Tissues were collected from gastrointestinal endoscopy during annual physical examinations or from curative resections of colorectal cancer. The collected lesions underwent evaluation through haematoxylin and eosin (H&E) staining as well as immunohistochemical staining for pathological analysis. FFPE or fresh tissues, and blood samples were subjected to whole-exome sequencing (WES). WES sequencing was performed on multiple regions of adenocarcinoma (n = 8), villous adenoma (n = 10), tubular adenoma (n = 9) and blood samples obtained from 9 patients belonging to 7 Chinese FAP families.

Whole-exome sequencing was conducted utilizing the Agilent whole-exome capture kit (SureSelectXT Human All Exon 50 Mb), as described previously (Coffey et al., 2011). Briefly, total DNA was extracted from the collected specimens using a standard DNA extraction protocol. The extracted genomic DNA was sonicated to generate fragments of 150–200 bp. Subsequently, the fragmented DNA underwent processing and preparation for sequencing using a whole-exome library preparation kit, following the manufacturer’s guidelines. Multiple libraries, each with distinct barcode adaptors, were pooled together for the whole-exome sequence capture step.

To verify the APC germline mutation sites, Sanger sequencing was employed. The primers utilized for Sanger sequencing are summarized in online Supplementary Table S1.

Heterozygous mutations were first clustered on the basis of their variant allele fraction using sciClone (Miller et al., 2014) to identify the founding clones and subclones that were subsequently analyzed using ClonEvol (Dang et al., 2017) to infer clonal evolution models.

The study cohort consisted of 9 patients diagnosed with FAP, with 3 patients (FAP07, FAP08, and FAP09) from the same family. A total of 36 samples were collected, including 9 peripheral blood samples, 9 tubular adenomas, 10 villous adenomas, and 8 carcinomas (Table 1). Among the patients, inherited germline mutation sites were identified in 8 out of 9 individuals. These mutations include FAP01 p.S1281* (COSM19212), FAP03 p.S384Tfs*19, FAP04 p.E1538* (COSM6041693), FAP05 p.Q1062* (COSM3696862), and FAP07-FAP09 p.V677Sfs*3. The pedigree analysis of FAP patients were shown in Figure 1.

A novel germline mutation, c.2028dupA (p.V677Sfs3), was identified in the family of FAP09 (Figure 1G, H). This mutation has not been previously reported in published literature or any existing databases. Genetic screening was conducted for p.V677Sfs3 on 5 individuals spanning four generations of this family. FAP09, the proband, had a grandfather (I1) who passed away from colon cancer at the age of 45, but it remains unclear if he carried the pathogenic mutation. The proband’s mother (FAP08,Ⅱ2) and uncle (FAP07, II1) were found to have the heterozygous p.V677Sfs3 mutation, and colonoscopy revealed multiple polyps in their cases. However, their sister (Ⅲ2) had a normal colonoscopy and did not carry the p.V677Sfs3 mutation. Similarly, their nephew (IV1) exhibited a normal phenotype and did not exhibit any genetic mutation. Hence, the p.V677Sfs3 mutation segregates independently from the phenotype in his family.

According to the guidelines established by the American College of Medical Genetics and Genomics (ACMG) (Sirohi et al., 2020), we have assessed the pathogenicity of the p.V677Sfs*3 variant as follows: The p.V677Sfs*3 variant is located in exon 14 of the APC gene, where an additional base has been inserted, leading to a frameshift mutation that results in premature termination of protein translation. This is considered a loss-of-function variant, meeting the criteria for strong evidence of pathogenicity (PVS1). Upon searching the 1,000 Genomes, gnomAD, and ESP databases, the p.V677Sfs* variant has not been identified in the general population, which aligns with the pathogenicity criterion PM2. The frequency of the variant is significantly higher in the affected population compared to the control population, which satisfies the pathogenicity criterion PS4. There is clear evidence of co-segregation of the mutation with the disease phenotype within families, which meets the pathogenicity criterion PP1. Based on the classification standards for pathogenic variants established by the ACMG, the APC p.V677Sfs*3 variant meets the pathogenicity criteria PVS1, PM2, PS4, and PP1, and is therefore classified as a pathogenic variant.

FAP09 underwent curative resection of colorectal cancer in 2016, but liver metastasis was detected in 2018. Unfortunately, FAP09 eventually succumbed to CRC recurrence 1 year after the discovery of liver metastasis. Since 2016, both FAP07 and FAP08 have been undergoing regular annual colonoscopic adenoma treatment. Tissue samples were collected from different historical stages of patients within this family to elucidate the evolutionary history of adenomas progressing to colorectal cancer in the FAP family.

The somatic mutation landscape of the cohort was investigated (see somatic mutations list in Supplementary Data Sheet S2). Firstly, mutations present in the tissue specimens were subtracted from blood mutations. Next, these resulting mutations were further filtered based on their similarities to reproductive mutations in public databases and algorithms to isolate somatic mutations. A total of 1,484 somatic mutations were detected in eight carcinomas, 1,499 somatic mutations in 10 villous adenomas, and 2,919 somatic mutations in nine tubular adenomas (Figure 2A). It is worth noting that there were few common mutations shared between any two tissue samples within the same patient, indicating that the majority of somatic variants are random mutations (Supplementary Figure S1). Figure 2B illustrates the total number of exonic somatic variants (SVNs) in individual tissue samples. The mean quantities of SVNs were 108.0 (27, 206.5) in tubular adenomas, 122.0 (51.5, 309.8) in villous adenomas, and 290.5 (89, 388.5) in FAP-CRC. There is a tendency that the quantities of SVNs increased from tubular, villous to FAP-CRC. However, due to the limited sample size within each group and the substantial variability observed, traditional analysis of variance (ANOVA) did not reveal statistically significant differences (p = 0.50). To avoid bias in further analysis, tubular adenomas in FAP3 and FAP08, which had abnormally large numbers of SNVs, were excluded. The mutation burdens (MBs), measured as the number of mutations per mega-base, were analyzed for FAP-CRC and adenomas. FAP-CRCs exhibited a median mutation burden of 5.8 single nucleotide alterations (SNAs) per Mb, which was roughly twice as high as S-CRC (TCGA-CRCs, 2.5 SNAs/Mb). Additionally, both tubular adenomas and villous adenomas displayed considerable mutation burdens, with medians of 2.8 and 3.1 SNAs, respectively, suggesting the accumulation of somatic mutations from the early stages of cancer transformation (Figure 2C). Although statistical significance was not achieved due to the limited sample size, FAP-CRCs showed a relatively higher occurrence of TP53 mutations compared to S-CRC (Figure 2D).

By examining the whole-exome sequencing (WES) data from villous, tubular, and adenocarcinoma samples and comparing them with the oncogenic and tumor suppressor genes listed in the COSMIC database, we identified 24 genes that were shared among these samples (Figure 3A). In addition to the conventional recurrently mutated genes observed in FAPs, such as APC, RAS, and TP53, we also discovered novel recurrently mutated genes in the homologous recombination repair pathway. These genes include SUZ12, KMT2C, BCLAF1, DROSHA, RUNX1, SMAD4, RGPD3, and ARID1B. The presence of mutations in these genes suggests a state of genome instability.

The frequency of APC mutations, including protein-truncating mutations, was found to be the highest among all the lesions from patients with FAP (15 out of 25; mutation frequency = 69%) (Figure 3B). The occurrence of the second hit of APC may occur in the early phases of carcinogenesis since there was no significant difference in APC mutation frequency observed among tubular adenomas, villous adenomas, and carcinomas. KRAS missense mutations were found in carcinomas (3 out of 8; mutation frequency = 37.5%), in villous adenomas (5 out of 10; mutation frequency = 50%) and in tubular adenomas (2 out of 9; mutation frequency = 22.2%). Interestingly, it seems that villous adenomas had a higher KRAS mutation rate than carcinomas, although this was not statistically significant due to the limited sample size. Nearly all FAP-CRC samples (7 out of 8; mutation frequency = 87.5%) showed TP53 mutations, including three truncating mutations and four missense mutations. Two villous adenomas with TP53 mutations were found to be complicated with regional canceration, indicating that TP53 mutations occurred during the malignant transformation stage.

FAP07, FAP08, and FAP09 belong to the same family and carry the pathogenic variation p.V677Sfs*3. Within this family, the second hit of APC gene was detected in 3 out of 4 villous adenomas, 2 out of 3 tubular adenomas, and 3 out of 5 carcinomas (Table 2; Figure 4A). Interestingly, a specific somatic mutation, APC p.R1450*, was found in certain adenomas of all three patients. Mutation analysis revealed that RAS mutations, including KRAS and NRAS, were only detected in the tissues of carcinomas. Notably, the presence of the same mutation panel, with NRAS p.G12S and TP53 p.G245S, was identified in 18R10487 (carcinoma from curative resection of colorectal cancer) and 18R10489 (carcinoma from liver metastases) (Figure 4B). Except for divergent mutations in NRAS and TP53, the two samples shared numerous somatic mutations, indicating a common origin.

In the mutation burden analysis, the mutation burden for tubular adenomas was calculated as 6.28 (6.17, 6.53), for villous adenomas it was 10.06 (5.13, 4.54), and for carcinomas it was 10.07 7.7, 11.75). There are significant difference among them (p = 0.01) in one way ANOVA analysis. In the analysis of homologous recombination deficiency (HRD), the HRD score for tubular adenomas was 6.28 (ranging from 5.0 to 15.0), for villous adenomas it was 15.50 (ranging from 12.0 to 24), and for carcinomas it was 22.0 (ranging from 21.0 to 45.0). There is a significant difference between tubular adenoma and cancer with p = 0.03 by t-test. Although the statistical difference between tubular adenoma and villous adenoma is not significant, it may be due to the limited number of specimens. There is a clear tendency of progressively increasing mutation burden and HRD score from tubular adenomas to villous adenomas and ultimately to carcinomas, indicating a state of genome instability.

To understand the development and evolution from adenomas to carcinomas in FAP, phylogenetic trees of tissue samples from various sites were assessed. The TP53 mutation was identified as the primary driver gene for malignant transformation, except in the case of FAP.07. Remarkably, the conventional driver mutations (TP53, RAS, and SMAD4) consistently emerged simultaneously rather than being acquired sequentially after the initiation of cancer growth, as exemplified in FAP.02, FAP.04, and FAP.09 (Figure 5). Based on the current data, a linear evolution relationship from tubular adenomas to villous adenomas and ultimately to carcinomas cannot be determined. The malignant transformation from adenomas to carcinomas appears to occur abruptly, accompanied by mutations in driver genes.

In the case of FAP.09, depth sequencing was performed on five carcinomas, allowing for accurate quantification of mutant allele frequencies and estimation of tumor clonal population sizes. Through the analysis of mutation clustering, four distinct clones with different sets of mutations were identified in the carcinomas of patient FAP.09 (Figure 6). The sample labeled as 19R03643 was presumed to be the initial clone, with median mutant allele frequencies of 34.98%, 5.38%, 12.42%, and 40.65% observed for clusters 0 to 3, respectively. Clone 0 serves as the “founding” clone, from which the other subclones are derived, and is presumed to be present in nearly all tumor cells. Clones 1, 2, and 3 are believed to have evolved from clone 0. It is likely that a single cell from clone three acquired a specific set of mutations, giving rise to clone 4, which subsequently became the dominant clone possessing metastatic capabilities. The NRAS p.G12S and TP53 p.G245S mutations are the most probable candidate genes in clone 4, providing a significant selective advantage for proliferation and metastasis.