Long non-coding RNAs PTENP1, GNG12-AS1, MAGI2-AS3 and MEG3 as tumor suppressors in breast cancer and their associations with clinicopathological parameters

3.4.1Age, stage and grade

No association was determined between the expression data and the age.

Concerning the tumor stage, we determined a marginally significant association for GNG12-AS1 (p= 0.062741), which revealed a significant lower the expression the more advanced the stage trend (p𝑡𝑟𝑒𝑛𝑑= 0.04101). For details see Figs 2 and 3.

Concerning the clinical stage, lower levels of GNG12-AS1 were significantly associated with more advanced stages (p= 0.044762, p𝑡𝑟𝑒𝑛𝑑= 0.01188), revealing significant differences between clinical stage 1 and stages 3 and 4, and clinical stage 2 and clinical stage 4 (p< 0.05).

We determined significant associations between the lncRNA expression and the grade for three lncRNAs. Significant differences were observed particularly between grades 1 and 3 (GNG12-AS1, p= 0.0433, PTENP1, p= 0.0094 and MAGI2-AS3, p= 0.0094), the latter also revealing a difference between grades 1 and 2 (p= 0.0322). A lower lncRNA expression was recorded for the more advanced grades in all these associations. The course of the lncRNAs expression changes in relation to the tumor grade is depicted in Fig. 4.

Figure 4.

Relative expression of long non-coding RNAs compared among different tumor grades of breast cancer (Validation experiment). Notes: A clustered multiple-comparison graph showing mean lncRNA expression data (log-transformed CNRQ) as bar charts. Error bars indicate 95% CI for the mean.

3.4.2Multifocality and lymph node metastasis (LNM)

A significant association between higher expression and multifocality was determined for UCA1 (p= 0.0206) (Figure S1), while the low expression of MEG3 was marginally significantly (p= 0.0672) associated with this clinical parameter. Only a marginally significant association between LNM positivity and higher expression was noted for UCA1 (p= 0.0709).

3.4.3Ki-67 proliferation marker, estrogen (ER) and progesterone (PR) receptors

Both MEG 3 and PTENP1 indicated that their lower expression was linked significantly with the Ki-67-positive samples when the two common cut-offs 20% (MEG3, p= 0.0328, Figure S2; PTENP1, p= 0.0447, Figure S3) and 15% (MEG3, p= 0.0401; PTENP1, p= 0.0132) were applied.

Figure 5.

Progression-free survival in relation to PTENP1 expression levels. Notes: Univariate Kaplan-Meier survival curves for progression-free survival (PFS) related to low and high concentrations of PTENP1 in breast cancer tumors. Mean PFS for the low expression subgroup was 3.3 years (171.7 weeks), and for the high expression subgroup it was 4.9 years (257 weeks).

Figure 6.

Progression-free survival in relation to GNG12-AS1 expression levels. Notes: Univariate Kaplan-Meier survival curves for progression-free survival (PFS) related to low and high concentrations of GNG12-AS1 in breast cancer tumors. Mean PFS for the low expression subgroup was 3.5 years (183.8 weeks), and for the high expression subgroup it was 4.6 years (240 weeks).

ER positivity (10% cut-off) was significantly linked with the elevated expression of NRSN2-AS1 (p= 0.0499, Figure S4), and also marginally significantly for GNG12-AS1 (p= 0.0766). A similar (but only marginally significant) pattern was observed for the dual positivity of ER and PR (10% cut-off) and the increased expression of these two lncRNAs (GNG12-AS1, p= 0.0611; NRSN2-AS1, p= 0.0738).

A marginally significant association was noted for PR negativity (10% cut-off) and the lower expression of GNG12-AS1 (p= 0.0682).

3.6.1TCGA-based data

Tumors versus controls

First, we used a combined TCGA GTEX data set as the primary dataset based on TCGA RNAseq data for breast cancer available for all six investigated lncRNAs. This dataset includes 1092 primary invasive breast cancer samples (including both ductal and lobular carcinomas), 7 metastatic, 113 adjacent (benign) and 179 normal tissue (from patients without cancer) samples.

Figure 7.

Relative expression of lncRNAs in primary tumors, metastatic samples, benign (“adjacent normal”) and normal mammary tissues in TCGA GTEX data set focused on invasive breast cancer samples.. Notes: A box of the box-plot is drawn from the 1st to 3rd quartile (the 25th and 75th percentiles). A horizontal line within a box plot represents the median. Horizontal lines are drawn at the highest value and the lowest expression value. Circles and squares represent outside values.

The expression differences in this dataset generally corresponded with our results concerning both down- and upregulations for all tested lncRNAs (see Table S5 and Fig. 7). Interestingly, PTENP1 expression was downregulated comparing primary tumors with adjacent normal tissue (similarly as in our study) but upregulated between primary tumor and normal tissue. This pattern was unique among the tested lncRNAs. However, it should be noted that this dataset contains data for all invasive breast cancer specimens.

Next, using GDC TCGA Breast Cancer (BRCA) dataset we selected invasive ductal carcinomas and compared their lncRNAs expression among four different stages groups (I-IV) available for 748 samples. Two investigated lncRNAs (MAGI2-AS3 and GNG12-AS1) evinced significant results. MAGI2-AS3 had the significant different expression between stage I and stages II, III and IV (p= 0.003178) and a significant trend to lower expression in more advanced stages (p𝑡𝑟𝑒𝑛𝑑= 0.01595), GNG12-AS1 showed a difference between stage I and stages II and III (p= 0.000079) with a similar trend (p𝑡𝑟𝑒𝑛𝑑= 0.00002). Nevertheless, no significant associations were observed in Kaplan-Meier survival analyses focused on overall survival using this data set. The lncRNA expression alterations among the different stages is presented in Figure S7.

Finally, we explored TCGA and TARGET Pan-Cancer dataset including invasive breast cancer samples using Xena hub. The expression of lncRNAs was compared according to BRCA-PAM50 characteristics (50-gene signature) of 834 samples designated as Her2, Luminal A, Luminal B, Normal and Basal subtypes [18]. As a result, significant differences were observed for all but one lncRNA (UCA1) among different subtypes (Table S5).

For example, apparently higher expression levels between normal subtype (along with luminal A) and other subtypes were demonstrated for GNG12-AS1, MEG3 and MAGI2-AS3. PTENP1 expression was lowest in the basal subtype (and luminal B) samples. NRSN2-AS1 evinced highest expression in luminal B (and luminal A) samples and the lowest in normal samples. As HER2-enriched and basal-like subtypes are considered more aggressive diseases, and luminal A should have a better prognosis than luminal B subtype, these results also supported our findings in the present study. For details see Table S6.

3.6.2Geo datasets

Eight eligible Geo datasets were selected to analyze microarray expression data using breast cancer cells (GSE25976, GSE31812, GSE24460) or tissues (GSE5764, GSE23988, GSE22093, GSE3893, GSE27447). In the GSE22093 dataset, we found a significant association of grade and MEG3 expression (n= 79, lower expression in grade 3 as compared with grade 2 (p= 0.009736)). In the GSE25976 dataset, measurements of the UCA1 expression showed increased levels in metastatic variants (231BoM-1833, 231BrM-2a) of the parental cell line MDA-MB-231. The remaining datasets had only limited data as regards sample numbers and availability for the tested lncRNAs as well as a specific functional focus, therefore no significant results were obtained.

4.1.1Potential tumor suppressor lncRNAs

The data obtained in this study from the screening and validation experiments confirmed that PTENP1, GNG12-AS1, MAGI2-AS3 and MEG3 may have, or provide support for, potential tumor suppression roles in breast cancer. Of these 4 lncRNAs, the most remarkable underexpression (minus 8-fold) was noted for PTENP1 in the comparison of the tumors with the benign samples. This lncRNA also performed well in terms of its diagnostic performance, evincing the highest values for AUC (0.958), sensitivity (86.21%) and specificity (100%). It was also possible to infer its tumor suppressor role from the association between its lower expression and the highest grade. Further, it evinced a lower expression in the Ki-67 positive samples (both cut-offs, i.e. 15% and 20%). Ki-67 is known to be a proliferation marker that worsens patient outcomes and, previously, we found it to be associated with the levels of certain microRNAs in the plasma samples of breast cancer patients [19]. Moreover, the higher expression of PTENP1 in the survival analysis was significantly associated with improved overall survival (the marginally significant difference in the OS was 10.2 months), as well as with enhanced progression-free survival (a significant difference in the PFS, i.e. 19.6 months). Several studies that focused on PTENP1 published to date provide additional support for the role of PTENP1 as a tumor suppressor in breast cancer. Using experimental cell lines and animal models, Yndestad et al. [20] investigated the effects of PTENP1 transduction on ER-positive MCF7 and T47D and ER-negative MDA-MB-231 and C3HBA breast carcinoma cells, and determined divergent effects. PTENP1 upregulation acted to decrease the PTEN gene expression and to accelerate MCF7 tumor growth in vivo in both ER-positive cell lines. On the other hand, the upregulation of PTENP1 resulted in an increased PTEN gene expression and reduced metastatic potential in the two ER-negative cell lines. Moreover, PTENP1 inhibited the growth rate of ER-negative C3HBA murine breast cancer xenografts [20]. Recently, Yi et al [21] analyzed MDA-MB-231 cells treated with YPB, OPB and control peptides. YPB and OPB peptides reduced breast cancer cell viability and migration, and induced apoptosis. Additionally, both peptides inhibited xenograft tumor growth in nude mice. Finally, the authors discovered that PTENP1 showed markedly reduced H3K27me3 signal. The treatment by YPB and OPB peptides resulted in PTENP1 and PTEN transcripts upregulation in MDA-MB-231 cells and also in the mouse xenograft samples. Experimental reduction of PTENP1 led to opposite effects and enhanced cell viability. The results confirmed that YPB and OPB peptides may decrease EZH2 recruitment to the promoters of target genes, such as PTENP1 and enhance anticancer activities reducing breast cancer cell proliferation and xenograft tumor formation.

Shi et al. [22] determined lower PTENP1 levels in breast cancer tissues and cell lines and its experimental overexpression in BC cell lines (MCF-7 and MDA-MB-231), increased cell survival, colony formation, migration and invasion, but decreased apoptosis. Gao et al. [23] reported underexpressed PTENP1 (and PTENP) in breast cancer tumors compared to the adjacent tissues, as well as in MDA-MB-231 and ADR resistant BC cell MCF7/ADR and T47D/ADR cell lines. The lower expression of both PTENP1 and PTEN also correlated with advanced stage and poor prognosis. Experimentally overexpressed PTENP1 in BC cell lines resulted in limited breast cancer cell viability and reduced proliferation, migration and invasion capabilities. The reduced expression of Ki-67 (as in our clinical results) was also noted; thus, all the results demonstrated the tumor suppressive role of PTENP1 in BC cells [23]. Similarly, Li et al. [24] demonstrated the lower expression of PTENP1 (and PTEN) in breast cancer tissues, as well as of PTENP1 in all the BC cell lines (BT-20, MCF-7, MDA-MB-231 and T-47D). The authors also noted the inhibition of proliferation, migration and invasion in transfected cancer cells with increased PTENP1. With respect to the first analysis of PTENP1 in human breast cancer, Yndestad et al. [25] found no association between PTENP1 and the chemotherapy or survival responses. The published data also suggests a tumor suppressor role for other cancers. PTENP1 has been observed to be downregulated in the sera of patients with gastric cancer [26] and, with the exception of carcinomas, in osteosarcoma [27].

GNG12-AS1 comprises a further lncRNA that we identified as a potential tumor suppressor due to its evincing a lower expression (around minus 3-fold) in the tumors. Several indications served to support this role, i.e. a trend toward its low expression in more advanced stages and its underexpression in the grade 3 samples; the low expression level of GNG12-AS1 was linked to both worsened overall survival (a marginally significant difference of 5.4 months) and progression-free survival (a significant difference of 12.9 months).

It has been demonstrated previously that the transcription of DIRAS3 and GNG12-AS1 is coordinately downregulated in breast cancers [28]. Functionally, GNG12-AS1 is known to be a nuclear lncRNA and is transcribed in antisense orientation to the tumor suppressor DIRAS3 [29]. Their experimental data suggested that GNG12-AS1 may have a dual function in terms of controlling cell cycle progression. The concomitant transcriptional upregulation of DIRAS3 was induced by GNG12-AS1 silencing, thus indicating transcriptional interference. Moreover, as observed in breast cancer cell lines, GNG12-AS1 has the potential to reduce cell migration after its knockdown, but independent of DIRAS3. The authors further focused on the mesenchymal epithelial transition factor (MET), which is known to regulate cell migration and invasion, and determined that GNG12-AS1 regulates MET signaling independently of DIRAS3, which indicates that GNG12-AS1 transcripts are cell migration inhibitors [29]. No other data is available that indicates or elucidates the impact of GNG12-AS1 on breast cancer.

Concerning our study, MEG3 was underexpressed (around minus 3.8-fold) in the tumors, which suggests its role as a further tumor suppressor lncRNA in breast cancer. It was found (at lower levels) to be significantly associated with the Ki-67 positive samples (both cut-offs, i.e. 15% and 20%) and marginally significantly associated with multifocality. According to the literature, this lncRNA is known for its tumor suppressor roles in breast cancer, which is consistent with our research results. It has been shown that it may act to inhibit cell proliferation and invasion in breast cancer cell lines, as well as in vivo tumorigenesis and angiogenesis in a nude mouse xenograft model [30]. Similar findings were reported by Zhu et al. [31], according to whom MEG3 suppressed cell proliferation and glycolysis and induced apoptosis breast cancer cells. MEG3 was further shown to be a molecular sponge for miR-21 (suppressing the expression of this oncogenic miRNA); moreover, an in vivo experiment also proved that overexpressed MEG3 acted to inhibit tumor growth in breast cancer by suppressing miR-21. Similarly, recent findings revealed that the knockdown of DNMT1 inhibited the progression of breast cancer cells by enhancing the MEG3 expression through demethylation [32]. Recently, Pan et al. [33] focused on similar aspects and confirmed that MEG3 was epigenetically methylated in primary breast tissues and cells, while being unmethylated in normal breast tissues and cells. As the reduced MEG3 expression resulted from the promoter methylation, inhibition of DNA methylation reversed MEG3 expression, inhibited cell proliferation and promoted cell apoptosis. MEG3 expression was negatively correlated with DNMT1 and DNMT1 knockdown increased MEG3 expression and inhibited tumor growth in mice tumor model. Overall, the results supported tumor suppressor role of MEG3 in breast cancer.

Adopting the data mining approach, Cui et al. [34] showed that DNA methylation is a factor that contributes to decreased MEG3 expression in breast cancer samples, and the positive correlation between the MEG3 expression and the estrogen receptor (ER) and the progesterone receptor (PR) status were noted. Moreover, higher levels of MEG3 were associated with enhanced overall survival (OS), relapse-free survival (RFS), distant metastasis-free survival (DMFS), and disease-specific survival (DSS) in breast cancer [34]. A similar impact on survival was reported in an earlier study [35]. Interestingly, MEG3 was seen to be highly expressed in the triple negative metastatic human Hs578T breast cancer cell line [36]. The inhibitory effects of MEG3 on cancer progression have been reported in several other reports concerning breast cancer (e.g. [37, 38, 39, 40, 41]) and other cancers such as squamous cell carcinoma of the head and neck, and lung, gastric, hepatocellular and colorectal carcinomas (see [34, 42]).

MAGI2-AS3 was found to be the second most highly underexpressed lncRNA in the tumors (minus 5.9-fold) in this study. This lncRNA was associated with the grade (its lower expression in more advanced grades). Several other reports have indicated its downregulation in breast cancer samples and its exerting of tumor-inhibitory functions support the view of this lncRNA as a tumor suppressor in breast cancer. Yang et al. [43] showed that MAGI2-AS3 significantly inhibited breast cancer cell growth and simultaneously enhanced the expression of Fas and Fas ligand (FasL). Du et al. [44] reported the inhibitory effects of MAGI2-AS3 on the migration and invasion of breast cancer cells, while its increased expression inhibited miR-374a and enhanced the expression of PTEN. Similar effects were observed by Xu et al. [45] who revealed MAGI2-AS3 as an inhibitor of cell proliferation and migration, while promoting apoptosis in MCF-7 breast cancer cells. The authors also showed that overexpressed MAGI2-AS3 is able to upregulate MAGI2 and inhibit the Wnt/β-catenin pathway, and that the DNA demethylase TET1 inhibitor is able to reverse the overexpression of MAGI2-AS3 [45]. The higher expression of MAGI2-AS3 may be associated with the enhanced relapse-free survival of triple-negative breast cancer [46]. Recently, Zhang et al. [47] demonstrated that MAGI2-AS3 encodes ORF5 polypeptide (MAGI2-AS3-ORF5) which acted as an anti-tumor peptide to hamper BRCA cell viability, proliferation, and migration. Further, MAGI2-AS3-ORF5 may interact with ECM-associated proteins and modulate breast cancer cell migration.

Other published studies on MAGI2-AS3 in various cancers revealed similar tumor suppressor functions, e.g. for non-small cell lung cancer [48], hepatocellular carcinoma [49], and glioma [50].

The opposite, i.e. oncogenic functions of MAGI2-AS3 were suggested in connection with nasopharyngeal carcinoma [51]. This study not only found that MAGI2-AS3 was elevated in NPC cell lines (especially CNE1 and SUNE1), but also determined that it promoted cell proliferation, migration and EMT processing, and even contributed to the cisplatin resistance of NPC cells, thus clearly indicating its oncogenic functions [51]. A similar report was also published for cervical squamous cell carcinoma where it was found to be overexpressed and to experimentally promote cell proliferation [52]. Both reports indicated that this lncRNA may exert divergent functions in various cancers and the associated factors should be considered in future studies.

4.1.2Potential oncogenic lncRNAs

Two lncRNAs, i.e. NRSN2-AS1 and UCA1 were included in the validation experiment as potential oncogenic lncRNAs. Both these lncRNAs evinced ambiguous results with respect to the level of upregulation (usually below 2-fold) and their diagnostic performance. Regarding the clinicopathological data, the higher expression of UCA1 was associated with multifocality and marginally significantly with lymph node metastasis. Elevated levels of NRSN2-AS1 were linked to ER positivity.

Very little research data has been published to date on the newly-discovered lncRNA NRSN2-AS1. Xu et al. [53] reported that NRSN2-AS1 was upregulated in tumor tissues (however, according to the presented graphs, the fold-differences do not appear too large) and promoted cell proliferation, migration and invasion in esophageal squamous cell carcinoma [53]. According to another study, this lncRNA was also upregulated in ovarian cancer and its knockdown contributed to the inhibition of the migration and invasion of OV cells [54]. Conversely, NRSN2-AS1 was reported to be downregulated in hepatocellular carcinoma; however, the exact level of this alteration has not been published [55]. Further research will be needed to examine the exact functions of NRSN2-AS1 for breast cancer and other cancers.

The track record of UCA1 in the literature is extensive and suggests oncogenic functions in breast cancer. For example, Tuo et al. [56] examined the interaction of UCA1 with miR-143 and found that upregulated UCA1 is able to modulate breast cancer cell growth and apoptosis via the downregulating of this miRNA. Li et al. [57] revealed that UCA1 may upregulate PTP1B and downregulate the miR-206 expression, while enhancing cell proliferation in breast cancer cells. UCA1 may also be enhanced in tamoxifen resistant LCC2 cells and their released exosomes, and the exosome-mediated transfer of UCA1 may significantly increase tamoxifen resistance in ER-positive MCF-7 cells [58]. The activity of UCA1 in tamoxifen resistance was also reported by Li et al. [59]. The enhancing effects of UCA1 were reported regarding both paclitaxel resistance [60] and doxorubicin resistance [61]. However, the impact of UCA1 on resistance to various therapeutics had already been noted in several previous investigations (see [62]).

Xiao et al. [63] experimentally applied UCA1 knockdown in the MDA-MB-231 breast cancer cell line (cells with a strong invasion capability) revealing both impaired mesenchymal properties and a reduced number of invading cells. Conversely, UCA1 upregulation enhanced the invasiveness of breast cancer cells.

Similarly, the oncogenic functions of UCA1 have also been noted in other cancers, e.g. lung cancer (e.g. [64]), cervical cancer (e.g. [65]), ovarian cancer (e.g. [66]) and a variety of other cancers (see [67]).