Skip to main content
NCBI home page
Annals of Oncology logoLink to Annals of Oncology
. 2018 Aug 23;29(10):2061–2067. doi: 10.1093/annonc/mdy337

Colorectal premalignancy is associated with consensus molecular subtypes 1 and 2

K Chang 1,2,#, J A Willis 3,#, J Reumers 4, M W Taggart 5, F A San Lucas 6, S Thirumurthi 7,8, P Kanth 9, D A Delker 9, C H Hagedorn 10, P M Lynch 7,8, L M Ellis 2,11, E T Hawk 1, P A Scheet 2,6, S Kopetz 2,12, J Arts 4, J Guinney 13, R Dienstmann 13,14,✉,3, E Vilar 1,2,8,12,✉,3
PMCID: PMC6225810  PMID: 30412224

Abstract

Background

Gene expression-based profiling of colorectal cancer (CRC) can be used to identify four molecularly homogeneous consensus molecular subtype (CMS) groups with unique biologic features. However, its applicability to colorectal premalignant lesions remains unknown.

Patients and methods

We assembled the largest transcriptomic premalignancy dataset by integrating different public and proprietary cohorts of adenomatous and serrated polyps from sporadic (N = 311) and hereditary (N = 78) patient populations and carried out a comprehensive analysis of carcinogenesis pathways using the CMS random forest (RF) classifier.

Results

Overall, transcriptomic subtyping of sporadic and hereditary polyps revealed CMS2 and CMS1 subgroups as the predominant molecular subtypes in premalignancy. Pathway enrichment analysis showed that adenomatous polyps from sporadic or hereditary cases (including Lynch syndrome) displayed a CMS2-like phenotype with WNT and MYC activation, whereas hyperplastic and serrated polyps with CMS1-like phenotype harbored prominent immune activation. Rare adenomas with CMS4-like phenotype showed significant enrichment for stromal signatures along with transforming growth factor-β activation. There was a strong association of CMS1-like polyps with serrated pathology, right-sided anatomic location and BRAF mutations.

Conclusions

Based on our observations made in premalignancy, we propose a model of pathway activation associated with CMS classification in colorectal carcinogenesis. Specifically, while adenomatous polyps are largely CMS2, most hyperplastic and serrated polyps are CMS1 and may transition into other CMS groups during evolution into carcinomas. Our findings shed light on the transcriptional landscape of premalignant colonic polyps and may help guide the development of future biomarkers or preventive treatments for CRC.

Keywords: adenomatous polyps, serrated adenomas, colorectal cancer, consensus molecular subtypes, systems biology, bioinformatics

Key Message

Consensus molecular subtyping (CMS) of a large cohort of sporadic and hereditary colorectal polyps revealed CMS2 and CMS1 as major subtypes in premalignancy. Our results will assist in the design of prospective studies to evaluate the role of transcriptomic profiling of resected polyps in identifying individuals at highest risk for recurrence of advanced polyps or carcinoma development.

Introduction

The canonical genetic pathway underlying malignant transformation of colonic mucosa was well characterized by the work of Vogelstein et al. [1]. Specifically, they described a step-wise cascade of somatic mutations in tumor suppressor genes (e.g. APC, TP53, SMAD4), oncogenes (e.g. KRAS, PIK3CA) and additional epigenetic aberrations that are now strongly implicated in colorectal cancer (CRC) initiation and progression. Yet, subsequent work has also revealed a heterogeneous molecular landscape of CRC at additional levels. For example, at the structural level, it has been shown that chromosomal instability (CIN) in the context of these driver mutations helps to promote tumor invasion. Indeed, imbalances in chromosome number (aneuploidy) and loss-of-heterozygosity are seen in 85% of invasive CRC tumors [2]. Furthermore, approximately 15% of early-stage colorectal tumors are characterized by DNA mismatch repair (MMR) deficiency secondary mainly to epigenetic inactivation of MLH1, thus giving rise to hypermutation and microsatellite instability (MSI) [3]. At the transcriptomic level, Guinney et al. used large-scale gene expression-based profiling of primary CRC tumors to identify four distinct consensus molecular subtypes (CMS) with distinguished biological features: (i) CMS1, characterized by hypermutation, MSI, enrichment for BRAFV600E mutation and strong immune activation; (ii) CMS2, a ‘canonical’ subtype with marked WNT and MYC signaling activation and epidermal growth factor receptor dependence; (iii) CMS3, enriched for KRAS mutations and evident metabolic dysregulation; and (iv) CMS4, a mesenchymal subtype, with prominent transforming growth factor (TGF)-β activation, an immunosuppressive microenvironment, stromal invasion and angiogenesis activation. Notably, CMS tumor classification was shown to stratify CRC patients into distinct prognostic subgroups [4].

Up to 5% of all CRC cases arise in the setting of well-defined inherited syndromes, such as familial adenomatous polyposis (FAP) and Lynch syndrome (LS), among others. FAP results from germline mutations in APC, which constitutively activates WNT/β-catenin-mediated transcription, driving the transformation of intestinal crypts to conventional precursor lesions (tubular, tubulovillous or villous adenomas). FAP-related adenomas have a spectrum of molecular features similar to CIN-positive CRCs [1]. LS-associated CRC results from germline mutations in the MMR genes (MLH1, MSH2, MSH6, PMS2 and EPCAM), with conventional adenomas representing the majority of precursor lesions [5]. Even though MSI is present in only half of LS-associated polyps at diagnosis, MMR becomes universally detected in more advanced precursor lesions, when larger in size [6]. An alternative route of colorectal carcinogenesis is the serrated pathway, which may account for up to one-third of all CRCs. Serrated/hyperplastic polyposis syndrome is characterized by numerous sessile serrated adenomas (SSA), predominantly located in the right side of the colon, in addition to hyperplastic polyps (HP) [7]. SSA carcinogenesis pathway is associated with high CpG island methylation phenotype (CIMPhi) and BRAFV600E mutations as the major driving mechanisms in both sporadic and familial cases [8, 9].

Collectively, such molecular insights of sporadic and familial CRC have been crucial in the development of clinical standards for managing both early-stage and advanced disease. On the contrary, applications toward disease risk prediction and targeted prevention still remain limited. This is due to a relative paucity of information regarding the full spectrum of genetic, epigenetic and transcriptomic changes in benign colon polyps or pre-malignant adenomas. We hypothesize that filling this knowledge gap, particularly at the transcriptomic level, may lead to new approaches for CRC disease prevention not only in the general population but also among high-risk groups such as those with hereditary CRC. Therefore, to more broadly understand the transcriptomic landscape of premalignant polyps, we have applied the CMS tumor classification on several cohorts of sporadic and hereditary adenomas with gene expression data. We hypothesize that: (i) most FAP-related and sporadic conventional adenomatous polyps (AP) have a CMS2-like (epithelial canonical) phenotype; (ii) LS polyps display a CMS1-like (MSI-Immune) phenotype; (iii) SSA and HP are enriched for both CMS1-like and CMS4-like (mesenchymal) phenotypes.

Materials and methods

Study design

The objective of this study was to carry out a comprehensive analysis of carcinogenesis pathways of premalignant colorectal polyps in the context of the CMS classification of CRC. We collected different cohorts of adenomatous (N = 301) and serrated polyps (N = 28 HPs and 60 SSAs) from sporadic and hereditary populations from a variety of sources: (i) an original institutional cohort from The University of Texas MD Anderson Cancer Center, (ii) eleven publicly available datasets from prior publications (GSE10714 [10], GSE19963, GSE4183 [11], GSE45270 [12], GSE8671 [13], GSE79462 [14], GSE46513 [15], GSE76987 [16], GSE88945 [17], GSE106500 [18], GSE108317) and (iii) proprietary data from Janssen Oncology (GSE117606, GSE117607) (supplementary Tables S1 and S2, available at Annals of Oncology online). No blinding was used for analyses. All primary data presented in this article is provided in supplementary Tables, available at Annals of Oncology online.

Next-generation sequencing datasets and Affymetrix array datasets

RNA-sequencing (RNA-seq) was carried out using Illumina HiSeq2000 and 4000 platforms in a total of 31 colorectal polyps. Methods for collection, extraction of RNA, processing of the raw data and bioinformatics analyses have been described previously [17–19]. Affymetrix microarrays were carried out in a total of 216 colorectal polyps obtained by Janssen Oncology (supplementary Tables S1 and S2, available at Annals of Oncology online).

Gene expression data processing and normalization and consensus molecular subtyping

We constructed aggregated gene expression matrices for each data platform before downstream analysis. Then, to correct for systematic batch differences generated by three different platforms (i.e. RNA-seq, Affymetrix HGU133 Plus 2.0, Affymetrix HT HGU133 Plus PM), we applied the ComBat [20] method across multiple studies within each platform using SVA [21] R package (supplementary Figure S1A–C, available at Annals of Oncology online), including polyp type as a covariate. For Affymetrix data, we used the arrayQualityMetrics [22] R package to exclude outliers based on gene expression distribution and either distances between array or MA plot. For RNAseq data, we used a method based on Euclidean distances between samples. Outlier samples were excluded if their summed distances were >2.5 SD. Lastly, we removed genes that were not common across the three aggregated expression matrices. We carried out CMS tumor classification on the polyp samples using the random forest (RF) predictor in CMSclassifier R package [4]. The decision to use an RF versus a single-sample predictor classifier was made a priori based on several assumptions (see supplementary Materials and Methods, Figure S6 and Tables S10–S11, available at Annals of Oncology online). CMS classification was assigned to the subtype with highest posterior probability.

Targeted mutations analysis

Targeted panel sequencing on KRAS and BRAF was only carried out by Janssen Oncology in the Avaden cohorts. Complete details of each of the methods sections can be found in supplementary Materials and Methods, available at Annals of Oncology online.

Results

Consensus molecular subtyping of a large cohort of polyps revealed CMS2 and CMS1 as major subtypes in premalignancy

We first analyzed the distribution of CMS groups only across polyps obtained from different clinical contexts (sporadic versus hereditary syndrome) and pathologic subtypes (AP, HP and SSA). Overall, the majority of sporadic polyps (n = 311) were classified as either CMS2 (69.5%) or CMS1 (21.9%), while CMS3 (5.1%) and CMS4 (1.6%) classifications were less abundant. Furthermore, within sporadic polyps, the majority of AP (80%) were classified as CMS2, whereas the majority of HP (57.1%) and SSA (76.5%) were classified as CMS1 (Figure 1A, see supplementary Table S3, available at Annals of Oncology online).

Figure 1.

Figure 1.

Open in a new tab

Circos plots presenting the distributions of consensus molecular subtyping (CMS) groups in sporadic (A) and hereditary polyps (B). AP, adenomatous polyps; FAP, familial adenomatous polyposis; HP, hyperplastic polyps; LS, Lynch syndrome; SPS, serrated polyposis syndrome.

Similarly, hereditary polyps (n = 78) were mostly distributed between CMS2 (52.6%) and CMS1 (38.5%). CMS3 (2.6%) and CMS4 (6.4%) again accounted for a small percentage of total hereditary polyps (Figure 1B). AP from a hereditary background were predominantly (86.7%) CMS2. Surprisingly, the majority (86.7%) of AP from LS patients were also classified as CMS2, which is in contrast to our a priori assumption. HP (71.4%) and SSA (96.2%) from a hereditary background were mostly classified as CMS1. Overall, these results suggest that the CMS2 (canonical) and CMS1 (MSI-Immune) molecular subtypes play dominant roles in early conventional adenomas and serrated polyps, respectively.

Pathway enrichment analysis of CMS showed immune activation and classical WNT and MYC targets as dominant signatures in premalignancy

We carried out GSEA using previously described biological pathways and expression signatures pertinent to CRC carcinogenesis [4] (supplementary Table S9, available at Annals of Oncology online). CMS1-like polyps were characterized by significant enrichment of genes involved in immune and stromal infiltration as well as pathways implicated in immune cytotoxicity. They also showed strong activation in JAK-STAT and MAPK signaling (supplementary Figures S2, S7–S9 and Tables S4–S7, available at Annals of Oncology online). CMS2-like polyps displayed strong enrichment for WNT and MYC targets, which are classical carcinogenesis pathways in CRC (supplementary Figures S2, S7–S9 and Tables S4–S7, available at Annals of Oncology online). For the small number polyps classified as CMS3, we did not observe significant enrichment for glutamine and fatty acid pathways, thus making their activation a molecular feature that arise in advanced adenomas or carcinomas (supplementary Figures S2, S7–S9 and Tables S4–S7, available at Annals of Oncology online). Lastly, although the number of CMS4 polyps was small, they showed significant enrichment of mesenchymal and stromal signatures along with TGFβ activation (supplementary Figures S2, S7–S9 and Tables S4–S7, available at Annals of Oncology online). Additional exploratory analyses were carried out by stratifying polyps within each CMS group according to high-risk pathologic features (size >1 cm or presence of high-grade dysplasia) and did not identify any differential signals due to the relatively small sample sizes (supplementary Figures S3 and S4, available at Annals of Oncology online). Taken together, our results confirm that immune activation and classical carcinogenesis pathways were the main transcriptomic events in colorectal premalignancy.

Associations of CMS with polyp location and KRAS and BRAF mutations

We next explored CMS distributions across various clinical-pathologic and molecular features (supplementary Table S8, available at Annals of Oncology online). We found that CMS1 and CMS2 polyps have similar proportions in both males and females (supplementary Figure S5A, available at Annals of Oncology online). No statistically significant association was found between the presence of high-grade dysplasia/carcinoma in situ and CMS classification (supplementary Figure S5B and Table S8, available at Annals of Oncology online). Interestingly, CMS1 polyps were more frequently presented in right colon in both sporadic (P < 0.01) and hereditary (P < 0.0001) cohorts. On the contrary, CMS2 polyps are more frequently presented in the left colon in both sporadic (P < 0.005) and hereditary (P < 0.005) cases (supplementary Figure S5C, available at Annals of Oncology online). These results suggest that CMS1 carcinomas may be largely derived from HP and SSA that are often found in the right colon. Next, we investigated mutations associated with CMS groups and found that BRAFV600E was more frequently present in CMS1 polyps (P < 0.0001). Furthermore, KRAS codon 12 and 13 mutations showed a trend to occur less frequently among CMS1 compared with CMS2 polyps that did not reach statistical significance. Due to the small number of CMS3 polyps in our sample, we did not observe significant over-representation of KRAS (supplementary Figure S5D–E, available at Annals of Oncology online).

Discussion

In the United States, the overall incidence of CRC has steadily decreased over the past decade [23] owing to increased utilization of screening colonoscopy. Yet, CRC still remains the third most common newly diagnosed malignancy in both men and women. Furthermore, despite its decreased overall incidence, an alarming trend toward younger age-of-onset (<55 years old) has also emerged [24]. These observations highlight an important challenge to better understand the molecular diversity of colonic polyps and to develop targeted approaches for disease prevention. Toward this end, we carried out a large-scale transcriptomic analysis of both sporadic and familial colon polyps by applying the CMS framework.

Taking together our results along with previously published reports, we propose of a new model for pathway activation driving premalignancy (Figure 2). First, we found that FAP, LS and sporadic conventional adenomas display an epithelial canonical CMS2 phenotype with strong WNT and MYC downstream targets enrichment. Interestingly, MMR deficiency has been observed in only half of LS polyps at diagnosis but becomes nearly universal in advanced adenomas and carcinomas [6]. The majority of LS adenomas included in our study were early lesions (low grade), which likely explains the predominance of CMS2-associated signaling in these samples. Nonetheless, our data presented here in combination with our recent immunoprofiling of LS polyps leads us to hypothesize that LS polyps transition from an epithelial phenotype (CMS2-like) with some degree of immune activation at early stages to a complete MSI-Immune (CMS1-like) phenotype at advanced stages with further development of dysplasia and complete loss of MMR functioning [18]. However, additional correlative studies of MMR deficiency and the immune microenvironment in advanced LS lesions are warranted to explore this hypothesis using ex vivo organoid models and longitudinal samples. Second, we found that HP and SSA were both enriched for MSI-Immune CMS1 phenotype. While these polyps displayed strong enrichment of immune and JAK-STAT activation, they also showed enrichment for TGFβ activation and stromal gene signatures, which are features that previously had been described in CMS4 carcinomas [4]. However, our results are consistent with prior studies showing that TGFβ activation plays an important role in colorectal carcinogenesis. Using human organoid cultures and genome editing technology, Fessler et al. have shown that the genetic background of premalignant lesions dictates the dominating response to TGFβ, changing it from a largely apoptotic response in WNT pathway-activated conventional tubular adenomas to a dominant epithelial–mesenchymal transition response in BRAFV600E-mutated SSA [14]. Taken together, these findings lead us to hypothesize that TGFβ signaling alone is not sufficient to determine whether pre-malignant lesions progress into CMS4 versus CMS1 carcinomas.

Figure 2.

Figure 2.

Open in a new tab

Hypothetical model of pathway activation driving the consensus molecular subtyping (CMS) classification in adenomatous (top) and serrated polyps (bottom). Well-established associations based on the literature are presented in solid lines [4, 14, 18] and those that are hypothetical or derived from the data reported in this article are presented in dashed lines. AP, adenomatous polyps; CIN, chromosomal instability; EMT, epithelial-to-mesenchymal transformation; HP, hyperplastic polyps; MSI, microsatellite instability; SSA, sessile serrated adenomas; TGF, transforming growth factor; MAPK, mitogen-activated protein kinase.

Finally, it is interesting to consider how transcriptional profiling of pre-malignant lesions by themselves, without combining them with carcinoma samples, might capture clinically relevant phenotypes and alter standard clinical practice. For example, in the work by Guinney et al., CMS4 carcinomas were associated with an aggressive clinical phenotype defined by a higher proportion of diagnoses at advanced stage and worse outcomes following surgery and adjuvant chemotherapy [4]. In addition, CMS4 carcinomas harbored a distinct transcriptional profile characterized by up-regulation of genes linked to epithelial-to-mesenchymal transformation (EMT). Other studies have contributed to a model of cancer evolution in which EMT increases the metastatic potential of carcinomas through enhanced cell migration and invasion [25–27]. Extrapolating from these data, we hypothesize that CMS4-like pre-malignant lesions progress more rapidly in the adenoma-to-carcinoma transformation pathway. We further speculate that patients who are found to have a CMS4-like polyp on screening colonoscopy may benefit from closer surveillance of at-risk normal mucosa than would otherwise be pursued based on current surveillance guidelines. Certainly, several key studies are needed to explore this hypothesis and firmly establish the prognostic utility of pre-malignant CMS classification. Toward this end, analysis of longitudinal patient cohorts with sufficient clinical outcomes data (e.g. diagnosis of advanced adenoma or carcinoma) would be highly valuable. Although current CMS RF classifier has robust performance on archived tissue specimens, the classifier contains more than 200 genes and, therefore, a new classifier requiring fewer genes would be ideal in a clinical setting. Recently, a CMS classifier using 38 genes derived from Nanostring platform shown to be suitable for FFPE samples [28].

We acknowledge that our study has several limitations. First, although we have demonstrated that CMS classification is technically feasible for premalignant tissue, it is also important to recognize that the classifier used in our study was derived specifically from carcinomas. As a result, it is possible that this approach may not adequately or meaningfully capture the transcriptomic landscape of adenomas. Indeed, the optimal approach toward transcriptional subtyping of adenomas remains unclear and requires additional investigation. We suggest that a CMS classifier specifically derived for premalignant tissue would be a reasonable alternative approach and would provide an opportunity to identify the minimum number of genes necessary to adequately pre-malignant transcriptional diversity. Mapping the pre-malignancy CMS space onto the carcinoma CMS space in such a way that provides clinical and biological insights would itself present another challenge. For this reason, we decided not to carry out CMS classification of premalignancy by combining adenoma samples with carcinomas together in order to avoid biases intrinsic to gene expression normalization between two biologically different types of lesions that could dilute the transcriptomic signals coming from low abundant dysplastic cells in premalignancy. A second limitation of our study is that we did not include comprehensive analysis of somatic single-nucleotide mutations or copy number alterations in all the polyp samples. These additional analyses would have allowed us to correlate various known CRC drivers with CMS classification of polyps. Third, we did not have information on the MSI status of the polyps in our study. We attempted to assess the MSI status of polyps using gene signatures derived from carcinoma data and hierarchical clustering but we were unable to observe distinct groups of samples. These limitations are primarily driven by the diminutive size of polyps and the requirement to prioritize tissue for gene expression analysis. Lastly, to help maximize the number of adenomas included in our study, we opted to classify samples according to the CMS subtype with the highest posterior probability. By comparison, setting a minimum threshold of 0.5 posterior probability reduced the number of analyzable samples significantly but would not change the overall conclusions of our study.

Overall, to the best of our knowledge, our study is the largest investigation of transcriptional drivers in colorectal premalignancy. Our results show that pathway dependencies of different CMS groups originally described in carcinomas are indeed recapitulated in adenomas, thus opening the door to more personalized development of targeted chemopreventive strategies for polyps, particularly in hereditary syndromes.

Acknowledgements

We thank the staff of the Microarray and Sequencing Core Facility at UTMDACC for the assistance with RNA sequencing.

Datasets

RNA-seq and array files have been already deposited in GEO. The following links can be used by Reviewers to access the data for re-analysis with tokens for those ones that are not public yet:

GSE88945 (3 FAP samples):

https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE88945 GSE106500 (13 FAP and 11 LS samples):

https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE106500 GSE108317 (5 LS samples):

https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE108317 GSE10714 (16 samples):

https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE10714 GSE19963 (5 samples):

https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE19963 GSE4183 (14 samples):

https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE4183 GSE45270 (13 samples):

https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE45270 GSE8671 (31 samples):

https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE8671

GSE79462 (16 samples):

https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE79462

GSE46513 (7 samples):

https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE46513

GSE76987 (40 samples):

https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE76987

Avaden 1 and 2 (216 samples) GSE117606 and GSE117607:

https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE117606

https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE117607

Funding

This work was supported by R21 CA208461 (US National Institutes of Health/National Cancer Institute), The University of Texas MD Anderson Cancer Center Colorectal Cancer Moonshot Program and a gift from the Feinberg Family to EV.; Cancer Prevention Educational Award R25T CA057730 (US National Institutes of Health/National Cancer Institute) to KC; Research Training in Academic Medical Oncology T32 Award CA009666-23 (U.S. National Institutes of Health/National Cancer Institute) to JAW; and P30 CA016672 (US National Institutes of Health/National Cancer Institute) to the University of Texas Anderson Cancer Center Core Support Grant; RD was supported in part by the Merck KGaA, Darmstadt, Germany (Grant for Oncology Innovation 2015).

Disclosure

EV has a consulting role with Janssen Research and Development. All remaining authors have declared no conflicts of interest.

Supplementary Material

Supplementary Figures
Click here for additional data file. (3.8MB, pdf)
Supplementary Tables
Click here for additional data file. (288.6KB, pdf)
Supplementary Data
Click here for additional data file. (59.6KB, docx)

References

  • 1. Fearon ER. Molecular genetics of colorectal cancer. Annu Rev Pathol Mech Dis 2011; 6(1): 479–507. [DOI] [PubMed] [Google Scholar]
  • 2. The Cancer Genome Atlas Network. Comprehensive molecular characterization of human colon and rectal cancer. Nature 2012; 487(7407): 330–337. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 3. Vilar E, Tabernero J.. Molecular dissection of microsatellite instable colorectal cancer. Cancer Discov 2013; 3(5): 502–511. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 4. Guinney J, Dienstmann R, Wang X. et al. The consensus molecular subtypes of colorectal cancer. Nat Med 2015; 21(11): 1350–1356. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 5. Liljegren A, Barker G, Elliott F. et al. Prevalence of adenomas and hyperplastic polyps in mismatch repair mutation carriers among CAPP2 participants: report by the colorectal adenoma/carcinoma prevention programme 2. J Clin Oncol 2008; 26(20): 3434–3439. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6. Yurgelun MB, Goel A, Hornick JL. et al. Microsatellite instability and DNA mismatch repair protein deficiency in Lynch syndrome colorectal polyps. Cancer Prev Res (Phila) 2012; 5(4): 574–582. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 7. Syngal S, Brand RE, Church JM. et al. ACG clinical guideline: genetic testing and management of hereditary gastrointestinal cancer syndromes. Am J Gastroenterol 2015; 110(2): 223–262. quiz 63. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8. Rex DK, Ahnen DJ, Baron JA. et al. Serrated lesions of the colorectum: review and recommendations from an expert panel. Am J Gastroenterol 2012; 107(9): 1315–1329. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9. Huang CS, O'Brien MJ, Yang S. et al. Hyperplastic polyps, serrated adenomas, and the serrated polyp neoplasia pathway. Am J Gastroenterol 2004; 99(11): 2242–2255. [DOI] [PubMed] [Google Scholar]
  • 10. Galamb O, Sipos F, Solymosi N. et al. Diagnostic mRNA expression patterns of inflamed, benign, and malignant colorectal biopsy specimen and their correlation with peripheral blood results. Cancer Epidemiol Biomarkers Prev 2008; 17(10): 2835–2845. [DOI] [PubMed] [Google Scholar]
  • 11. Galamb O, Gyorffy B, Sipos F. et al. Inflammation, adenoma and cancer: objective classification of colon biopsy specimens with gene expression signature. Dis Markers 2008; 25(1): 1–16. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12. De Sousa EMF, Wang X, Jansen M. et al. Poor-prognosis colon cancer is defined by a molecularly distinct subtype and develops from serrated precursor lesions. Nat Med 2013; 19(5): 614–618. [DOI] [PubMed] [Google Scholar]
  • 13. Sabates-Bellver J, Van der Flier LG, de Palo M. et al. Transcriptome profile of human colorectal adenomas. Mol Cancer Res 2007; 5(12): 1263–1275. [DOI] [PubMed] [Google Scholar]
  • 14. Fessler E, Drost J, van Hooff SR. et al. TGFbeta signaling directs serrated adenomas to the mesenchymal colorectal cancer subtype. EMBO Mol Med 2016; 8(7): 745–760. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15. Delker DA, McGettigan BM, Kanth P. et al. RNA sequencing of sessile serrated colon polyps identifies differentially expressed genes and immunohistochemical markers. PLoS One 2014; 9(2): e88367.. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 16. Kanth P, Bronner MP, Boucher KM. et al. Gene signature in sessile serrated polyps identifies colon cancer subtype. Cancer Prev Res (Phila) 2016; 9(6): 456–465. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17. Crespo M, Vilar E, Tsai SY. et al. Colonic organoids derived from human induced pluripotent stem cells for modeling colorectal cancer and drug testing. Nat Med 2017; 23(7): 878–884. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18. Chang K, Taggart MW, Reyes-Uribe L. et al. Immune profiling of premalignant lesions in patients with Lynch syndrome. JAMA Oncol 2018; 4(8): 1085–1092. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 19. Borras E, San Lucas FA, Chang K. et al. Genomic landscape of colorectal mucosa and adenomas. Cancer Prev Res (Phila) 2016; 9(6): 417–427. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 20. Johnson WE, Li C, Rabinovic A.. Adjusting batch effects in microarray expression data using empirical Bayes methods. Biostatistics 2007; 8(1): 118–127. [DOI] [PubMed] [Google Scholar]
  • 21. Leek JT, Johnson WE, Parker HS. et al. The sva package for removing batch effects and other unwanted variation in high-throughput experiments. Bioinformatics 2012; 28(6): 882–883. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 22. Kauffmann A, Gentleman R, Huber W.. arrayQualityMetrics–a bioconductor package for quality assessment of microarray data. Bioinformatics 2009; 25(3): 415–416. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 23. Siegel RL, Miller KD, Jemal A.. Cancer statistics, 2018. CA Cancer J Clin 2018; 68(1): 7–30. [DOI] [PubMed] [Google Scholar]
  • 24. Bailey CE, Hu CY, You YN. et al. Increasing disparities in the age-related incidences of colon and rectal cancers in the United States, 1975-2010. JAMA Surg 2015; 150(1): 17–22. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 25. Pereira L, Mariadason JM, Hannan RD. et al. Implications of epithelial-mesenchymal plasticity for heterogeneity in colorectal cancer. Front Oncol 2015; 5: 13.. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 26. Pino MS, Kikuchi H, Zeng M. et al. Epithelial to mesenchymal transition is impaired in colon cancer cells with microsatellite instability. Gastroenterology 2010; 138(4): 1406–1417. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 27. Polyak K, Weinberg RA.. Transitions between epithelial and mesenchymal states: acquisition of malignant and stem cell traits. Nat Rev Cancer 2009; 9(4): 265–273. [DOI] [PubMed] [Google Scholar]
  • 28. Ragulan C, Eason K, Nyamundanda G. et al. Analytical validation of multiplex biomarker assay to stratify colorectal cancer samples into molecular subtypes. bioRxiv 2018. doi: 10.1101/174847. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplementary Figures
Click here for additional data file. (3.8MB, pdf)
Supplementary Tables
Click here for additional data file. (288.6KB, pdf)
Supplementary Data
Click here for additional data file. (59.6KB, docx)

Articles from Annals of Oncology are provided here courtesy of Oxford University Press

RESOURCES