Genetic Pathways in the Evolution of Morphologically Distinct Colorectal Neoplasms¹

Colorectal cancer requires genomic instability for its development. More than four-fifths of colorectal cancer evolve as a consequence of chromosomal instability, a poorly understood form of genomic instability originally proposed by Fearon and Vogelstein (1) that results in chromosomal aberrations and loss and gain of genetic material. A progression of loss of wild-type tumor suppressor genes has been consistently identified in colorectal cancer, involving the APC gene (chromosome 5q), a chromosome 18q allele at or near the DCC/SMAD4 locus, and the p53 gene (chromosome 17p) (2, 3). Additionally, mutational activation of K-RAS has been found in >50% of adenomas and colorectal cancers (2, 4, 5). These accumulated genetic changes permit escape from the growth restraints of tumor suppressors, and accelerate growth (6). A second form of genomic instability that is operative in hereditary nonpolyposis colorectal cancer and one-fifth of all sporadic cancers involves inactivation of the DNA mismatch repair (MMR)³ system (7). Among sporadic colorectal cancers with microsatellite instability (MSI), a phenotypic marker for inactivation of DNA MMR, the predominant alteration is the epigenetic inactivation of the DNA MMR gene hMLH1 (chromosome 3p) by promoter hypermethylation (8, 9). Among hereditary nonpolyposis colorectal cancer patients, germ-line mutation of the DNA MMR genes hMSH2 (chromosome 2p) and hMLH1 are most commonly detected (7). Unlike tumors with chromosomal instability, microsatellite-unstable tumors have rare loss and gain of genetic material. Instead, bi-allelic mutational inactivation occurs within exons of genes that have a microsatellite sequence present (10, 11, 12). The complement of tumor suppressor genes inactivated in tumors with MSI is different when compared to tumors with chromosomal instability, but overlapping pathways that are affected in both forms of genomic instability might be present.

It is not clear how these two different pathways of tumor development translate into the morphological heterogeneity observed among the pathological lesions in the colon. Most small polyps in the colon are adenomas, and the likelihood of finding cancer increases with the size of the neoplasm (13, 14, 15). It is currently thought that most colorectal cancers arise from preexisting adenomas, although some small proportion may emerge de novo, i.e., in the absence of identifiable preexisting adenoma (16, 17, 18, 19, 20, 21, 22). It is now recognized that some adenomas are not exophytic and are indeed flat, with the height of the adenoma not greater than twice the thickness of the adjacent normal mucosa (23, 24, 25, 26, 27, 28). Similarly, colorectal carcinomas may be morphologically classified either as polypoid or flat. Most of the polypoid cancers have adenomatous elements on the periphery when found at an early stage, leading to the suggestion that these have arisen from preexisting adenomas. It has been suggested that the less common flat-type cancers correspond to de novo tumors, which contain no observable adenomatous component (23, 24, 25, 26, 27, 28). Flat colorectal lesions are more difficult to identify by the usual diagnostic techniques (29), and if these develop through distinct genetic pathways, it might be of importance as genetic diagnostic approaches are developed to screen for colorectal neoplasia.

In this study, we originally hypothesized that flat and exophytic adenomas might be the precursor lesions to cancer arising de novo and polypoid cancers, respectively. To test this hypothesis, we examined the genetic makeup of these tumors in the context of the two understood pathways for colorectal cancer development.

A total of 94 sporadic colorectal tumors, including 34 neoplastic polyps and 60 cancers, were investigated. The 34 neoplastic polyps (including 15 small exophytic adenomas, 6 large exophytic adenomas, and 13 small flat adenomas) were collected from the Karolinska Hospital (Stockholm, Sweden). The size of small adenomas was ≤10 mm in diameter. Flat adenomas spread superficially, and the height of a flat adenoma is not greater than twice the thickness of the adjacent normal mucosa. Twenty-five early flat-type cancers arising de novo, namely, cancer without adenomatous components, were collected from the University of Tsukuba (Tsukuba City, Japan). An additional 25 early polypoid-type cancers with adenoma and 10 advanced cancers were used for comparison from tissue archives in the laboratory. “Early cancer” indicates cancer invasion limited to the submucosa, and “advanced cancer” indicates cancer invasion into the muscularis propria or serosa. “Flat-type cancer” indicates cancer in which lesions were flat against the surrounding mucosa or minimally elevated with a depression. The protruding lesions (sessile, semipedunculated, or pedunculated) were classified as “polypoid-type carcinoma.” The histopathological diagnosis was determined with reference to the WHO classification (30). All dysplastic specimens were excised endoscopically. The cancers were obtained either surgically or endoscopically.

DNA was extracted from formalin-fixed, paraffin-embedded tissues sections. One section was stained with H&E. This reference slide was used to select areas for microdissection using a sterile scalpel blade under a dissecting microscope. Genomic DNA was isolated from the paraffin-embedded microdomains removed from the slides as previously described (3, 31). Normal tissue was obtained from histologically normal mucosa and/or normal lymph nodes.

The DNA was amplified by a heminested PCR protocol. PCR amplification of exon 1 of a K-RAS containing codons 12 and 13 was first performed by the following primers: forward 5′-CGTCCACAAAATGATTCTGAATTAGCTGTATC-3′ and reverse 5′-CCTTATGTGTGACATGTTCTAATATAGTCAC-3′. Thirty-five cycles (92°C for 30 s and 67°C for 30 s) were performed, followed by a 10-min extension at 72°C. Initial PCR products were diluted and further amplified with a new forward primer, 5′-AGGCCTGCTGAAAATGAC-3′, and the same reverse primer described above. Thirty-five cycles (92°C for 25 s, 55°C 25 s, and 72°C for 25 s) were performed followed by a 10-min extension at 72°C. The 104-bp amplicons were then dot blotted onto nylon filters (Hybond-N; Amersham, Buchinghamshire, United Kingdom) and hybridized with radiolabeled oligomer primers representing all possible mutations at codon 12 and the GAC mutation of codon 13. Plaque hybridization and direct sequencing were performed to confirm the presence of K-RAS mutations at codons 12 and 13 that were detected by dot blot hybridization. Briefly, PCR products containing K-RAS mutations were cloned into a plasmid vector, pUC18. After amplification, the plasmid was plated and lifted on nylon filters. The plaque lifts were hybridized with the oligomer probes as described above. The subcloned plasmid was then analyzed by direct sequencing.

Ten sets of polymorphic microsatellite sequences that are tightly linked to known tumor suppressor genes and DNA MMR genes were used. DNA was amplified by PCR using ³²P-end-labeled primers at microsatellite loci linked to the hMSH2 locus on 2p16 (CA-5 and D2S123) (32, 33), hMLH1 locus on 3p23-21.3 (D3S1611 and D3S1561) (34), APC/MCC locus on 5q21 (D5S107 and D5S346) (3, 35), p53 locus on 17p13 (D17S513 and p53 intron 1) (36, 37), and DCC/SMAD4 locus on 18q21.3 (D18S35A and 18qDCC-TA) (31, 38). PCR was carried out in a total volume of 5 μl containing 1× PCR buffer (Life Technologies, Inc., Gaithersburg, MD), 0.125 pmol of a labeled primer, 0.125 pmol of the other (nonlabeled) primer, 3 μl of template DNA, 0.25 units of Taq DNA polymerase (Life Technologies, Inc.), and 200 μm of each deoxynucleotide (Life Technologies, Inc.). Thirty-four cycles of 93°C for 1 min, 55°C for 2 min, and 72°C for 2 min were performed with an initial denaturation step of 94°C for 2 min, and a final extension step of 72°C for 10 min. The PCR products were analyzed on an 8% polyacrylamide gel containing 7.5 M urea and 0.5× Tris-borate-EDTA buffer. The PCR was repeated more than twice to ensure that the results were reproducible in each case. Assessment of LOH was assigned when a tumor allele showed at least a 50% reduction in the relative intensity of one allele in neoplastic tissue compared with the matched normal DNA.

We used the stringent criteria for determination of MSI from a National Cancer Institute Workshop (39). MSI-H was defined by a novel bandshift or allele at ≥30% of microsatellite loci tested when compared to non-neoplastic tissue from the same patient. Our assay to determine MSI was based on PCR amplification of a panel of 10 microsatellite primer pairs linked to tumor suppressor genes listed above. Samples positive for MSI were not included as informative at that locus in the LOH analyses.

Statistical analysis was performed using the χ² test, Fisher’s exact test, or Student’s t test. A P < 0.05 was considered to be statistically significant.

The DNA encompassing codons 12 and 13 of the K-RAS proto-oncogene was successfully amplified from all microdissected tissue. Forty-five of 94 (48%) colorectal neoplasms had mutations of K-RAS at codon 12 and/or codon 13 (Tables 1 and 2). The mutations were found in 38% (13 of 34) of adenomas and 53% (32 of 60) of cancers (Table 3). No significant difference was found in the K-RAS mutation frequency between polyps and cancers. On the other hand, only 15% (2 of 13) of flat adenomas had K-RAS mutations, whereas 52% (11 of 21) of exophytic adenomas had the mutations (P < 0.03, Fisher’s exact test). Interestingly, bi-allelic mutations of K-RAS were found in one-third (5 of 15) of exophytic adenomas, but not at all in flat adenomas (Table 4). The frequency of K-RAS mutations in early colon cancer with adenoma (68%) was significantly higher than that in early colon cancer de novo (28%; P < 0.01, χ² test; Table 3). The frequency of bi-allelic mutations in early colon cancer with adenoma (28%) was also higher than that of early colon cancer de novo (8%). There was no significant difference in the tumor size between early cancer arising de novo and cancer with adenoma (Table 4).

Twenty-six of 94 (28%) tumors had a single K-RAS allele mutation and 17 of 94 (18%) tumors had bi-allelic mutations (Table 4). Two cases had three and four mutated alleles, patient 57C and 35C, respectively (see Table 2). These multiple mutations were studied by using PCR followed by hybridization of plasmid plaques with oligonucleotide probes specific for each mutation (Fig. 1,A). Mutations detected after cloning and sequencing (Fig. 1,B) were identical to those found by the dot-blotting hybridization in these two cases. Overall, the most frequent mutational pattern found was a codon 12 GGT (glycine) to a codon 12 GAT (aspartic acid) transition (32% of a total of 73 mutations identified), followed by codon 12 AGT (serine) transition (19% of mutations identified). In 11 of the 73 (15%) mutations, K-RAS was mutated to codon 12 GTT (valine) and in 10 (14%), K-RAS was mutated at codon 13 to GAC (aspartic acid). The proportion of the mutational pattern was similar between benign and malignant tumors (Table 4).

Fig. 2 demonstrates typical examples of the LOH (Fig. 2,A) and MSI (Fig. 2,B) at the hMSH2 locus on chromosome 2p, hMLH1 locus on 3p, APC/MCC locus on 5q, p53 locus on 17p, and DCC/SMAD4 locus on 18q. The informativity and LOH results for all genes studied are individualized in Tables 1 and 2 and summarized in Table 3. LOH of chromosome 2p was found in 4 of 52 (8%) informative colon cancers, but not for any polyps. No significant difference was found in the frequency of LOH at 2p between polyps and cancers. LOH at chromosome 3p was found in 3 of 28 (11%) informative adenomas and 11 of 48 (23%) informative cancers. No significant difference was found in LOH at 3p between benign tumors and malignant tumors; however, the frequency of 3p LOH in early cancers arising de novo (33%) was significantly higher than that in early cancer with adenoma (5%; P < 0.03, Fisher’s exact test). LOH of chromosome 5q was found in 9 of 33 (27%) informative adenomas and 20 of 50 (40%) informative cancers. This frequency of 5q LOH is consistent with assessments of 5q LOH in cancers by our group and others (2, 3). No significant difference was found in the frequency of LOH at 5q between polyps and cancers. No LOH at 17p (0/32) was found in any informative benign cases similar to previous observations (2, 3), whereas 24 of 51 (47%) informative colorectal cancers had LOH at 17p (P < 0.0001, χ² test). The frequency of LOH at 17p in early cancer de novo (44%) was similar to that in early cancer with adenoma (46%). No significant difference was found in depth of invasion, differentiation, and location of the cancer. LOH at 18q was found in 4 of 32 (16%) and 30 of 59 (51%) informative adenomas and cancers, respectively. The incidence of LOH at 18q in cancer was significantly higher than that in polyps (P = 0.0003; Table 3). This frequency of 18q LOH is similar to previous observations (2). The frequency of cancer with LOH at any locus (48 of 60) was significantly higher than that in polyps (13 of 33; P < 0.0001).

Using the stringent criteria for determining MSI-H from a recent National Cancer Institute Workshop (39), none of the sporadic adenomas, whether flat or exophytic, demonstrated MSI-H (Table 1). We identified five cancers from the 60 analyzed (8.3%) that had MSI-H. Four of the MSI-H cancers were de novo, and one was an advanced cancer (Table 2). None of the early cancers with adenoma were MSI-H. There was no significant difference in the incidence of MSI-H between adenomas and cancers. There was a trend toward significance for MSI-H occurrence in de novo cancers compared with early cancers with adenoma (P = 0.055, Fisher’s exact test; Table 3). Additionally, there was a trend toward significance for MSI-H occurrence in well-differentiated cancers of any type (P = 0.052, Fisher’s exact test; Table 4). None of the MSI-H de novo cancers had LOH at 17p, but three of the four had LOH at 3p (Table 2).

Microsatellite instability-low (MSI-L), defined as the occurrence of a novel allele in <30% but >0% of markers tested when compared to normal tissue (39), was identified in 6 of 13 flat adenomas and 8 of 21 exophytic adenomas. Additionally, MSI-L tumors, which are reported to be similar in clinical and histological characteristics to microsatellite-stable tumors, were found in 13 of 25 de novo cancers, 9 of 25 of early cancers with adenoma, and 4 of 10 advanced cancers. There was no statistical difference in MSI-L occurrence between adenomas and cancers or between the subtypes of adenomas or subtypes of cancers.

We found no significant difference in K-RAS mutation frequency between benign polyps and malignant tumors, which suggests that K-RAS mutations are relatively early events in colorectal tumorigenesis, as previously described (1, 2). On the other hand, exophytic adenomas and cancers with adenoma had a significantly higher K-RAS mutation frequency than flat adenomas and de novo cancers, respectively, despite the similarity in overall tumor size. This finding suggests that K-RAS mutations are correlated with the morphological growth pattern of colorectal neoplasms and that exophytic adenomas may acquire these mutations at an early stage. The association of mutant K-RAS with an exophytic neoplasm seems to require simultaneous inactivation of the APC gene or its pathway. Mutant K-RAS has been observed in nondysplastic aberrant crypt foci, many of which may never progress or grow (40, 41). Thus, a solitary activating mutation in K-RAS may not be sufficient to sustain neoplastic growth.

Our data confirm the previous observation that codon 12 mutations of K-RAS occur more frequently in exophytic tumors compared to endophytic (nonpolypoid) tumors (40, 41, 42, 43, 44). The K-RAS gene encodes a 21-kDa guanosine triphosphatase protein (p21^ras) which mediates signaling events regulating cell proliferation, and a single mutationally activated p21^ras protein continuously induces cell proliferation (6). Increased cell growth activity induced by activated K-RAS mutations in the context of multistep carcinogenesis appears to be necessary for exophytic growth in colorectal neoplasia (40, 41, 44). Consistent with this concept, flat dysplastic lesions found in inflammatory bowel disease have a low rate of K-RAS mutations (45, 46, 47, 48) and rarely have mutations in APC (45, 49, 50, 51).

Bi-allelic mutations of K-RAS were present in exophytic tumors more frequently than flat tumors, which suggests that different genetically determined clonal expansions developed within the exophytic neoplasm. In two cases, three and four mutations of K-RAS were detected. Exposure to environmental mutagens can cause multiple mutations of a single cancer-related gene (52), each of which might contribute to separate clonal expansions in exophytic tumors. We suggest that each K-RAS mutation originates from different evolving subclones and that each may be detected with differing frequencies from a single tissue sample of an exophytic tumor. In this study, we examined neoplasms in which a potentially dominant or aggressive subclone might have not overtaken the entire structure of the tumor, thus allowing detection of the clonal heterogeneity. In other tumors, either one clone may have developed or sampling error at the time of microdissection may have missed other clones.

The frequency of colorectal cancers with allelic loss events was significantly higher than that of benign tumors, whereas the frequency of K-RAS mutations and the occurrence of MSI were not significantly different between cancers and benign tumors. Forty-seven percent (24 of 51) of cancers had LOH at 17p, while LOH at 17p was completely absent in all benign tumors. The incidence of LOH at 18q in cancers (51%) was also significantly higher than that in benign tumors (13%). There was no significant difference in LOH occurrence at chromosome loci 2p, 3p, and 5q between benign tumors and malignant tumors. These findings suggest that LOH at 17p and 18q are important events for malignant transformation of a benign lesion. In particular, our data are consistent with previous data suggesting that LOH at 17p is necessary for a benign lesion to become a malignant neoplasm (3).

LOH at chromosome 3p was significantly associated with cancers arising de novo. It is of interest that of the four cases of cancer arising de novo with MSI-H, three were associated with LOH at 3p (Table 2). However, the three other cancers arising de novo with LOH at 3p were MSI-L, which behaves like microsatellite-stable tumors clinically and are not associated with inactivation of the DNA MMR system. There are several genes on chromosome 3 that are involved in the pathogenesis of colorectal cancer, including hMLH1 and transforming growth factor β receptor II. One, both, or neither of these genes might play a role in cancers arising de novo.

We found no significant difference in the incidence of MSI between benign and malignant lesions. Five of 60 cancers (8.3%) demonstrated MSI-H, which is slightly lower than the frequency seen in previous studies (53, 54). This may be due to selection of these types of tumors which may have a different pathogenesis, or that these lesions may be early enough in their pathogenesis that sufficient mutations at microsatellite loci are not frequent enough for detection. Sporadic adenomas have a much lower frequency of MSI than their exophytic malignant counterparts (3% versus 15–20%) (55), and de novo cancers and early cancers with adenoma may be more similar to adenomas in this regard. Some MSI-H tumors demonstrated LOH at chromosome 3p, the location of hMLH1, while no LOH occurred at chromosome 2p, the location of hMSH2. LOH at chromosome 3p could explain the onset of MSI-H as an alternative mechanism to hypermethylation of the hMLH1 promoter, the predominant mechanism in sporadic MSH-H cancers (8, 9). Although MSI can be disguised as LOH electrophoretically, this would not be expected at both chromosome 3p loci analyzed. Furthermore, none of the de novo colorectal cancers with MSI-H had allelic loss at the p53 locus, the most important determinant of malignant onset, which suggests that these cancers with MSI develop in the absence of loss of the p53 tumor suppressor locus. This finding is consistent with a previous study that p53 gene mutations were detected less frequently from MSI tumors (56).

We propose the following genetic and morphological correlative scheme. Most cancers (80–85%) arise through LOH mechanisms. LOH at chromosome 5q may be required for both exophytic and flat adenoma development. K-RAS mutations (in association with inactivation of APC on chromosome 5q) are necessary for exophytic adenoma development and are essentially absent from flat adenomas. All tumors in this pathway subsequently have LOH events at both chromosome 18q and 17p, mediating malignant transformation. In the case of exophytic polyps, the result is a polypoid cancer. In the case of a flat adenoma, the result may be a cancer arising de novo, which overgrows the flat adenoma to the point of no recognizable remaining adenomatous tissue. Cancers arising de novo may additionally develop from flat adenomas with LOH of chromosome 3p, which might progress through other LOH events at 18q and 17p, or may progress through MSI mechanisms without LOH at 18q or 17p. The MSI mechanism, responsible for about 15% of colorectal tumor formation, is operative in some polypoid cancers as well. Because of the minimal overlap seen with MSI tumors and LOH at 17p in this study and others (56), it is likely that these two mechanisms are independent pathways for cancer development.

In summary, de novo colon cancers (i.e., flat cancers without any adenomatous component) developed with a unique pattern of genetic abnormalities that may include LOH of chromosome 3p in some cases, but excludes mutational activation of K-RAS. The lack of mutated K-RAS mutation is also seen in flat adenomas. It is possible that some flat adenomas may progress to de novo cancers with LOH of chromosome 3p. Our study indicates that colorectal neoplasms may develop along different genetic pathways that may determine the morphological variations observed in the human colon.