Abnormal promoter DNA hypermethylation of the integrin, nidogen, and dystroglycan genes in breast cancer

Cell transmembrane receptors and extracellular matrix components play a pivotal role in regulating cell activity and providing for the concerted integration of cells in the tissue structures. We have assessed DNA methylation in the promoter regions of eight integrin genes, two nidogen genes, and the dystroglycan gene in normal breast tissues and breast carcinomas (BC). The protein products of these genes interact with the basement membrane proteins LAMA1, LAMA2, and LAMB1; abnormal hypermethylation of the LAMA1, LAMA2, and LAMB1 promoters in BC has been described in our previous publications. In the present study, the frequencies of abnormal promoter hypermethylation in BC were 13% for ITGA1, 31% for ITGA4, 4% for ITGA7, 39% for ITGA9, 38% for NID1, and 41% for NID2. ITGA2, ITGA3, ITGA6, ITGB1, and DAG1 promoters were nonmethylated in normal and BC samples. ITGA4, ITGA9, and NID1 promoter hypermethylation was associated with the HER2 positive tumors, and promoter hypermethylation of ITGA1, ITGA9, NID1 and NID2 was associated with a genome-wide CpG island hypermethylated BC subtype. Given that ITGA4 is not expressed in normal breast, one might suggest that its abnormal promoter hypermethylation in cancer is non-functional and is thus merely a passenger epimutation. Yet, this assumption is not supported by our finding that it is not associated with a hypermethylated BC subtype. ITGA4 acquires expression in a subset of breast carcinomas, and methylation of its promoter may be preventive against expression in some tumors. Strong association of abnormal ITGA4 hypermethylation with the HER2 positive tumors (p = 0.0025) suggests that simultaneous presence of both HER2 and integrin α4 receptors is not beneficial for tumor cells. This may imply HER2 and integrin α4 signaling pathways interactions that are yet to be discovered.

, gastric and bladder tumors 11 . Integrin α4β1 may promote melanoma cell adhesion to the vascular endothelium. Higher-level ITGA4 expression may suppress the detachment and invasion of cancer cells and, on the other hand, may promote their dissemination by increasing their interaction with surface ligands of endothelial cells 25 . Like integrins, dystroglycan also acts as a cell surface receptor. Dystroglycan consists of two subunits, α and β, which are synthesized as a propeptide from one mRNA; the propeptide is then cleaved proteolytically into two noncovalently associated subunits. Dystroglycan interacts with various extracellular matrix components, such as laminins, perlican, and agrin. In muscle cells, dystroglycan ensures physical contacts between the cytoskeleton and the basement membrane. Dystroglycan is also expressed in other cells and plays a role in epithelial morphogenesis, cell adhesion, and signal transduction 2 .
Studies of dystroglycan in breast tissue have shown its importance for maintaining the cell polarity, tissue architectonics, and β-casein secretion 26 . Diminished function of α-dystroglycan has been associated with a more aggressive tumor phenotype in BC cell lines 27 .
Nidogens are another family of essential components of the extracellular matrix. The family includes only two members, nidogens 1 and 2. Structurally, nidogens are 150-kDa sulfated glycorpoteins that consist of three globular domains. Nidogen surface presents binding sites for extracellular matrix proteins, such as laminins, collagens, and integrins. Nidogen-1, which is also known as entactin-1, interacts with integrins, perlican, and fibulin and can form a triple complex with collagen IV and laminin 111. The latter interaction underlies the formation of all basement membranes. Mesenchymal and myoepithelial cells express nidogen-1 in breast tissues. Nidogen-1 is thought to modulate the signals transmitted by laminin 111, in particular, to regulate β-casein expression 28 . Nidogen-2 interacts with the same matrix molecules as nidogen-1 with the exception of fibulin. In addition, nidogen-2 binds with endostatin and tropoelastin. Inactivation of nidogen-2 has led to a higher frequency of melanoma metastasis to the lung in experiments with mice 29 .
Nidogens play a role in epithelial morphogenesis by regulating cell adhesion and maintaining the basement membrane integrity and indirectly affect signal transduction 30 . Although nidogens are involved in these processes, little is still known today on their role in tumorigenesis. Negrini et al. 31 have observed that the promoter regions of NID1 and NID2 are abnormally methylated in 67 and 29% of rectal cancer cases, respectively. Methylation of the two regions has additionally been detected in more than 90% of cases in gastric cancer. Abnormal methylation of the promoter regions in NID1 and NID2 is accompanied by lack of their expression in the MCF7, MDA-MB-231, and BT-20 BC cell lines. A demethylating agent has been shown to restore their expression in MDA-MB-231 cells 31 . It has recently been reported that endothelial cell-derived nidogen-1 inhibits migration of SK-BR-3 breast cancer cells 32 .
Methylation of the promoter region is one of the mechanisms that regulate gene expression. In this work, we have assessed methylation status of the promoter regions of the integrin, dystroglycan, and nidogen genes, and have evaluated the association of their abnormal hypermethylation with expression and epigenetic subtypes of breast cancer. ITGA1, ITGA2, ITGA3, ITGA4, ITGA6, ITGA7, ITGA9, ITGB1, NID1, NID2, and DAG1 gene promoters in normal and malignant breast tissues and cell lines. Data for the integrin, nidogen, and dystroglycan genes 5′-cytosine-phosphate-guanine-3′ (CpG) islands assessed by methylation-sensitive restriction enzyme digestion PCR (MSRE-PCR) are summarized in Table 1. Based on the results of MSRE-PCR, the genes were classified into two categories: genes with CpG islands nonmethylated in normal breast tissues but prone to abnormal hypermethylation in BC (ITGA1, ITGA4, ITGA7, ITGA9, NID1, and NID2), and genes with CpG islands found to be nonmethylated in all breast tissues, no matter normal or cancerous (ITGA2, ITGA3, ITGA6, ITGB1, and DAG1).  Nonmethylated status of the promoter CpG islands of the ITGA3, ITGA4, ITGA6, ITGA9, ITGB1, NID1, NID2 and DAG1 genes in normal breast tissues is also supported by XmaI-RRBS previously performed by us on the same six autopsy samples 34 . ITGA1, ITGA2 and ITGA7 gene promoters are not covered with XmaI-RRBS data.

DNA methylation at
DNA methylated status of the regions assessed by MSRE-PCR was confirmed by a validating method of bisulfite sequencing by Sanger. Bisulfite Sanger sequencing was performed for the promoter of ITGA7, which evaded from ENCODE RRBS and XmaI-RRBS, to confirm its nonmethylated status in normal and tumor samples; and for one of the promoters (NID2) that demonstrated differential methylation by MSRE-PCR, to validate nonmethylated status in normal samples and abnormal methylation in cancer. ITGA7 and NID2 promoter regions were bisulfite sequenced in one normal sample and four tumor samples and the results confirmed MSRE-PCR findings in all cases. DNA methylation within a DAG1 gene intron. Alongside with the gene promoters' methylation information, our XmaI-RRBS results 34 provide information on the methylation status of CpG-rich fragments of an intron separating the two coding exons of DAG1, one of the genes of interest in this study. The results presented on Fig. 2 demonstrate overwhelming prevalence of methylated alleles in BC samples and BC cell lines, as well as in normal breast tissues, indicating no or negligible evolutionary advantage of this area hypomethylation for breast cancer cells. Of note, this region is not covered in the available ENCODE project RRBS dataset 33 . Associations of hypermethylation of the ITGA1, ITGA4, ITGA9, NID1, and NID2 gene promoters and characteristics of breast tumors. Further analysis was focused on the ITGA1, ITGA4, ITGA9, NID1, and NID2 genes in which we have identified CpGs nonmethylated in normal breast tissues but prone to abnormal hypermethylation in BC, with hypermethylation found in no less than 5% of samples from our collection.
A multiple correspondence analysis was carried out in order to suggest clinical and genetic associations between the methylation status of ITGA1, ITGA4, ITGA9, NID1, and NID2 and clinical/morphological characteristics of the tumors, including the disease stage, tumor grade, tumor type, as well as estrogen receptor, progesterone receptor, and HER2 expression assessed by immunohistochemistry (IHC). By multiple correspondence analysis, potential association with abnormal methylation of ITGA1, ITGA4, ITGA9, NID1, and NID2 was observed only for the HER2 expression level (Fig. 3).
Representations of BC samples with hypermethylated promoters in groups with different levels of HER2 expression are shown on Fig. 4. The association between the methylated gene status and high HER2 expression was confirmed by the χ 2 and Fisher's exact tests for ITGA4, ITGA9, and NID1 (Table 2). www.nature.com/scientificreports/ We have also identified positive association of the CpG island hypermethylated status of the ITGA1, ITGA9, NID1 and NID2 genes in tumors with their attribution to the hypermethylated epigenomic BC subtype determined for a subset of samples from the same collection by genome-wide bisulfite sequencing of the CpG islands by XmaI-RRBS. Interestingly, we found no such association for the ITGA4 gene (Table 3).
For the ITGA4, ITGA9, NID1 and NID2 genes for which XmaI-RRBS data are available, we have calculated the relative fractions of methylated and nonmethylated alleles in tumor samples, in respect to HER2 expression in tumors. The results are shown on

Discussion
We have previously examined DNA methylation at the CpG islands of all laminin subunit genes and have shown that the promoter regions of LAMA1, LAMA2, and LAMB1 are affected by abnormal methylation in BC with the frequencies of 29%, 26% and 16% respectively 5 . The present study focuses on the genes for eleven proteins that are capable of interacting with the LAMA1, LAMA2, and LAMB1 products.
We have studied the methylation status of eight integrin gene promoters in normal and BC cells, and observed abnormal hypermethylation in BC for ITGA1, ITGA4, ITGA7, and ITGA9.  Table) HER2 The nonmethylated gene status is designated as "0" at the gene symbol; methylated as "1" at the gene symbol. Multiple correspondence analysis allows investigating the structure of objects and variables that characterize them, initially included in the multi-input table, where objects (samples) are located in rows, and grouping variables (N) in columns, which are object coordinates in the N-dimensional space. The goal of the multiple correspondence analysis is to represent the contents of the   www.nature.com/scientificreports/ ITGA1 gene promoter was found to be abnormally hypermethylated in 13% (19/146) samples from our BC tissue collection. ITGA1 is normally expressed in multiple tissues including breast. It was recently shown that ITGA1 is a differentially expressed-aberrantly methylated gene in breast cancer, providing a possible independent functional role of its hypermethylation in BC 35 . In our study, ITGA1 hypermethylation was not associated with HER2 tumor status, providing further evidence of its independent role in BC tumorigenesis.
The ITGA4 gene promoter showed a high abnormal methylation frequency in our BC tissue collection (48/156 samples, 31%). Abnormal hypermethylation of this gene has previously been detected in 41% of tumor samples and was associated with the HER2 positive immunohistochemistry status and poor differentiation of tumor cells 14 . In the present study we also demonstrate high frequency of abnormal DNA methylation at the ITGA4 gene promoter CpG island, also associated with the HER2 positive immunohistochemistry status. Do et al. 14 speculate that such results support a previous study that suggested integrin α and HER2 are involved in the same signaling pathway, referring to the publication describing a novel mechanism for integrin-mediated Ras activation in breast carcinoma cells 36 . Yet, it should be kept in mind that the mentioned mechanism was described for integrin α6, as part of the α6β4 molecule. Both α6 and β4 integrin genes, ITGA6 and ITGB4, are well expressed in the normal mammary tissues, unlike ITGA4, whose expression in normal breast is negligible (the gene expression data were obtained from the GTEx Portal https ://www.gtexp ortal .org on 11/11/20). With this fact alone one www.nature.com/scientificreports/ should be cautious when drawing direct conclusions regarding common mechanisms of action of integrins α4 and α6. Further, the mechanism of integrin α6β4 action implies its involvement in HER2 expression. Depletion of β4 by shRNA reduced HER2 protein level without affecting ERBB2 (HER2 gene) mRNA level and reexpression of β4 increased HER2 protein level 36 . These results show that integrin α6β4 positively regulates HER2 through translational control, and we thus should anticipate high α6β4 in the HER2 positive breast tumors. In the case of ITGA4 in breast cancer, we witness the opposite situation, where the fraction of HER2 positive tumors is significantly enriched with samples harboring ITGA4 methylation. If ITGA4 is silenced by DNA methylation, as it is suggested by the results of several studies 11,14,24 , then integrin α4 should be low in HER2 positive tumors. Indeed, the question to discuss in the context of ITGA4 abnormal hypermethylation in BC, is what the reason may be for the gene negligibly expressed in the normal tissue to become hypermethylated in the tumor of the same organ. We have recently suggested based on the examples of other extracellular matrix related molecules 37 , that cancer-specific abnormal hypermethylation of the genes that are not expressed in the same normal tissue might be a "passenger" epigenetic event reflecting merely a CIMP (CpG island methylator phenotype) of a tumor. Yet, while this may be an explanation for other genes found abnormally hypermethylated in BC samples in our study, as far as their hypermethylation well correlates with the CIMP phenotype (Table 3), this is not the case for ITGA4 which is significantly more frequently found in tumors with moderate genome-wide levels of CpG islands methylation. Still, ITGA4 hypermethylation is strongly associated with HER2 hyperexpression, and the association is more pronounced than for the other genes ( Table 2). In terms of tumor evolution, this finding might reflect advantage for the cells in which elevation of ITGA4 expression was occasionally "preventively prohibited" by its promoter methylation. Like any other CpG island, that of the ITGA4 gene may acquire abnormal methylation in the process of cancer progression given general deregulation of epigenetic processes in tumors. If under any circumstances the mechanisms of ITGA4 transcription in a tumor are activated (the activating mechanisms may be, for example, ectopic expression of a transcription factor, or activation of an otherwise inactive topologically associating domain), it would be more effective in the cells with nonmethylated ITGA4 promoter then in those with the hypermethylated one. Then cells with nonmethylated ITGA4 alleles would produce integrin α4 protein the function of which in cancer is not yet well understood but, judging by depletion of nonmethylated ITGA4 alleles from the HER2 positive tumors, simultaneous presence of both HER2 and integrin α4 receptors is not beneficial for tumor cells (Fig. 6). This may suggest importance of HER2 and integrin α4 signaling pathways interactions that are yet to be discovered. Important though is the fact that in the HER2 positive tumors we never find hypermethylation of more than 50% of ITGA4 alleles (Figs. 5, 7) and BC samples with completely nonmethylated ITGA4 also exist, which suggests that simultaneous expression of HER2 and integrin α4 is not lethal, but rather renders some evolutionary disadvantage for the tumor cells that acquire such expression pattern. ITGA7 promoter CpG island was abnormally hypermethylated in 4% (6/156) samples in our study. Although The Cancer Genome Atlas data assessed by TCGA Wanderer 38 indicate that ITGA7 methylation at the region that we assessed tend to downregulate its expression, the number of hypermethylated samples is too small in our collection to draw any conclusions on the associations with clinical and molecular characteristics of the tumors.
In the present study, the ITGA9 gene showed a high (39%) frequency of abnormal methylation in our collection of BC samples. This supports a previous report where in breast tumors, ITGA9 expression was downregulated or totally absent in 44% of cases and intact or upregulated in 45% of cases, and the frequency of ITGA9 promoter methylation was 90% in the former group; while treatment with the demethylating agent 5-aza-dC restored integrin α9 expression in ITGA9-negative cells of the MCF7 line 18 .

Conclusions
In this study, we have identified high frequencies of cancer specific abnormal hypermethylation of the parts of promoter regions of integrin ITGA1, ITGA4, ITGA9, and nidogen NID1, NID2 genes in breast cancer. Although their abnormal hypermethylation is cancer specific and frequent in breast tumors, which makes them attractive candidates for diagnostic markers, in the plain of tumor biology we suggest that this hypermethylation might be mere a "passenger" epigenetic event reflecting tumor CpG island methylator phenotype (CIMP). For the only gene among listed, ITGA4, abnormal hypermethylation is not correlated with CIMP, but is strongly correlated with high levels of HER2 expression in breast tumors. This finding may suggest importance of HER2 and integrin α4 signaling pathways interactions that are yet to be discovered. Data on possible involvement of receptors known to be overexpressed in breast cancer in regulation of integrin genes expression are gradually accumulating. It has very recently been suggested that estrogen receptor α is involved in the regulation of ITGA8 methylation in estrogen receptor-positive breast cancer 39 . Further elaboration of this subject promises new exciting insights in cancer epigenomics.

Materials and methods
Clinical material. We examined 156 BC samples, 156 matched samples of morphologically normal adjacent tissue, five BC cell lines (ZR711, HS578T, BT474, T47D, and MCF7), and six autopsy samples of normal breast tissues. Biological material was obtained from the Blokhin Russian Cancer Research Center, Gertsen Moscow Research Cancer Institute, Research Centre for Medical Genetics, Institute of Gene Biology, and Russian Scientific Center of Roentgenoradiology. Altogether, 323 samples have been analyzed in this study. All BC and matched samples were obtained at surgery of cancer cases none of which underwent neoadjuvant chemotherapy. All tissue samples were fresh-frozen. Fragments of tissues for DNA analysis were examined macro-and micromorphologically: expression of ER, PR, and HER2 receptors was evaluated by immunohistochemistry. Negative HER2 status was diagnosed in 72 samples (46%); HER2 "1+", "2+" and "3+" positivity was documented for 34 (22%), 31 (20%), and 19 (12%) samples, respectively. This study was conducted in accordance with the Declaration of Helsinki, and was approved by the Bioethics Committee of Research Centre for Medical Genetics. Written informed consent was obtained from each participant of this study.
DNA isolation and methylation-sensitive restriction enzyme digestion. Genomic DNA was isolated by standard phenol-chloroform extraction. The DNA digestion mixture for the methylation-sensitive restriction enzyme digestion contained 1.5 μg of genomic DNA, 10 units of HpaII restriction endonuclease, and 2 μL of a SEBufferY buffer (10×) (SibEnzyme, Russia). Deionized water was added to bring the final volume to 20 μL, and the mixture was incubated at 37 °C for 16 h. Figure 6. A hypothetical mechanism for the tumor microevolution advantage of the methylated ITGA4 gene promoter in HER2 positive breast tumors. The hypothesis is that co-expression of the HER2 and integrin α4 receptors might reduce microevolution fitness of cancer cells; "preventive prohibition" of the ITGA4 gene expression by its promoter methylation provides advantage for the cells in cases when under any circumstances the mechanisms of ITGA4 transcription in a tumor are activated (see detailed explanation in the text). TF, transcription factor; TAD, topologically associating domain.

Methylation-sensitive restriction enzyme digestion PCR (MSRE-PCR) assays. Triplex MSRE-
PCR assays were used for each locus under study, where one fragment was amplified from the target gene, another one served as a positive PCR control (a constitutively methylated region of the CUX1 gene 5 ), and a third one was used to check the completeness of DNA hydrolysis (a constitutively nonmethylated region of SNRK 5 ). Target loci with distinctly different amplicon lengths may be combined in the single-tube multiplex reactions (Fig. 5). The nucleotide sequences of the primers are shown in Table 4. PCR reactions were performed as described earlier 5 . The MSRE-PCR products were resolved by electrophoresis in 8% polyacrylamide gel and stained with silver nitrate (Fig. 7).
Bisulfite sequencing by sanger. The results of the analysis of promoter methylation of target genes obtained by MSRE-PCR were verified with bisulfite sequencing of corresponding fragments. For bisulfite conversion, genomic DNA was denatured in NaOH (at a final concentration of 0.3 M) at 65 °C for 15 min. DNA was modified using sodium bisulfite and hydroquinone taken at final concentrations 2 M and 0.5 M, respectively, for 15 h at 55 °C. Modified DNA was purified using Wizard DNA Cleanup system (Promega, USA) according to the manufacturer's instructions. PCR reactions were performed as described earlier 5 . PCR products were sequenced with an ABI3100 genetic analyzer using terminating dideoxynucleotides according to the protocol for ABI Prism 3100 Genetic Analyzer (Thermo Fisher Scientific, Waltham, Massachusetts, USA). The nucleotide sequences of the primers are shown in Table 5.

Validation of MSRE-PCR results by RRBS. For the validation of MSRE-PCR results by RRBS, two
RRBS datasets were used, one from the ENCODE project 33 , and another from our previous XmaI-RRBS study 34 performed on a subset of 64 BC samples, five BC cell lines and six normal breast samples matching the samples from the collection described here. XmaI-RRBS was performed as described by us earlier 40 . A representative example of XmaI-RRBS results visualized with Integrative Genomic Viewer 41 (IGV) is shown on Fig. 8.     Red is for methylated cytosines and blue is for nonmethylated cytosines in the CpG context. Reference genome sequence is at the bottom. Symbols "-" and "-3-" seen within the reads reflect spurious single nucleotide and trinucleotide deletions, correspondingly, that arise as technical errors in detecting the lengths of homopolynucleotide tracts (poly-T and poly-A) arising as a result of bisulfite DNA conversion.