Overall gene copy number alterations

In the overall cohort, the median number of genes with total copy number alterations per patient was 1728, with a median of 1,252.5 genes with a gain (range: 0–40,285), 0 with LOH (range: 0–18,552), and 1.5 with CN-LOH (range: 0–5577). In subgroup 1, the median number of genes with total copy number alterations per patient was 40,277.5, with a median of 40,277.5 genes with a gain (range: 37,735–40,285), 0 with LOH (range: 0–747), and 0 with CN-LOH (range: 0–1699). In subgroup 2, the median number of genes with total copy number alterations per patient was 1662.5, with a median of 1,063 genes with gains (range: 0–15,116), 0 with LOH (range: 0–18,552), and 2 with CN-LOH (range: 0–5577).

Hierarchical clustering analysis based on gene copy number alterations in IFN, LGD, and HGD cases with the MSS phenotype

We evaluated gene copy number alterations (GCNAs) in lieu of SCNAs to determine copy number changes, given that the number of individual genes mapping to corresponding loci (GCNAs) is easier to calculate than the number of allelic loci (SCNAs). The IFN, LGD, and HGD cases examined were classified as MSS. We performed hierarchical clustering analysis based on the GCNAs, including gains, LOH, and CN-LOH, to examine differences in genetic alterations among the IFN, LGD and HGD cases with the MSS phenotype (Fig. 2).

Fig.2

Hierarchical cluster analysis based on somatic copy number alterations in 108 gastric intramucosal neoplasias with the MSS phenotype

We classified the IFN, LGD, and HGD cases with the MSS phenotype into two subgroups according to GCNA frequency: high (subgroup 1) and low (subgroup 2) (Fig. 2), and examined the clinicopathological characteristics of each (Supplementary Table 1). The frequencies of IFN and LGD cases were statistically higher in subgroup 2 versus 1. However, no significant difference existed in the frequency of HGD between subgroups. There were no statistical differences in the other clinicopathological findings examined between subgroups.

We compared the numbers of genes with gains, LOH, CN-LOH, and total GCNAs between subgroups. Significant differences in the average number of genes with copy number gains were found (p < 0.0001) (Fig. 3a). However, the numbers of genes with LOH and CN-LOH were similar between subgroups (Fig. 3b and c). In addition, there was a significant difference in the number of genes with total GCNAs between subgroups 1 and 2 (p < 0.0001) (Fig. 3d).

Fig.3

Comparison of the numbers of total GCNAs, gains, copy-neutral LOH, and LOH in 108 gastric intramucosal neoplasia cases with MSS. a Comparison of the number of genes with gains between subgroups 1 and 2. b Comparison of the number of genes with LOH between subgroups 1 and 2. c Comparison of the number of genes with CN-LOH between subgroups 1 and 2. d Comparison of the number of genes with total SCNAs between subgroups 1 and 2. Detailed GCNA data for subgroups 1 and 2 are shown

Somatic gene number alterations in IFN, LGD, and HGD cases with the MSS phenotype

We compared the numbers of genes with gains, LOH, CN-LOH, and total GCNAs among lesion types (IFN, LGD and HGD). Significant differences in the average number of genes with gains between IFN and HGD or LGD were found (p < 0.01; p < 0.01) (Fig. 4a). The number of genes with CN-LOH was significantly higher in IFN or HGD than in LGD. However, the number of genes with LOH was similar among lesion types (Fig. 4b and c). Moreover, significant differences existed in the total number of GCNAs among the three lesion types (p < 0.01; p < 0.001) (Fig. 4d).

Fig. 4

Comparison of the numbers of total GCNAs, gains, CN-LOH, and LOHs in 108 gastric intramucosal neoplasia cases with MSS. a Comparison of the total number of genes with gains among the IFN, LGD, and HGD cases. b Comparison of the number of genes with LOH among the IFN, LGD, and HGD cases. c Comparison of the number of genes with CN-LOH among the IFN, LGD, and HGD cases. d Comparison of the number of genes with total SCNAs among the IFN, LGD, and HGD cases. Detailed GCNA data for IFN, LGD and HGD are shown

Gain events detected in > 30% of cases were located at 3p21.1, 3p14.2–3p14.1, 3p13, 4p13–4q35.2, 8p23.3–8q24.3 in the IFN cases (Supplementary Table 2) versus 1p36.33–1p36.11, 1p35.2–1p31.1, 1p12–1q44, 2p25.3–2q37.3, 3p26.3–3q29, 4p16.3–4p15.32, 4q31.23–4q34.1, 5p15.33–5q11.1, 6p25.3–6q27, 7p22.3–7q36.3, 8p23.3–8q24.3, 9q13–9q34.3, 10p15.3–10q26.3, 11q12.2–11q25, 12p13.33–12p13.32, 12q14.1–12q15, 13q11–13q34, 15q21.3–15q26.3, 16p13.3–16q24.3, 17q11.1–17q25.3, 18p11.32–18p11.21, 18q11.1–18q23, 19q11–19q13.2, 20p13–20q13.33 and 21q21.2–21q22.11 in the HGD cases (Supplementary Table 3). In addition, no LOH or CN-LOH events were detected in more than 30% of cases (Supplementary Table 2, 3).

There was a significant difference in the frequency of gain events (> 30% of cases) between IFN and LGD at 4p13–4q35.2, but no difference (> 30% of cases) in that of LOH or CN-LOH (Supplementary Table 4). In contrast, significant differences existed in the frequencies of multiple gain events (present in > 30% of cases) between IFN and HGD cases, but no such difference in LOH or CN-LOH was observed (Supplementary Table 5). We searched for candidate oncogenes located at 4p13–4q35.2 using the Cancer Gene Census. In this chromosomal region, we identified Rap1 GTPase-GDP dissociation stimulator 1 (RAP1GDS1) and lymphoid enhancer binding factor 1 (LEF1), which have oncogenic functions. The frequencies of RAP1GDS1 and LEF1 mutations were significantly higher in IFN than in LGD (Table 2). There was no significant difference in the frequency of gain events between IFN and HGD.

Table 2 Comparison of mutations in 4p13–4q35.2, the locus encompassing RAP1GDS1 and LEF1, between IFN and LGD

Next, we examined the interaction networks closely associated with RAP1GDS1 and LEF1, respectively, using the STRING database (https://string-db.org). Of the candidate signaling pathways found, Rap1 signaling (potentially downstream of the MAPK and PI3K/AKT pathways) and Wnt signaling were associated with RAP1GDS1 and LEF1, respectively (Supplementary Fig. 1).

Mutation analysis of IFN, LGD, and HGD using NGS

We examined 28 genes included on a customized gene panel, in which whole exons of APC and TP53 were included, but only hot spots were included for the other 26 genes. The frequency of TP53 mutation was significantly higher in HGD than in LGD (10 of 32 versus 1 of 39 cases, respectively), but no significant difference existed between IFN and HGD. Although the APC mutation frequency was higher in LGD and HGD (7 of 32 and 6 of 39 cases, respectively) than in IFN (1 of 30), the difference was not statistically significant. APC nonsense (LGD: 4; HGD: 6) and frameshift (LGD: 2; HGD: 1) mutations were common in LGD and HGD. TP53 mutations in exons 4–8 were observed. TP53 and missense and transition mutations were common in each lesion examined. The frequency of KRAS mutations was lower in IFN, LGD, and HGD. Detailed mutation results are shown in Supplementary Tables 6–10.