Abstract
Nonclustered protocadherins (PCDH) family is a group of cell–cell adhesion molecules. We have found differentially methylated genes in the nonclustered PCDHs family associated with Helicobacter pylori (H. pylori) infection in prior genome-wide methylation analysis. To further investigate the methylation and expression of nonclustered PCDHs encoding genes in H. pylori--related gastric carcinogenesis process, four candidate genes including PCDH7, PCDH10, PCDH17, and PCDH20 were selected, which were reported to be tumor suppressors for digestive cancers. A total of 747 participants with a spectrum of gastric lesions were enrolled from a high-risk population of gastric cancer. Promoter methylation levels of four genes were significantly higher in H. pylori–positive subjects than the negative group (all P < 0.001). Elevated methylation levels of PCDH10 and PCDH17 were observed with the increasing severity of gastric lesions (both Ptrend < 0.001). In the protein expression analysis, PCDH17 expression was inversely associated with gastric lesions; the OR [95% confidence interval (CI)] was 0.49 (0.26–0.95) for chronic atrophic gastritis (CAG), 0.31 (0.15–0.63) for intestinal metaplasia, and 0.38 (0.19–0.75) for indefinite dysplasia and dysplasia, compared with superficial gastritis. In addition, PCDH10 expression was significantly lower in CAG (OR, 0.40; 95% CI, 0.24–0.68). The inverse association between methylation and protein expression of PCDH10 and PCDH17 was further supported when we explored the methylation and mRNA expression in The Cancer Genome Atlas database (all P < 0.001). Our study found elevated promoter methylation and decreased expression of PCDH10 and PCDH17 in advanced gastric lesions, suggesting that elevated PCDH10 and PCDH17 methylation may be an early event in gastric carcinogenesis. Cancer Prev Res; 11(11); 717–26. ©2018 AACR.
Introduction
Helicobacter pylori (H. pylori) is a major risk factor of gastric cancer, whereas eradication of H. pylori infection may reduce the risk of advanced gastric lesions and gastric cancer (1–6). Aberrant DNA methylation accumulated in precancerous tissues after H. pylori infection has been proposed to underlie gastric carcinogenesis. Consistent with this, our previous study comparing the genome-wide methylation profiling before and after H. pylori eradication has reported a number of H. pylori–associated aberrant methylated genes (7).
Protocadherins (PCDH) are a group of cell–cell adhesion molecules and can be divided into “clustered” and “non-clustered” PCDHs based on the genomic structure of their encoding genes (8). Clustered PCDHs are encoded as a large cluster in the genome and are epigenetically dysregulated in Wilms' tumor and brain disorder (9, 10). In contrast, nonclustered PCDHs are scattered across the genome and may serve as tumor suppressors in many human cancers. In our genome-wide methylation profiling analysis, 13 of 16 nonclustered PCDHs encoding genes were found differentially methylated after H. pylori eradication (P < 0.05). Among them, several nonclustered PCDHs encoding genes have been reported as tumor suppressors for digestive cancers. For example, hypermethylation and decreased expression of PCDH7, PCDH10, and PCDH17 have been observed in gastric cancer and several other cancers (11–14). PCDH20 was reported to be downregulated and methylated in hepatocellular carcinoma and other cancers (15–17).
Whether methylation and expression of these genes may be involved in H. pylori–related gastric carcinogenesis is still unclear. We therefore further examined the methylation in the promoter and protein expression of PCDH7, PCDH10, PCDH17, and PCDH20 in subjects with different H. pylori infection status and a spectrum of precancerous gastric lesions, including superficial gastritis (SG), chronic atrophic gastritis (CAG), intestinal metaplasia (IM), indefinite dysplasia (Ind DYS), and dysplasia (DYS) from a high-risk population of gastric cancer in Linqu County. The association between potentially differentially methylated genes and mRNA expression was further examined by accessing The Cancer Genome Atlas (TCGA) database.
Materials and Methods
Study population
In 2002, we launched an endoscopic screening survey in 12 villages selected randomly in Linqu County, Shandong Province (2). A total of 2,638 subjects completed endoscopic examination. The detailed procedures and histopathologic criteria have been described elsewhere (2, 18). Briefly, five biopsies were obtained from standard sites of gastric mucosa (19). Histopathologic diagnosis for each biopsy specimen was classified as normal, SG, CAG, IM, Ind DYS, DYS, or gastric cancer by three senior pathologists independently according to the Updated Sydney System and Padova International Classification (20). Each subject was assigned a global diagnosis based on the most severe diagnosis among any of the 5 biopsy specimens.
For the current study, a total of 747 subjects, with 555 H. pylori–positive subjects and 192 H. pylori–negative subjects, were selected at random based on the global diagnosis of precancerous gastric lesions, including SG (n = 181), CAG (n = 175), IM (n = 169), and Ind DYS/DYS (n = 222). H. pylori antibody assays were used to determine H. pylori infection status (21). Genomic DNA for each subject was extracted from the formalin-fixed paraffin-embedded biopsy tissues as previously reported (22). The study was approved by the Institutional Review Board of Peking University School of Oncology and University of Hong Kong. Written-informed consent was obtained from each participant.
Selection of candidate genes and promoter methylation assays
The selection of candidate genes was based on the prior epigenome-wide association study in 6 pairs of gastric mucosa samples before and after H. pylori eradication by using Infinium HumanMethylation 450K BeadChip (7). Based on this methylation array, we evaluated the association between H. pylori eradication and methylation levels of all the 16 nonclustered PCDHs, in which 173 relevant CpG sites in the promoter regions were assessed. We found 46 CpG sites of 13 genes differentially methylated after H. pylori eradication (P < 0.05, Supplementary Table S1). In our literature research, five genes, including PCDH7, PCDH8, PCDH10, PCDH17, and PCDH20, have been reported as tumor suppressors for digestive cancers (11, 14, 15, 23, 24) and were then selected as the candidate genes for our study. As the detection of PCDH8 methylation failed in the denaturing high-performance liquid chromatography (DHPLC) assay, we restricted our current study to PCDH7, PCDH10, PCDH17, and PCDH20. Both methylated and unmethylated PCDH7, PCDH10, PCDH17, and PCDH20 were amplified by a universal primer set without CpG under thermal cycle conditions (Supplementary Table S4). The PCR products of methylated and unmethylated selected genes were separated by DHPLC at the corresponding partial denaturing temperature. The peak areas corresponding to the methylated and unmethylated PCR products were used to calculate the percentage of methylated copies (proportion of hypermethylated copies = methylated peak area/total peak area) for each sample. We used the percentage of methylated copies as the measure for methylation level. MKN45 or RKO cell line was used as positive control.
Immunohistochemical staining
We conducted immunohistochemical analysis to determine the protein expression status of genes that were differentially methylated according to precancerous gastric lesions in the above analysis, including PCDH10 and PCDH17. Details of the immunohistochemical staining based on paraffin-embedded tissues are shown in the Supplementary File.
Statistical analysis
ORs with corresponding 95% confidence intervals (CI) were calculated by unconditional logistic regression to evaluate the association of DNA methylation with major characteristics. These characteristics have been associated with progression of gastric lesions and risk of gastric cancer previously (21, 25, 26) and may change methylation status based on current literature (27, 28) and our prior research (7). Participants were classified into hypermethylation (median or higher) and hypomethylation (lower than median) groups for each gene. We also used multivariate-adjusted unconditional logistic regression analysis to evaluate the association between DNA methylation or expression and risk of each gastric lesion, with SG as the reference, adjusting for the major characteristics. For the analysis on methylation of each gene, participants were classified into four groups based on the quartiles of methylation percentage. We also examined the association between methylation and protein expression of PCDH10 and PCDH17 by using logistic regression analysis. To evaluate the changing trend in risk for gastric lesions or DNA expression with increasing methylation levels, linear trend test was applied, using the values of 1 to 4 assigned to increasing methylation quartiles (Q1–Q4), respectively. Stratified analysis was conducted to evaluate the association of PCDH10, PCDH17 methylation, and gastric lesions by H. pylori infection, and the heterogeneity of ORs among strata was evaluated using Q statistics.
A value of P < 0.05 (two tailed) was considered statistically significant. All the statistical analyses were conducted using SPSS program (version 17.0; SPSS) or R software (version 3.4.1).
DNA methylation and mRNA expression of PCDH10 and PCDH17 in TCGA data
As our endoscopic screening yielded limited number of subjects with gastric cancer, we accessed TCGA data for the methylation and expression of PCDH10 and PCDH17 in gastric cancers. We selected all 45 CpG sites within the genomic region of PCDH10 and PCDH17 (21 CpGs for PCDH10 and 24 CpGs PCDH17) from the Illumina HumanMethylation450 DNA methylation data in TCGA Stomach Adenocarcinoma (Provisional) study. The mRNA expression (RNA Seq V2 RSEM) data were retrieved at cBioPortal (http://www.cbioportal.org/public-portal/index.do; refs. 29, 30). Spearman correlation analysis was performed on the association between methylation and mRNA expression by using the Log2-transformed values of mRNA levels (n = 317 gastric cancer samples).
Result
Participant information
Major characteristics of 747 subjects according to H. pylori infection status are shown in Table 1. Significant differences in the distribution of age, pathologic diagnosis, smoking, and drinking status were identified between H. pylori–positive and –negative subjects (all P < 0.05). The quantitative methylation assays were finally completed in 747 subjects for PCDH7, 711 subjects for PCDH10, 564 subjects for PCDH17, and 687 subjects for PCDH20.
Selected characteristics of study subjects according to H. pylori infection status
. | Total . | H. pylori negative . | H. pylori positive . | . |
---|---|---|---|---|
. | (n = 747) . | (n = 192) . | (n = 555) . | Pa . |
Age, median (IQR) | 50.0 (45.0–55.0) | 52.0 (47.0–56.0) | 49.0 (44.0–54.0) | 0.001b |
Gender (%) | 0.1c | |||
Female | 331 (44.3) | 76 (39.6) | 255 (45.9) | |
Male | 416 (55.7) | 116 (60.4) | 300 (54.1) | |
Smoking (%) | 0.04c | |||
No | 404 (54.1) | 91 (47.4) | 313 (56.4) | |
Yes | 337 (45.1) | 98 (51.0) | 239 (43.1) | |
Missing | 6 (0.8) | 3 (1.6) | 3 (0.5) | |
Drinking (%) | <0.001c | |||
No | 432 (57.8) | 87 (45.3) | 345 (62.2) | |
Yes | 267 (35.7) | 85 (44.3) | 182 (32.8) | |
Missing | 48 (6.5) | 20 (10.4) | 28 (5.0) | |
Baseline pathology (%) | <0.001c | |||
SG | 181 (24.2) | 75 (39.1) | 106 (19.1) | |
CAG | 175 (23.4) | 45 (23.4) | 130 (23.4) | |
IM | 169 (22.6) | 38 (19.8) | 131 (23.6) | |
Ind DYS/DYS | 222 (29.7) | 34 (17.7) | 188 (33.9) |
. | Total . | H. pylori negative . | H. pylori positive . | . |
---|---|---|---|---|
. | (n = 747) . | (n = 192) . | (n = 555) . | Pa . |
Age, median (IQR) | 50.0 (45.0–55.0) | 52.0 (47.0–56.0) | 49.0 (44.0–54.0) | 0.001b |
Gender (%) | 0.1c | |||
Female | 331 (44.3) | 76 (39.6) | 255 (45.9) | |
Male | 416 (55.7) | 116 (60.4) | 300 (54.1) | |
Smoking (%) | 0.04c | |||
No | 404 (54.1) | 91 (47.4) | 313 (56.4) | |
Yes | 337 (45.1) | 98 (51.0) | 239 (43.1) | |
Missing | 6 (0.8) | 3 (1.6) | 3 (0.5) | |
Drinking (%) | <0.001c | |||
No | 432 (57.8) | 87 (45.3) | 345 (62.2) | |
Yes | 267 (35.7) | 85 (44.3) | 182 (32.8) | |
Missing | 48 (6.5) | 20 (10.4) | 28 (5.0) | |
Baseline pathology (%) | <0.001c | |||
SG | 181 (24.2) | 75 (39.1) | 106 (19.1) | |
CAG | 175 (23.4) | 45 (23.4) | 130 (23.4) | |
IM | 169 (22.6) | 38 (19.8) | 131 (23.6) | |
Ind DYS/DYS | 222 (29.7) | 34 (17.7) | 188 (33.9) |
Abbreviation: IQR, interquartile range.
aP value for each covariate was estimated among participants without missing value on that covariate.
bTwo-Sample Kolmogorov–Smirnov test.
cPearson χ2 test without missing value.
Association of DNA methylation levels and H. pylori infection
We evaluated whether the four candidate genes may be differentially methylated by H. pylori infection status. As shown in Fig. 1A, the methylation levels of these four genes were significantly higher in H. pylori–positive group compared with negative group (all P < 0.05). Multivariate logistic regression analysis showed that hypermethylated PCDH7 and PCDH20 occurred more frequently in H. pylori–positive subjects than negative ones (OR, 4.43; 95% CI, 2.93–6.70 for PCDH7; OR, 2.34; 95% CI, 1.59–3.44 for PCDH20; Supplementary Table S2).
The methylation levels of selected candidate genes in different H. pylori infection status and gastric lesions. A, For comparison of DNA methylation levels in different H. pylori infection status, P values were calculated by Wilcoxon rank-sum tests between H. pylori–negative and –positive groups. B–E, For comparison of the methylation levels of PCDH7, PCDH10, PCDH17, and PCDH20 in different gastric lesions, the P values were calculated by the Dwass–Steel–Critchlow–Fligner rank test between SG subjects and CAG, IM, Ind DYS/DYS groups.
The methylation levels of selected candidate genes in different H. pylori infection status and gastric lesions. A, For comparison of DNA methylation levels in different H. pylori infection status, P values were calculated by Wilcoxon rank-sum tests between H. pylori–negative and –positive groups. B–E, For comparison of the methylation levels of PCDH7, PCDH10, PCDH17, and PCDH20 in different gastric lesions, the P values were calculated by the Dwass–Steel–Critchlow–Fligner rank test between SG subjects and CAG, IM, Ind DYS/DYS groups.
Association between DNA methylation levels and precancerous gastric lesions
As shown in Fig. 1B–E, the methylation levels of PCDH10 and PCDH17 were increased significantly from SG and CAG to advanced gastric lesions including IM and Ind DYS/DYS subjects (both P < 0.001). By dividing the methylation levels into four grades according to the quartile cut-points, we found the risk of advanced gastric lesions increased significantly with elevated methylation levels of PCDH10 and PCDH17 (all Ptrend < 0.001, Table 2). Compared with PCDH10 methylation levels in the lowest quartile (Q1), the ORs (95% CIs) of IM were 2.41 (1.25–4.65), 3.15 (1.63–6.09), and 3.95 (1.98–7.87), and the ORs (95% CIs) of Ind DYS/DYS were 2.59 (1.32–5.09), 3.02 (1.50–6.09), and 4.75 (2.33–9.70) for methylation levels in the second to fourth quartiles respectively. Similar trend was observed for PCDH17.
Associations between DNA methylation levels and risk of precancerous gastric lesions
. | SG . | CAG . | IM . | Ind DYS/DYS . | ||||||
---|---|---|---|---|---|---|---|---|---|---|
Methylation quartile levels . | N . | N . | ORa (95% CI) . | Ptrend . | N . | ORa (95% CI) . | Ptrend . | N . | ORa (95% CI) . | Ptrend . |
PCDH7 | ||||||||||
Q1 = 0% | 74 | 50 | 1.00 | 55 | 1.00 | 76 | 1.00 | |||
Q2 0–<5.71% | 31 | 31 | 1.21 (0.64–2.29) | 27 | 1.15 (0.60–2.19) | 29 | 0.89 (0.43–1.82) | |||
Q3 5.71%–<15.65% | 41 | 44 | 1.42 (0.78–2.57) | 46 | 1.33 (0.74–2.38) | 56 | 1.52 (0.83–2.78) | |||
Q4 ≥15.65% | 35 | 50 | 1.82 (0.97–3.43) | 0.06 | 41 | 1.27 (0.68–2.38) | 0.4 | 61 | 1.46 (0.79–2.70) | 0.1 |
PCDH10 | ||||||||||
Q1 <17.57% | 63 | 52 | 1.00 | 27 | 1.00 | 35 | 1.00 | |||
Q2 17.57%–<43.89% | 36 | 42 | 1.40 (0.77–2.55) | 44 | 2.41 (1.25–4.65) | 56 | 2.59 (1.32–5.09) | |||
Q3 43.89%–<73.48% | 34 | 38 | 1.30 (0.70–2.42) | 50 | 3.15 (1.63–6.09) | 56 | 3.02 (1.50–6.09) | |||
Q4 ≥73.48% | 26 | 37 | 1.63 (0.85–3.14) | 0.8 | 45 | 3.95 (1.98–7.87) | <0.001 | 70 | 4.75 (2.33–9.70) | <0.001 |
PCDH17 | ||||||||||
Q1 = 0% | 73 | 69 | 1.00 | 42 | 1.00 | 48 | 1.00 | |||
Q2 0%–<7.40% | 17 | 13 | 0.70 (0.31–1.59) | 6 | 0.50 (0.18–1.74) | 14 | 1.00 (0.42–2.38) | |||
Q3 7.40%–<41.53% | 30 | 37 | 1.37 (0.73–2.58) | 37 | 1.99 (1.02–3.87) | 37 | 1.83 (0.92–3.64) | |||
Q4 ≥41.53% | 16 | 19 | 0.95 (0.43–2.09) | 0.7 | 53 | 5.09 (2.48–10.42) | <0.001 | 53 | 3.65 (1.76–7.56) | <0.001 |
PCDH20 | ||||||||||
Q1 <1.53% | 39 | 37 | 1.00 | 40 | 1.00 | 55 | 1.00 | |||
Q2 1.53%–<7.12% | 47 | 41 | 1.07 (0.57–2.02) | 35 | 0.81 (0.42–1.55) | 49 | 1.45 (0.72–2.89) | |||
Q3 7.12%–<16.78% | 37 | 46 | 1.35 (0.71–2.57) | 37 | 1.00 (0.52–1.94) | 52 | 1.25 (0.63–2.48) | |||
Q4 ≥16.78% | 34 | 39 | 1.18 (0.60–2.32) | 0.5 | 52 | 1.52 (0.79–2.93) | 0.2 | 47 | 1.27 (0.64–2.52) | 0.6 |
. | SG . | CAG . | IM . | Ind DYS/DYS . | ||||||
---|---|---|---|---|---|---|---|---|---|---|
Methylation quartile levels . | N . | N . | ORa (95% CI) . | Ptrend . | N . | ORa (95% CI) . | Ptrend . | N . | ORa (95% CI) . | Ptrend . |
PCDH7 | ||||||||||
Q1 = 0% | 74 | 50 | 1.00 | 55 | 1.00 | 76 | 1.00 | |||
Q2 0–<5.71% | 31 | 31 | 1.21 (0.64–2.29) | 27 | 1.15 (0.60–2.19) | 29 | 0.89 (0.43–1.82) | |||
Q3 5.71%–<15.65% | 41 | 44 | 1.42 (0.78–2.57) | 46 | 1.33 (0.74–2.38) | 56 | 1.52 (0.83–2.78) | |||
Q4 ≥15.65% | 35 | 50 | 1.82 (0.97–3.43) | 0.06 | 41 | 1.27 (0.68–2.38) | 0.4 | 61 | 1.46 (0.79–2.70) | 0.1 |
PCDH10 | ||||||||||
Q1 <17.57% | 63 | 52 | 1.00 | 27 | 1.00 | 35 | 1.00 | |||
Q2 17.57%–<43.89% | 36 | 42 | 1.40 (0.77–2.55) | 44 | 2.41 (1.25–4.65) | 56 | 2.59 (1.32–5.09) | |||
Q3 43.89%–<73.48% | 34 | 38 | 1.30 (0.70–2.42) | 50 | 3.15 (1.63–6.09) | 56 | 3.02 (1.50–6.09) | |||
Q4 ≥73.48% | 26 | 37 | 1.63 (0.85–3.14) | 0.8 | 45 | 3.95 (1.98–7.87) | <0.001 | 70 | 4.75 (2.33–9.70) | <0.001 |
PCDH17 | ||||||||||
Q1 = 0% | 73 | 69 | 1.00 | 42 | 1.00 | 48 | 1.00 | |||
Q2 0%–<7.40% | 17 | 13 | 0.70 (0.31–1.59) | 6 | 0.50 (0.18–1.74) | 14 | 1.00 (0.42–2.38) | |||
Q3 7.40%–<41.53% | 30 | 37 | 1.37 (0.73–2.58) | 37 | 1.99 (1.02–3.87) | 37 | 1.83 (0.92–3.64) | |||
Q4 ≥41.53% | 16 | 19 | 0.95 (0.43–2.09) | 0.7 | 53 | 5.09 (2.48–10.42) | <0.001 | 53 | 3.65 (1.76–7.56) | <0.001 |
PCDH20 | ||||||||||
Q1 <1.53% | 39 | 37 | 1.00 | 40 | 1.00 | 55 | 1.00 | |||
Q2 1.53%–<7.12% | 47 | 41 | 1.07 (0.57–2.02) | 35 | 0.81 (0.42–1.55) | 49 | 1.45 (0.72–2.89) | |||
Q3 7.12%–<16.78% | 37 | 46 | 1.35 (0.71–2.57) | 37 | 1.00 (0.52–1.94) | 52 | 1.25 (0.63–2.48) | |||
Q4 ≥16.78% | 34 | 39 | 1.18 (0.60–2.32) | 0.5 | 52 | 1.52 (0.79–2.93) | 0.2 | 47 | 1.27 (0.64–2.52) | 0.6 |
Abbreviations: Q1, quartile 1; Q2, quartile 2; Q3, quartile 3; Q4, quartile4.
aUnconditional logistic regression analysis, adjusted for other characteristics (age, gender, H. pylori infection, drinking, and smoking status).
In the stratified analysis, we found significant association between PCDH10 or PCDH17 hypermethylation and increased risk of advanced gastric lesions (IM/Ind DYS/DYS vs. SG/CAG) both in H. pylori–negative and –positive groups (Table 3). No significant heterogeneity was found in the associations among H. pylori–positive and –negative subjects (P = 0.9 for PCDH10; P = 0.3 for PCDH17).
Stratified analysis on the association between DNA methylation status and precancerous gastric lesions by H. pylori infection
. | H. pylori negative . | H. pylori positive . | . | ||||
---|---|---|---|---|---|---|---|
. | SG/CAG . | IM/Ind DYS/DYS . | ORa (95% CI) . | SG/CAG . | IM/Ind DYS/DYS . | ORa (95% CI) . | Pbheterogeneity . |
PCDH7c | 0.2 | ||||||
Hypo | 94 (78.3) | 54 (75.0) | 1.00 | 92 (39.0) | 133 (41.7) | 1.00 | |
Hyper | 26 (21.7) | 18 (25.0) | 1.76 (0.81–3.82) | 144 (61.0) | 186 (58.3) | 0.95 (0.67–1.36) | |
PCDH10c | 0.9 | ||||||
Hypo | 71 (64.0) | 35 (48.6) | 1.00 | 122 (56.2) | 127 (40.8) | 1.00 | |
Hyper | 40 (36.0) | 37 (51.4) | 2.34 (1.15–4.78) | 95 (43.8) | 184 (59.2) | 1.81 (1.26–2.61) | |
PCDH17c | 0.3 | ||||||
Hypo | 65 (67.0) | 13 (31.0) | 1.00 | 107 (60.5) | 97 (39.1) | 1.00 | |
Hyper | 32 (33.0) | 29 (69.0) | 4.27 (1.85–9.83) | 70 (39.5) | 151 (60.9) | 2.61 (1.73–3.94) | |
PCDH20c | 0.6 | ||||||
Hypo | 70 (63.1) | 59 (72.8) | 1.00 | 94 (45.0) | 120 (42.0) | 1.00 | |
Hyper | 41 (36.9) | 22 (27.2) | 0.90 (0.44–1.84) | 115 (55.0) | 166 (58.0) | 1.15 (0.79–1.67) |
. | H. pylori negative . | H. pylori positive . | . | ||||
---|---|---|---|---|---|---|---|
. | SG/CAG . | IM/Ind DYS/DYS . | ORa (95% CI) . | SG/CAG . | IM/Ind DYS/DYS . | ORa (95% CI) . | Pbheterogeneity . |
PCDH7c | 0.2 | ||||||
Hypo | 94 (78.3) | 54 (75.0) | 1.00 | 92 (39.0) | 133 (41.7) | 1.00 | |
Hyper | 26 (21.7) | 18 (25.0) | 1.76 (0.81–3.82) | 144 (61.0) | 186 (58.3) | 0.95 (0.67–1.36) | |
PCDH10c | 0.9 | ||||||
Hypo | 71 (64.0) | 35 (48.6) | 1.00 | 122 (56.2) | 127 (40.8) | 1.00 | |
Hyper | 40 (36.0) | 37 (51.4) | 2.34 (1.15–4.78) | 95 (43.8) | 184 (59.2) | 1.81 (1.26–2.61) | |
PCDH17c | 0.3 | ||||||
Hypo | 65 (67.0) | 13 (31.0) | 1.00 | 107 (60.5) | 97 (39.1) | 1.00 | |
Hyper | 32 (33.0) | 29 (69.0) | 4.27 (1.85–9.83) | 70 (39.5) | 151 (60.9) | 2.61 (1.73–3.94) | |
PCDH20c | 0.6 | ||||||
Hypo | 70 (63.1) | 59 (72.8) | 1.00 | 94 (45.0) | 120 (42.0) | 1.00 | |
Hyper | 41 (36.9) | 22 (27.2) | 0.90 (0.44–1.84) | 115 (55.0) | 166 (58.0) | 1.15 (0.79–1.67) |
Abbreviations: Hyper, hypermethylated; hypo, hypomethylated.
aUnconditional logistic regression analysis, adjusting for age, gender, drinking, and smoking status.
bP value for heterogeneity was evaluated using Q statistics.
CHypo and hypermethylation for each gene was classified based on the median of methylation levels in all subjects, which was 5.71% for PCDH7, 43.89% for PCDH10, 7.40% for PCDH17, and 7.12% for PCDH20.
Protein expression levels of PCDH10 and PCDH17 in gastric mucosa
Considering the significant associations of PCDH10 or PCDH17 methylation levels with H. pylori infection and precancerous gastric lesions, we further explored the protein expression status of these two genes (Supplementary Fig. S1; Table 4). For both genes, the frequency of positive expression frequencies appeared decreased in advanced gastric lesions. Compared with SG, expression of PCDH17 was inversely associated with advanced gastric lesions; the OR (95% CI) was 0.49 (0.26–0.95) for CAG, 0.31 (0.15–0.63) for IM, and 0.38 (0.19–0.75) for Ind DYS/DYS. In addition, PCDH10 expression status was significantly lower in CAG compared with SG subjects (OR, 0.40; 95% CI, 0.24–0.68).
Associations between PCDH10 and PCDH17 protein expression levels and risk of precancerous gastric lesions
. | SG . | CAG . | IM . | Ind DYS/DYS . | |||
---|---|---|---|---|---|---|---|
. | N (%) . | N (%) . | OR (95% CI)a . | N (%) . | OR (95%CI)a . | N (%) . | OR (95%CI)a . |
PCDH10 | |||||||
Negative | 78 (52.7) | 107 (70.4) | 1.00 | 87 (61.7) | 1.00 | 115 (60.5) | 1.00 |
Positive | 70 (47.3) | 45 (29.6) | 0.40 (0.24–0.68) | 54 (38.3) | 0.68 (0.42–1.11) | 75 (39.5) | 0.62 (0.37–1.05) |
PCDH17 | |||||||
Negative | 79 (69.9) | 89 (77.4) | 1.00 | 100 (86.2) | 1.00 | 102 (79.7) | 1.00 |
Positive | 34 (30.1) | 26 (22.6) | 0.49 (0.26–0.95) | 16 (13.8) | 0.31 (0.15–0.63) | 26 (20.3) | 0.38 (0.19–0.75) |
. | SG . | CAG . | IM . | Ind DYS/DYS . | |||
---|---|---|---|---|---|---|---|
. | N (%) . | N (%) . | OR (95% CI)a . | N (%) . | OR (95%CI)a . | N (%) . | OR (95%CI)a . |
PCDH10 | |||||||
Negative | 78 (52.7) | 107 (70.4) | 1.00 | 87 (61.7) | 1.00 | 115 (60.5) | 1.00 |
Positive | 70 (47.3) | 45 (29.6) | 0.40 (0.24–0.68) | 54 (38.3) | 0.68 (0.42–1.11) | 75 (39.5) | 0.62 (0.37–1.05) |
PCDH17 | |||||||
Negative | 79 (69.9) | 89 (77.4) | 1.00 | 100 (86.2) | 1.00 | 102 (79.7) | 1.00 |
Positive | 34 (30.1) | 26 (22.6) | 0.49 (0.26–0.95) | 16 (13.8) | 0.31 (0.15–0.63) | 26 (20.3) | 0.38 (0.19–0.75) |
aUnconditional logistic regression analysis, adjusted for age, gender, H. pylori infection, drinking, and smoking status.
We also examined the protein expression of PCDH10 and PCDH17 according to methylation levels. Although we observed suggestive evidence on decreasing PCDH10 and PCDH17 expression from the lowest to the highest methylation grades (41.4%, 25.0% to 38.9%, 17.6%, respectively; Supplementary Table S3), no statistical significance was found (Ptrend = 0.6 for PCDH10 and 0.1 for PCDH17).
Correlation between DNA methylation levels and mRNA expression of PCDH10 and PCDH17 in the TCGA data
As a limited number of gastric cancer tissue samples was collected in our endoscopic screening survey, we retrieved methylation and mRNA expression of PCDH10 and PCDH17 from 317 gastric adenocarcinoma samples available in TCGA database. Spearman correlation analysis demonstrated a significant inverse correlation between PCDH10 or PCDH17 promoter methylation and mRNA expression in tumor tissues (Fig. 2), with cg06667761 (r = −0.47, P < 0.001) and cg15112032 (r = −0.51, P < 0.001) having the highest correlation efficient. All CpG loci within the same genomic region of PCDH10 (cg00945238, cg02114924, and cg07665387) as targeted in our DHPLC-based analysis were significantly inversely associated with mRNA expression in tumor tissues (P < 0.001).
Correlation of methylation levels and mRNA levels of PCDH10 (cg06667761) and PCDH17 (cg15112032) in the TCGA cohort of STAD (n = 317). The X axis presents the methylation levels of PCDH10 and PCDH17 in GC subjects. The Y axis presents the log2 mRNA levels of PCDH10 and PCDH17 in GC subjects. R, correlation coefficient.
Correlation of methylation levels and mRNA levels of PCDH10 (cg06667761) and PCDH17 (cg15112032) in the TCGA cohort of STAD (n = 317). The X axis presents the methylation levels of PCDH10 and PCDH17 in GC subjects. The Y axis presents the log2 mRNA levels of PCDH10 and PCDH17 in GC subjects. R, correlation coefficient.
Discussion
Our prior genome-wide methylation analysis found methylation levels for 13 of 16 nonclustered PCDHs encoding genes changed significantly after H. pylori eradication. We therefore further examined four genes (PCDH7, PCDH10, PCDH17, and PCDH20), which may function as tumor suppressors for digestive cancers as reported previously (11, 14, 15, 23), in our current high-risk population-based study. In this study, promoter methylation of PCDH7, PCDH10, PCDH17, and PCDH20 was significantly higher in H. pylori–positive than in –negative group. Moreover, we found that increased promoter methylation and decreased expression of PCDH10 and PCDH17 were significantly associated with precancerous gastric lesions. An inverse association between methylation and expression of PCDH10 and PCDH17 was supported in TCGA database.
Several hospital-based studies have reported epigenetic dysregulation of PCDH7, PCDH10, PCDH17, and PCDH20 in gastric or other digestive cancers. Low expression of PCDH7 was found in gastric cancer compared with normal subjects and may be correlated with poor prognosis, but PCDH7 methylation change is unknown in gastric cancer yet (23). PCDH20 hypermethylation has been reported in hepatocellular carcinoma and other cancers (15–17), but no evidence is available on gastric cancer yet. Hypermethylation and decreased expression of PCDH10 and PCDH17 have been found in gastric and several other digestive cancer tissues (11, 13, 14, 31–34). However, published studies focused on the comparison of aberrant methylation in cancer and adjacent nontumor tissue, while lacking information on methylation changes in the H. pylori–related precancerous gastric lesions.
Although a role of nonclustered PCDHs encoding genes in carcinogenesis of digestive cancers has been reported (11, 14, 15, 23, 24), how H. pylori infection may alter PCDH methylation is unknown. In our previous genome-wide methylation profiling analysis, we found that methylation levels of 13 nonclustered PCDHs encoding genes decreased after H. pylori eradication. Specifically, the methylation levels for 43 of 46 CpG sites in nonclustered PCDHs genes decreased after H. pylori eradication, which supports that long-term H. pylori infection status might significantly affect methylation status in precancerous gastric lesions. However, direct evidence is required to confirm how long-term H. pylori infection status may have changed methylation in gastric lesions. In current study, we therefore further examined the association between H. pylori infection, as determined by ELISA and methylation of four nonclustered PCDHs encoding genes, and further found that methylation levels of these genes were significantly higher in H. pylori–positive subjects, which is consistent with our genome-wide methylation profiling. Indeed, previous studies have suggested that H. pylori eradication could reverse the aberrant methylation (7, 35–37). These data support an association between H. pylori infection and promoter hypermethylation. Further studies are required to explore the underlying mechanisms and to determine whether H. pylori infection is a factor leading to promoter hypermethylation.
We examined the methylation of PCDH7, PCDH10, PCDH17, and PCDH20 in subjects with a spectrum of precancerous gastric lesions. PCDH10 and PCDH17 hypermethylation was found in advanced gastric lesions (IM or Ind DYS/DYS) than in SG or CAG, suggesting that the aberrant alteration of PCDH10 and PCDH17 methylation may be an early event in the process of gastric carcinogenesis. Interestingly, we found the methylation level of PCDH10 is much higher than other PCDHs. No prior literatures have reported the methylation levels of PCDH10 and other PCDH genes simultaneously in one study. Further studies are required to confirm the relatively higher methylation levels of PCDH10 than other PCDHs in different populations.
We evaluated protein expression status of PCDH10 and PCDH17 in gastric lesions and observed a trend in decreasing expression of PCDH10 and PCDH17 with the severity of precancerous gastric lesions. The downregulated expression of tumor-suppressor genes by promoter methylation is crucial in carcinogenesis (38). We therefore assessed how promoter methylation of PCDH10 and PCDH17 changed the protein expression and observed decreased protein expression of two genes by promoter methylation, but the association was not statistically significant. Further based on TCGA data, our analyses showed that PCDH10 and PCDH17 promoter methylation significantly decreased the mRNA expression in tumor tissues, supporting a dysregulated expression of PCDH10 and PCDH17 by promoter methylation.
Inconsistent associations of mRNA and protein with methylation are not completely unexpected, given the complicated regulatory processes that occur between transcription and translation. Several reasons may help explain. First, the mRNA detected might correspond to different physical location in the gene than the targeted location of the protein assay. Second, multiple CpG sites included in the DHPLC-targeted chromosome regions might hide the effect of individual CpG site. We designed our targeted chromosome regions based on both prior array analysis and literature research so may not contain many CpG sites covered in Illumina HumanMethylation450. No overlapping CpG loci were found between the Illumina HumanMethylation450 platform and our targeted genomic region for PCDH17. Third, the association between methylation and mRNA expression in tumor tissues might have been confounded by other changes caused by cancer itself, but we do not have power to examine the association in nontumor tissues (n = 2 only with both methylation and expression data).
Although H. pylori infection increased methylation of these genes in our analysis, we did not find significant heterogeneity in the association between methylation and risk of gastric lesions across H. pylori–positive and –negative participants, demonstrating that PCDH10 and PCDH17 hypermethylation might be a risk predictor for advanced gastric lesions potentially independent of H. pylori infection as well.
In a high-risk population-based study, we evaluated the methylation level and expression of PCDH genes in a spectrum of gastric lesions with a reasonable sample size. We collected information on H. pylori infection and major characteristics, which allowed for detailed analyses according to H. pylori infection and gastric lesions adjusting for major confounders. Our study also has some limitations. First, we only had access to a limited number of gastric cancer tissues and were not able to examine the association of methylation and expression with risk of gastric cancer based on our own samples. Second, TCGA database only has two nongastric cancer tissues with methylation data, which restricted our ability to compare the methylation between tumor and nontumor tissues. Third, the DHPLC method we used for methylation assay has the advantage of integrating signals from multiple CpG sites but may possibly dilute the effect of individual CpG site. Fourth, the median methylation levels of each gene were used as cutoff values to classify the participants into hyper- and hypomethylation groups, but the extrapolation of the cutoff values to other studies should be approached with caution. Fifth, we cannot distinguish between current and past H. pylori infection by using serological method. However, our goal was to examine how long-term H. pylori infection may have changed the methylation status. It may be interesting to evaluate the association between current H. pylori infection and methylation, but it would not be possible to test the temporal relationship in that analysis.
In conclusion, our high-risk population-based study focused on four PCDHs with reported tumor-suppressing effect for digestive cancers and showed that methylation of these genes increased in H. pylori–positive subjects. The advanced gastric lesions had elevated promoter methylation and decreased protein expression of PCDH10 and PCDH17. Our findings provide evidence that methylation of PCDH10 and PCDH17 in H. pylori–related gastric cancer precursors, suggesting the possibly early involvement of epigenetic alteration of PCDH10 and PCDH17 in gastric carcinogenesis, which may have implications for risk assessment of gastric cancer.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Authors' Contributions
Conception and design: S. Wu, W.-C. You, K.-F. Pan
Development of methodology: S. Wu, Y. Zhang
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S. Wu, L. Zhang, J.-L. Ma, T. Zhou, W.-D. Liu, W.-C. You, K.-F. Pan
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S. Wu, Z.-X. Li, W.-Q. Li
Writing, review, and/or revision of the manuscript: S. Wu, Y. Zhang, W.-Q. Li, W.-C. You, K.-F. Pan
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): Y. Zhang, T. Zhou, Z.-X. Li, K.-F. Pan
Study supervision: W.-C. You, K.-F. Pan
Acknowledgments
We thank all the participants who participated in this study and donated samples.
This work was supported by National Basic Research Program of China (973 Program: 2010CB529303), National Key Technology Research and Development Program (2015BA13B07), National Natural Science Foundation of China (81572774), Science Foundation of Peking University Cancer Hospital (2017-6), and Beijing Municipal Administration of Hospitals' Ascent Plan (DFL20181102).
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.