Epigenetic regulation of LINC00473 by promoter methylation in colorectal cancer

CpGI hypermethylation in the promoter of LINC00473 has been recently described in the colorectal cancer cell line HCT-116 using the Infinium HumanMethylation450 (450K) microarray (Illumina) [14]. To confirm this epigenetic feature, we first analyzed the methylation status of LINC00473 in HCT-116 cells in comparison with normal colon mucosa by bisulfite genomic sequencing of multiple clones (Fig. 1A). Importantly, the CpGI promoter region of LINC00473 that we analyzed contains CpGs included in the 450K array. As expected, we observed hypermethylation of all the CpGs analyzed in HCT-116 cells compared to unmethylated normal colon mucosa. The effect of the CpGI hypermethylation on gene expression (Fig. 1B) was analyzed by qRT-PCR, showing the downregulation of LINC00473 in HCT-116 cells with respect to unmethylated normal colon mucosa. Importantly, the use of the demethylating agent 5-aza-2’-deoxycytidine (AZA) in HCT-116 cells restored the expression of LINC00473 (Fig. 1B, C). Thus, we extended the methylation and expression analysis to other common colorectal cancer cell lines (Fig. 1D) using 450K array and RNA-seq datasets obtained from Gene Expression Omnibus (GEO): GSE49143 and GSE138734, respectively. These analyses confirmed CGI promoter hypermethylation and downregulation of LINC00473 in all CRC cell lines analyzed (COLO-205, HCC-2998, HCT-116, HCT-15, HT-29, KM12, and SW-620) in comparison with normal colon mucosa.

Fig. 1

Epigenetic silencing of the LINC00473 in colorectal cancer cells. A Bisulfite genomic sequencing analysis of LINC00473 promoter CpG island in the colorectal cancer cell line HCT-116 and normal tissue. Locations of CpG dinucleotides (vertical lines) and the TSS (long black arrow) are shown. Ten single clones are represented for each sample. The presence of unmethylated and methylated CpGs is indicated by white and black squares, respectively. B DNA methylation-associated transcriptional silencing of LINC00473 in the colorectal cancer cell line HCT-116. Expression levels of LINC00473 were determined by qRT-PCR in the methylated cancer cell line HCT-116 and in colorectal normal tissues (N = 3). C Restored LINC00473 expression in the methylated cancer cell line HCT-116 after AZA treatment analyzed by qRT-PCR. Values were determined from triplicates and are expressed as the mean ± SEM. D Methylation and expression analysis of LINC00473 in common colorectal cancer cell lines and normal colon mucosa using 450K array and RNA-seq public datasets. Expression levels obtained by RNA-seq were expressed as transcripts per million (TPMs). TSS, transcription start site; NC; normal colon mucosa

Methylation status of the LINC00473 promoter in colorectal tumors

After confirming the epigenetic regulation of LINC00473 by promoter methylation in CRC cell lines, we decided to evaluate whether this epigenetic alteration was a general event in primary colorectal tumors. Thus, we evaluated the methylation status of the LINC00473 promoter and the expression level of the corresponding gene by using a 450K array and RNA-seq, respectively, from a TCGA dataset (Cohort 1) of primary colorectal tumors (from TNM stage I to IV) and matched normal tissues (controls). As expected from our results with CRC cell lines, this analysis revealed a significantly higher methylation level of LINC00473 in colorectal tumors than in controls (Fig. 2A), which was consistent across all TNM tumor stages (Fig. 2B). In addition, LINC00473 promoter hypermethylation was associated with a significant downregulation of its expression levels in primary colorectal tumors (Additional file 1: Fig. S1). Consistent with this, promoter methylation of LINC00473 showed a significant inverse correlation (r = − 0.19; p = 0.0009) with its expression levels in CRC patients from the TCGA dataset (Cohort 1).

Fig. 2

Methylation of LINC00473 in colorectal cancer tissues. Methylation analysis of LINC00473 promoter in tissues from primary CRC and matched normal colorectal mucosa (controls) of two independent cohorts analyzed by 450K array (Cohort 1) and pyrosequencing (Cohort 2). A, B Methylation status of LINC00473 in (A) all CRC patients (N = 293) and (B) in those CRC patients (N = 283) with available clinical information on their TNM tumor stage (Cohort 1). C ROC curve analysis evaluating the methylation of LINC00473 for the detection of CRC in tissue samples in Cohort 1. D, E Validation of the methylation status of LINC00473 in (A) all CRC patients (N = 180) and (B) in those CRC patients (N = 178) with available clinical information on their TNM tumor stage in Cohort 2. F ROC curve analysis to validate the methylation of LINC00473 for the detection of colorectal tumors in Cohort 2. Horizontal lines represent mean methylation levels of LINC00473. P, p value analyzed by Mann–Whitney U test or ROC curve; AUC, area under the ROC curve; Ctrl, controls; CRC, colorectal cancer

Next, we generated receiver operating characteristic curves (ROCs) to evaluate the robustness of the methylation status of the LINC00473 promoter for CRC diagnosis across all tumor stages. This analysis revealed a significantly high CRC detection accuracy with an area under the receiver operating characteristic curve (AUC) of 0.941 (95% CI 0.915–0.966, p < 0.0001) (Fig. 2C), a sensitivity of 91% (CI 95% 87–94%), a specificity of 100% (CI 95% 91–100%), a positive predictive value (PPV) of 100% and a negative predictive value (NPV) of 58%. Moreover, ROC curve analyses yielded a very high diagnostic accuracy across all tumor stages separately (Additional file 2: Fig. S2). Subsequently, these results were confirmed in an independent cohort (Cohort 2) of primary colorectal tumors (from TNM stage I to IV) and matched non-tumor controls using bisulfite pyrosequencing (Fig. 2D–F). This analysis also showed a significantly higher methylation level of the LINC00473 promoter in primary colorectal tumors than in controls (Fig. 2D), which was constant across all TNM tumor stages (Fig. 2E). As expected, the LINC00473 methylation level identified primary colorectal tumors with an AUC of 0.893 (95% CI 0.851–0.935, < 0.0001) (Fig. 2F), a sensitivity of 78% (CI 95% 71–84%), a specificity of 98% (CI 95% 92–100%), a PPV of 99% and an NPV of 67%. The detection accuracy was constant across all CRC stages separately, as shown in Additional file 3: Fig. S3.

Methylation status of the LINC00473 promoter in tissue from precancerous colorectal lesions

DNA methylation is an epigenetic mechanism that can be deregulated in precancerous colorectal lesions [16]. To confirm this feature, we analyzed the methylation status of the LINC00473 promoter by bisulfite pyrosequencing in tissues from premalignant colorectal polyps, CRC and matched normal colorectal mucosa (controls) (Cohort 3) (Fig. 3). As expected, the methylation of LINC00473 was significantly higher in polyps and in CRC than in healthy controls, while no differences were found between polyps and CRC (Fig. 3A). Importantly, from a clinical viewpoint, we also found significant differences between LINC00473 methylation levels in non-ACP (N-ACP) compared to ACP (Fig. 3B) and in N-ACP compared to CRC (Fig. 3C) but not in controls compared to N-ACP or in ACP compared to CRC (Additional file 4: Fig. S4A-B). Next, we generated ROC curves to evaluate the robustness of this methylation biomarker to detect CRC or premalignant colorectal polyps, and we observed a high capacity of LINC00473 to differentiate controls from polyps (AUC = 0.840, CI 95% 0.657–1.00, p = 0.0047; sensitivity = 83%, CI 95% 52–98%; specificity = 92%, CI 95% 62–100%; PPV = 91%; NPV = 85%) and controls from CRC (AUC = 0.917, CI 95% 0.794–1.00, p = 0.0005; sensitivity = 83%, CI 95% 52–98%; specificity = 100%, CI 95% 74–100%; PPV = 100%; NPV = 86%) (Additional file 4: Fig. S4C-D). More importantly, the methylation status of LINC00473 was able to accurately detect ACP (AUC = 0.992, CI 95% 0.967–1, p = 0.0001; sensitivity = 100%, CI 95% 63–100%; specificity = 94%, CI 95% 70–100%, PPV = 89%; NPV = 100%) (Fig. 3D) and AN (AUC = 0.947, CI 95% 0.873–1.00, p < 0.0001; sensitivity = 85%, CI 95% 62–97%; specificity = 100%, CI 95% 79–100%, PPV = 100%; NPV = 84%) (Fig. 3E).

Fig. 3

Methylation levels of LINC00473 in tissue precancerous colorectal lesions. A Methylation levels of LINC00473 promoter in tissues from premalignant colorectal polyps, CRC and matched normal colorectal mucosa (controls) by pyrosequencing (Cohort 3). B, C Methylation levels of LINC00473 promoter in tissues from N-ACP, ACP and CRC (Cohort 3) analyzed by pyrosequencing. D ROC curve analysis evaluating the methylation of LINC00473 promoter for the detection of ACPs with respect to the combination of controls and N-ACPs (Cohort 3). E ROC curve to evaluate the methylation of LINC00473 promoter for the detection of AN with respect to the combination of controls and N-ACPs (Cohort 3). Horizontal lines represent mean methylation levels of LINC00473. P, p value analyzed by Mann–Whitney U test or ROC curve; AUC, area under the ROC curve; Ctrl, controls; P, polyps; CRC, colorectal cancer; N-ACP, non-advanced colorectal polyps; ACP, advanced colorectal polyps; AN, advanced neoplasia

To better appreciate the translational potential of our previous findings, we asked whether this methylation biomarker could be successfully validated using pyrosequencing in an additional cohort (cohort 4) composed of colorectal polyps and normal mucosa from non-cancer patients (controls). In addition to confirming methylation differences between controls and polyps (Additional file 5: Fig. S5A), with a larger number of N-ACPs, in this cohort, we were able to detect significantly higher methylation levels in controls than in N-ACPs (Additional file 5: Fig. S5B). Importantly from a clinical viewpoint, we confirmed significantly higher methylation levels of LINC00473 in ACPs than in N-ACPs (Additional file 5: Fig. S5C). Furthermore, ROC curves demonstrated the high accuracy of the methylation status of LINC00473 to detect polyps (AUC = 0.831, CI 95% 0.738–0923, p = 0.0009; sensitivity = 70%, CI 95% 57–81%; specificity = 100%, CI 95% 69–100%, PPV = 100%; NPV = 36%) (Additional file 5: Fig. S5D) and, even more importantly, to detect ACPs (AUC = 0.776, CI 95% 0.665–0.888, p < 0.0001; sensitivity = 71%, CI 95% 54–85%; specificity = 79%, CI 95% 62–91%, PPV = 78%; NPV = 73%) (Additional file 5: Fig. S5E).

Additionally, we also assayed for a possible effect of age on LINC00473 methylation levels in tissue samples; however, no significant effect was found (p > 0.05). Furthermore, we also did not find any significant difference (p > 0.05) in methylation levels according to the sex of the individuals (data not shown).

Diagnostic potential of methylation of the LINC00473 promoter to detect colorectal cancer and precancerous lesions in plasma cell-free DNA

Beyond tissue samples, promoter hypermethylation of several genes in plasma cfDNA of patients with ACPs or CRC has also been described [5]. Based on this fact, we analyzed the methylation of the LINC00473 promoter in plasma cfDNA of a cohort of self-declared healthy controls and CRC patients by qMSP (Cohort 5). The result of this analysis showed significantly higher methylation levels in CRC than in controls (Fig. 4A). In addition, the methylation levels of LINC00473 in cfDNA exhibited a very high diagnostic accuracy to detect CRC patients with an AUC of 0.881 (CI 95% 0.776–0983, p < 0.0001), a sensitivity of 81% (95% CI 61–93%), a specificity of 100% (95% CI 88–100%), a PPV of 100% and an NPV of 85% (Fig. 4B). Furthermore, we also analyzed the plasma cfDNA of a cohort of self-declared healthy controls and patients with ACPs presenting at least one polyp > 10 mm previously confirmed by colonoscopy (Cohort 6). Of note, this assay revealed significantly higher methylation levels of LINC00473 in cfDNA of ACPs than in controls (Fig. 4C), showing a high diagnostic accuracy to detect ACPs (AUC = 0.836; 95% CI 0.722–0.949, p < 0.0001) with a sensitivity of 79% (95% CI 58–93%), a specificity of 88% (95% CI 73–97%), a PPV of 83% and an NPV of 86% (Fig. 4D).

Fig. 4

Methylation of LINC00473 in cfDNA from patients with advanced colorectal polyps and colorectal cancer. A Methylation status of LINC00473 promoter in plasma cfDNA of CRC patients analyzed by qMSP (Cohort 5). B ROC curve analysis evaluating the methylation of LINC00473 for the detection of CRC in plasma cfDNA (Cohort 5). C, Methylation status of LINC00473 promoter in plasma cfDNA of patients with ACP analyzed by qMSP (Cohort 6). D ROC curve analysis evaluating the methylation of LINC00473 for the detection of ACP in plasma cfDNA (Cohort 6). E Validation of the methylation status of LINC00473 promoter in plasma cfDNA of patients with ACP by ddPCR analysis (Cohort 7). F, ROC curve analysis to validate the methylation of LINC00473 for the detection of ACP and CRC in plasma cfDNA (Cohort 7). Horizontal lines represent mean methylation levels of LINC00473. P, p value analyzed by Mann–Whitney U test or ROC curve; AUC, area under the ROC curve; Ctrl, controls. ACP, advanced colorectal polyps; CRC, colorectal cancer.

To confirm the feasibility of using methylation of the LINC00473 promoter for the noninvasive early detection of CRC, we retrospectively analyzed an independent cohort of plasma cfDNA samples (Cohort 7) obtained either prior to a scheduled colonoscopy as part of standard CRC screening or prior to colonic surgery for primary tumors (Fig. 4E, F). Due to the need for very sensitive methodologies for the early detection of cancer in liquid biopsy [5], we used ultrasensitive droplet digital PCR (ddPCR) for the methylation analysis of LINC00473. Consistent with our previous results in tissue and cfDNA, the methylation levels of LINC00473 were significantly higher in ACP and CRC than in confirmed healthy controls (Fig. 4E). In addition, no significant differences were found between the methylation status of LINC00473 of ACP and CRC (Fig. 4E). Of note, methylation of the LINC00473 promoter showed high accuracy for the detection of ACP, with an AUC of 0.721 (95% CI 0.553–0.890, p = 0.0235), a sensitivity of 76% (95% CI 50–93%), a specificity of 63% (95% CI 38–84%), a PPV of 65% and an NPV of 75% (Fig. 4F). Similarly, methylation of LINC00473 was able to identify CRC patients with a high AUC of 0.833 (95% CI 0.709–0.958, p = 0.0003), a sensitivity of 90% (95% CI 70–99%), a specificity of 63% (95% CI 38–84%), a PPV of 73 and an NPV of 86. In addition, LINC00473 showed the ability to detect AN with an AUC of 0.783 (95% CI 0.662–0.904, p = 0.0005), a sensitivity of 84% (95% CI 69–94%), a specificity of 63% (95% CI 38–84%), a PPV of 82% and an NPV of 67% (Fig. 4F).

Similar to our previous analysis in tissue samples, we also found no significant effect of age (p > 0.05) on LINC00473 methylation levels in cfDNA samples. In addition, no significant difference (p > 0.05) in methylation levels according to the sex of the individuals was found (data not shown).

Clinical utility of LINC00473 promoter methylation for the noninvasive detection of colorectal cancer during the follow-up of metastatic patients

After confirming its utility in the detection of precancerous lesions and CRC, we decided to evaluate whether the methylation status of the LINC00473 promoter in plasma cfDNA could be useful to diagnose the presence of CRC in the palliative setting. For this purpose, we analyzed the methylation of the LINC00473 promoter in plasma cfDNA by ddPCR at various clinically relevant time points from a cohort of six randomly selected mCRC patients (Cohort 8), whose disease evolution was evaluated according to a standard clinical practice (serial serum carcinoembryonic antigen (CEA) determinations and computed tomography scans). As shown in Additional file 6: Fig. S6, the plasma cfDNA methylation levels of LINC00473 decreased with effective therapy and increased with disease progression. Notably, in some cases (e.g., cases #1, #4, and #5), LINC00473 methylation preceded CEA in detecting the presence of CRC. Together, these six cases represent proof of concept of the utility of LINC00473 methylation as a potential biomarker to detect the presence of CRC during the follow-up of metastatic patients.