stranded DNA breaks (DSB). Mutations leading to homologous recombination deficiency increase the degree of genetic instability and the emergence of other mutations that potentially promote tumor progression. Using human and mouse cells and DNA damage induced by ionizing radiation, Moureau et al. (2016) provide evidence that p53 is associated with the regulation of DSB repair [41]. Patients with somatic and germline BRCA1/2 mutations, the leading players of the HR pathway, have sensitivity to platinum-based chemotherapy and PARP inhibitors, an alternative to treat IBC patients. IBC models demonstrated that PARP inhibitors improved radiotherapy's effectiveness and provided the preclinical rationale for the phase II randomized trial for IBC patients [42]. The molecular classification of a large cohort of IBC cases according to mutations and DNA methylation data may help identify biomarkers and design therapeutic strategies to improve patient outcomes. Several epigenetic drugs are under investigation in clinical trials in different tumor types and could be an efficient alternative to treat IBC patients.

We also identified 4369 DMRs mapped on known genes (2392 DMR hypomethylated and 1977 hypermethylated). Two genes showing DMRs were selected for data confirmation using BS-P: BCAT1 and TBX15. Increased methylation levels of 16 CpGs (gene body) of BCAT1 in IBC were compared with normal samples. BCAT1 expression levels of TCGA-BRCA stages III-IV samples were significantly increased comparing normal versus TNBC and non-TNBC versus TNBC. BCAT1 levels have been positively associated with tumor progression and worse prognosis in invasive breast cancer and TNBC [43]. In a pan-cancer study with 16,847 samples, Li et al. (2022) [44] explored the expression, potential mechanisms, and clinical significance of BCAT1. BCAT1 was predominantly expressed in ERα-negative/HER-2-positive breast cancer and TNBC, and its overexpression enhanced the capacity of antiestrogen-sensitive cells to grow in the presence of antiestrogens [44]. The silencing of BCAT1 in an orthotopic TNBC xenograft model resulted in a massive tumor volume reduction in vivo [44]. Increased BCAT1 expression has been associated with unfavorable prognosis in several tumor types, including BC, and it has been suggested its potential as an immunotherapeutic cancer marker.

Ten CpGs of the DMR mapped on the TBX15 (T-box transcription factor 15) gene were also confirmed as hypermethylated in IBC compared with normal samples. Also, TN IBC samples presented increased hypermethylation compared with non-IBC and normal samples. TBX15 encodes a transcript factor associated with mesodermal differentiation, skeleton development, and adipocyte differentiation. Aberrant TBX15 methylation has been described in various tumors, such as prostate [45], ovarian [46], and hepatocellular carcinomas [47]. Recently, a study analyzed the functionality of DNA methylation in a DMR region of TBX15 immediately upstream and downstream of the unmethylated promoter region in myoblasts and skeletal muscle [48]. The authors showed that hypermethylated DMR suppresses enhancer/extended promoter activity and down modulates gene expression without silencing it [48]. Importantly, the TBX15 DMR overlaps with cis-regulatory elements, and the DMPs validated by BS-P are located close to the eight best candidates’ functional single nucleotide polymorphisms (SNPs) for obesity-trait or osteoporosis risk [49]. Significantly lower TBX15 expression level was observed in TNBC (TCGA-BRCA) than in non-TNBC and normal tissues. Moreover, IBC showed significantly decreased expression levels compared with normal samples (GSE45581). We showed a significant association between TBX15 methylation levels and obesity. BMI is an independent and positively associated risk factor for IBC [3]. TBX15 is a novel adipose master transcription factor [50]. Its knockdown in human primary preadipocytes resulted in changes in the expression of 130 network genes, including the key adipose transcription factors. We suggest that TBX15 is a key gene regulated by DNA methylation that potentially contributes to obesity in IBC patients.

Bisulfite pyrosequencing was also used to evaluate two CpGs mapped on CXCL12. CXCL12 belongs to the chemokine family, and its expression level has been associated with poor prognosis in TNBC [50]. We observed that TN IBC showed lower DNA methylation levels than non-TN IBC and normal samples. In BC at advanced stages from TCGA-BRCA dataset, CXCL12 expression levels were higher in TNBC than in non-TNBC. Increased levels of CXCL12 promote the accumulation and amplification of myeloid-derived suppressor cells, M2-phenotype macrophages, and regulatory T cells, which contributes to the induction of an immunosuppressive/anergic tumor-permissive environment to benefit tumor formation in TNBC [50]. The authors targeted the delivery of antagonists and the disruption of CXCL12/CXCR4 signaling in the tumor tissue, which promoted depletion of the stromal barrier and relieved immunosuppression, and the T cell infiltration into the tumor site, enhancing the efficacy of immune checkpoint blockade therapy in TNBCs [50].

The enrichment analysis of the internal DMP showed pathways previously reported in IBC biology, such as focal adhesion, extracellular matrix interaction, regulation of actin cytoskeleton, and cell adhesion molecules. These pathways are related to cell–cell contact between cancer cells or cancer cells and tumor microenvironment cells required for forming the IBC emboli structure and the tumor cell clusters with high metastatic potential [10]. The PI3K-AKT signaling pathway is involved in breast cancer progression and associated with cell motility and actin reorganization in IBC [51]. Alterations in MAPK pathway genes were associated with IBC aggressiveness, increased cell proliferation, stemness, and metastatic potential [51]. RAS signaling pathway initiates cell motility and focal adhesions in aggressive BC [52]. These data suggest that IBC aggressiveness features may be controlled by methylation mechanisms in critical genes closely related to cancer progression and worse prognosis.

We also found pathways related to nervous system development and neuronal signaling encompassing nine differentially methylated genes (CDNF, NBEA, NTM, PACSIN1, PPP2R2B, RAB7A, RASL10A, ROBO3, and VANGL1). The nerve-cancer crosstalk suggests an association with intratumoral neural infiltration. The nervous system modulates angiogenesis, the tumor microenvironment, bone marrow, immune functions, and inflammatory pathways to influence metastases [53]. These neural signaling altered by DNA methylation in IBC may relate to tumor progression and metastasis.


We found a distinct DNA methylation profile in triple-negative inflammatory breast tumors. The methylation pattern of CXCL12, BCAT1, and TBX15 genes was confirmed by bisulfite pyrosequencing. We suggest that TBX15 is a potential marker regulated by DNA methylation and related to obesity in IBC. Also, nervous system development and neuronal signaling pathways were involved in IBC and should be further investigated to amplify the therapeutic options of this aggressive and morbid disease.

Availability of data and materials

Data are reported in detail. Further data are available upon academical request.



American College of Medical Genetics and Genomics


Body mass index


Beta MIxture Quantile dilation


Confidence interval




Differentially expressed genes


Differentially methylated probes


Differentially methylated regions


Double-stranded DNA breaks


Epidermal growth factor receptor


Estrogen receptor


Gene Expression Omnibus


Genotype-Tissue Expression


Homologous recombination


Hazard Ratio


Inflammatory breast cancer


Kyoto Encyclopedia of Genes and Genomes


Likely pathogenic variant


Odds Ratio


Pathogenic variant


Pathological complete response


The Cancer Genome Atlas


Attenuated transforming growth factor β


Tumor microenvironment


Triple-negative breast cancer


Targeted next-generation sequencing


Variance stabilizing transformation


Variants with uncertain significance


  1. van Uden DJ, van Laarhoven HW, Westenberg AH, de Wilt JH, Blanken-Peeters CF. Inflammatory breast cancer: an overview. Crit Rev Oncol Hematol. 2015;93:116–26.

    Article  PubMed  Google Scholar 

  2. Fouad TM, Ueno NT, Yu RK, et al. Distinct epidemiological profiles associated with inflammatory breast cancer (IBC): a comprehensive analysis of the IBC registry at The University of Texas MD Anderson Cancer Center. PLoS ONE. 2018;13(9): e0204372.

    Article  PubMed Central  PubMed  Google Scholar 

  3. Schairer C, Laurent CA, Moy LM, et al. Obesity and related conditions and risk of inflammatory breast cancer: a nested case-control study. Breast Cancer Res Treat. 2020;183(2):467–78.

    Article  CAS  PubMed Central  PubMed  Google Scholar 

  4. Van Laere SJ, Ueno NT, Finetti P, et al. Uncovering the molecular secrets of inflammatory breast cancer biology: an integrated analysis of three distinct affymetrix gene expression datasets. Clin Cancer Res. 2013;19(17):4685–96.

    Article  PubMed Central  PubMed  Google Scholar 

  5. Woodward WA, Krishnamurthy S, Yamauchi H, et al. Genomic and expression analysis of microdissected inflammatory breast cancer. Breast Cancer Res Treat. 2013;138(3):761–72.

    Article  CAS  PubMed Central  PubMed  Google Scholar 

  6. Bertucci F, Ueno NT, Finetti P, et al. Gene expression profiles of inflammatory breast cancer: correlation with response to neoadjuvant chemotherapy and metastasis-free survival. Ann Oncol. 2014;25(2):358–65.

    Article  CAS  PubMed  Google Scholar 

  7. Zare A, Postovit LM, Githaka JM. Robust inflammatory breast cancer gene signature using nonparametric random forest analysis. Breast Cancer Res. 2021;23(1):92.

    Article  CAS  PubMed Central  PubMed  Google Scholar 

  8. Ross JS, Ali SM, Wang K, Khaira D, et al. Comprehensive genomic profiling of inflammatory breast cancer cases reveals a high frequency of clinically relevant genomic alterations. Breast Cancer Res Treat. 2015;154(1):155–62.

    Article  CAS  PubMed  Google Scholar 

  9. Faldoni FLC, Villacis RAR, Canto LM, et al. Inflammatory breast cancer: clinical implications of genomic alterations and mutational profiling. Cancers. 2020;12(10):2816.

    Article  CAS  PubMed Central  PubMed  Google Scholar 

  10. Lim B, Woodward WA, Wang X, et al. Inflammatory breast cancer biology: the tumour microenvironment is key. Nat Rev Cancer. 2018;18(8):485–99.

    Article  CAS  PubMed  Google Scholar 

  11. Faldoni FLC, Rainho CA, Rogatto SR. Epigenetics in inflammatory breast cancer: biological features and therapeutic perspectives. Cells. 2020;9:1164.

    Article  CAS  PubMed Central  PubMed  Google Scholar 

  12. Gillman AS, Gardiner CK, Koljack CE, et al. Body mass index, diet, and exercise: testing possible linkages to breast cancer risk via DNA methylation. Breast Cancer Res Treat. 2018;168(1):241–8.

    Article  CAS  PubMed  Google Scholar 

  13. Cheng Y, He C, Wang M, et al. Targeting epigenetic regulators for cancer therapy: mechanisms and advances in clinical trials. Signal Transduct Target Ther. 2019;17(4):62.

    Article  Google Scholar 

  14. Van der Auwera I, Yu W, Suo L, et al. Array-based DNA methylation profiling for breast cancer subtype discrimination. PLoS ONE. 2010;5(9): e12616.

    Article  PubMed Central  PubMed  Google Scholar 

  15. Amin MB, Edge SB, Greene F, et al. AJCC cancer staging manual. 8th ed. New York: Springer. Breast Neoplasms; 2017. p. 1032.

    Google Scholar 

  16. Do Canto LM, Barros-Filho MC, Rainho CA, et al. Comprehensive analysis of DNA methylation and prediction of response to neoadjuvant therapy in locally advanced rectal cancer. Cancers (Basel). 2020;12(11):3079.

    Article  PubMed  Google Scholar 

  17. Gu Z, Eils R, Schlesner M. Complex heatmaps reveal patterns and correlations in multidimensional genomic data. Bioinformatics. 2016;15(18):2847–9.

    Article  Google Scholar 

  18. Aryee MJ, Jaffe AE. Corrada-Bravo, et al. Minfi: A flexible and comprehensive Bioconductor package for the analysis of Infinium DNA methylation microarrays. Bioinformatics. 2014;1363–1369.

  19. Teschendorff AE, Marabita F, Lechner M, et al. A betamixture quantile normalization method for correcting probe design bias in Illumina Infinium 450 k DNA methylation data. Bioinformatics. 2013;29:189–96.

    Article  CAS  PubMed  Google Scholar 

  20. Leek JT, Johnson WE, Parker HS, et al. The sva package for removing batch effects and other unwanted variation in high-throughput experiments. Bioinformatics. 2012;28:882–3.

    Article  CAS  PubMed Central  PubMed  Google Scholar 

  21. Ritchie ME, Phipson B, Wu D, et al. limma powers differential expression analyses for RNA-sequencing and microarray studies. Nucleic Acids Res. 2015;43: e47.

    Article  PubMed Central  PubMed  Google Scholar 

  22. Peters TJ, Buckley MJ, Statham AL, et al. De novo identification of differentially methylated regions in the human genome. Epigenetics Chromatin. 2015;27(8):6.

    Article  Google Scholar 

  23. Guembarovski AL, Guembarovski RL, Hirata BKB, et al. CXCL12 chemokine and CXCR4 receptor: association with susceptibility and prognostic markers in triple negative breast cancer. Mol Biol Rep. 2018;45(5):741–50.

    Article  CAS  PubMed  Google Scholar 

  24. Mounir M, Lucchetta M, Silva TC, et al. New functionalities in the TCGAbiolinks package for the study and integration of cancer data from GDC and GTEx. PLoS Comput Biol. 2019;15(3): e1006701.

    Article  PubMed Central  PubMed  Google Scholar 

  25. Love MI, Huber W, Anders S. Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. Genome Biol. 2014;15(12):550.

    Article  PubMed Central  PubMed  Google Scholar 

  26. Durinck S, Moreau Y, Kasprzyk A, et al. BioMart and Bioconductor: a powerful link between biological databases and microarray data analysis. Bioinformatics. 2005;21(16):3439–40.

    Article  CAS  PubMed  Google Scholar 

  27. Nishiyama A, Nakanishi M. Navigating the DNA methylation landscape of cancer. Trends Genet. 2021;37(11):1012–27.

    Article  CAS  PubMed  Google Scholar 

  28. Yang X, Han H, De Carvalho DD, et al. Gene body methylation can alter gene expression and is a therapeutic target in cancer. Cancer Cell. 2014;26(4):577–90.

    Article  CAS  PubMed Central  PubMed  Google Scholar 

  29. Neri F, Rapelli S, Krepelova A, et al. Intragenic DNA methylation prevents spurious transcription initiation. Nature. 2017;543(7643):72–7.

    Article  CAS  PubMed  Google Scholar 

  30. Wang Q, Xiong F, Wu G, et al. Gene body methylation in cancer: molecular mechanisms and clinical applications. Clin Epigenetics. 2022;14(1):154.

    Article  PubMed Central  PubMed  Google Scholar 

  31. Li J, Gonzalez-Angulo AM, Allen PK, et al. Triple-negative subtype predicts poor overall survival and high locoregional relapse in inflammatory breast cancer. Oncologist. 2011;16(12):1675–83.

    Article  CAS  PubMed Central  PubMed  Google Scholar 

  32. Masuda H, Baggerly KA, Wang Y, et al. Comparison of molecular subtype distribution in triple-negative inflammatory and non-inflammatory breast cancers. Breast Cancer Res. 2013;15(6):R112.

    Article  PubMed Central  PubMed  Google Scholar 

  33. Lehmann BD, Bauer JA, Chen X, et al. Identification of human triple-negative breast cancer subtypes and preclinical models for selection of targeted therapies. J Clin Invest. 2011;121(7):2750–67.

    Article  CAS  PubMed Central  PubMed  Google Scholar 

  34. Dano D, Lardy-Cleaud A, Monneur A, et al. Metastatic inflammatory breast cancer: survival outcomes and prognostic factors in the national, multicentric, and real-life French cohort (ESME). ESMO Open. 2021;6(4): 100220.

    Article  CAS  PubMed Central  PubMed  Google Scholar 

  35. Almansour NM. Triple-negative breast cancer: a brief review about epidemiology, risk factors, signaling pathways, treatment and role of artificial intelligence. Front Mol Biosci. 2022;25(9): 836417.

    Article  Google Scholar 

  36. Biswas T, Efird JT, Prasad S, et al. Inflammatory TNBC breast cancer: demography and clinical outcome in a large cohort of patients with TNBC. Clin Breast Cancer. 2016;16(3):212–6.

    Article  PubMed  Google Scholar 

  37. Jiang YZ, Ma D, Suo C, et al. Genomic and transcriptomic landscape of triple-negative breast cancers: subtypes and treatment strategies. Cancer Cell. 2019;35(3):428-440.e5.

    Article  CAS  PubMed  Google Scholar 

  38. Olivier M, Langerød A, Carrieri P, et al. The clinical value of somatic TP53 gene mutations in 1,794 patients with breast cancer. Clin Cancer Res. 2006;12(4):1157–67.

    Article  CAS  PubMed  Google Scholar 

  39. Maiti A, Drohat AC. Thymine DNA glycosylase can rapidly excise 5-formylcytosine and 5-carboxylcytosine: potential implications for active demethylation of CpG sites. J Biol Chem. 2011;286(41):35334–8.

    Article  CAS  PubMed Central  PubMed  Google Scholar 

  40. Duffy MJ, Synnott NC, O’Grady S, et al. Targeting p53 for the treatment of cancer. Semin Cancer Biol. 2022;79:58–67.

    Article  CAS  PubMed  Google Scholar 

  41. Moureau S, Luessing J, Harte EC, et al. A role for the p53 tumour suppressor in regulating the balance between homologous recombination and non-homologous end joining. Open Biol. 2016;6(9): 160225.

    Article  PubMed Central  PubMed  Google Scholar 

  42. Michmerhuizen AR, Pesch AM, Moubadder L, et al. PARP1 inhibition radiosensitizes models of inflammatory breast cancer to ionizing radiation. Mol Cancer Ther. 2019;18(11):2063–73.

    Article  CAS  PubMed Central  PubMed  Google Scholar 

  43. Song Y, Zhao B, Xu Y, Ren X, Lin Y, Zhou L, Sun Q. Prognostic significance of branched-chain amino acid transferase 1 and CD133 in triple-negative breast cancer. BMC Cancer. 2020;20(1):584.

    Article  CAS  PubMed Central  PubMed  Google Scholar 

  44. Li GS, Huang HQ, Liang Y, et al. BCAT1: A risk factor in multiple cancers based on a pan-cancer analysis. Cancer Med. 2022;11(5):1396–412.

    Article  CAS  PubMed Central  PubMed  Google Scholar 

  45. Kron K, Liu L, Trudel D, et al. Correlation of ERG expression and DNA methylation biomarkers with adverse clinicopathologic features of prostate cancer. Clin Cancer Res. 2012;18(10):2896–904.

    Article  CAS  PubMed  Google Scholar 

  46. Gozzi G, Chelbi ST, Manni P, et al. Promoter methylation and downregulated expression of the TBX15 gene in ovarian carcinoma. Oncol Lett. 2016;12(4):2811–9.

    Article  CAS  PubMed Central  PubMed  Google Scholar 

  47. Gao F, Liang H, Lu H, et al. Global analysis of DNA methylation in hepatocellular carcinoma by a liquid hybridization capture-based bisulfite sequencing approach. Clin Epigenetics. 2015;7(1):86.

    Article  PubMed Central  PubMed  Google Scholar 

  48. Ehrlich KC, Lacey M, Baribault C, et al. Promoter-adjacent DNA hypermethylation can downmodulate gene expression: TBX15 in the muscle lineage. Epigenomes. 2022;6(4):43.

    Article  CAS  PubMed Central  PubMed  Google Scholar 

  49. Zhang X, Ehrlich KC, Yu F, et al. Osteoporosis- and obesity-risk interrelationships: an epigenetic analysis of GWAS-derived SNPs at the developmental gene TBX15. Epigenetics. 2020;15(6–7):728–49.

    Article  PubMed Central  PubMed  Google Scholar 

  50. Lu G, Qiu Y, Su X. Targeting CXCL12-CXCR4 signaling enhances immune checkpoint blockade therapy against triple negative breast cancer. Eur J Pharm Sci. 2021;157: 105606.

    Article  CAS  PubMed  Google Scholar 

  51. van Golen KL, Bao LW, Pan Q, et al. Mitogen activated protein kinase pathway is involved in RhoC GTPase induced motility, invasion and angiogenesis in inflammatory breast cancer. Clin Exp Metastasis. 2002;19:301–11.

    Article  PubMed  Google Scholar 

  52. Kleer CG, Griffith KA, Sabel MS, et al. RhoC-GTPase is a novel tissue biomarker associated with biologically aggressive carcinomas of the breast. Breast Cancer Res Treat. 2005;93:101–10.

    Article  CAS  PubMed  Google Scholar 

  53. Wang W, Li L, Chen N, et al. Nerves in the tumor microenvironment: origin and effects. Front Cell Dev Biol. 2020;8: 601738.

    Article  PubMed Central  PubMed  Google Scholar 

Download references


The authors are grateful to the patients who agreed to participate in this study. We also acknowledge Barretos Cancer Biobank, Barretos, SP, Brazil for providing human specimens. We thank Luisa Matos do Canto, Mariana Bizarro dos Reis, and Bianca Campos Troncarelli Flores for their technical assistance.


Open access funding provided by University of Southern Denmark. This study was supported by grants from Region of Southern Denmark Research Fund, Denmark, National Institute of Science and Technology in Oncogenomics (FAPESP #2008/57887-9 and CNPq #573589/08-9), Brazil. F.L.C.F. received a fellowship from National Council for Scientific and Technological Development, CNPq).

Author information

Authors and Affiliations



F.L.C.F. and S.R.R. contributed to conception and design, L.C.F., D.B., C.P.S., I.V.V.S., F.A.M. contributed to acquisition, all authors contributed to analysis and data interpretation, F.L.C.F., C.A.R., S.R.R. drafted the manuscript, and S.R.R. supervised the study and project administration and funding acquisition. All authors read and approved the final version of the manuscript.

Corresponding author

Correspondence to Silvia Regina Rogatto.

Ethics declarations

Ethics approval and consent to participate

The study was conducted according to the guidelines of the Declaration of Helsinki, approved by the Human Research Ethics Committee of Barretos Cancer Hospital (CEP# 37779220.5.1001.5437), and informed consent was obtained from all subjects involved in the study.

Consent for publication

Not applicable.

Competing interests

The authors declare that they have no competing interests.

Additional information

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary Information

Figure S1

Workflow representative of the strategy used in our DNA methylation analysis in inflammatory breast cancer (IBC). (A) DNA methylation comparing our data (internal dataset), breast cancer stages III and IV from TCGA, and Van der Auwera et al. (2010) study revealed 385 shared differentially methylated probes (DMPs). Gene expression of IBC evaluated with the Agilent microarray platform (GSE45581, available in Gene Expression Omnibus) revealed 2480 differentially expressed genes (DEGs). Integrative analysis between DMPs of the internal dataset and DEGs (GSE45581) with significant Pearson correlation resulted in 151 DMPs. (B) Strategy used to select candidates investigated by bisulfite pyrosequencing. Four selected CpGs differentially methylated of the TBX15 gene were altered in the TCGA-BRCA dataset; three CpGs of the CXCL12 gene were differentially methylated in triple-negative cases compared with non-triple-negative IBC cases, and BCAT1 presented four CpGs differentially methylated and was differentially expressed (integrative analysis)

Table S1

Primer sets used in the bisulfite pyrosequencing reaction for CpGs mapped in the differentially methylated regions of TBX15, CXCL12, and BCAT1.

Table S2

Genetic variants found in TP53, PIK3CA, homologous recombination and mismatch repair genes among the 29 IBC cases assessed by t-NGS.

Table S3

Differentially methylated regions identified in the analysis of 24 inflammatory breast carcinomas.

Table S4

Body mass index (Kg/m2) and overall survival among tumor clusters.

Table S5

Shared differentially methylated probes and delta-Beta values among our internal dataset, 27 k study (Van der Auwera et al., 2010) [14] and Advanced Breast Cancer (TCGA-BRCA dataset). Significant threshold with p values < 0.05.

Table S6

KEGG pathway enrichment analysis revealed five pathways shared among DMPs of our internal dataset, 27 K Study (Van der Auwera et al., 2010), and TCGA-BRCA.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit The Creative Commons Public Domain Dedication waiver ( applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Faldoni, F.L.C., Bizinelli, D., Souza, C.P. et al. DNA methylation profile of inflammatory breast cancer and its impact on prognosis and outcome. Clin Epigenet 16, 89 (2024).

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: