Clinical Features and Genomic Characterization of Post-Colonoscopy Colorectal Cancer

INTRODUCTION: Some colorectal cancers (CRCs) may be missed during colonoscopies. We aimed to determine the clinicopathological, biological, and genomic characteristics of post-colonoscopy CRCs (PCCRCs). METHODS: Of the 1,619 consecutive patients with 1,765 CRCs detected between 2008 and 2016, 63 patients with 67 PCCRCs, when colonoscopies were performed 6–60 months before diagnosis, were recruited. After excluding patients with inflammatory bowel disease, familial polyposis syndrome, CRCs that developed from diminutive adenomatous polyps, and recurrent CRCs after endoscopic resection, 32 patients with 34 PCCRCs were enrolled. The lesions' clinicopathological features, mismatch repair proteins (MMRs), and genomic alterations were investigated. RESULTS: The overall PCCRC-5y rate, rate of intramucosal (Tis) lesions, and rate of T1 or more deeply invasive cancers were 3.7% (66/1,764), 3.9% (32/820), and 3.6% (34/944), respectively. Thirty-three patients' MMRs were investigated; 7 (21%) exhibited deficient MMRs (dMMRs), comprising 4 with T2 or more deeply invasive cancers and 5 whose lesions were in the proximal colon. Twenty-three tumors' genomic mutations were investigated; PIK3CA had mutated in 5 of 6 T2 or more deeply invasive cancers, of which, 4 were located in the proximal colon. Two patients with dMMRs and BRAFV600E mutations had poor prognoses. Sixty-one percent (17/28) of the macroscopic type 0 lesions were superficial. All superficial Tis and T1 PCCRCs were detected <24 months after the negative colonoscopies. They were distributed throughout the colon and rectum. DISCUSSION: PCCRCs may be invasive cancers in the proximal colon that exhibit dMMRs and/or PIK3CA mutations or missed early CRCs especially superficial lesions.


Patient enrollment
This retrospective, single-center study recruited 1,619 consecutive patients with 1,765 CRCs diagnosed at Hiroshima University Hospital between January 2008 and December 2016. The colonoscopy dates and results were retrieved from electronic medical records and colonoscopy reports. Patients who had $1 negative colonoscopies 6-60 months before a CRC diagnosis were extracted as patients with PCCRC. To investigate the characteristics of PCCRCs, which had an undetermined etiology, and evaluate the reason for them having been missed during the colonoscopies, PCCRCs with definitive causes or obvious carcinogenic factors were excluded. Therefore, the exclusion criteria were as follows: inflammatory bowel disease, familial polyposis syndrome, CRCs developed from previously detected diminutive adenomatous polyps, and recurrent CRC after ER. The study was performed in accordance with the Declaration of Helsinki, and its protocol was approved by Hiroshima University's Institutional Review Board (E-1371).
Definitions "Negative colonoscopy" was defined as a colonoscopy that was negative for CRC or a colonoscopy wherein all CRCs had undergone ER. PCCRC was defined as a CRC with a history of $1 negative colonoscopies performed 6-60 months before its diagnosis to minimize contamination with diagnostic examinations (23). CRC with a history of colonoscopy performed ,6 or .60 months before diagnosis was defined as a "detected CRC." "Multiple negative colonoscopies" refer to patients with histories of $1 negative colonoscopies performed ,60 months before the last negative colonoscopy. The PCCRC rate was calculated as the number of PCCRCs (including cases of inflammatory bowel disease, familial polyposis syndrome, CRCs developed from previously detected diminutive adenomatous polyps, and recurrent CRC after ER)/the total number of CRCs (PCCRCs 1 detected CRCs) 3 100 (18), which was described as the PCCRC-5y rate. The PCCRC rate with a colonoscopy interval of 6-36 months was also evaluated, and this was described as the PCCRC-3y rate (35).
All colonoscopies were performed by endoscopic experts with experience performing .1,000 total colonoscopies. A Tis lesion was defined as intramucosal cancer according to the Japanese Society for Cancer of the Colon and Rectum's (31) and the TNM Classification of Malignant Tumors (32) criteria, which correspond with high-grade dysplasia in the World Health Organization classification (36).

Evaluation of clinicopathological characteristics
Data were collected from medical records, electronically stored colonoscopic images and reports, and histology reports that described the patient's age and sex; treatment options; death from cancer; history of adenoma; synchronous/metachronous CRCs; family history of CRC; time to a CRC diagnosis from the previous colonoscopy; history of multiple negative colonoscopies; the cancer location, namely, the proximal colon, including the cecum and ascending/transverse colon, distal colon, including the descending/sigmoid colon, and rectum; the lesion's macroscopic classification; histologic types; T stage; TNM staging; and lymph node or other organ metastases.

Tissue capture and DNA extraction
Tumors and normal colorectal mucosa were visually dissected from several 10-mm FFPE sections. The DNA was extracted using the GeneRead DNA FFPE kit (Qiagen GmbH, Hilden, Germany), eluted into buffer (40 mL), and quantified using the Qubit dsDNA HS kit (Thermo Fisher Scientific, Waltham, MA) and a Qubit 1.0 fluorometer (Thermo Fisher Scientific). The quantities and quality of the FFPE-derived DNA were ascertained by determining the normalized DNA integrity score obtained from the quantitative polymerase chain reaction using an NGS FFPE QC Kit (Agilent Technologies).
Target enrichment and next-generation sequencing DNA from the tumors and normal colorectal mucosa was fragmented into 150-200 base pairs using a restriction enzyme and the SureSelect XT HS Kit (Agilent Technologies) and the XT Low Input Enzymatic Fragmentation Kit (Agilent Technologies); the fragments were used to construct libraries according to the manufacturer's instructions. The DNA ($109 ng) (range 109-2,910 ng) was prepared for sequencing. The exons of 90 oncogenes and the associated introns of 35 fusion oncogenes were enriched using the SureSelect-XT HS NCC oncopanel (Agilent Technologies) (see Table, Supplementary Digital Content 1, http://links.lww.com/ CTG/A401). The pooled libraries underwent quality control checks using the High Sensitivity D1000 ScreenTape Assay (Agilent Technologies) and the 2200 TapeStation System (Agilent Technologies). Sequencing was performed with paired-end reads using the HiSeq X platform (Illumina, San Diego, CA).

Variant detection
The sequencing reads were aligned to the hg19 reference sequence and analyzed using SureCall Software, version 4.1 (Agilent Technologies). To improve mapping quality before variant calling, polymerase chain reaction duplicates were removed using molecular barcodes and SureCall Software, version 4.1. Pairedend and single-end analyses using SureCall Software, version 4.1, were used to identify single nucleotide variants and insertions/ deletions (indels) in the tumors. Called variants in the normal colorectal mucosa were also considered germline mutations. To reduce the false-positive rate, we set cutoff values for somatic mutations in tumors, as follows: read depth .20 and forward/ reverse balance between 0.25 and 0.75. We configured the Sure-Call SNP caller using the SureSelect default settings, as follows: variant score threshold: 0.3, minimum quality for base: 30, variant Clinical and Translational Gastroenterology call quality threshold: 100, minimum allele frequency: 0.1, and minimum number of reads supporting the variant allele: 10. Variants that were repeat sequences registered in the RepeatMasker program (University of California Santa Cruz, Santa Cruz, CA) and called as replacements or clearly identified as sequencing errors in the Integrative Genomic Viewer (Broad Institute of Massachusetts Institute of Technology and Harvard, Cambridge, MA) were excluded from all samples as somatic mutation candidates. The somatic mutations were classified as follows: category I: frameshift indels or nonsense mutations; category II: missense mutations; and others, including synonymous changes, mutations located at introns, or intergene mutations (see Tables, Supplemental Digital  Content 2

RESULTS
Of the 1,619 patients with 1,765 CRCs, 63 patients with 67 CRCs were classified as having PCCRCs. Of the 67 PCCRCs, 33, 18, and 16 were Tis, T1, and T2 or more deeply invasive cancers, respectively. One patient had 2 synchronous Tis PCCRCs. The overall PCCRC-5y rate was 3.7% (66/1,764). The rate of Tis, T1, and T2 or more deeply invasive cancers was 3.9% (32/820), 4.8% (18/377), and 2.8% (16/567), respectively (Table 1). After excluding patients who met the exclusion criteria, 32 patients with 34 PCCRCs were analyzed in the study ( Figure 1). Of the 34 eligible PCCRCs, 18, 8, and 8 were Tis, T1, and T2 or more deeply invasive cancers, respectively. Table 2 shows the clinicopathological features of the PCCRCs. More men (79%) than women (21%) had PCCRCs. Approximately 50% of PCCRCs were located in the proximal colon. Of the macroscopic type 0 PCCRCs, 61% (17/28) were superficial lesions. Most patients had early cancer, and all the PCCRCs were resectable with ER alone, additional surgery after ER, or initial surgery. Eight PCCRCs were metachronous cancers, 4 of which had a history of a colectomy, and 3 had histories of ER at a different site of the colorectum. Table 3 shows the clinicopathological features of the PCCRC according to the T stage. T2 or more deeply invasive cancers were absent in the rectum and were more likely to be in the proximal colon than Tis and T1 cancers. All the T2 or more deeply invasive cancers were diagnosed .12 months after the negative colonoscopies, while 3 of 18 Tis and 3 of 8 T1 cancers were diagnosed within 12 months of previous negative colonoscopies. Sixty-two percent (21/34) of patients had a history of multiple negative colonoscopies. Cecal intubation was achieved in all negative colonoscopies; however, poor bowel preparation was observed in 2 of 18 negative colonoscopies performed in patients with Tis cancers. The 3-year survival rates, determined using the Kaplan-Meier method, for patients with Tis, T1, and T2 or more deeply invasive cancers were 94%, 75%, and 60%, respectively; 2 patients who died from T2 or more deeply invasive cancers had poorly differentiated or mucinous adenocarcinomas with peritoneal disseminations.

Biological characteristics
Thirty-three tumors were analyzed immunohistochemically for MLH1, MSH2, MSH6, and PMS2; one T1 tumor was excluded because an FFPE block containing the tumor was unavailable. Seven PCCRCs were dMMRs, all of which showed absent MLH1 and PMS2 (Table 3). dMMRs were observed in 50%, 29%, and 6% of the T2 or more deeply invasive, T1, and Tis cancers, respectively. We purified DNA from 33 paired PCCRCs and normal tissues, but 10 patients' DNAs were of insufficient quality and/or quantity for sequencing. Therefore, we analyzed 23 PCCRCs with DNA of sufficient quality. Figure 2 shows the mutation landscapes of 23 PCCRCs. BRAF and KRAS were mutated in 3 (13%) and 12 (52%) cases, respectively. Of the 3 cases with BRAF mutations, 2 had missense BRAF V600E mutations coexisting with dMMRs. No cases of Tis or T1 PCCRCs had PIK3CA mutations, and 5 of 6 T2 or more deeply invasive PCCRCs had PIK3CA missense mutations. Next-generation sequencing of DNA from 3 cases with dMMRs showed hypermutations with large numbers of somatic mutations (tumor mutational burden .10). A patient had suspected Lynch syndrome (case 16): One had 2 first-degree relatives with CRC, and developed CRC before 50 years, which had dMMRs but not the BRAF mutation; their MMR gene germline mutations were not investigated.
Relationship among the colonoscopy intervals and the clinical and biological characteristics Figure 3a shows the intervals between the PCCRC diagnoses and previous colonoscopies in the context of the T stage, tumor morphology, and MMR immunohistochemistry in all cases. The intervals between the previous negative colonoscopies and the diagnoses of superficial Tis and T1 PCCRCs were ,24 months and mostly around 12 months. None of the superficial lesions exhibited dMMRs. Compared with the PCCRCs exhibiting pMMRs, those exhibiting dMMRs tended to be detected later when they had invaded more deeply. One T3 PCCRC detected within 24 months of a negative colonoscopy exhibited pMMRs (case 9); this was the only tumor with an MYC missense mutation. Figure 3b shows the distribution of the PCCRCs within the colon and rectum. Cancers exhibiting dMMRs were more likely to be located in the proximal colon. Superficial lesions were distributed evenly throughout the entire colon and rectum. Less superficial lesions, including type 2/3 lesions and type 0 lesions with polypoid morphologies, were more likely to be located in the sigmoid colon. Figure 4 presents endoscopic images of superficial PCCRCs. Many of the Tis and T1 cancers had superficial appearances and coloring that was similar to that of the surrounding normal mucosa and were considered difficult to detect. Both the PCCRCs with dMMRs and BRAF V600E mutations were located in the proximal colon, had peritoneal disseminations, and caused the patients' deaths (see Figure

DISCUSSION
This study's findings revealed the clinicopathological and biological characteristics of PCCRCs. Although most PCCRCs are considered lesions that were missed at previous colonoscopies (13,20), this is difficult to prove scientifically; a missed lesion may be a rapidly developing cancer. To investigate this issue, we included Tis lesions, excluded CRCs that had recurred after ER, and examined the MMR status and genomic landscapes of PCCRCs as well as the relationship with morphology and location.
In this study, the PCCRCs were more likely to be detected in men, which differs from previous reports (14,(18)(19)(20)(21). Unlike the Tis and T1 lesions, the T2 or more deeply invasive cancers were more likely to be located in the proximal colon, which concurs with previous reports (13,14,(16)(17)(18)(20)(21)(22)(23)(24)(25)(26)(27)(28)(29). Many Tis and T1 cancers appeared superficial, their coloring was similar to that of the surrounding normal mucosa, and they were detected soon after negative colonoscopies. Therefore, it is highly likely that the early-stage PCCRCs were missed lesions irrespective of their locations and were detected through high-quality colonoscopies. However, less superficial lesions, regardless of their T stage, were detected frequently in the sigmoid colon where folds and bends cause blind spots. Therefore, colonoscopists must take care not to miss less superficial lesions, especially in the sigmoid colon. Clearly, inadequate colonoscopies and poor endoscopic skills affect the frequencies at which lesions are missed (13)(14)(15)19,21,22,24). Furthermore, in this study, poor bowel preparation was observed in 11% of the previous negative colonoscopies in patients with Tis cancers.
Regarding the biological features of PCCRCs, 40% of T1 or more deeply invasive cancers exhibited dMMRs, which aligns with previously reported rates of 24%-32% (23,26,(28)(29)(30); these lesions were more likely to be located in the proximal colon. Interestingly, only 6% of the Tis cancers exhibited dMMRs, which was a considerably lower rate than the rates for T1 or more deeply invasive cancers. This indicates that most of the Tis PCCRCs were slowgrowing lesions, missed during previous colonoscopies. By contrast, 4 of 8 T2 or more deeply invasive cancers located in the proximal colon exhibited dMMRs; these may have been rapidly advancing cancers. Two of these lesions were mucinous or poorly differentiated adenocarcinomas, had BRAF V600E mutations (suggesting carcinogenesis through the serrated pathway) (34), and were associated with poor prognoses; mutation analyses failed for the other 2 T2 or more deeply invasive cancers. Shaukat et al. (30) also reported that PCCRCs with BRAF mutations were more likely to be located in the proximal colon, be mucinous or poorly differentiated adenocarcinomas, have a high microsatellite instability, and be associated with poor prognoses. Thus, PCCRCs that progress through the serrated pathway grow rapidly and are associated with poor prognoses. Sessile serrated lesions, which are precursors of CRCs, also appear superficial, and their coloring is similar to that of the surrounding normal mucosa, and are easily missed, as To date, studies have investigated a few gene mutations, e.g., KRAS and BRAF. In this study, the genomic mutations were analyzed exhaustively, and some characteristic genomic alterations were detected. We confirmed that PCCRCs with dMMRs were associated with hypermutations. One T3 PCCRC with pMMRs, detected ,24 months after a negative colonoscopy (case 9), had an MYC missense mutation, which is inactivated by Wnt activation and the inactivation of transforming growth factor (TGF)-b signaling, and is strongly associated with CRC proliferation; therefore, it might be associated with tumor progression (41). PIK3CA encodes p110a, which is a catalytic subunit of PI3K that is associated with the proliferation of CRCs. PIK3CA mutated in 5 of 6 T2 or more deeply invasive PCCRCs, which was more frequent than that reported for CRC (27%) by The Cancer Genome Atlas (42). In addition, 4 of 5 tumors with PIK3CA mutations were located in the proximal colon, and PIK3CA mutations were not detected in the Tis/T1 PCCRCs; therefore, invasive PCCRCs in the proximal colon were strongly associated with PIK3CA mutations. KRAS and BRAF mutations were found in 52% (12/23) and 13% (3/23) of the tumors, respectively. Previous studies have shown that KRAS and BRAF mutations were present in 23%-29% (9,23,27,29) and 17%-28% (9,23,27,29,30) of PCCRCs, respectively; these rates did not differ from those in detected CRCs (9,27,29,30). The frequency of the KRAS mutation was higher in this study than in other studies, which may be due to fewer patients.
This study had some limitations. First, this was a single-center retrospective study with a relatively small number of patients. Although a large multicenter study could generate superior epidemiological results, it may generate less accurate information about the colonoscopies and lower quality colonoscopies. In this study, we could determine whether PCCRCs are related to missed lesions by investigating detailed and accurate data. Second, the genomic analyses were performed on a limited 23 of 34 PCCRCs because most others had insufficient quality and/or quantity DNA. Tumor sample storage was not consistent, and many tumors had been resected .5 years before they were analyzed. Third, the analysis of the patients with CRC during this study was based on the policy that diminutive polyps may be followed without resection. Therefore, PCCRCs in this study included many lesions that developed from previously detected diminutive polyps, which ought to be excluded from the analysis. Fourth, the number of PCCRCs in this study may have been underestimated compared with those in other population-based studies because we could not account for patients who had undergone previous colonoscopies in other hospitals. Finally, the adenoma detection rate of each colonoscopist, which can have a great impact on PCCRC (12), could not be evaluated. Although a large-scale prospective study with uniform colonoscopy intervals and high-quality colonoscopy is required, our study investigated the clinicopathological features, dMMR immunohistochemistry, genomic mutations of PCCRCs, and the interrelationship among these characteristics, and its findings indicated that PCCRCs are associated with missed lesions and rapidly growing cancers.
In conclusion, PCCRCs can be invasive cancers with dMMRs and/or PIK3CA mutations, which can grow rapidly, are associated with poor prognoses, and are located in the proximal colon, or missed early lesions that are either superficial lesions that can be located anywhere in the colon and rectum or less superficial lesions in the sigmoid colon. To prevent PCCRC, superficial lesions should not be missed during colonoscopy. Although these findings must be validated in prospective large-scale clinical trials, we believe that our findings inform the development of new CRC surveillance methods and the identification of a group of CRCs with a poor prognosis.