Context.—Colorectal carcinoma is the third most common cause of cancer death in males and females in the United States. Rectal adenocarcinoma can have distinct therapeutic and surgical management from colonic adenocarcinoma owing to its location and anatomic considerations.
Objective.—To determine the oncologic driver mutations and better understand the molecular pathogenesis of rectal adenocarcinoma in relation to colon adenocarcinoma.
Design.—Next-generation sequencing was performed on 20 cases of primary rectal adenocarcinoma with a paired lymph node or solid organ metastasis by using an amplicon-based assay of more than 2800 Catalogue of Somatic Mutations in Cancer (COSMIC)–identified somatic mutations.
Results.—Next-generation sequencing data were obtained on both the primary tumor and metastasis from 16 patients. Most rectal adenocarcinoma cases demonstrated identical mutations in the primary tumor and metastasis (13 of 16, 81%). The mutations identified, listed in order of frequency, included TP53, KRAS, APC, FBXW7, GNAS, FGFR3, BRAF, NRAS, PIK3CA, and SMAD4.
Conclusions.—The somatic mutations identified in our rectal adenocarcinoma cohort showed a strong correlation to those previously characterized in colonic adenocarcinoma. In addition, most rectal adenocarcinomas harbored identical somatic mutations in both the primary tumor and metastasis. These findings demonstrate evidence that rectal adenocarcinoma follows a similar molecular pathogenesis as colonic adenocarcinoma and that sampling either the primary or metastatic lesion is valid for initial evaluation of somatic mutations and selection of possible targeted therapy.
Colorectal carcinoma is the third most common cause of cancer death in males and females in the United States.1 The American Cancer Society estimates that 93 090 cases of colon cancer and 39 610 cases of rectal cancer are expected to occur in 2015.1 Currently, patients with rectal adenocarcinoma may undergo distinct therapeutic and surgical management as compared to colonic adenocarcinoma,2,3 owing to its location and anatomic considerations. Therapeutic management for rectal adenocarcinoma, unlike colon cancer, frequently includes neoadjuvant chemotherapy and radiation for locally advanced tumors, as this has been shown to decrease rates of local recurrence in rectal cancers.4–6 To better understand the molecular basis of rectal adenocarcinoma, we sought to determine the somatic mutation profile of primary rectal adenocarcinomas and compare it to the somatic mutation profile of paired metastases, as well as to known mutations in colon cancer.
Comprehensive studies of the molecular alterations in colorectal carcinoma have been performed by The Cancer Genome Atlas (TCGA).7 However, analysis of solely primary rectal adenocarcinomas with paired lymph node and solid organ metastases by next-generation sequencing (NGS) has not been previously examined, to our knowledge. The Catalogue of Somatic Mutations in Cancer (COSMIC) database encompasses a curated database of clinically meaningful mutations in cancer, which have an increasing role in predicting prognosis and therapeutic response. Systematic examination of patient-matched primary and metastatic rectal adenocarcinoma NGS data and identified COSMIC mutations could provide insight into the biologic progression of rectal adenocarcinoma. Also, these data may allow flexibility in clinical workup and molecular analysis by potentially using DNA samples from primary and/or metastatic sites in rectal adenocarcinoma.
MATERIALS AND METHODS
A retrospective search of our clinical and surgical pathology databases for patients with rectal adenocarcinoma from February 2009 to January 2014 was performed. The tumors selected for this study were designated as rectal both clinically and by corresponding pathologic examination, defined as the inferior margin of the tumor being less than 16 cm from the anal verge. Twenty patients with primary rectal adenocarcinoma and synchronous or metachronous metastases present in either a locoregional lymph node or a solid organ were identified. Tumor deposits were excluded. The procedures followed were approved by the institutional review board (IRB(2)0214-0045).
The hematoxylin-eosin (H&E)–stained slides of the primary tumor and metastatic lesion were reviewed. One case (No. 6) showed significant tumor heterogeneity in the primary tumor with components of tubulovillous adenoma (TVA), low-grade adenocarcinoma, and high-grade adenocarcinoma. Each of these components, as well as the metastasis, was microdissected and individually analyzed, thus resulting in analysis of 42 samples. Selection of tumors for analysis by NGS was based on reviewing H&E-stained slides to select a single block with the highest tumor burden, as previously described.8
DNA was extracted from 42 formalin-fixed, paraffin-embedded (FFPE) tissue samples by using the QIAamp DNA FFPE Tissue Kit (Qiagen, Valencia, California) and the Agencourt AMPure XP Kit (Beckman Coulter, Brea, California). Manual microdissection was performed on areas with highest tumor density and proper morphology as designated by consensus of 2 of the authors (S.M.C. and M.R.S.) on the H&E-stained slides. This designated area also had tumor percentage determined by consensus of 2 of the authors (S.M.C. and M.R.S.). The visual tumor percentage within the dissected area ranged from 5% to 90%. Tumor regression grading was assigned on cases that received neoadjuvant chemoradiation, according to the American Joint Commission on Cancer Staging Handbook, 7th edition.9
The Ion AmpliSeq Library Kit 2.0 and the Ion AmpliSeq Cancer Hotspot Panel v2 (Life Technologies, Grand Island, New York) were used for library preparation, according to the manufacturer's instructions. The assay was previously validated in our Molecular Diagnostics Laboratory. It uses 10 ng of DNA to analyze more than 2800 mutations across 50 known oncogenes and tumor suppressor genes including ABL1, AKT1, ALK, APC, ATM, BRAF, CDH1, CDKN2A, CSF1R, CTNNB1, EGFR, ERBB2, ERBB4, EZH2, FBXW7, FGFR1, FGFR2, FGFR3, FLT3, GNA11, GNAS, GNAQ, HNF1A, HRAS, IDH1, IDH2, JAK2, JAK3, KDR, KIT, KRAS, MET, MLH1, MPL, NOTCH1, NPM1, NRAS, PDGFRA, PIK3CA, PTEN, PTPN11, RB1, RET, SMAD4, SMARCB1, SMO, SRC, STK11, TP53, and VHL. Recovered DNA was quantified via the Qubit dsDNA HS Assay Kit (Life Technologies). For specimen multiplexing, Ion Xpress Barcode Adaptors (Life Technologies) were used. Each amplified and barcoded library was then diluted in nuclease-free water to a concentration of 100 pg/mL. Template preparation was performed by using the Ion Chef and the Ion PI IC 200 Kit (Life Technologies), according to the manufacturer's instructions. The 42 samples were loaded onto 2 separate Ion PI chips (Life Technologies) for sequencing on the Ion Proton, using the Ion PI Sequencing Kit (Life Technologies). Manual library preparation and the Ion OneTouch2 System (Life Technologies) and template-positive Ion PI Ion-Sphere particles were also used.
The NGS uniformity and mapped reads were reviewed (supplemental digital content, Supplemental Tables 1 and 2, is available at www.archivesofpathology.org in the June 2016 table of contents). A read depth of 100× amplicon coverage was the standard cutoff for inclusion in the final analysis. However, samples with a mean read depth of less than 100× reads per amplicon were reviewed manually and included at the discretion of the pathologist (B.P.P.). For automated variant calling, a 4.5% variant frequency cutoff was used. In cases with discrepancy in mutation profile between primary and metastatic lesion, sequences were directly visualized by accessing the Integrated Genomic Viewer version 2.3 (http://www.broadinstitute.org/igv/projects/current/igv_mm.jnlp, Accessed January to February 2015).
The mutational analysis was performed by using the Torrent Suite Software version 4.2 (Life Technologies) via the Proton-Somatic-Low Stringency Variant Caller PlugIn, using default settings. The mutations identified were confirmed by reference to the COSMIC database v72.10–12 Silent and germline polymorphisms were excluded. The Integrated Genomic Viewer was used for visualization confirmation of each variant call and to exclude sequencing artifacts such as false-positive mutations identified owing to homopolymer regions (http://www.broadinstitute.org/igv/projects/current/igv_mm.jnlp, Accessed January to February 2015).
The NGS data was generated from 16 primary/metastatic rectal adenocarcinoma pairs. Most of the metastases were to locoregional lymph nodes (13 of 16, 81%). The remaining metastases were pulmonary (2 of 16, 13%) and hepatic (1 of 16, 6%). All cases were at least stage 3A on presentation, and most were low grade (13 of 16, 81%). The average patient age at presentation was 57 years (range, 35–81 years) and 7 were female. More than half of the patients (9 of 16, 56.3%) had documented neoadjuvant chemoradiation (Table 1). The tumor regression scores in the cases with neoadjuvant chemoradiation were grade 2 (4 of 9 cases, 44%) and grade 3 (5 of 9 cases, 56%).
The average primary tumor size was 2.8 cm (range, 0.8–5.5 cm) and most were locally advanced: pT3 (12 of 16, 75%) or pT4b (1 of 16, 6%). The remaining 3 cases were pathologic stage pT1 or pT2. There was similar average tumor percentage in the primary rectal adenocarcinomas (52% tumor) as in the metastases (50% tumor), with an overall range of 5% to 90% tumor percentage per case.
Most rectal adenocarcinoma cases had identical somatic mutations in the primary tumor and the metastatic tissue (13 of 16, 81%). These cases included predominantly locoregional lymph node metastases (11 cases) as well as both of the pulmonary metastases (2 cases). The mutations identified, listed in order of frequency, included TP53, KRAS, APC, FBXW7, GNAS, BRAF, NRAS, PIK3CA, FGFR3, and SMAD4 (Figure 1). Every case had at least 1 identified mutation, and the average mutations per case were 2.4 (range, 1–4). For the patients with correlating mutations between primary tumor and metastasis, most had received neoadjuvant chemoradiation (8 of 13, 62%) as opposed to not (5 of 13, 38%).
Three cases had additional mutations detected in the metastatic tumor as compared to the primary rectal adenocarcinoma (Table 2). One of these patients received neoadjuvant chemoradiation while the other 2 did not. The patient who received neoadjuvant chemoradiation presented with a pathologic stage pT3 tumor with paired liver metastasis that had APC c.4348C>T p.R1450* gene mutation (COSM13127) in the metastatic tumor but not the primary tumor. This mutation falls within the mutation cluster region in the APC gene that encompasses the most common somatic mutations in sporadic colorectal carcinoma.13 Mutations in the APC gene affect the β-catenin–dependent Wnt signaling pathway and elevate levels of β-catenin, contributing to dysregulated gene transcription.13,14 However, when the preneoadjuvant therapy biopsy was analyzed, the same APC gene mutation as in the metastasis was detected in the primary tumor. A KRAS c38G>A p.G13D (COSM532) mutation was also present in both the primary tumor and metastasis in this first case (Table 1).
The second case had a lymph node metastasis with newly detected APC c.4126_4127insT p.Y1376fs*10 (COSM18910) and TP53 c.844C>T p.R282W (COSM10704) mutations. The APC mutation in codon 1376 is also within the most common somatic mutation cluster of colorectal carcinoma in the APC gene, as described above. The TP53 mutation identified in this case (COSM10704) is a missense mutation. Missense mutations in p53 may contribute to colorectal carcinogenesis by multiple mechanisms, including increasing tumor cell proliferation and angiogenesis, and decreasing apoptosis and tumor hypoxia.13 These p53 missense mutations are also predicted to affect autophagy and microRNA processing.13 This case had a low number of mapped reads (Supplemental Tables 1 and 2, case 2); however, an APC mutation was detected and confirmed by direct visualization. This case was also the only pathologic stage pT1 tumor included in the cohort and the patient did not receive neoadjuvant chemoradiation.
In the third case, the patient did not receive neoadjuvant chemoradiation and the primary tumor had histologic heterogeneity with 3 separate histologic patterns observed and independently analyzed by NGS, including TVA, low-grade adenocarcinoma, and high-grade adenocarcinoma. The lymph node metastasis showed only low-grade adenocarcinoma histologic profile. The TVA had both a KRAS c.35G>A p.G12D (COSM521) and APC c.4474delG p.A1492fs*15 (COSM19340) mutation. This KRAS mutation is the most frequent KRAS mutation detected in metastatic colorectal adenocarcinoma (comprising 144 of 400 cases [36%] of all KRAS mutations in one study).15 The low- and high-grade adenocarcinoma components both carried the APC mutation, but also a TP53 c.535C>A p.H179N mutation (COSM44156) was detected. The KRAS mutation was not detected in the invasive components. The lymph node metastasis had the TP53 mutation, but the APC and KRAS mutations were not detected (Figure 2).
In this study, we demonstrate that most rectal adenocarcinoma metastases have similar somatic mutation profiles as their paired primary tumors. This supports selecting either specimen in the molecular diagnostics laboratory to identify mutations that could be targeted for therapy. This is significant as only minimal primary tumor may remain after neoadjuvant treatment. Therefore, a lymph node or solid organ metastasis could provide a sample for NGS analysis in cases that show progression or persistent tumor post therapy. A similar previous study using the iPLEX assay (Sequenom, San Diego, California) to analyze 84 pairs of primary and metastatic colorectal cancers found that KRAS, NRAS, BRAF, PIK3CA, and TP53 mutation status were highly concordant (>90% concordance for all 5 analyzed genes) in primary and metastatic sites.16 These findings lend further support to our current study, which narrowed its focus to rectal adenocarcinoma alone and expanded the scope of analyzed gene mutations.
Extensive study into the genetic profile of colorectal adenocarcinoma by TCGA identified the 8 most frequently mutated genes (APC, TP53, KRAS, PIK3CA, FBXW7, SMAD4, TCF7L2, and NRAS)7 ; mutations in all of these genes (except for TCF7L2) were also identified in our rectal adenocarcinoma–based study. This finding provides molecular-based evidence supporting the close link between rectal adenocarcinoma oncogenic drivers and that of the better characterized colon adenocarcinoma. There has been some debate over the similarity of molecular drivers of colon versus rectal adenocarcinoma in the literature, with significant molecular differences in mutation status and tumor-associated markers reported in some studies.17,18 Other studies have shown similar molecular profiles between colon and rectum except for the BRAF mutation frequency.19 A large study of 1443 colorectal cancers found that there is a gradual shift in molecular characteristics along the length of the bowel, particularly in relationship to the frequency of BRAF mutations, microsatellite instability–high status, and CpG island methylator phenotype (CIMP)-high tumors.20 The TCGA analysis found no difference between colon and rectal carcinomas in copy number, CIMP, messenger RNA, and microRNA in nonhypermutated tumors.7
APC mutations were the third most commonly detected gene mutation in our study, after TP53 and KRAS mutations (Figure 1). Inactivation of APC is one of the most frequent mutation events in colorectal carcinoma.13,21 The classic colorectal adenoma-carcinoma sequence begins with inactivation of APC at 5q21. This is followed by mutations in KRAS and subsequent p53 mutations with loss of heterozygosity.21 Therefore, it was initially surprising that 2 of our cases had newly detected APC mutations only in the metastatic component, as it is traditionally thought of as an early event in the colorectal carcinogenesis. Direct visualization of these 2 cases showed that in 1 case, an APC mutation was present in the primary tumor (Table 2, case 2; and Supplemental Tables 1 and 2). In the case with no APC mutation in the primary tumor resection, the patient had received neoadjuvant chemoradiation. The analysis of the preneoadjuvant therapy biopsy specimen did reveal the APC mutation, and therefore the same APC mutation discovered in the metastatic disease was also present in all primary tumors.
The relatively higher frequency of KRAS mutations in our current study (Figure 1) is likely explained by the cancer stem cell data recently published by Moon et al.22 They found that oncogenic KRAS mutations increased the cancer stem cell characteristics of colorectal cancer cells that already had an APC mutation, promoting progression and metastasis.22 KRAS mutations have also been reported as more frequent in cases with liver metastasis and more aggressive behavior,23 and KRAS status has been touted as a negative predictive biomarker in colorectal cancer, particularly when stratification with BRAF and microsatellite instability status are incorporated.24,25 As we included only metastatic rectal cancers in this study, this might explain the relatively higher relative levels of KRAS mutations.
KRAS mutational analysis is recommended by National Comprehensive Cancer Network Guidelines Version 3.2015 for metastatic colorectal cancer.26 In one of our cases, a KRAS mutation was detected within the TVA and not within the adjacent primary invasive adenocarcinoma or metastasis (Figure 2). This recommendation guides therapeutic management, and if there is a known KRAS or NRAS mutation, cetuximab or panitumumab (anti–epidermal growth factor receptor therapy) is not recommended.26 Therefore, determining KRAS mutation status is important, owing to its influence on both therapy and prognosis. The lack of KRAS mutation detection in the invasive adenocarcinoma and metastasis may be due to several possibilities: (1) low percentage cellularity tumor DNA, (2) intratumoral heterogeneity, or (3) neoadjuvant therapy effect and/or tumor subclonal expansion.
The first possibility, low percentage of tumor DNA, is common in rectal adenocarcinoma and can create low-cellularity specimens for molecular analysis. Neoadjuvant therapy for rectal adenocarcinoma is frequent and was used in more than half of the patients in our study. The effect of low tumor cellularity as problematic for KRAS mutation detection was investigated in a recent study by Dudley et al.27 They found KRAS mutation status could be confounded (falsely detected as wild type) in low-cellularity specimens analyzed by pyrosequencing.27 Another study by Boissière-Michot et al,28 investigating rectal adenocarcinomas (and also using primarily pyrosequencing), also found that the KRAS mutation was no longer detected in some of their cases after therapy. However, although we had low cellularity tumor percentage in some of the cases in our study, the case with KRAS mutation in TVA with a subsequent KRAS wild type had good tumor percentage estimation by visual analysis (90% in primary tumor, 40% in metastasis; Table 1), making this a less likely explanation for the KRAS mutation status change.
The second possibility is that intratumoral heterogeneity affected the KRAS status determination. Intratumoral heterogeneity has been well described in other tumor types such as melanoma, with variable frequency of BRAF V600E mutation in different areas of primary melanoma as well as within paired metastases.29 Pertinent to our current case, in the recent study by Dudley et al27 they separately analyzed 3 distinct areas in a TVA for KRAS mutations. In 2 of 3 separate areas within the TVA, the KRAS status was wild type. However, in a third area, a 23% p.G12D KRAS mutation was identified.27 This G12D mutation is the most commonly identified KRAS mutation in metastatic colorectal carcinoma15 and was also the mutation identified in the TVA in our case. Therefore, intratumoral heterogeneity could explain the difference of KRAS mutation status between the TVA and invasive tumor/metastasis.
The third possibility, neoadjuvant therapy effect, brings up interesting issues addressed in recent literature. A study by Demes et al30 of rectal adenocarcinoma cases found a change in KRAS mutation status from wild type to mutated in 2 of 25 patients after receiving neoadjuvant radiochemotherapy. Providing a possible explanation for this phenomenon is the mathematical analysis performed in an article by Diaz et al,31 which found that the mechanism of resistance to anti–epidermal growth factor receptor inhibitors in colorectal cancers is due to expansion of preexisting rare cells of KRAS-mutated subclones within the tumor.31 This mechanism could account for an initial molecular analysis giving a KRAS wild-type result (tumor subclone below the level of detection) with subsequent KRAS mutation detection with progression and/or metastasis. The patient in our study had an initial KRAS mutation detected with subsequent wild-type KRAS (the opposite of the studies described above). However, this patient did not receive neoadjuvant chemoradiation. The lack of detection with progression from the in situ to invasive tumor could potentially be due to the selection of another (KRAS wild-type) subclone within the cancer, unrelated to neoadjuvant therapy.
This article represents the findings from a pilot study and thereby has limitations. One limitation is the sample size. Larger-scale, prospective studies are needed to confirm these findings. In addition, there were only 3 cases with solid organ metastasis, and most cases were synchronous metastases resected or biopsied at the time of resection of the primary tumor. These factors could limit the potential differences that tumor cells may acquire after metastatic spread. Also, our analysis looked only at selected somatic mutations in rectal adenocarcinoma. Potential differences in the molecular profile outside of these common somatic mutations could exist as well as differences in protein expression, copy number change, and gene translocations. In addition, this study included cases with lower tumor percentages than reported in other studies, which could affect detection of somatic mutations with low allele frequency.
In conclusion, most rectal adenocarcinomas and paired metastases in our study showed an identical mutation profile. Based on the similar mutation profile of primary and metastatic tumors, use of either appears valid for molecular analysis in a clinical setting. However, multiple factors may contribute to differences in the somatic mutation profiles by NGS for rectal adenocarcinoma, and careful histologic review and optimization of tumor percentage is needed. Additionally, similar mutations were found in our current study of rectal adenocarcinoma as have previously been reported for colon adenocarcinoma. A prospective, larger-scale study is needed to substantiate these preliminary findings.
Funding was provided for this work by the Department of Pathology and Genomic Medicine at Houston Methodist Hospital.
The authors have no relevant financial interest in the products or companies described in this article.
Supplemental digital content is available for this article at www.archivesofpathology.org in the June 2016 table of contents.