HD Chromoendoscopy Coupled with DNA Mass Spectrometry Profiling Identifies Somatic Mutations in Microdissected Human Proximal Aberrant Crypt Foci

This article is featured in Highlights of This Issue, p. 813

Screening colonoscopy has been established as an effective strategy to reduce the risk of colorectal cancer (1). However, recent evidence suggests that protection may be most successful in the distal colorectum as the incidence of proximal colon cancer has not been significantly reduced by the implementation of widespread screening colonoscopy (2). The frequency of interval colon cancers, or those cancers developing between screening procedures, raises the possibility that proximal tumors in particular may have accelerated growth characteristics that limits the effectiveness of colonoscopy (3). This lack of protection afforded by colonoscopy, in particular within the proximal colon, underscores the need for advanced endoscopic techniques that may be used to identify small precursor lesions that would otherwise be missed by traditional screening approaches.

The multistage colon cancer model provides a paradigm for understanding the role of early mutations in cancer progression through defined histologic stages (4). Aberrant crypt foci (ACF) are morphologically identifiable mucosal abnormalities, a subset of which may be precancerous and contribute to a “field defect” within the mucosa (5). With the use of dye-spray (e.g., indigo carmine or methylene blue) and high-definition (HD) magnifying colonoscopy, the in situ identification of ACF has become relatively straightforward (6, 7). Although their utility as a surrogate marker of colon cancer has been challenged (8, 9), we recently reported that elevated numbers of distal ACF observed during index colonoscopy predicts the development of advanced neoplasia within a 5-year screening interval (10).

The following study was undertaken to more accurately define the molecular alterations that are present within colonic ACF. Using DNA mass spectrometry (DNA-MS) combined with laser capture microdissection (LCM), we report a highly sensitive method to interrogate the mutational spectrum of microscopic biopsies following their removal from the human colon. A limited number of somatic mutations have been identified, including mutations to APC^R876* and FLT3^I836M, as well as an insertion/deletion within the EGFR gene, underscoring the biologic significance of ACF within the context of colorectal cancer and in particular within the proximal colon.

All subjects underwent a total screening colonoscopy at the University of Connecticut Health Center (UCHC) in accordance with the institutional policies. Patients who met the Amsterdam criteria for familial adenomatous polyposis (FAP) or hereditary nonpolyposis colorectal cancer (HNPCC) were excluded from the study. This study was performed only following the institutional review board approval and receipt of written informed consent from the subjects.

ACF were identified in situ and biopsied from grossly normal-appearing colonic mucosa during HD, close-focus magnifying chromoendoscopy. The proximal colon from the cecum to the right hepatic flexure, in addition to the distal 20-cm of the colorectum, were sprayed with a freshly prepared solution of 1% indigo carmine. ACF were visualized and photographed using a HD colonoscope (Olympus, PCF-190; Olympus Corp.) with visualization from 2 to 100 mm at ×60 magnification. A finding was accepted as an ACF if five or more crypts had an increased lumen diameter (1.5–2×), thick crypt walls, or abnormally shaped lumens relative to the surrounding mucosa. In addition, the lesion must be less than 5 mm in diameter to be considered an ACF. Biopsies were immediately embedded in freezing medium (OCT), flash frozen, and stored at −80°C (11).

Frozen sections were stained with hematoxylin and eosin (H&E) and routine histologic analyses were performed by two independent, board-certified human gastrointestinal pathologists blinded to clinical findings according to our previously established criteria (11). Dysplastic ACF are characterized histologically by enlarged upper cryptal regions of irregular shape with stratified, elongated nuclei and a general dysplastic appearance. Hyperplastic ACF are characterized according to the same criteria applied to hyperplastic polyps (12). These hyperplastic ACF are subclassified into serrated and distended (nonserrated) pathologies as previously described (11). Briefly, serrated ACF are defined as ACF that show stellate luminal shape upon cross-section with a prominent component of columnar crypts with microvessicular cytoplasm. Distended ACF lack serration, prominently feature goblet cells, and frequently exhibit tufting of the surface epithelium.

A Veritas microdissector was used to capture approximately 5,000 cells (or the equivalent of approximately 1 mm² of collected tissue area) from 12-μm thick frozen serial sections of ACF prepared on PEN membrane glass slides (Applied Biosystems). Genomic DNA was extracted from aberrant crypts (isolated from surrounding stroma and normal mucosa by LCM) using the PicoPure DNA extraction kit (Applied Biosystems). Samples were then cleaned using the DNA Wizard protocol (Promega).

Mutation screening was performed at the Yale Center for Genome Analysis using the Sequenom MassARRAY DNA-MS approach. This method uses a matrix-assisted laser desorption ionization-time of flight mass spectrometry (MALDI-TOF MS) platform to detect single-base mutations with increased sensitivity (13). Extracted DNA was amplified using multiplexed OncoCarta PCR primers targeting 105 mutations across 22 known tumor suppressor and proto-oncogenes (v.3.0; Sequenom Inc.; Supplementary Table S1). The DNA concentration per reaction approached 1 ng of genomic DNA (Supplementary Fig. S1), a concentration that is within the limits of detection for the Sequenom platform (13). DNA-MS was carried out as previously described (13). Briefly, target regions are amplified through multiplex PCR reactions. Shrimp alkaline phosphatase is added to the reaction to dephosphorylate unincorporated dNTPs and an extension reaction mix is added. A post-PCR primer extension reaction creates a single-nucleotide extension using mass-modified terminators and the reactions are desalted using 6 mg of CLEAN Resin. Reaction products are then printed onto a SpectroCHIP to be read by MALDI-TOF MS. Mass spectrometry data are then analyzed using Typer4.0 software to identify sample genotype with respect to the assayed mutation.

Somatic mutations were confirmed by extracting DNA from whole blood using the DNAeasy Blood & Tissue Kit standard protocol (Qiagen) and sequenced by GENEWIZ, Inc. To determine mutation status in adenomas, DNA was extracted from two 10-μm FFPE curls using the QIAamp DNA FFPE Tissue Kit standard protocol and sequenced (GENEWIZ, Inc.) Primer sequences for PCR amplification and Sanger sequencing were: EGFR F: 5′-CCCCAGCAATATCAGCCTTA-3′, R: 5′-ATGAGGTACTCGTCGGCATC-3′; APC F: 5′-GCATGCTCAAGAACCTCATTC-3′, R: 5′-TAGGTCGGCTGGGTATTGAC-3′; FLT3 F: 5′-AGAACTGCAGCCACCATAGC-3′, R: 5′-ACCTCATGATCTGCCCTCCT-3′ (Integrated DNA Technologies). Sanger sequencing for BRAF and KRAS mutations was performed as previously described (14).

Immunohistochemistry (IHC) was performed on frozen tissue sections prepared at 7-μm thickness. Sections were treated with 3% hydrogen peroxide, blocked, and incubated in anti-β-catenin primary antibody (1:2,000; Sigma-Aldrich). Sections were incubated in ImmPRESS anti-mouse Ig (Vector Laboratories). Signal detection was achieved using a 3,3′-diaminobenzidine solution (Vector Laboratories) and tissues were counterstained with methyl green nuclear stain.

Using HD chromoendoscopy, we and others have previously demonstrated the ability to identify ACF within the distal colon (6, 7). We have recently expanded our analysis to include the identification of ACF within the proximal colon, a region of the colon that includes the cecum to the splenic flexure (Fig. 1A). A total of 96 patients have been screened and 88 proximal ACF and 1,010 distal ACF have been identified (Fig. 1B). The majority of the subjects were undergoing a screening colonoscopy (n = 70 of 96; 73%) and approximately half of the subjects (n = 47 of 96; 49%) had at least one polyp identified and removed during the procedure. Histologic evaluation reveals that proximal ACF harbor the full spectrum of established histologic abnormalities present within distal ACF, including hyperplasia (serrated or distended) and dysplasia (Fig. 1C).

ACF within the distal colon share many genetic and histologic abnormalities associated with more advanced colonic neoplasia. For example, hyperplastic ACF commonly harbor BRAF or KRAS mutations (14), resulting in increased ERK activation (11). The presence of ACF with dysplastic histologic features, however, is uncommon within the distal colon, although in one case a dysplastic ACF was found to carry a novel somatic mutation to the APC gene (14). A high frequency of APC mutations has been reported previously in dysplastic ACF (15), but the inclusion of patients with FAP disease in this study may have exaggerated the prevalence of APC mutations. In an earlier study, we found that, a significant percentage of ACF (∼40%) does not harbor mutations to BRAF or KRAS (11). Thus, in the present study, a broader genetic screen was undertaken to identify additional mutations that may contribute to early neoplastic changes in the colon.

Comprehensive mutation profiling of early neoplasia may provide insight into future colon cancer risk (16). However, there are a number of technical challenges associated with the application of high-throughput screening strategies to the analysis of small mucosal biopsy specimens. In addition, the number of morphologically abnormal colonic crypts within an ACF biopsy is limited, necessitating the use of LCM to enrich for subpopulations of aberrant and normal-appearing colonocytes. Recent advances in the DNA-MS technology using MALDI-TOF have provided a high-throughput, multiplexed approach to allow somatic mutational profiling with as little as 1 ng of input genomic DNA (13). To examine the feasibility of performing DNA-MS analyses on microdissected ACF samples, DNA was extracted from approximately 5,000 colonocytes and subjected to the MassARRAY PCR standard protocol. Amplification of genomic loci was achieved with less than 1 ng of input genomic DNA (Supplementary Fig. S1). A total of 26 microdissected ACF were then subjected to DNA-MS analyses. This group included seven proximal ACF, 19 distal ACF, and four normal colon biopsies.

As shown in Fig. 2A, a proximal, serrated hyperplastic ACF was positive for the in-frame insertion/deletion, c.2239_2251delTTAAGAGAAGCAAinsC (AA:L747_ 751> P), in the EGFR gene (19.6% deletion: 80.4% wild-type) as detected by MassARRAY. This mutation was not present in the subject's blood, confirming that it was somatically acquired. Although EGFR mutations occur in approximately 6% of colorectal cancers (COSMIC; n = 3,102), they are much more frequent in lung adenocarcinomas (25.2%; COSMIC; n = 26,293). This specific mutation has not been previously detected in colorectal cancer and accounts for <1% of lung EGFR mutations, but is associated with a gain of EGFR function (17).

In addition, a distal, distended hyperplastic ACF was positive for a FLT3 c.2508C>G (I836M) mutation (24.0% mutant: 76.0% wild-type) and confirmed by Sanger sequencing to be somatic (Fig. 2B). FLT3 mutations have widely been implicated in cases of acute myeloid leukemia (23.5%; COSMIC; n = 63,213) and mutations to FLT3 occur in 4.8% of colorectal cancers (COSMIC; n = 1,003). To our knowledge, the c.2508C>G mutation found in this study has never before been reported in colorectal cancer. The I836M missense mutation has been shown previously to cause constitutive activation of the receptor tyrosine kinase and subsequent downstream signaling (18). Constitutive activation of these two receptor tyrosine kinases, not previously implicated in ACF formation, may contribute to the aberrant activation of MAPK signaling reported earlier in hyperplastic lesions (11).

Although the commercially available panel used in this study does not include KRAS or BRAF, we have previously reported a high frequency of mutations to these proto-oncogenes in distal hyperplastic ACF using conventional PCR-based sequencing (11, 14). Thus, to establish the presence of BRAF^V600E and KRAS (codons 12 and 13) mutations, Sanger sequencing was performed on seven randomly selected ACFs that had no other somatic mutations detected by the OncoCarta panel. Four of the seven ACFs were positive for KRAS mutations and none of the ACFs had BRAF^V600E mutations.

An APC^R876* mutation, a c.2626C>T (30.4% mutant T: 69.6% wild-type C), was identified in a proximal dysplastic ACF (Fig. 3A). To determine the function of this rare mutation, immunolocalization of the Wnt target protein, β-catenin, was performed on the dysplastic ACF. As shown in Fig. 3B, there was an increase in nuclear accumulation of β-catenin in the dysplastic colonocytes compared with adjacent normal crypts. Although mutations to APC are frequently detected in colorectal cancers (42%; COSMIC; n = 3,982), the APC^R876* mutation is relatively rare, accounting for approximately 2.5% of all APC mutations. The c.2626C>T nonsense mutation is most commonly somatically acquired (19), although it has occasionally been detected in FAP pedigrees (19–21). In the present study, however, this mutation was confirmed to be somatic, as it was not detected in a matched blood sample (Fig. 3C), thereby ruling out the possibility of underlying FAP disease. Interestingly, this mutation was not detected in any of the three synchronous tubular adenomas removed during the chromoendoscopy procedure (Fig. 3C).

Sporadic forms of colorectal cancer are most likely the result of early somatic mutations resulting in the clonal expansion of initiated progenitor cells. ACF have been shown to be monoclonal and as such are likely the direct consequence of this initial clonal expansion (22). Polyps and adenomas are presumed to arise from these lesions following the accumulation of multiple genetic alterations within this initiated cell population. Clearly, additional somatic mutations must occur to drive neoplastic progression (4). Additional changes have been identified in early stages of colorectal cancer that contribute to carcinogenesis, including the acquisition of aberrant promoter methylation (7) and microsatellite instability (23). Therefore, it is reasonable to anticipate that only a small percentage of ACF may sustain multiple oncogenic mutations at this early precancerous stage. In fact, in the present study there was no evidence for the presence of multiple somatic mutations, at least within the cancer panel that was tested.

The in situ identification of colonic ACF, and in particular those from the proximal colon, is dependent upon the use of dye-spray and HD magnifying endoscopy, and we believe that in some cases this additional clinical effort is warranted. For example, the identification of the APC mutation, APC^R876*, associated with the formation of aggressive, invasive carcinomas (24) in a commonly missed mucosal lesion underscores the potential preventive benefit of implementing this clinical approach to larger patient populations. It is possible that ACF that harbor this and other significant somatic mutations have the capacity to rapidly progress to a more advanced neoplasia, perhaps representing a subset of “missed lesions” that may be responsible for interval colon cancers. Expanded molecular classification of proximal ACF with a DNA-MS panel designed specifically for colorectal cancer, as well as establishing detailed associations of proximal ACF with known colorectal cancer risk factors, will be necessary to firmly establish their premalignant potential and utility as a surrogate marker for future cancer risk.

No potential conflicts of interest were disclosed.

Conception and design: D.A. Drew, D.W. Rosenberg

Development of methodology: D.A. Drew, D.W. Rosenberg

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): D.A. Drew, T.J. Devers, M.J. O'Brien, N.A. Horelik, J. Levine, D.W. Rosenberg

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): D.A. Drew, T.J. Devers, D.W. Rosenberg

Writing, review, and/or revision of the manuscript: D.A. Drew, T.J. Devers, D.W. Rosenberg

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): D.A. Drew, N.A. Horelik, D.W. Rosenberg

Study supervision: T.J. Devers

The authors thank Dr. T.V. Rajan for providing a detailed pathologic analysis of ACF. The authors also thank Dr. Shi Yang at Boston Medical Center for performing KRAS and BRAF sequencing services.

This work was supported by the State of Connecticut Department of Public Health, Biomedical Research Application #2012-0913 (to D.W. Rosenberg) and the NIH 1RO1CA159976 (to D.W. Rosenberg).