Genomic Landscape of Early-stage Colorectal Neoplasia Developing From the Ulcerative Colitis Mucosa in the Japanese Population

Abstract

Backgrounds

Colorectal neoplasias (CRN)s developing from the ulcerative colitis (UC) mucosa include both colitic and sporadic neoplasias. Although several genomic analyses of advanced colitis-associated cancer are available, such studies do not distinguish between colitic and sporadic cases, and the early-stage genomic alterations involved in the onset of colitic cancer remain unclear. To address this, we performed a genomic analysis of early-stage CRN developing from the UC mucosa (CRNUC).

Methods

We extracted DNA from 36 early-stage CRNUCs (T1 cancer, 10; dysplasia, 26) from 32 UC patients and performed targeted sequencing of 43 genes commonly associated with colitis-associated cancer and compared the results with sequencing data from the Japanese invasive colitis-associated cancer.

Results

The most frequently mutated gene in the CRNUC cohort was APC (mutated in 47.2% of the cases), followed by TP53 (44.4%), KRAS (27.8%), and PRKDC (27.8%). None of the TP53 mutations occurred at any of the hotspot codons. Although the TP53 mutations in The Cancer Genome Atlas of Colorectal Cancer were dispersed throughout the gene, those detected here in CRNUC cases were concentrated in the amino terminal part of the DNA-binding domain. Interestingly, the mutations in KRAS and TP53 were mutually exclusive in CRNUC, and CRNUCs with KRAS mutations had histologically serrated lesions in the gland duct. Mayo endoscopic subscore was higher in TP53-mutated CRNUCs and lower in KRAS-mutated CRNUCs.

Conclusions

Our findings suggest that early-stage CRNUC can be classified into 2 groups: those developing through the carcinogenic pathway via TP53 mutations and those developing through the carcinogenic pathway via KRAS mutations.

INTRODUCTION

Patients with ulcerative colitis (UC) have an increased risk for developing colorectal cancer (CRC), known as colitis-associated cancer (CAC).¹ High-risk UC patients are placed under close colonoscopic surveillance to facilitate earlier CRC detection, thereby improving prognosis.^2–4

Advances in the molecular genetics of cancer have led to the characterization of the genomic landscape of numerous types of tumor. These findings allowed the improvement of diagnosis and prognosis of cancer, expanding the possibilities for targeted therapies.^5–7 In the field of inflammatory bowel disease, the biological changes underlying the occurrence of CAC have not been fully elucidated. Nonetheless, current data suggest that the genomic variations leading to the development of CAC are different from those linked to sporadic CRCs. Chronic inflammation due to inflammatory bowel disease has been hypothesized to lead to genomic variations that increase the risk of CAC, and several studies reporting genomic analyses of CAC have deepened our molecular understanding of this disease.^{8, 9}

With the increasing availability of next-generation sequencing, exploration of the genomic landscape has become more feasible. Two recent studies comparing mutations in cancer-related genes between CACs and sporadic CRCs^{8, 9} found several recurrent mutations, including prevalent somatic mutations in TP53, APC, and KRAS, with a high frequency of TP53 mutations ranging from 63% to 89%. Additionally, IDH1, a new potential therapeutic target, was identified as a recurrent mutated gene in CACs that is more often reported in patients with Crohn’s disease than in those with UC.⁹

Although several studies of advanced-stage CAC, including the 2 aforementioned analyses, have been reported, only a few studies have performed genomic analysis on early-stage CAC. Contrarily, there has been an increasing number of reports available on genomic analysis of early-stage cancers of other organs such as the esophagus and stomach, providing useful information for early detection/diagnosis and treatment.^{10, 11}

In recent years for UC patients with CRN, the Surveillance for Colorectal Endoscopic Neoplasia Detection and Management in Inflammatory Bowel Disease Patients: International Consensus Recommendations (SCENIC) guidelines (2015) recommend that after complete removal of endoscopically resectable dysplastic lesions, patients should also undergo a surveillance colonoscopy rather than colectomy.¹² Although detection and diagnosis of CAC with colonoscopy is important, random biopsies are occasionally required because it is often difficult to identify CAC by endoscopy alone. However, improvements in the diagnostic capabilities using narrow-band imaging and magnifying chromoendoscopy have enabled the detection of an increased number of CAC.^{13, 14} Latest findings have also demonstrated the usefulness of endoscopic resection such as endoscopic submucosal dissection for CAC.¹⁵ By performing a genomic analysis of CAC specimens resected endoscopically, detailed information on CACs can be obtained, which may further increase diagnostic accuracy and thereby facilitate successful treatment.

Moreover, distinguishing colitic cancers originating from the inflamed mucosa from sporadic cancers originating in a noninflammatory environment is crucial for determination of the proper treatment, which directly affects prognosis.^{16, 17} However, in previous reports on CAC, the term “CAC” referred to colorectal neoplasia (CRN) developing from the UC mucosa and included both colitic and sporadic neoplasia. To date, no method of classification has been developed to distinguish between the 2 tumor types, hindering the treatment.^18–20 In this study, we aimed to examine the genetic and histopathologic features of early-stage colorectal neoplasia developing from the UC mucosa (CRNUC) to provide classification criteria to distinguish between colitic and sporadic neoplasia in the UC mucosa.

MATERIALS AND METHODS

Patients and Sample Preparation

Thirty-two consecutive UC patients with 36 early-stage CRNUC cases and who underwent surgery or endoscopic resection at the Hiroshima University Hospital from June 1998 to July 2018,were retrospectively enrolled in the study. Of the 36 CRNUC cases, 10 had T1 cancer, 19 had high-grade dysplasia (HGD), and 7 had low-grade dysplasia (LGD). Clinicopathological features of each CRNUC case are shown in Supplementary Table 1.

Histopathological Assessment

All specimens were fixed in 10% formalin, cut into 2-mm sections, embedded in paraffin, serially sectioned, stained with hematoxylin and eosin, and examined microscopically. Histopathology was determined based on the Vienna classification²¹ and the histological system proposed by Riddell et al.²⁰ If necessary, a combination of p53 immunostaining and hematoxylin and eosin staining was used for the diagnosis of CRNUC, along with Ki-67 immunostaining. In sporadic neoplasia, Ki-67-positive cells are mainly distributed at the superficial zone of the mucosal layer, and tumor cells differentiate toward the basal side of the mucosa; this is known as the “top-down pattern.” However in colitic neoplasia, Ki-67-positive cells are mainly found at the basal side of the mucosa, and tumor cells differentiate toward the superficial side of the mucosa; this is known as the “bottom-up pattern.” Colorectal neoplasia developing from the UC mucosa in which Ki-67 cells were not distributed biased to either the superficial zone or basal side of the mucosa but were distributed over the entire mucosa was classified as “diffuse pattern.” ²² In all cases, dysplasia or cancer was confirmed by 1 or more gastrointestinal pathologists.

Tissue Collection and DNA Extraction

The pathological tumor tissues (cancer and/or dysplasia) and the nontumor tissues surrounding the tumor were dissected from several 10-μm thick slides prepared using formalin-fixed paraffin-embedded (FFPE) specimens, which were deparaffinized, stained, and dehydrated via the Arcturus Paradise PLUS Reagent System (Thermo Fisher Scientific, Waltham, MA, USA). Dissections were performed via the Laser Capture Microdissection System (Leica LMD 6500) in accordance with the pathological diagnosis (Supplementary Fig. 1). DNA was extracted from these tissues using the GeneRead DNA FFPE Kit (Qiagen, Valencia, CA, USA), and DNA concentrations were determined using the Qubit 1.0 Fluorometer (Life Technologies, Grand Island, NY, USA). The quantity and quality of the FFPE-derived DNA samples were determined by calculating the normalized DNA integrity scores (ΔΔCq) via quantitative polymerase chain reaction (qPCR) analysis using the Agilent NGS FFPE QC Kit (Agilent Technologies, Santa Clara, CA, USA). In samples with ΔΔCq less than 1, more than 40 ng of DNA based on Qubit was used; contrarily, in samples with ΔΔCq greater than 1, more than 40 ng of DNA based on qPCR quantification was used. There were no samples that did not satisfy these criteria.

Target Enrichment and Next-generation Sequencing

DNA extracted from tumors and nontumor mucosa was fragmented into 150 to 200 bp sequences using a SureSelect XT HS and XT Low Input Enzymatic Fragmentation (Agilent Technologies) and used for library construction according to the manufacturer’s instructions. The amount of DNA was measured using a TapeStation D1000 (Agilent Technologies) before hybridization, and good quality was determined if the prepared library was greater than 1000 ng. If a sample was less than the recommended 1000 ng, at least 500 ng or more was used in the hybridization step. All samples could use more than 500 ng DNA in the next hybridization step. We selected 43 genes (Supplementary Table 2) that were previously reported to be strongly associated with Japanese CACs²³ and performed target capturing using the SureSelect XT Target Enrichment System (Agilent Technologies). Sequencing libraries were generated using the Agilent Haloplex HS Custom Kit (Agilent Technologies) following the manufacturer’s instructions. The resulting pooled libraries were analyzed for quality control using the High Sensitivity D1000 Screen Tape System and the 2200 TapeStation Instrument (Agilent Technologies). Sequencing was performed with paired-end reads via the HiSeq X platform (Illumina, San Diego, CA, USA).

Variant Detection

Sequencing reads were aligned to hg38 human reference sequence and analyzed using the SureCall Software version 4.1 (Agilent Technologies). Polymerase chain reaction duplicates were removed by molecular barcode using the SureCall Software version 4.1 to improve the mapping quality before variant calling. Paired-end and single-end analysis in the SureCall Software version 4.1 was used to identify single nucleotide variants and insertions/deletions (indels) in tumors. Called variants were considered germline mutations if they were also present in the colorectal nontumor mucosa. To reduce the false-positive rate, we set the cutoff values for somatic mutations in tumors as read depth greater than 20 and a forward/reverse balance between 0.25 and 0.75. We also configured the SureCall SNP caller using the SureSelect default settings: variant score threshold at 0.3, minimum quality for base at 30, variant call quality threshold at 100, minimum allele frequency at 0.05, and minimum number of reads supporting variant allele at 10. Variants that were (1) repeated sequences registered in UCSC’s repeat masker, (2) called as replacements, or (3) clearly identified as sequence errors in the Integrated Genomic Viewer (Broad Institute) were excluded as somatic mutation candidates in all sample types. We classified somatic mutations into 3 categories as follows: category 1, frameshift indels or nonsense mutations; category 2, missense mutations; or category 3, synonymous changes or mutations located in introns.

Immunohistochemistry of p53

Paraffin-embedded human CRC tissue was cut into 2 to 3–μm sections and mounted on positively charged slides. Antigen retrieval was conducted with Tris-EDTA buffer (pH 9.0) in a microwave oven at 800 W for 5 minutes and at 150 W for 10 minutes. The slides were then incubated with the primary antibody antihuman p53 (M3636, Dako, Tokyo, Japan) at a dilution of 1:500 for 2 hours at room temperature. The bound antibodies were detected using the EnVision system (Dako, Copenhagen, Denmark). After immunostaining, the slides were counterstained with hematoxylin. Histopathological evaluation was performed by expert pathologists including 1 or more gastrointestinal pathologists (KA, FS).

Statistical Analysis

Values have been expressed as medians with ranges or as percentages. The frequency of genomic mutations in CRNUC was compared with that of genomic mutations in the invasive CAC cases analyzed by Fujita et al.²³ Comparisons were performed using the Fisher exact test and the Wilcoxon/Kruskal–Wallis test. We also examined the concordance rate in CRNUC specimens between presenting as p53 immunohistochemistry-positive and harboring TP53 mutations. The concordance rate was calculated by the kappa value. We examined the correlation among TP53, KRAS, and other genes that were mutated in early-stage CRNUC by Pearson correlation coefficient. Pearson correlation coefficient, also referred to as Pearson’s r, is a statistic that measures linear correlation between 2 variables. It has a value between +1 and −1; a value of +1 is total positive linear correlation, 0 is no linear correlation, and −1 is total negative linear correlation. All statistical analyses were performed using the JMP software version 13 (SAS Institute Inc., Cary, NC, USA), and P values less than 0.05 were considered to indicate significant differences.

Ethical Considerations

The study was performed in accordance with the Declaration of Helsinki, and the study protocol was approved by the human ethics review committee of Hiroshima University (E-1399).

RESULTS

Clinicopathological Features of the Patients and Neoplasia

Table 1 shows the clinicopathological features of the CRN in the UC patients. We identified 32 consecutive UC patients who had 36 early-stage CRNUCs and who underwent surgery or endoscopic resection at the Hiroshima University Hospital during the study period. Metachronous neoplasias were observed in 3 of the 32 UC patients (CRNUC10 and 14, CRNUC20, 21 and 22, and CRNUC25 and 27 were the multiple lesions in patient 10, patient 19, and patient 22, respectively). The median (range) age at diagnosis was 65 (35‒85) years, and half of the patients were males. The median (range) disease duration was 17 (0–43) years. All cases were of pancolitis or left-sided colitis, and the clinical course of most patients was either chronic or relapsing-remitting. The neoplasia locations were cecum/ascending colon (n = 3), transverse colon (n = 2), descending colon (n = 4), sigmoid colon (n = 9), and rectum (n = 18). Of the 36 lesions, 34 were located within the colitic segment, and 2 were located outside the colitic segment. The macroscopic type and endoscopically determined lesion borders were distributed as follows: 24 superficial lesions and 12 polypoid lesions, and 21 lesions with clear endoscopic border and 15 lesions with unclear endoscopic border. Mucosal resection or submucosal dissection was endoscopically performed in 15 cases, and surgery was performed in 21 cases. Of the 15 endoscopic resection cases, additional surgery was performed in 1 case, in which the endoscopic resection was not curative (submucosal deep invasion). The histological findings of the neoplasias were 10 lesions with T1 cancer, 19 lesions with HGD, and 7 lesions with LGD.

Landscape of Somatic Mutations in CRNUC

We performed targeted sequencing of 43 genes that are associated with the development of CACs in Japanese patients.²³ Mutations from each cancer/dysplastic tissue were compared with those in the nontumor mucosa counterpart. The median read depth was 120 (after excluding PCR duplicates), and 99.94% of the target bases were covered by more than 10 independent readings (Supplementary Tables 3 and 4). A total of 171 somatic mutations, including 132 nonsynonymous mutations, 23 synonymous mutations, and 16 indels, were identified (Supplementary Table 5). The number of somatic mutations per sample ranged from 0 to 23 (mean: 4.75), and at least 1 somatic mutation was detected in 35 of the 36 samples. Further, somatic mutations were found in 28 of the 43 target genes. As shown in Figure 1 and Supplementary Table 5, we identified 11 genes that were frequently mutated, and each gene was mutated in more than 4 patients (at least 10% of all cases). The most frequently mutated gene in the CRNUC samples was APC (mutated in 47.2% of cases), followed by TP53 (44.4%), KRAS (27.8%), PRKDC (27.8%), and CSMD3 (22.2%).

Association Between Somatic Mutations and Tumor Histological Progression in CRNUC

We examined the 28 genes that were somatically mutated in the 36 CRNUC samples to determine their association with tumor histological progression (T1, HGD, or LGD). However, there was no statistically significant difference in mutations between T1 and dysplasia or between T1/HGD and LGD (data not shown).

Genomic Mutation Patterns of Metachronous CRNUC Arising From the Same UC Patients

Details of the genomic mutations in metachronous CRNUC arising from the same UC patients are shown in Supplementary Table 5. We compared each of the mutation patterns. When somatic mutations in each metachronous lesion in the same UC patient were compared, the positions of mutations occurring in genes involved in carcinogenesis, such as TP53, KRAS, BRAF, and APC, were found to be different between each metachronous lesion. This result suggested that even CRNUC originating from the same patient has a multicentric pattern of carcinogenesis.

Characteristics of Somatic Mutations in CRNUC Compared With Those in Invasive CAC

The mutation rates of each of the 43 genes were compared with the Japanese CAC sequence data from a previous report²³ (Supplementary Fig. 2). Approximately 80% of the CACs in this data set consisted of invasive cancers in stages 2, 3, and 4. Although it is known that TP53 mutations occur in the early stages of CAC carcinogenesis,²⁴ the mutation frequency in our CRNUC samples was less than that in invasive CAC (44% vs 79%, P < 0.01). Because TP53 mutations are known to occur later in the progress of sporadic neoplasia carcinogenesis, this result suggests that CRNUC cases examined in this study include many early-stage sporadic neoplasias. Alternatively, it is also possible that the invasive CAC cohort includes sporadic cancers at advanced stages because these usually have mutations in TP53. In contrast, somatic mutations in APC, PRKDC, CSMD3, and CREBBP were significantly more frequent in CRNUC than those in invasive CACs.

Differences in TP53 Mutations Between CRNUC, Invasive CAC, and Cancer Genome Atlas–Colorectal Cancer

Of the 17 TP53 mutations observed in the CRNUC cohort, 14 were missense mutations, and 3 were nonsense mutations (Supplementary Table 5). None of the TP53 mutations were located at any of the DNA-binding domain hotspot codons such as 173, 175, 248, 273, and 282 found in the Cancer Genome Atlas-Colorectal Cancer (TCGA-CRC)²⁵ or at sites reported in invasive CAC cohorts. Moreover, in contrast with the distribution of mutations in the TP53 gene in TCGA-CRC, which were located diffusely across the entire gene, almost all TP53 mutations (15 of 17) observed in CRNUC were found to concentrate in the amino terminal part of the DNA-binding domain (amino acids 100‒200, Fig. 2).

Immunohistochemistry of p53 and Ki-67

To determine whether the mutational status was consistent with the level of protein, we examined the level of p53 protein by immunohistochemistry. In 1 case, the amount of available tissue sample was not sufficient for analysis, so the p53 immunostaining status was evaluated in only 35 CRNUC cases (Fig. 3, Supplementary Fig. 3). Of the 15 CRNUC samples with TP53 mutations, 12 were positive or weakly positive, and 3 were negative for p53. In contrast, out of 20 CRNUC without TP53 mutations, 18 were negative, and 2 were weakly positive. The concordance rate between p53 immunochemistry and TP53 mutations was high (kappa value = 0.706). Moreover, the 3 cases that had TP53 nonsense mutations showed complete loss of p53 in immunohistochemistry analysis, which may be considered to be due to the loss of p53 function. These results suggest a correlation between p53 protein expression and TP53 mutation status.

We evaluated the relationship between Ki-67 immunostaining patterns and TP53 mutations by performing Ki-67 immunostaining, one of the most popular methods for differentiating colitic neoplasia associated with inflammatory carcinogenesis and sporadic neoplasia (Fig. 3, Supplementary Fig. 4). The same one case (CRNUC22) of Ki-67 immunostaining could not be evaluated because of insufficient amount of tissue. Six of the 10 T1 cancers had a diffuse pattern, making it difficult to classify bottom-up and top-down patterns as the neoplasia progressed to invasive cancer. The bottom-up pattern with TP53 mutation was observed in 8 cases and the top-down pattern in 2 cases; however, the bottom-up pattern without TP53 mutation was observed in 4 cases and the top-down pattern in 11 cases. Colorectal neoplasias developing from the UC mucosa with TP53 mutation tend to have a bottom-up pattern, suggesting an association between TP53 mutations and colitic neoplasias to a certain extent.

Comparison of Somatic Mutation Frequency in Each Gene With and Without TP53 Mutation

We investigated the somatic mutation frequency of 42 genes in cases with and without TP53 mutations (Fig. 4). Only KRAS showed a significant difference in the somatic mutation frequency; these mutations were observed in 1 out of 16 cases with TP53 mutations and in 9 of 20 cases without TP53 mutations (P < 0.05). This result suggests that KRAS and TP53 mutations are mutually exclusive in CRNUC. We examined the correlation among TP53, KRAS, and 9 genes that were mutated in at least 10% of all cases by Pearson correlation coefficient. As a result of this examination, a strong negative correlation between TP53 and KRAS was observed (r = −0.43, P = 0.0089); however, there was no correlation between TP53 or KRAS and other genes. Moreover, in 26 cases with dysplasia, KRAS mutations were observed in 9 out of 17 cases without TP53 mutations, but not in 9 cases with TP53 mutations. All CRNUC cases with KRAS mutations had ducts with histologically serrated lesions (Fig. 5). Three of the 15 CRNUC cases with TP53 mutations but no KRAS mutations had serrated lesions, and 2 of the 11 CRNUC cases with wild-type TP53 and KRAS had serrated changes. Of these 5 CRNUC with no KRAS mutations but with serrated changes, the BRAF V600E somatic mutation was found in 2. Relationships with clinicopathological features are shown in Figure 6. Mayo endoscopic subscore (MES) was significantly higher in cases with TP53 mutations and significantly lower in those with KRAS mutations. Mayo endoscopic subscore is the value assessed at the time of the preoperative thorough examination by colonoscopy, and 24 patients underwent surgery within 1 month of the thorough examination. Eight other patients had undergone surgery within 3 months of the thorough examination, and 4 patients had undergone surgery more than 3 months later. There were no cases in which the treatment of UC was changed or intensified between thorough examination and surgery. Of the 24 cases in which colonoscopy had been performed at our hospital more than 1 year before the preoperative thorough examination, the MES value did not change for 22 cases and improved from 1 to 0 for 2 cases. The proportion of rectal/sigmoid colonic lesions was significantly higher in cases with TP53 mutations and significantly lower in those with KRAS mutations. We also examined the association of TP53 and KRAS mutations with endoscopic image findings including the macroscopic type, the lesion border, the presence or absence of visible dysplasia, and the features of serrated lesions such as traditional serrated adenoma and sessile serrated adenoma/polyp—but no significant associations were found.

DISCUSSION

In this study, using genomic analyses, we were able to differentiate between colitic and sporadic neoplasia originating in the UC mucosa. The common CRC carcinogenesis model (ie, adenoma-carcinoma sequence)²⁶ proposes that p53 anomalies occur between the late adenoma phase and the carcinoma phase, whereas the colitic cancer carcinogenesis model (ie, dysplasia-carcinoma sequence) proposes that such anomalies occur during the development of dysplasia.²⁷ We therefore focused on TP53 mutations because their presence varies by the carcinogenesis pathway.²⁴

Treatment for CRN in UC patients varies according to the tumor type. Polypoid dysplasia can be adequately treated by endoscopic resection, provided that the lesion can be completely excised and no other evidence of nonpolypoid or invisible dysplasia exists elsewhere in the colon. However, for superficial neoplasia, colectomy is generally recommended because the prognosis after endoscopic resection has not been fully investigated. This may be due to the difficulty in distinguishing between “true” colitic neoplasia and sporadic neoplasia. In fact, previous studies have reported that CAC includes sporadic neoplasia.^{8, 9, 23, 28} As a result, occasionally colectomy is performed for sporadic neoplasia, which is an excessive treatment for this tumor type, and endoscopic resection and surveillance colonoscopy are performed for colitic neoplasia, which can be considered as an undertreatment for such cases. Distinguishing between these 2 neoplasias by collecting tissue from patients with CRN, as performed in this study, can prevent unnecessary colectomy and promote the accurate choice of endoscopic resection or at least partial colectomy and surveillance colonoscopy where proper, resulting in an increase in the patient’s quality of life. Consequently, it is clinically advantageous to classify neoplasias that occur in UC mucosa as colitic or sporadic, but no reliable way to do so has been devised to date.^18–20 In endoscopic diagnosis, the presence of an apparently flat mucosa,^{29, 30} indicating dysplasia, around a dysplastic lesion is considered to be useful for the distinction between the 2 neoplasia types, but the definition remains obscure. Because absolute markers for distinction such as additional endoscopic observations have not been established, it has been challenging to confidently distinguish between colitic and sporadic neoplasias in some cases.^{28, 31}

Immunohistochemistry of p53 has been widely used as one of the classification methods in practice for these tumors. In general, the p53 that can be identified by immunostaining is mutated. Accordingly, overexpression of p53 protein can be used as an indication of p53 gene abnormality. Although immunohistochemistry of p53 is a valuable diagnostic tool in CACs, its utility in the diagnosis of inflammatory carcinogenesis has not been completely established.³² Therefore, we examined the relationship between the presence or absence of TP53 mutations and the status of p53 by immunohistochemistry, and as a result, we found a correlation. Moreover, all cases containing TP53 nonsense mutations were completely negative for p53 in immunohistochemistry as a result of the loss of p53 function. This demonstrates that there are a few early-stage CRNUC cases wherein the presence or absence of TP53 mutation cannot be predicted by immunohistochemical analysis alone.

Next, by analyzing the mutations sites, we found that the distribution of TP53 mutations in CRNUC was different from that in TCGA-CRC, in which the mutations were found throughout the TP53 gene as with other malignancies.²⁵ Similar to that in invasive CAC, most of the TP53 mutations in CRNUC cases were located in the DNA-binding domain of the protein.^{8, 23} However, unlike TCGA-CRC and invasive CAC, almost all TP53 mutations observed in CRNUC were found to be concentrated in the amino terminal part of the DNA-binding domain, and none occurred in any of the hotspot codons. Approximately 90% of TP53 mutations in human cancers encode missense mutant proteins that span ~190 different codons localized in the DNA-binding domain of the gene and protein.³³ These mutations produce a protein with a reduced capacity to bind to a specific DNA sequence that regulates the p53 transcriptional pathway. A specific few of these mutations, so-called hotspot mutations, are localized in codons that account for ~28% of the total p53 mutations; these alleles seem to be selected for preferentially human cancers of many tissue types. The hotspot mutant alleles produce a protein that has a highly altered structure. The hotspot mutations arise at selected sites in the gene because of a specific DNA sequence, such as a methylated cytosine residue in a CpG dinucleotide, which has a higher mutation rate changing C to T nucleotides. Along with the observed change in mutant p53 proteins, which produce a loss of function, some mutant proteins have an allele-specific gain of function that promotes cancer. Therefore, hotspot mutations in the TP53 gene contribute significantly to the development of human cancer, regardless of organ type, and may play a critical role in carcinogenesis. Thus, the TP53 mutations, which were not found in any of the hotspots, were considered to be different from those in common cancer development. Contrarily, TP53 mutations in early-stage CRNUC may have been caused by tissue-specific changes, supporting the involvement of UC-induced mucosal inflammation in the carcinogenesis of CRNUC with TP53 mutations. In future, function analysis of TP53 mutations in CRNUC that were found to be concentrated in the amino terminal part of the DNA-binding domain is necessary.

An alternative explanation might be that the colorectal mucosa of CRNUC with TP53 mutations had severe inflammation because MES was significantly higher in the TP53-mutated cases. This result suggests that the majority of the cases with mutations in TP53 may be CRNUC-derived from the inflammatory carcinogenic pathway (ie, inflammation-dysplasia-carcinoma sequence). Therefore, examination of the genomic landscape of sporadic and colitic neoplasias based on the presence or absence of TP53 mutations revealed that mutations in KRAS and TP53 were mutually exclusive. Moreover, the exclusionary relationship was more pronounced in cases of dysplasia, which is an earlier stage of carcinogenesis. These outcomes suggest that there may be 2 distinct pathways arising from either TP53 or KRAS mutations in the early stages of CRNUC development. Actually, MES in cases with KRAS mutations were significantly lower than those in cases without KRAS mutations. These data, in contrast to the data on those with mutated TP53, support the notion that CRNUC cases with KRAS mutations have less inflammation. Moreover, because many CRNUCs with KRAS mutations were located in right-sided colon lesions, the presence or absence of histological serrated lesions in CRNUC was examined to investigate the correlation between serrated lesions and CRNUC with KRAS mutations. As a result, all cases with KRAS mutations were histologically accompanied by serrated changes, suggesting a strong correlation between KRAS mutations and serrated lesions in CRNUC. Additionally, all CRNUC with the BRAF V600E mutation had serrated lesions. In fact, it has been reported that serrated colorectal lesions with KRAS and BRAF mutations occur in UC patients at a constant rate.^34–37 Therefore, early-stage CRNUC can be classified into a group with true colitic neoplasias with TP53 mutations that develop via inflammatory carcinogenesis,and a group with sporadic neoplasias with KRAS/BRAF (V600E) mutations in which inflammation is not involved in tumorigenesis.

In the inflammation-dysplasia-carcinoma sequence, the carcinogenic pathway of colitic neoplasias, TP53 mutations precede premalignant UC mucosa in the early phase, followed by microsatellite instability and methylation, and then KRAS and APC mutations in the late phase, which progress to HGD and cancer.^{24, 32, 38} In the adenoma-carcinoma sequence, the most major pathway of sporadic neoplasias, APC mutations precede early onset, followed by microsatellite instability, methylation and KRAS mutations, and TP53 mutations in the late phase, which progress to invasive cancer.^{23, 26} In the serrated pathway, the carcinogenic pathway of serrated lesions in sporadic neoplasias, KRAS or BRAF mutations first precede in the early phase, followed by methylation, including CpG island methylation phenotype, and TP53 mutations in the late phase, leading to the progression to traditional serrated adenoma, sessile serrated adenoma/polyp, and then cancer.^{39, 40} Differences in the timing of several genomic mutations due to these carcinogenic pathways may make it possible to classify CRNUC. That is, by examining genomic mutations in CRNUC at the time of dysplasia—the especially early stage of carcinogenesis—it is possible to classify them into 2 groups: the inflammatory carcinogenic pathway originated by TP53 mutations and the noninflammatory carcinogenic pathway originated by KRAS or BRAF mutations. In fact, in the present study, we believe that both TP53 and KRAS mutations were exclusively related to each other and that all KRAS-mutated cases had serrated changes, supporting the aforementioned considerations. In addition, improvements in endoscopy and the surveillance skills of endoscopists could lead to earlier detection of CRNUC, which could contribute to an increase in the number of cases in which our data are applied.

In our study, the APC mutation frequency was lower in CRNUC than that in sporadic CRC as previously reported^{8, 9, 23} but was significantly higher than that in invasive CAC in contrast to previous reports.^{8, 9, 23}APC is a key gene in the adenoma-carcinoma sequence and is mutated at an early stage in sporadic CRC.²⁶ Periodic surveillance colonoscopy is recommended for longstanding UC patients due to high risk of developing CRC.^{3, 4} Sporadic cancers have several well-circumscribed and polypoid lesions, so they are likely to be detected before progressing to advanced-stage cancer. However, colitic cancers have poorly circumscribed and superficial lesions, so it is highly likely that they are already advanced cancers at the time of detection.⁴¹ Given this background, it is possible that our early-stage CRNUC cohort contained more sporadic cancers with APC mutations than those in the invasive CAC cohort. However, unlike TP53 and KRAS mutations, we could not find an association between carcinogenic pathways and APC mutations. There have been several reports of TP53 mutations and positive p53 immunohistochemistry in nondysplastic areas of the colonic mucosa of longstanding UC patients, and the cohort studied here also included several cases in which APC and TP53 mutations overlapped.^{42, 43} In other words, it is possible that some CRNUCs may follow a carcinogenic pathway via APC mutations despite the TP53-mutated mucosa.

This study had some limitations. First, it was a retrospective, single-center study. Therefore, the number of cases that could be used was limited, and the scale of this study was relatively small. In our study, we performed genetic analysis of early-stage CRNUC, but we did not have data on early-stage sporadic CRCs for comparison. The invasive CAC cohort used for another comparison was also cited from a previous report.²³ Additionally, all patients in our CRNUC cohort were UC patients, but previously reported samples included patients with Crohn’s disease. This might have caused a slight difference in the results of our analysis. In light of these considerations, the data from our current study may be insufficient to recommend its application to clinical practice. Therefore, larger scale prospective studies are required for clinical applications. In practice, it is necessary to perform genomic analysis of preoperative biopsy specimens using a small targeted panel, including TP53 and KRAS, to classify them as colitic and sporadic neoplasia, determine a treatment strategy based on this analysis, and evaluate their long-term prognosis.

In conclusion, our gene analysis suggests that early-stage CRNUC can be classified into 2 groups: those with TP53 mutations that develop through the inflammatory carcinogenesis pathway and those with KRAS mutations that develop through the serrated pathway as the noninflammatory carcinogenesis pathway. Histological evaluation of CRNUC, including p53 immunohistochemistry and serrated lesions, may also help predict the presence of TP53 and KRAS mutations, except in cases of CRNUC with TP53 nonsense and BRAF mutations. This suggests that TP53 and RAS/RAF mutations in tissues obtained by an endoscopic biopsy or resection may be useful biomarkers for the diagnosis of early-stage CRNUC.

Conflicts of interest: The authors have no conflicts of interest to declare.

Author Contribution: YU and SO contributed to the methodology of the study. KM, YU, and RY contributed to the formal analysis. KM and SO contributed to study investigation. RY, RH, SO, KA, FS, and ST helped with resources for the study. KM curated data and helped write, review, and edit the article. KM and YU contributed to writing the original draft of the article. YU supervised the study. ST and KC coordinated project administration. YU, ST and KC acquired funding for the study. This work supported by Management Expenses Grants from ST. This work was supported by the Program of the network-type Joint Usaga/Research Center for Radiation Disaster Medical Science of Hiroshima University, Nagasaki University, and Fukushima Medical University. All authors have read and approved this version of the manuscript.

REFERENCES