Association of cyclin-dependent kinase inhibitor 2B antisense RNA 1 gene expression and rs2383207 variant with breast cancer risk and survival

The expression signature of deregulated long non-coding RNAs (lncRNAs) and related genetic variants is implicated in every stage of tumorigenesis, progression, and recurrence. This study aimed to explore the association of lncRNA cyclin-dependent kinase inhibitor 2B antisense RNA 1 (CDKN2B-AS1) gene expression and the rs2383207A>G intronic variant with breast cancer (BC) risk and prognosis and to verify the molecular role and networks of this lncRNA in BC by bioinformatics gene analysis. Serum CDKN2B-AS1 relative expression and rs2383207 genotypes were determined in 214 unrelated women (104 primary BC and 110 controls) using real-time PCR. Sixteen BC studies from The Cancer Genome Atlas (TCGA) including 8925 patients were also retrieved for validation of results. CDKN2B-AS1 serum levels were upregulated in the BC patients relative to controls. A/A genotype carriers were three times more likely to develop BC under homozygous (OR = 3.27, 95% CI 1.20–8.88, P = 0.044) and recessive (OR = 3.17, 95% CI 1.20–8.34, P = 0.013) models. G/G homozygous patients had a higher expression level [median and quartile values were 3.14 (1.52–4.25)] than A/G [1.42 (0.93–2.35)] and A/A [1.62 (1.33–2.51)] cohorts (P = 0.006). The Kaplan–Meier curve also revealed a higher mean survival duration of G/G cohorts (20.6 months) compared to their counterparts (A/A: 15.8 and A/G: 17.2 months) (P < 0.001). Consistently, BC data sets revealed better survival in cohorts with high expression levels (P = 0.003). Principal component analysis (PCA) showed a deviation of patients who had shorter survival towards A/A and A/G genotypes, multiple lesions, advanced stage, lymphovascular invasion, and HER2+ receptor staining. Ingenuity Pathway Analysis (IPA) showed key genes highly enriched in BC with CDKN2B-AS1. The findings support the putative role of CDKN2B-AS1 as an epigenetic marker in BC and open a new avenue for its potential use as a therapeutic molecular target in this type of cancer.

recent pregnancy, or lactation in the last 2 years and/or associated malignancy were included as controls.

Histopathological and immunohistochemical assessment
Specimens from BC tissue were histopathologically analyzed after the operations. Assessment of pathological grade and clinical stage were performed using the Elston and Ellis modification of the Scarff-Bloom-Richardson system and the American Joint Committee on Cancer (AJCC) tumor-lymph node-metastasis (TNM) staging system [25]. Assessment of hormonal receptors for molecular subtyping was investigated [26]. Estrogen receptor (ER) and progesterone receptor (PR) evident nuclear staining in ≥ 10% of the tumor cells was reported as positive (+ve), and if absent or < 10% staining was recorded as negative (−ve) [27]. At the same time, human epidermal growth factor receptor 2 (HER2)/neu expression was semi-quantified by the following membrane scoring system: 0, no staining or membrane staining in < 10% of tumor cells; 1+, faint or partly stained membranes in ≥ 10% of tumor cells; 2+, weak to moderate complete membrane staining in ≥ 10% of tumor cells; 3+, strong complete membrane staining in ≥ 10% of tumor cells; and accordingly, samples were classified into different molecular subtypes [22]. Then patients were subdivided into four molecular subgroups: luminal A, luminal B, human epidermal growth factor receptor 2 (HER2 + ), and triple-negative BC, as previously detailed [28,29].

Clinical assessment and prognostic evaluation
Clinical features, risk factor assessment, and investigations were evaluated. The Nottingham Prognostic Index (NPI) and the Immunohistochemical Prognostic Index (IHPI) were applied as previously described [28], and patients were classified accordingly to have a good, moderate, and poor prognosis. The follow-up period was extended for up to 3 years for overall survival (OS) and disease-free survival (DFS) assessment.

Sample collection
Venous blood (5 ml) samples were collected into plain and ethylenediamine tetra-acetic acid (EDTA) Vacutainers. The samples were preserved after collection for 30 min to 2 h in the refrigerator to allow blood clotting. After that, all samples were centrifuged for 10 min. The obtained serum from plain tubes was divided into aliquots and stored at − 80 °C. One freeze-thaw cycle was carried out for the samples. The EDTA tubes were used for genomic DNA analysis.

Nucleic acids extraction
From the serum samples, total RNAs were isolated via the Qiagen miRNeasy Serum/ Plasma Kit (Qiagen, Clinilab Co., Catalog no. 217184), and from the buffy coat of EDTA blood samples, the genomic DNA was isolated via the QIAamp DNA Blood Mini kit (Catalog No. 51104; Qiagen) following the instructions of the vendors. The purity/concentrations of nucleic acids were assessed using the NanoDrop ND-1000 spectrophotometer (NanoDrop Tech., Inc. Wilmington, DE, USA). Furthermore, the integrity of the nucleic acids was tested by agarose gel electrophoresis.

Reverse-transcription and quantitative polymerase chain reaction analysis
The Minimum Information for Publication of Quantitative Real-Time PCR Experiments (MIQE) guideline was followed during the quantitative real-time reverse-transcription polymerase chain reaction (qRT-PCR) runs [30]. The first step of the amplification process was the reverse transcription (RT) of 1 µg of total RNA to yield complementary DNA (cDNA) using a High-Capacity cDNA RT Kit (P/N 4368814, Applied Biosystems, Foster City, California, USA). The second step is the real-time PCR reaction using a specific TaqMan probe for the lncRNA CDKN2B-AS1 (Assay no. Hs04406279_m1), compared to the housekeeping gene glyceraldehyde 3-phosphate dehydrogenase (GAPDH) for normalization of the data. In each run, appropriate negative controls were applied: no template and no reverse transcriptase controls. The final volume of the reaction (20 µl) contained RT products (1.33 µl), 2× TaqMan Universal PCR Master Mix (10 µl), and TaqMan RNA assay (1 µl). The StepOne Real-Time PCR machine (Applied Biosystems) was set up as follows: 10 min at 95 °C, followed by 40 cycles of 15 s at 92 °C and 60 s at 60 °C [31].

CDKN2B-AS1 expression data analysis
The quantification cycle (C q ) is the cycle number at which the fluorescence passed a fixed threshold [30]. The relative amount of the study lncRNA to GAPDH in patients compared to controls was calculated using the formula: [32].

Allelic discrimination analysis
TaqMan Genotyping PCR Master Mix, No. UNG (4440043), and TaqMan single nucleotide polymorphism (SNP) Genotyping Assay Mix (assay IDC__15789010_20, Catalog# 4351379, Thermo Fisher Scientific) were used for the rs2383207 real-time allele discrimination assay. Briefly, the extracted DNA (20 ng) was diluted to 11.25 μl with DNasefree water, then added to the reaction mix containing TaqMan Master mix (12.5 μl) and TaqMan SNP Genotyping Assay (20×) Mix (1.25 µl). The PCR program was described previously [33]. The investigators were not aware of the sample status (patient versus control) during the genotyping. In each run, non-template and TaqMan enzyme negative controls were included. Ten percent of samples were tested in duplicate with a 100% concordance rate for genotype calls. The SDS software version 1.3.1 (Applied Biosystems) was applied for genotype calling.

Genomic and functional analysis of CDKN2B-AS1 gene
Gene locus analysis was performed in the Ensembl Genome Browser (http:// ensem ble. org), a database that annotates genes, transcripts, and genomic variations, computes multiple alignments, predicts regulatory function, and collects disease-related data. Variant analysis of CDKN2B-AS1 and the impact of each were determined in the Varsome web application (https:// varso me. com/), a search engine, aggregator, and impact analysis tool for human genetic variation. It displays a detailed annotation of the queried variant, including multiple notations, predicted pathogenicity status from various tools, and genomic context. Genetic variants of the CDKN2B-AS1 gene with known pathogenicity were confirmed in ClinVar and PubMed. The predicted and tested CDKN2B-AS1-related SNPs' impact on gene expression was determined from the literature and the LDHap project, a haplotype map project of the human genome that describes the common genetic patterns' variation associated with human disease. Ensembl.org and varsome. com were used to analyze the structural and functional impact of the selected rs2383207 intron variant. The Integrative Genomics Viewer (http:// igv. org/) web application was run to view the locations and Genome-Wide Association Studies (GWAS)-associated phenotypic traits in the CDKN2A/B locus.
The Gene-Tissue Expression (GTEx) Portal (http:// gtexp ortal. org) and BioGPS database (http:// biogps. org) were used to identify the gene expression pattern of the gene across diverse normal human tissues from the U133plus2 Affymetrix microarray experiments. The expression level of CDKN2B-AS1 in BC tissues with different molecular subtypes was determined from Expression Atlas (www. ebi. ac. uk/), an open resource that assists in finding information about gene and protein expression. Localization of the lncRNA was detected in the Compartment subcellular localization database (www. compa rtmen ts. jense nlab. org/).
To retrieve high-throughput experiments in public repositories, 16 BC studies (total number of patients = 8925) in The Cancer Genomic Atlas (TCGA) were downloaded from the cBioPortal for Cancer Genomics database (http:// cBiop ortal. org) to identify the rate of genetic alterations in the CDKN2B-AS1 genome and aberrant expression of the gene in BC. The online Kaplan-Meier plotter program (http:// kmplot. com) was utilized to plot Kaplan-Meier survival curves for CDKN-AS1 expression from the datasets stored in this database [34]. Next, to identify the predictive role of the CDKN2B-AS1 gene in the therapeutic response, a receiver operator characteristics (ROC) plotter tool (http:// rocpl ot. org/) was used to link gene expression and response to therapy using transcriptome-level data of 3104 BC patients [35]. Lnc2Cancer version 2.0, a manually curated database, was used to identify experimentally supported associations between lncRNA and human cancer. It provides information on circulating, drug-resistant, prognostic lncRNAs in cancer [36].
The Gene Ontology Annotation (GOA) database was used to define the gene ontology terms related to the CDKN2B-AS1 gene. Next, the Automatic Cancer Hallmarks Analytics Tool (CHAT) was used for classification of the PubMed literature according to the cancer taxonomy hallmarks. Breast cancer and CDKN2B-AS1 (data are shown as NPMI; normalized pointwise mutual information) were compared. The LncMAP lncRNA Modulator Atlas in Pan-cancer (http:// bio-bigda ta. hrbmu. edu. cn/ LncMAP/ index. jsp) was used to determine putative transcription factors (TFs). NetworkAnalyst (www. netwo rkana lyst. ca) version 3.0 was used to construct gene regulatory networks (GRN) for the upstream regulators of the CDKN2B-AS1 gene. The TF targets derived from the JASPAR TF binding site profile database and inferred from integrating the literature curated Chip-X database were identified. Also, the literature curated regulatory interaction information was collected from the RegNetwork repository [37]. IPA was employed to identify the most relevant spatial molecular interactions within the cell, predict the direction of the downstream effect in BC, and the activation and inhibition of upstream TFs. Finally, ENCORI for the RNA interactome (http:// starb ase. sysu. edu. cn/) with default settings presented the ceRNA (competing endogenous RNA) networks from multiple interactions of miRNA targets in pan-cancer supported by CLIP-seq data. The results were examined at a P-value ≤ 0.01 and the false discovery rate (FDR) ≤ 0.01 [38].

Statistical analysis
When G power-3 software (http:// www. gpower. hhu. de/) was used for study power estimation with a medium effect size, specified study design, and allowable error rate (alpha error = 0.05), it yielded 95% study power. Data were expressed as number and percentage or mean ± standard deviation for categorical and continuous data. Data were checked for normality, outliers, and skewness and were transformed in case of violation of the normality assumption. Appropriate non-parameter tests were then applied and presented as median and quartiles. Two-sided chi-square, Fisher's exact, Student t-test, one-way ANOVA, Mann-Whitney U, or Kruskal-Wallis tests were used. Construction of contingency tables was performed for genotype distribution and allelic frequency comparison. The Hardy-Weinberg equilibrium (HWE) was applied using the χ 2 test to assess genotype distribution in all study subjects. Six genetic association models were analyzed as previously described [39]. Odds ratios (ORs) with 95% confidence intervals (CIs) and corresponding P-values were computed by logistic regression analysis and adjusted by the patients' demographic, clinical, and pathological characteristics. The statistical significance was set at P-value < 0.05. All statistical analyses were executed using SPSS version 26.0 and GraphPad Prism version 8.0. SNPStats (https:// www. snpst ats. net/ start. htm) was applied for genetic analysis. Kaplan-Meier curves were plotted for CDKN-AS1 genotypes with overall survival. In RStudio, correlation plots were constructed using the corrplot package, while principal component analysis (PCA) was performed using the factoextr and FactoMineR packages.

Baseline characteristics of the study participants
The mean age of BC women was 44.3 ± 12.2 years. Only one-third (30.9%) had a positive family history of cancer. Two-thirds (63.6%) were overweight/obese. Most of the patients had a sedentary lifestyle (90%); nevertheless, only 11.8% had night shifts at work, and 7.3% of women had a prior history of breast problems. About 64.5% had early onset of menarche. A few of them had late first gravida (4.5%), nullipara (12.7%), and breastfed (16.4%), while one-third of patients were at the post-menopausal state at the time of diagnosis.

CDKN2B-AS1 expression profile
Based on screening 52 cases (who had available extracted RNA matched with the DNA samples for further genotype-expression correlation analysis) and 104 controls, circulatory CDKN2B-AS1 was upregulated in cancer patients compared to controls (P < 0.001). The median log fold change of CDKN2B-AS1 in the BC cohort was 1.82 (IQR: 1.38-3.71) relative to the controls.

Allele and genotype frequencies of CDKN2B-AS1(rs2383207) variant
A total of 110 cases and 104 controls were genotyped. Genotype frequencies followed the Hardy-Weinberg equilibrium in controls (P = 0.80) and cases (P = 0.06). Minor allele frequency (A allele) was 31% in the overall population and 26% in controls. Apart from the African population, allele frequencies were similar to those of different populations in the 1000 Genome Project Phase 3 (http:// ensem bl. org) (Fig. 2a). In our study population, half of the patients carried G/G genotype (107 cohorts), while A/G heterozygotes exceed a third (83 patients, 39%). A higher frequency of the A allele was observed in BC patients (77, 35%) compared to controls (54, 26%), P = 0.042 (Fig. 2b). Similarly, A/A homozygosity was the most frequent genotype among patients, representing 16% (18 patients) compared to 6% (6 patients), P = 0.047 (Fig. 2c).

CDKN2B-AS1 genotypes modulate lncRNA expression in the breast cancer cohort
Although different CDKN2B-AS1 genotypes in the control group showed no significant difference in gene relative expression (P = 0.069) (Fig. 2d), the G/G genotype was associated with a higher expression level than A/G in the BC cohort; median and quartile values were 3.14 (1.

Association of CDKN2B-AS1 with clinicopathological features
Apart from body weight, in which overweight and obesity were more prevalent among G/G cohorts (P = 0.004), no significant differences in BC risk factors among patients with different genotypes were found ( Table 2).

Fig. 2
Genotype and allele frequencies of CDKN2B-AS1 (rs2383207) variant and association with gene expression. a-c A two-sided Chi-square test was used. d, e Spearman's correlation analysis was done, revealing the coefficient correlation value (r) and its P-value. Kruskal-Wallis (KW) test was applied to test for significant difference between cohorts with genotypes. Multiple comparison test was applied for pairwise comparison. Statistical significance was set at P-value < 0.05 Immunohistochemical analysis was conducted for tumor samples. Selected samples with positive estrogen, progesterone, and HER2 + receptors are illustrated in Fig. 3.
Regarding the pathological characteristics of BC tumors, no significant difference was observed among patients with various genotypes. However, the rs2383207*A variant was associated with poor survival. Higher frequencies of A/A and A/G genotypes were found in cohorts with short disease-free survival (P = 0.009) and overall survival (P < 0.001) ( Table 2). Consistently, the principal component analysis showed a deviation of patients who had shorter survival towards A/A and A/G genotypes, multiple lesions, advanced stage, lymphovascular invasion, and HER2 + receptor staining (Fig. 4a). Kaplan-Meier curve analysis also revealed a higher mean survival duration of G/G cohorts (20.6 months) compared to their counterparts (A/A: 15.8 months and A/G: 17.2 months), P < 0.001 (Fig. 4b).
In an association of the transcriptomic signature of CDKN2B-AS1 with clinicopathological features, the lower expression level (1.64 ± 1.85) was associated with poor pathological grade compared to well-differentiated and intermediately differentiated tumors (2.65 ± 1.60, P = 0.004) in breast cancer patients. However, there were no other significant associations with any other clinical or pathological parameters (Fig. 5).

Structural analysis of CDKN2B-AS1 gene
Although the CDKN2B-AS1 gene locus includes other protein-coding genes, i.e. CDKN2B (encoding p15ink4b) and CDKN2A (encoding p16ink4a and p14ARF) (Fig. 1b) with a key role in cell cycle inhibition, senescence, and stress-induced apoptosis, prior studies highlighted that SNPs in this locus act through effects on CDKN2B-AS1 itself. The gene has 28 splice variant linear and circular transcripts ranging from 602 bp up to 7173 bp and includes LINE, SINE, and Alu repetitive sequences. Some are tissue-specific, while others have distinct roles in cell physiology.

Variant analysis of CDKN2B-AS1 gene
The studied intronic variant rs2383207A>G in the CDKN2B-AS1 gene is located at chromosome 9: 22115959 (forward strand) 2161 bp away from the splice site with a minor allele frequency (G allele) of 0.31. This SNP overlaps 15 transcripts of the lncRNA and is predicted to be benign (Additional file 1: Table S2). Common disease genome-wide association studies (GWAS) have identified the CDKN2B-AS1 gene as a shared locus for genetic susceptibility to multiple cancers (Fig. 1d).

Transcriptomic profile of CDKN2B-AS1 gene
RNAseq analysis of 27 different normal tissues showed higher expression of the colon and small intestine and down-regulation in normal breast tissues (Fig. 6a). However, in BC tissues, overexpression was observed in all molecular subtypes (Fig. 6b). CDKN2B-AS1 accumulates in both the nucleus and cytoplasm. The family of linear CDKN2B-AS1 contains proximal (exon 1) and distal (exon 13b, 19) exons, and is enriched in the nucleus. However, the circular isoforms usually contain the middle exons (exons 5, 6, and 7), are enriched in the cytoplasm, and they differed markedly in their stability. Circular lncRNAs were reported to be associated with ribosome biogenesis and nucleolar stress, while nuclear isoforms are more likely to be involved in regulating gene transcription via chromatin modulation (Fig. 6c). Sixteen BC studies from the TCGA including 8925 patients were retrieved. Of these, 146 patients (2.1%) had a genetic alteration in the CDKN2B-AS1 gene (44 amplification and 102 deep deletions). Somatic mutations were not studied in these datasets. Higher expression levels were encountered in patients with more copies of the genes and vice versa. Patients with altered CDKN2B-AS1 had poorer survival than their counterparts (P = 0.031) (Fig. 6d-f ).

Prognostic and predictive role of CDKN2B-AS1 gene in breast cancer
Despite the up-regulation of the lncRNA in BC tissues and blood, a higher expression level of the gene was associated with better overall survival (Fig. 7a-e). The prognostic performance of lncRNA expression, represented as a ROC curve, is shown in Fig. 7f-h. The utility of CDKN2B-AS1 as a diagnostic and prognostic biomarker is illustrated in Table 3. Articles showing the association of CDKN2B-AS1 up-regulation with treatment resistance are shown in Table 4. The top 30 drug-lncRNA pairs were used to build the Drug-LncRNA Network. Of the drugs associated with CDKN2B-AS1, Lapatinib, a dual inhibitor of EGFR and HER2, is indicated for patients with advanced or metastatic BC treatment whose tumors overexpress HER2. Other tyrosine kinase inhibitors such as erlotinib and sorafenib were also connected to the lncRNAs. The single drug topotecan was also indicated in patients with BC (Fig. 8).

Functional enrichment analysis of CDKN2B-AS1 gene
Functional enrichment analysis revealed CDKN2B-AS1 as a key player in regulating gene expression and cell migration. Being within the INK4b-ARF-INK4a gene family, which encodes the p15, p14, and p16 tumor-suppressor proteins, respectively, it is transcriptionally silenced or homozygously deleted in various human cancers. These three proteins are implicated in apoptosis, senescence, and stem cell renewal via promoting anti-proliferative and pro-apoptotic activities of Rb1 and p53 (Fig. 9a).  We used CHAT to analyze PubMed literature on BC and our studied gene. Cell invasion and metastasis was the most common hallmark associated with BC literature, followed by sustaining proliferative signaling. CDKN2B-AS1 studies have two main hallmarks similar to those of the BC profile: sustaining proliferative signaling and genome instability and mutation (Fig. 9b). Upstream regulators enhancing transcription of the   CDKN2B-AS1 gene are presented in Fig. 9c. The LncMAP revealed that CDKN2B-AS1 could also bind to 12 TFs: E2F4, REST, BCL11A, RARA, ETS1, ESR2, GATA6, FLI1, EBF1, CTNNB1, STAT2, and ESRRA. As depicted in Fig. 9d, Ingenuity Pathway Analysis showed key genes highly enriched in BC with CDKN2B-AS1, namely (1) RELA proto-oncogene, (2) prostaglandin-endoperoxide synthase 2 (PTGS2), (3) interleukin 6 (IL6), (4) vascular endothelial growth factor A (VEGFA), and (5) tumor necrosis factor-alpha (TNF).

Discussion
The last two decades have witnessed remarkable developments in the era of non-coding RNAs [40]. Given their high stability in different storage and handling conditions, being easily monitored by repeated sampling and their circulatory levels mirroring those in tissue cancer, serum/plasma lncRNAs demonstrated features of particular relevance for ideal biomarkers [41].
In this study, we evaluated the impact of serum lncRNA CDKN2B-AS1 expression and variant signature in BC patients supported by in silico analyses followed by verifying the results on samples from TCGA. We found that CDKN2B-AS1 is upregulated in sera of BC patients compared with controls. This observation was in line with the oncogenic role this lncRNA plays in several cancers (Additional file 1: Table S3), which can support the potential role this type of lncRNA can play as a universal biomarker differentiating cancer from non-cancer patients. CDKN2B-AS1 is known to be involved in transcriptional repression through forming chromatin modification complexes that execute histone modifications at specific sites. It can interact with both chromobox 7 (CBX7), a polycomb repressor component within PRC-1, and SUZ12, a subunit of PRC2 [42]. Next, PRC2 downregulates INK4 expression (Fig. 9a) by inducing H3K27 tri-methylation. Meanwhile, PRC1 maintains the repressive chromatin structure by H2AK119 mono-ubiquitination [15,43].
CDKN2B-AS1 is mainly upregulated by the E2F1 in an ATM-dependent manner after DNA damage, leading to cell cycle arrest to allow for DNA repair [44]. Several known potent oncogenes that regulate CDKN2B-AS1 expression in various cancers have been reported in the literature, including MYC, RELA, and ERBB2 [12,45]. CDKN2B-AS1 expression is also regulated by interferon-gamma (IFNγ), tumor necrotic factor (TNFα) [46], and interleukin 1 beta (IL1B). These inflammatory mediators activate the nuclear factor kappa-B (NF-κB), a pro-proliferation and survival factor, pathway and form a complex with the YY1 transcription factor to create transcriptional regulatory loops [47]. CDKN2B-AS1 abundance can also be induced by exposure to hypoxia (a wellknown phenomenon associated with the tumor microenvironment). It can bind the aryl hydrocarbon receptor nuclear translocator (ARNT), which functions as a transcriptional regulator of the adaptive response to hypoxia [48]. Apart from the potential oncogenic role of CDKN2B-AS1 in BC, all the above mechanisms could support CDKN2B-AS1 upregulation in breast cancer in part as an adaptive response to DNA and cellular damage during the tumorigenesis process, which could explain the association of high expression of this type of lncRNA with survival and cancer grade; the findings were validated by the same results from the TCGA data set. Interestingly, using RNAscope, a recent study underscored that presence of CDKN2B-AS1 in different subcellular locations in breast tumors may affect its functionality in cancer progression [49]. Our in-silico analysis, including the Ingenuity Pathway Analysis, showed several key genes highly enriched with CDKN2B-AS1 in BC and could mediate part of its oncogenic role or impact the tumor microenvironment in this type of cancer. These genes include (1) the RELA (v-rel avian reticuloendotheliosis viral oncogene homolog A) gene encoding the NF-κB-p65 subunit (a ubiquitous TF held in the cytoplasm in an inactive state by a specific inhibitor), upon degradation of which the NF-κB translocates to the nucleus and activates specific gene expression, (2) PTGS2 encoding a major enzyme in prostaglandin (PTG) biosynthesis implicated in biosynthesis of prostanoids that are involved in inflammation and mitogenesis, (3) IL6, which functions in inflammation and B-cell maturation, (4) VEGFA, which stimulates proliferation and migration of vascular endothelial cells and is a key player in physiological/ pathological angiogenesis, and (5) the TNF gene that regulates cell proliferation, differentiation, apoptosis, lipid metabolism, and coagulation. It is worth noting that the transcriptomic abundance of CDKN2B-AS1 is not only modulated by epigenetic control through promoter transcriptional activity, promoter methylation, and splicing [50] but also post-transcriptionally regulated by miRNA sponging and RNA stability [12].
Using the ENCORI database, we identified the crosstalk of CDKN2B-AS1 lncRNA with mRNAs and transcribed pseudogenes in cancer. Putative lncRNA-ceRNA interactions in cancer represent a new layer of gene regulation (Table 5). Mounting data from the literature suggest that deregulated CDKN2B-AS1-miRNA interacting networks have also been implicated in carcinogenesis, thus representing another molecular mechanism of the lncRNA. CDKN2B-AS1 contains miRNA-binding domains in its sequence and therefore acts as a "sponge" to sequester miRNAs away from its mRNA targets. Through modulation of microRNA pathways, CDKN2B-AS1 can also be involved in post-transcriptional regulation ( Table 6). miRNA sponge mechanism was reported in BC patients with triple-negative receptors through interaction with microRNA-199a [51]. All the aforementioned cellular and genetic/epigenetic mechanisms support CDKN2B-AS1 involvement in breast carcinogenesis and suggest that it could be a biomarker for BC detection with other molecular panels.
As one of the well-known lncRNAs that occur in genomic loci that harbor many cancer-associated SNPs [12], the CDKN2B-AS1-related variant rs2383207 showed, for the first time, an association with BC risk in the present study. The A/A homozygous carriers were three times more liable to develop BC under dominant and recessive genetic models. They were associated with short disease-free/overall survival compared to their counterpart G/G homozygous carriers. Intriguingly, the biallelic study variant was also associated with the gene expression, as G/G genotype was associated with a higher expression level than other genotypes (A/G and A/A).
Studies indicate that SNPs in CDKN2B-AS1 can impact its expression and function with potential risk modification for diseases, including cancers [18,52]. As most of the polymorphisms in this region do not impact any protein sequence, they likely act by affecting the expression of a nearby gene in cis, as proposed by Cunnington et al. [18]. Furthermore, as our in silico analysis showed, the existence of key genes highly enriched in BC with CDKN2B-AS1 and upstream regulators enhancing the transcription of the CDKN2B-AS1 gene (as detailed in Fig. 8) collectively suggests a wide "regulatory panorama" for this type of lncRNA [53]. Notably, it is not proved that the risky allele (the A allele in our case) is associated with high gene expression to increase BC susceptibility. As demonstrated in our in silico analysis and reported previously, the CDKN2B-AS1 variants could act through "independent mechanisms" to increase/ decrease susceptibility to diseases [53], and more than one functional variant as well as others in disequilibrium with the specified study variant, collectively, could impact CDKN2B-AS1 expression [18]. In agreement with our results, the T allele of the rs2151280 variant (one of the CDKN2B-AS1 SNPs) was related to increased susceptibility to neurofibromatosis type 1, associated with gene downregulation [54]. Also, the melanoma-associated rs1011970-T variant was reported to be associated with CDKN2B-AS1 downregulation [18]. In contrast, the glioma-associated risky rs1063192-C allele was associated with increased CDKN2B-AS1 expression [16,18]. The presence of multiple regulatory elements and binding sites in the region of the studied variant suggests that its transcript, CDKN2B-AS1, with other genes in the same locus as CDKN2A/B, is subject to intricate "temporal and tissue-specific regulation" underlying the gene associations [53]. Based on the limited study cohort, future large-scale studies in other populations are warranted to confirm the studied variant's association with BC risk. Also, supporting in vitro studies are needed to confirm the impact of different rs2383207 genotypes on CDKN2B-AS1 expression. Finally, follow-up studies are warranted to assess the specified lncRNA's validity as a diagnostic/prognostic epigenetic marker and the response to chemotherapy in BC patients.

Conclusion
In summary, the study findings for the first time support the association of the CDKN2B-AS1 rs2383207 (A>G) variant with gene expression and BC risk in the study population. It might provide new insight into the molecular stratification of BC patients. Also, high levels of circulating CDKN2B-AS1 could discriminate BC