The study population consisted of four groups of patients identified through a retrospective review of pathology and gastroenterology databases at Cleveland Clinic and Brigham and Women’s Hospital between 2016 and 2023. The timeline was chosen to align with the updated Surveillance of Colorectal Endoscopic Neoplasia Detection and Management in IBD: International Consensus Recommendation (SCENIC) endoscopic classification of dysplasia introduced in endoscopy reports at the study institutions.7 Following approval by institutional review boards, consecutive cases where pathology material was available for analysis were reviewed to identify low-grade dysplastic polyps with conventional intestinal-type dysplasia arising either within or outside areas of known colitis as well as foci of invisible low-grade dysplasia. On joint consideration of morphological and molecular features (see details on ‘Sequencing analysis’ section), the following inclusion criteria were employed to select the study cohorts: (1) neoplastic cell content of at least 20%, (2) a yield of at least 50 ng of DNA from Formalin-fixed paraffin embedded (FFPE) sections and (3) mean target coverage of more than 100 times. Thus, from a series of 60 visible dysplastic polyps and 16 cases of invisible dysplasia, 35 polyps and 10 invisible dysplastic foci (all with conventional intestinal-type dysplasia) fulfilled these criteria and were chosen for further analysis (figure 1). None of the patients with IBD with visible dysplasia had synchronous invisible (flat) dysplasia detected along with their visible polypoid dysplasia.

Flow chart showing patient selection and inclusion criteria. IBD, inflammatory bowel disease. FFPE: Formalin fixed paraffin embedded

Cohort 1 consisted of 19 patients (22 polyps) with clinically, endoscopically and pathologically confirmed ulcerative colitis (UC) (n=16) or Crohn’s disease (CD; n=3), all of whom had at least one polyp located within areas of pathologically documented chronic active or inactive colitis, and no synchronous areas of invisible (flat) dysplasia or adenocarcinoma. In 11/19 patients, the mucosa adjacent to the visible dysplastic lesion showed chronic active colitis. All three patients with CD were confirmed to have colonic CD affecting >50% of their colon. While a recent study by Yaeger et al 5 has shown that there may be differences in molecular profiles of CRCs related to UC vs CD, these data are still emerging, and have not yet been validated in larger cohorts. Therefore, we chose to include patients with colonic CD in our study. Cohort 2 consisted of 9 UC patients with 13 polyps located outside (proximal to) areas of known colitis. Thus, all these nine patients had lesions isolated from the right side of the colon in a setting of left-sided UC (n=7) or UC proctitis (n=2). By definition, none of these nine patients had endoscopically or histologically documented IBD in the right colon prior to any medical therapy. All polyps in both cohorts consisted of low-grade dysplastic epithelium of the intestinal, tubular phenotype (tubular adenoma-like)17 that was histologically indistinguishable from a sporadic adenoma. Cohort 3 consisted of six sporadic tubular adenomas from five patients without history of polyposis or IBD. Lastly, cohort 4 consisted of 10 foci of low-grade invisible dysplasia of the intestinal, tubular subtype, from 5 UC patients. By definition, all the lesions analysed in cohort 1 were visible, polypoid dysplasia that included both pedunculated and sessile lesions based on SCENIC classification. They did not include any lesions that were endoscopically described as ‘thickened folds’, ‘depressions’ or ‘flat lesions’ that are forms of non-polypoid dysplasia. Similarly, cohort 4 consisted of ‘invisible’ dysplasia identified on random (non-targeted) samples of colonic mucosa without a visible lesion. For all groups, medical records including endoscopy reports were reviewed to determine demographic features, duration and extent of colitis, size, number, appearance and location of dysplasia. Of the 33 patients with IBD (cohorts 1, 2 and 4), none of the 28 patients in whom medical history and/or imaging was available for review, had a history of primary sclerosing cholangitis. All cases were re-reviewed by a gastrointestinal (GI) pathologist (DTP) and GI Pathology fellow (AC) to confirm the diagnosis and histological grade of dysplasia.

At the time of surveillance colonoscopy and per SCENIC guidelines, visible polypoid lesions were removed in total using standard techniques with snare or biopsy forceps7 and submitted for histological evaluation. H&E-stained slides were reviewed to confirm the degree of dysplasia (all low grade) according to previously published criteria18 and select tissue for DNA isolation. Cases were selected to maximise lesional DNA content (dysplastic epithelium >20% of overall cellularity) (figure 2).

The figure shows the process of selecting lesional tissue for macrodissection (area dotted in black interrupted lines) and DNA isolation from formalin-fixed paraffin-embedded sections. Cases were selected to have a minimum of 20% dysplastic cell content. This sample represents an example of a left-sided visible polypoid low-grade dysplasia of the conventional intestinal type (tubular adenoma-like) from a patient with history of UC for 25 years. The background mucosa (fragments not used for DNA isolation) shows crypt architectural distortion along with expansion of the lamina propria by lymphocytes and plasma cells, consistent with a chronic inactive colitis. UC, ulcerative colitis.

DNA was extracted from macrodissected regions of FFPE tissue using H&E-stained sections as a guide. Next-generation sequencing was performed using OncoPanel, a 447 gene hybrid-capture-based sequencing platform.19 20 Each specimen was analysed for mutations, small insertions/deletions, copy number changes, structural variants (including translocations), tumour mutational burden (TMB) and mismatch repair status. Mutations and small insertions/deletions were filtered to remove technical artefacts and synonymous variants present at >0.1% frequency in the Genome Aggregation Database (https://gnomad.broadinstitute.org/). All remaining alterations were assessed by a molecular pathologist (JN) for pathogenicity.

For the assessment of pathogenicity, each molecular alteration was considered in the context of the known biology of the given gene. Briefly, only known, activating mutations in oncogenes were considered pathogenic while any type of loss of function alteration in a tumour suppressor gene was considered pathogenic. For comparison with previously sequenced invasive or metastatic colorectal adenocarcinomas (regardless of their underlying biology sporadic, GI polyposis syndrome, Lynch syndrome, IBD) using the Dana-Farber Cancer Institute (DFCI) internal cBioPortal repository, mutations, structural variants and copy number changes of unknown significance were excluded from analysis. For copy number analysis, arm-level and chromosome-level alterations were each counted as individual events.

Categorical variables were described as frequency counts and percentages and compared with χ2 test or Fisher’s exact test; missing responses were noted but not used to calculate percentages. Continuous variables were described as median and range and compared with Student’s t-test and Mann-Whitney U test, as applicable. All statistical tests were two sided, and a p<0.05 was used to indicate statistical significance.

Numerous studies have confirmed that a decline in APC gene function occurs early in sporadic CRCs while this is a delayed event in IBD-associated colorectal carcinogenesis.24 In contrast, TP53 mutations are observed as early events in IBD-related CRCs and late events in sporadic CRCs. These data are mostly derived from tissue samples from patients with IBD with adenocarcinoma, lesions classified as ‘flat dysplasia’ and/or ‘dysplasia-associated lesions or masses (DALM)’, without information about the grade, size or location of the analysed lesions.9–11 25 Therefore, it is difficult to discern whether there are any differences in the molecular evolution of visible, polypoid, dysplastic lesions (formerly known as ‘adenoma-like DALM’) that arise within a segment of bowel affected by colitis versus those arising outside area of colitis, and if these alterations differ from invisible or ‘flat’ dysplasia. Furthermore, the existing molecular data are somewhat limited in that these studies: (1) only include high-grade dysplastic lesions or lack information about histological grade, (2) combine data on both flat/invisible dysplasia as well as polypoid lesions and ‘DALM’ and/or (3) do not always specify whether the lesions arose within or outside segments of bowel affected by IBD.

Using a targeted next-generation sequencing panel of 447 genes, we analysed 35 low-grade dysplastic polyps, all of the conventional, intestinal (tubular) phenotype, 22 of which arose in segments of colon affected by colitis, and compared them with 13 polyps that developed in regions of colon that were never affected by colitis, and 10 cases of invisible dysplasia. In addition, six sporadic adenomas (all low grade) from non-IBD patients were analysed. In general, polyps arising in areas of colitis showed a greater spectrum of mutations, including APC, KRAS, FBXW7, TP53, ARID1A and TCF7L2. In contrast, polyps outside areas of colitis as well as non-IBD sporadic adenomas showed a fairly limited mutational profile, with loss of function APC and additional CTNNB1 mutations. Invisible dysplasia also showed a limited mutational profile with alterations restricted to TP53, CTNNB1 and KRAS genes. APC mutations were the most frequent mutations in visible dysplastic polyps, with the highest frequency occurring in polyps outside colitis (92%), followed by sporadic adenoma in non-IBD patients (83%) and polyps arising within colitis (73%). This was significantly different from invisible dysplasia where no APC mutations were detected (p=0.0001). Instead, 50% of cases of invisible dysplasia harboured TP53 mutation, and this was significantly different from polyps within (p=0.02) and outside colitis (p=0.03). ARID1A, FBXW7, SMARCA4 and RNF43 mutations were exclusively seen in polyps within colitis. Based on these results, we conclude that there is a significant overlap in molecular alterations between visible low-grade dysplastic polyps with conventional intestinal-type dysplasia in patients with IBD and sporadic adenomas in non-IBD patients. Similar to the sporadic setting, APC alterations appear to play a major role in the development of visible low-grade dysplastic lesions in patients with IBD, regardless of whether or not the lesion arises in a segment of colon affected by colitis. These findings are indeed reassuring, and help in further confirming the conservative approach, namely polypectomy, rather than colectomy, that is employed for treating visible, low-grade dysplastic polyps encountered in patients with IBD.26 27 Our findings also indicate that similar to IBD-associated CRCs, TP53 mutations are an early event in the development of invisible low-grade dysplastic lesions compared with low-grade dysplastic polyps in patients with IBD, and patients with invisible low-grade conventional intestinal-type IBD-related dysplasia likely need more aggressive surveillance compared with visible, endoscopically resectable polyps. Although the finding of ARID1A, FBXW7 and RNF43 mutations in polyps arising within colitis suggests that an inflammatory milieu may play a role in pathogenesis of these lesions, further studies using larger numbers of patients will be needed to confirm this hypothesis.

Among genes involved in the WNT-signalling pathway, we found that the frequency of APC mutations in polyps from patients with IBD is much higher (92% in polyps outside colitis and 73% in polyps within colitis) than previously reported rates of 14%–38%14 28 29 These differences are probably related to variations in the type and sensitivity of the assays used to detect APC alterations, and the characteristics of the lesions analysed in prior studies. For instance, in previous studies, APC alterations were detected either by using an in vitro synthesised protein assay28 or by analysing loss of heterozygosity (LOH) of APC (D5S346).29 Similar to our study, a prior study by Odze et al also found that APC alterations were more common in lesions arising in segments of colon never affected by colitis (38%) compared with those arising within colitis (29%). However, it is difficult to compare these two studies since 30% of lesions within colitis and 45% of lesions outside colitis analysed in the Odze study were categorised as high-grade dysplasia, and thus included data from more advanced lesions. A recent study by Wanders et al interrogated the molecular profile of 16 IBD-associated dysplastic lesions (13 low grade, 3 high grade) and 118 sporadic adenomas (all low grade) from non-IBD patients (polypoid and non-polypoid) using a 48-gene NGS assay. The frequency of APC mutations in sporadic adenomas from non-IBD patients was 54% (45/83), but the frequency in low-grade dysplastic lesions from patients with IBD was quite low at 23% (3/13), compared with our study. What is notable is that while all the dysplastic lesions from patients with IBD were confirmed to be colitis-associated dysplastic lesions by endoscopists with an expertise in IBD, their endoscopic characteristics (visible vs invisible) were not documented. Since their cohort consisted of samples obtained from patients who had undergone colectomy for suspected IBD-related dysplasia, the data likely included mutational profile of lesions that were grossly invisible (‘flat dysplasia’). In our study, we were able to compare APC alterations between visible and invisible dysplasia, and found that none of the 10 foci of invisible dysplasia showed APC alterations. This is in agreement with prior studies which either documented lack of APC mutations30 or absence of LOH at APC locus31 in invisible low-grade dysplasia. All of these observations suggest that similar to the sporadic setting, APC alterations play a significant role at an early stage in the development of visible dysplastic polyps with conventional intestinal-type dysplasia in patients with IBD, regardless of the presence or absence of background colitis.

Prior studies have shown that LOH for TP53 and TP53 mutations occur early in IBD-associated neoplasia and range from 23% to 63% in dysplastic lesions.10 11 14 32 In our study, the frequency of TP53 mutations in visible low-grade dysplastic polyps located within and outside areas of colitis was quite similar—9% and 8%, respectively. However, this rate is much lower than what has been documented in the literature. We believe this discordance is due to the differences in the type of lesions analysed. For instance, previous studies included ‘DALMs’ and/or invisible dysplasia (‘flat dysplasia’) with both low-grade and high-grade cytomorphology found in random specimens collected from patients undergoing colectomy for neoplasia10 11 14 32 In a previous study by Fogt et al, LOH in TP53 was more frequently detected in ‘flat dysplasia’ (6 of 14 informative cases, including 4 high grade and 2 LGD) compared with polypoid dysplasia (2 of 12 informative cases). The authors also found that mucosa with chronic colitis surrounding ‘flat dysplasia’ had LOH of TP53 in 1/7 (14%; including 1 LGD) cases compared with 0/12 cases of polypoid dysplasia.33 Cumulatively, these results indicate that TP53 alterations occur much earlier in invisible dysplasia compared with visible, polypoid dysplasia. The finding of TP53 mutations in a significantly higher proportion (50%) of invisible dysplastic foci compared with visible dysplasia (9% in polyps within areas of colitis, 8% in polyps outside colitis) in our study provides additional evidence for differences in the genetic evolution of visible and invisible low-grade dysplastic lesions with conventional intestinal-type dysplasia in patients with IBD.

Some other genes that were unique to low-grade dysplastic polyps arising within a colitic background were FBXW7 (23%) and ARID1A (18%). These observations are similar to the Wanders study where FBXW7 mutations were found more commonly in IBD-associated low-grade dysplastic lesions (15% (2/13)) compared with sporadic adenomas (1% (1/72)); although it remains unclear whether the lesions from patients with IBD analysed in their study were endoscopically visible or invisible.14 The frequency of FBXW7 mutations in our invisible dysplasia cohort was 10%, a finding that has not yet been reported in invisible conventional intestinal-type dysplasia in patients with IBD. Our study is also the first to report ARID1A mutations in visible dysplastic polyps; all arising in a segment of colon affected by colitis. In fact, in three recent studies which studied differences in individual non-dysplastic colonic crypts isolated from patients with IBD and without IBD using whole genome sequencing, both ARID1A and FBXW7 showed evidence of positive selection within non-dysplastic mucosa.34–36 In our study, RNF43 mutation was observed in 5% of polyps arising within colitis. Somatic mutations in RNF43 have been reported in 6%–11% of IBD-related CRC.5 37 A previous study also found that RNF43 mutations were significantly associated with a longer duration of disease in patients with IBD.37 It is reasonable to speculate that an inflammatory milieu likely plays a role in accruing additional mutational events beyond APC alterations observed in visible dysplastic polyps with conventional intestinal-type dysplasia. Additional studies using larger cohorts are needed to fully elucidate the role of these aforementioned genes in IBD-related neoplastic progression.

More recently, Dhall et al evaluated tumour mutation burden and mutation profile of 24 low-grade dysplastic lesions from patients with IBD (group 1: 9 lesions arising in a background of normal mucosa, group 2: 8 lesions arising in a background of chronic active colitis and group 3: 7 lesions associated with synchronous cancers).38 By interrogating 409 genes, the authors found APC mutations were most frequent in group 1 (33%) while TP53 mutations were only observed in groups 2 and 3. Although the overall rate of APC mutations is lower than what is reported in our study, in a separate subanalysis of lesions arising in segments of colon unaffected by prior colitis, the frequency of APC mutations was found to be much higher (75%), and closer to the frequency noted in our study (92% for polyps arising outside colitis). Some of the differences in our observations can also be attributed to the morphological subtype of dysplasia analysed in these studies. We restricted our analysis to conventional intestinal low-grade dysplasia (tubular adenoma-like)17 while the Dhall study included non-conventional morphological subtypes of dysplasia such as hypermucinous, crypt cell dysplasia, basal crypt atypia and patients with multifocal low-grade dysplasia (especially in cohorts 2 and 3), all of which are more likely to be associated with invisible dysplasia and/or advanced neoplasia, and thus harbour a different risk of neoplastic progression.39

In order to better understand the differences between genomic alterations found in low-grade conventional intestinal-type dysplasia (visible and invisible) and IBD-related CRCs, we performed a comparative meta-analysis of genomic alterations found in colitis-associated CRC (CA-CRC) from four prior studies that employed whole exome sequencing or targeted NGS technology (table 5). TP53 mutations were significantly higher in CA-CRC compared with dysplastic polyps within colitis, outside colitis and sporadic adenomas from non-IBD patients (p<0.0001). This difference achieved borderline significance when TP53 mutation frequencies were compared between CA-CRCs and invisible dysplasia (p=0.05). As expected, APC mutations were significantly more common in dysplastic polyps compared with CA-CRCs (cohort 1 vs CA-CRCs: p=0.0001; cohort 2 vs CA-CRCs: p<0.0001; cohort 3 vs CA-CRCs: p=0.01). There was no significant difference in APC alterations between CA-CRCs and invisible dysplasia. These results further support the notion that similar to the sporadic setting, APC alterations likely play a major role in the early development of visible low-grade dysplastic polyps with conventional intestinal-type dysplasia in patients with IBD while TP53 alterations are acquired relatively later during neoplastic progression. The roles of APC and TP53 appear to be reversed during the development of invisible dysplastic foci with conventional intestinal-type dysplasia in patients with IBD, and their mutation profile is more similar to CA-CRCs.

Comparison of genomic alteration frequency in CA-CRC in four prior studies and low-grade dysplastic lesions analysed in our study

The major strength of our study includes analysis of clinically and pathologically well-characterised cohorts of visible polypoid low-grade dysplasia and invisible low-grade dysplasia of the conventional intestinal type (tubular adenoma-like). This cohort is also unique in that it compares molecular alterations of visible and invisible low-grade dysplastic lesions using advanced sequencing technology, allowing us to compare our data with prior studies that have used this technology to study differences in IBD-associated CRCs and sporadic CRCs. This study is not without limitations. First, the low number of cases likely contribute to type II statistical error and the lack of significant differences in the comparative analyses. However, as this study was restricted to low-grade dysplasia, a significant proportion of specimens (n=43) did not meet our strict inclusion criteria of isolating at least 50 ng of DNA, exhibiting mean target coverage of more than 100 times or demonstrating 20% neoplastic cell content on joint consideration of morphological and molecular features. While most prior studies in this field have only used DNA quantity as the inclusion criterion, we chose to include mean target coverage to improve the quality of our analysis. While trying to abide by these criteria, we acknowledge that the selection process led to a significant attrition in our original cohort size. For instance, although sporadic adenomas were readily available for analysis, when we attempted to increase the size of the cohort by sequencing an additional 10 sporadic adenomas, the amount and quality of DNA were found to be insufficient in 9 cases. Interestingly, despite the limited size of cohort 3, our data are in agreement with two prior studies by Wanders et al and van Lanschot et al which sequenced 100 and 27 sporadic adenomas, respectively; although with less stringent inclusion criteria and a limited panel of genes.14 40 Briefly, Wanders et al sequenced 100 sporadic polyps (LGD) and found the following mutation frequencies: APC—54% (n=83), KRAS—42% (n=66), BRAF—1%, CTNNB1—1%, FBXW7—1%, MET—2% and NRAS—2%.14 In the van Lanschot study which sequenced 27 tubular adenomas, the mutation frequencies were as follows: APC—30%, ATM—4%, CTNNB1—4%, NOTCH1—4%, ERBB2—4%, KIT—4% and SOM—4%.40 Second, we performed targeted sequencing, only covering small parts of the genome harbouring the most frequent cancer-associated mutations, thereby potentially missing alterations in other regions, copy number variations or epigenetic changes. Lastly, we had three patients with duration of disease less than 8 years (6 years, 4 years and 3 years). While duration of disease is a significant risk factor for the development of neoplasia in patients with IBD, all three patients had severe pancolitis and continued to have episodes of UC flare on follow-up. As extent of disease is also a known risk factor for neoplastic progression, we decided to include them in our study.

In summary, using advanced sequencing techniques, our study confirms that there is a significant overlap in molecular alterations between visible low-grade dysplastic polyps from patients with IBD and sporadic adenomas from non-IBD patients. Similar to sporadic setting, the high frequency of APC mutations in visible polyps with low-grade conventional intestinal-type dysplasia suggests that these alterations play an important role during the early phase of their development. This finding further justifies conservative management in the form of complete polypectomy for visible, dysplastic low-grade polyps. Additionally, our study validates the role of TP53 in the pathogenesis of early, low-grade, conventional intestinal-type invisible dysplasia in patients with IBD. The finding of ARID1A, FBXW7 and RNF43 mutations in a subset of visible low-grade dysplastic polyps arising in a colitic background implies a need for larger, prospective studies that explore the prevalence and natural history of such lesions.