|Home | About | Journals | Submit | Contact Us | Français|
Adjuvant chemotherapy improves survival among patients with stage III colon cancer, but no reliable molecular predictors of outcome have been identified.
We evaluated loss of chromosomal material (also called loss of heterozygosity or allelic loss) from chromosomes 18q, 17p, and 8p; cellular levels of p53 and p21WAF1/CIP1 proteins; and microsatellite instability as molecular markers. We analyzed tumor tissue from 460 patients with stage III and high-risk stage II colon cancer who had been treated with various combinations of adjuvant fluorouracil, leucovorin, and levamisole to determine the ability of these markers to predict survival.
Loss of heterozygosity at 18q was present in 155 of 319 cancers (49 percent). High levels of microsatellite instability were found in 62 of 298 tumors (21 percent), and 38 of these 62 tumors (61 percent) had a mutation of the gene for the type II receptor for transforming growth factor β1 (TGF-β1). Among patients with microsatellite-stable stage III cancer, five-year overall survival after fluorouracil-based chemotherapy was 74 percent in those whose cancer retained 18q alleles and 50 percent in those with loss of 18q alleles (relative risk of death with loss at 18q, 2.75; 95 percent confidence interval, 1.34 to 5.65; P=0.006). The five-year survival rate among patients whose cancer had high levels of microsatellite instability was 74 percent in the presence of a mutated gene for the type II receptor for TGF-β1 and 46 percent if the tumor did not have this mutation (relative risk of death, 2.90; 95 percent confidence interval, 1.14 to 7.35; P=0.03).
Retention of 18q alleles in microsatellite-stable cancers and mutation of the gene for the type II receptor for TGF-β1 in cancers with high levels of microsatellite instability point to a favorable outcome after adjuvant chemotherapy with fluorouracil-based regimens for stage III colon cancer.
Colorectal cancer is the second most common cause of death from cancer in the United States.1 Postoperative adjuvant chemotherapy improves the outcome in stage III (Dukes’ stage C) colon cancer and is now widely accepted as standard therapy.2,3 Many patients with stage II (Dukes’ stage B) disease are considered to be at high risk for recurrence and receive adjuvant therapy, although its benefit in such cases is uncertain. Markers that reliably predict survival are needed.2,4,5
The sequence of genetic alterations leading to colorectal cancer usually begins with the inactivation of the pathway involving the adenomatous polyposis coli tumor-suppressor gene and β-catenin.6,7 Subsequent changes often include the loss of portions of chromosomes, termed loss of heterozygosity or allelic loss, or the loss of whole chromosomes.8 In about 15 percent of cases of sporadic colorectal cancers, there are insertions or deletions of nucleotides within repeated sequences of DNA, termed microsatellite instability, due to defective repair of mismatched nucleotides.9–11 Tumors with high levels of microsatellite instability are characteristic of the hereditary non-polyposis colorectal cancer syndrome, but in most cases such tumors are sporadic.11–13 Neoplasms with high levels of microsatellite instability accumulate mutations in microsatellites within the coding region of certain genes,10,11,14 but loss of chromosomes is rare.8
Some of these genetic alterations are prognostic markers in colorectal cancer.15 Loss of heterozygosity at chromosome 18q indicates a poor prognosis.15–24 Other alterations that have been found to have prognostic value are allelic loss at chromosomes 17p, 1p, 3p, 4p, 5q, or 8p; changes in the levels of certain gene products, including the DCC (deleted in colorectal cancer) protein, p53, and p27Kip1; mutation of the ras gene; and increased expression of genes involved in fluoropyrimidine metabolism.15,25 In addition, colorectal cancers with high levels of microsatellite instability metastasize less often and have a better prognosis than microsatellite-stable cancers.15,26–33
Molecular alterations also have the potential to predict survival after chemotherapy.34–38 We examined a panel of molecular markers (listed in the Glossary) in specimens of colon cancer from patients enrolled by the Eastern Cooperative Oncology Group in two National Cancer Institute Gastrointestinal Intergroup clinical trials of adjuvant chemotherapy with fluorouracil-based regimens.
|Marker||Abnormality or Abnormal Gene*||Functions of Wild-Type Gene Product||Reported Prognostic or Predictive Value in Colorectal Cancer|
|Loss of heterozygosity at chromosome 18q||DCC||Netrin-1 receptor; caspase substrate in apoptosis; cell adhesion molecule; tumor suppression||Adverse prognostic marker|
|Smad4 (DPC4, MADH4)||Nuclear transcription factor in TGF-β1 signaling; regulation of angiogenesis; activator of WAF1 promoter; downstream mediator of Smad2; tumor suppression||Adverse prognostic marker|
|Smad2 (MADH2, JV18)||Endodermal differentiation; interacts with SKI protein||None|
|Loss of heterozygosity at chromosome 17p||p53 (TP53)||Transcription factor; regulator of cell-cycle progression after cellular stress, of apoptosis, of gene expression, and of DNA repair; tumor suppression||Adverse prognostic marker|
|Loss of heterozygosity at chromosome 8p||Unknown||Unknown||Adverse prognostic marker|
|High labeling index for p53 protein||p53 (TP53)||Same as for loss of heterozygosity at chromosome 17p||Adverse prognostic marker, adverse predictive marker|
|Increased labeling index for p21WAF1/CIP1 protein||WAF1 (CIP1, SDI1)||Cyclin-dependent kinase inhibitor; controller of cell-cycle progression||Favorable predictive marker|
|Microsatellite instability||Consequence of abnormal genes in mismatch-repair family||Repair of nucleotide mismatches||Favorable prognostic marker, favorable predictive marker|
|Mutation in gene for type II receptor for TGF-β1||TGF-β1 RII||Receptor for signaling in TGF-β1 pathway; inhibitor of colonic epithelial proliferation||None|
|Mutation in BAX gene||BAX||Proapoptosis||Adverse prognostic marker|
Specimens from 516 eligible patients enrolled in two randomized trials of adjuvant chemotherapy for colon carcinoma were studied. These patients had stage III cancer (Dukes’ stage C, with lymph-node metastasis) or high-risk stage II cancer (Dukes’ stage B2, with colonic obstruction, adherence to or invasion of adjacent organs, or tumor perforation and with en bloc resection of all visible disease, including regional peritoneal metastases). In one trial (Eastern Cooperative Oncology Group protocol E2284 [National Cancer Institute Gastrointestinal Intergroup INT 0035]), three treatments — fluorouracil plus levamisole, levamisole alone, and surgery alone — were compared.39 Patients in this trial were enrolled from February 1985 to October 1987, and the median duration of follow-up in surviving patients was 9.0 years. In the other trial (protocol E2288 [INT 0089]), four treatments — low-dose leucovorin plus fluorouracil, high-dose leucovorin plus fluorouracil, levamisole plus fluorouracil, and low-dose leucovorin plus fluorouracil plus levamisole — were compared.40 Patients were enrolled from August 1988 to July 1992, and the median duration of follow-up in surviving patients was 4.8 years.
Thus, of the seven cohorts in these two trials, five received fluorouracil-based chemotherapy. We studied survival in relation to the presence or absence of molecular markers in available tumor specimens from 460 patients in the five cohorts that received fluorouracil. There were no statistically significant differences in outcome among the patients in these five cohorts. The study population was representative of all the patients in the two clinical trials; the only statistically significant difference between patients included in this study and those not included was the frequency of regional metastases (6 percent vs. 9 percent, P=0.04) (Table 1).
Formalin-fixed, paraffin-embedded specimens were obtained through the Eastern Cooperative Oncology Group Pathology Coordinating Office. A data base of information on the patients was maintained at the Eastern Cooperative Oncology Group Statistical Center. Laboratory analysis of tumor specimens was performed without knowledge of the patients’ clinical data. Microdissection of tumor tissue and of non-neoplastic control tissue, when available, and preparation of DNA were performed as previously described.14,16,32,41
The microsatellite analysis depended on the type of tissue available (Fig. 1). In the case of 298 tumor specimens for which non-neoplastic control tissue was also available, allelic losses from chromosomes 18q, 17p, and 8p were evaluated with polymorphic markers, and microsatellite instability was determined with eight dinucleotide and two mononucleotide markers (additional information is available with the full text of the article at www.nejm.org).14,16,32,42 The 218 tumor specimens for which insufficient control tissue was available were tested for microsatellite instability with two mononucleotide markers that are rarely polymorphic and that do not require control tissue for evaluation. The mononucleotide repeat in the BAX gene14 was analyzed in cancers with high levels of microsatellite instability.
Allelic loss was determined by examining autoradiographs of DNA amplified by the polymerase chain reaction (PCR). Allelic loss was defined as a reduction in the intensity of the autoradiograph of one of the two alleles in the amplified tumor DNA to at least 50 percent of the level of DNA in non-neoplastic control tissue.14,16,42 Of the 298 cases for which control DNA was available, the results for 18q were interpretable in 279 (94 percent), for 17p in 223 (75 percent), and for 8p in 201 (67 percent). Tumors with high levels of microsatellite instability infrequently lose chromosomes, including 18q, 17p, and 8p,8,14 and were therefore categorized as having no allelic loss; these consisted of 40 additional tumors with no loss at 18q, 102 with no loss at 17p, and 102 with no loss at 8p.
Immunohistochemical analysis was performed as previously described.43,44 Three categories of p53 staining were defined for statistical analysis with the use of labeling-index cutoff points of 40 percent and 5 percent after quantitation by computer-assisted image analysis.43 After initial evaluation of immunohistochemical results for p21WAF1/CIP1, the labeling index was estimated and categorized as greater than 30 percent, 20 to 30 percent, 10 to 19 percent, 5 to 9 percent, and less than 5 percent of nuclei. Of the 516 analyzed tumors, results were interpretable for 445 (86 percent).
Changes in the electrophoretic mobility of DNA amplified by PCR were used to assess microsatellite instability.14,16,32,42 The number of markers with altered allelic sizes and the number of technically satisfactory markers were recorded for each tumor. In evaluating the 298 carcinomas for which control DNA was available, we ascertained microsatellite instability with the use of the interpretable markers among eight dinucleotide markers and two polyadenine tracts. Tumors with a shift in at least two markers and at least 30 percent of the interpretable markers were classified as having high levels of microsatellite instability, in accordance with international criteria.11 A low level of microsatellite instability was defined as a shift in only one dinucleotide marker. In this study, tumors with low levels of microsatellite instability were categorized as microsatellite-stable tumors.11
All the tumors with a shift in a mononucleotide marker had high levels of microsatellite instability when examined with the complete panel of markers, as reported previously.45,46 Therefore, in the 218 cases without control DNA that were evaluated with two mono-nucleotide markers, a shift in a marker was considered to indicate high levels of microsatellite instability.
Cases with missing results were included in all analyses that did not involve the missing data. The Cochran–Mantel–Haenszel (stratum-adjusted Pearson’s chi-square) test and Pearson’s chi-square test were used to analyze associations among categorical variables.47 Analysis of variance was applied to data on age and tumor size. Survival curves were estimated by the method of Kaplan and Meier,48 and differences were assessed by means of the stratified log-rank test.49 Proportional-hazards regression models were used for multivariable comparisons of time-to-event end points.50 All computations were performed with SAS software (version 6.12, SAS Institute, Cary, N.C.). All P values were calculated with two-sided tests of significance.
We first present the genetic abnormalities in the colon cancers we studied and then relate these molecular findings to the survival of the patients after adjuvant chemotherapy.
Loss of heterozygosity at chromosome 18q was observed in 155 of 319 cancers (49 percent). Of these 155 specimens, 143 (92 percent) had loss of all the analyzed 18q markers. Loss of heterozygosity at 17p was found in 166 of 325 tumors (51 percent). Of 309 tumors, 254 (82 percent) had allelic loss from both 18q and 17p or from neither (P<0.001 for concordance between the status of 18q and 17p). Loss of 8p alleles was found in 95 of 303 tumors (31 percent).
Of 445 cancers, 205 (46 percent) had a high labeling index for p53 protein, a finding consistent with a mutation of the p53 gene.51 In 204 of 288 tumors (71 percent), a high p53 labeling index was associated with allelic loss from 17p and a low index was associated with retention of 17p (P<0.001 for the concordance between p53 labeling index and 17p status). The p21WAF1/CIP1 protein was detected in 211 of 445 tumors (47 percent). In 276 of 445 cancers (62 percent), there was an inverse relation between p53 and p21WAF1/CIP1 labeling (P<0.001).
Of the 298 tumor specimens we evaluated for microsatellite instability, 62 (21 percent) had high levels of microsatellite instability. Low levels of microsatellite instability were found in 28 of 298 cancers (9 percent), which were categorized as the microsatellite-stable tumors.
Of the 218 tumor specimens for which control DNA was not available, high levels of microsatellite instability were found in 40 specimens (18 percent). Of the 516 specimens in the entire study that could be analyzed, 102 (20 percent) were classified as having high levels of microsatellite instability and 227 (44 percent) as having microsatellite stability; in 187 specimens (36 percent) the results of evaluation were indeterminate or unsatisfactory.
High levels of microsatellite instability and a high labeling index for p53 protein were found in 24 of 90 tumor specimens (27 percent), whereas 106 of 202 tumor specimens (52 percent) with microsatellite stability had a high p53 labeling index (P<0.001). Among the tumor specimens that had high levels of microsatellite instability and that were evaluated with dinucleotide and mononucleotide markers, mutation of the gene for the type II receptor for transforming growth factor β1 (TGF-β1) was present in 38 of 62 specimens (61 percent). None of the tumor specimens with low levels of microsatellite instability or with microsatellite stability had such a mutation. Mutation of the BAX gene was present in 22 of 60 cancers (37 percent) with high levels of microsatellite instability.
Among the 460 patients who were treated with fluorouracil-based chemotherapy, female sex, less advanced stage of disease, and the absence of regional metastases were significant favorable predictors of five-year disease-free survival and five-year overall survival after chemotherapy (Table 2). Younger age (≤65 years) and the presence of a well-differentiated adenocarcinoma, as compared with a moderately or poorly differentiated tumor, were also favorable predictors of five-year overall survival.
Because the efficacy of adjuvant chemotherapy is more firmly established for patients with stage III cancer than for those with stage II cancer, survival among those with stage II disease was analyzed separately. Of the molecular markers we tested, the status of 18q was significantly associated with both five-year disease-free survival and five-year overall survival after chemotherapy among patients with stage III cancer (Table 3). Patients with tumors that retained 18q had a five-year disease-free survival rate of 64 percent, as compared with 44 percent among those with loss of heterozygosity at 18q (P=0.002) (Fig. 2 and Table 3). The corresponding five-year overall survival rates were 69 percent with retention of 18q alleles and 50 percent with allelic loss at 18q (P=0.005) (Fig. 3 and Table 3). The 18q status also had predictive value in an analysis of the subgroup of patients with microsatellite-stable stage III carcinoma.
Mutation of the gene for the type II receptor for TGF-β1, a specific indicator of high levels of micro-satellite instability, was marginally associated with improved five-year overall survival (Table 3). High levels of microsatellite instability and alteration of the BAT-26 marker were also moderately associated with improved disease-free survival at five years (P=0.02 for both associations). Among the patients who had stage III cancer with both high levels of microsatellite instability and mutation of the gene for the type II receptor for TGF-β1, the rate of disease-free survival at five years was 79 percent, as compared with 40 percent among those whose tumors had high levels of micro-satellite instability and no mutation of this gene (P= 0.007) (Fig. 2 and Table 3). The corresponding rates of overall survival at five years were 74 percent and 46 percent (P=0.04) (Fig. 3 and Table 3).
There was no relation between survival after treatment with a particular regimen and the presence of any of the molecular markers. No marker had predictive value in the analysis of 121 patients with stage II cancer, possibly because of the small sample.
In proportional-hazards regression models that were adjusted for sex, age, the extent of spread, the presence or absence of regional metastases, and the presence or absence of obstruction, several variables —allelic loss at 18q, microsatellite stability, absence of mutation of the gene for the type II receptor for TGF-β1, and absence of an allelic shift in BAT-26 —were each independently associated with an increased risk of recurrence (Table 4). In models in which multiple markers were analyzed, microsatellite-stable cancers had a higher relative risk of recurrence than cancers with high levels of microsatellite instability and mutation of the gene for the type II receptor for TGF-β1 (relative risk, 2.60; 95 percent confidence interval, 1.36 to 4.95; P=0.004). After adjustment for multiple markers, loss of 18q alleles remained an indicator of recurrence and death (Table 4); in contrast, allelic loss at 17p was not predictive either in univariate analysis or after such adjustments (data not shown).
In this study, we showed that the status of chromosome 18q in tumors with microsatellite stability and of the gene for the type II receptor for TGF-β1 in tumors with high levels of microsatellite instability could be used to predict the likelihood of survival in patients with stage III colon cancer who received fluorouracil-based adjuvant chemotherapy. We do not know whether these markers reflect resistance or sensitivity to fluorouracil or inherent differences in the biologic characteristics of the tumors.
In several15–24 but not all15,29,52,53 previous studies, loss of heterozygosity at 18q was an indicator of a poor prognosis in patients with stage II cancer, patients with stage III cancer, or both groups. This loss usually involves the DCC gene, but there are numerous other genes in the deleted region. The product of the DCC gene is the netrin-1 receptor, which guides the migration of neuronal axons.54–57 In colon cancer, loss of DCC is associated with metastasis and an adverse prognosis.58–61 If it does not bind to netrin-1, the DCC protein triggers apoptosis.62 For this reason, loss of DCC as a result of loss of 18q could impair apoptosis, thereby conferring resistance to chemotherapy. The absence of an association between survival and loss of heterozygosity at 8p or 17p suggests that loss of heterozygosity at 18q is a specific marker for survival and not simply a reflection of generalized chromosomal instability.
Our study confirmed the concordance between allelic loss at chromosome 18q and allelic loss at 17p.16 Although alteration of p53 is a plausible predictive marker,15,33,34,63–69 we found no significant relation between survival and the status of the p53 gene or p53 protein. Another study, however, found a higher rate of seven-year survival after adjuvant therapy with fluorouracil and levamisole in patients who had cancer without increased levels of p53 protein than in those who had cancer with increased p53 levels.36 The explanation for these discrepant results in patients in the same clinical trial is not apparent.
The p21WAF1/CIP1 protein is a downstream effector of the p53 protein,70,71 and we found an inverse relation between p53 and p21WAF1/CIP1 in colon cancer, as has been reported previously.72,73 Despite the importance of p21WAF1/CIP1 for in vitro responses to chemotherapeutic agents35 and the report that increased levels of p21WAF1/CIP1 were associated with chemosensitivity of metastatic colorectal cancer,74 pretreatment levels of this protein were not related to survival in our study.
A mutation of the gene that encodes the type II receptor for TGF-β1 in cancers with high levels of microsatellite instability was associated with a favorable outcome, but the mechanism of this effect is uncertain. High levels of microsatellite instability improve the prognosis15,26–33 and may also increase the likelihood of survival after chemotherapy.37,38 Because cancers with high levels of microsatellite instability usually retain 18q alleles, loss of heterozygosity in such tumors is unlikely to be a determinant of outcome after adjuvant chemotherapy. The TGF-β1 pathway inhibits tumor proliferation by blocking the cell cycle late in the G1 (gap 1) phase,75,76 so continued proliferation due to mutation of the gene for the type II receptor for TGF-β1 could increase susceptibility to chemotherapy. However, colon-cancer cell lines that are deficient in mismatch-repair activity and that have high levels of microsatellite instability are relatively resistant to fluorouracil in vitro.77
We find that specific molecular markers in resected stage III colon cancer can be used to predict survival after adjuvant fluorouracil-based regimens. Prospective studies are needed to determine whether newer chemotherapeutic agents, such as irinotecan78 and oxaliplatin, would benefit patients with stage III cancer whose tumors have molecular markers associated with a reduced efficacy of fluorouracil-based regimens. Our study is a first step toward the goal of individualized cancer treatment based on the molecular characteristics of the tumor.
Supported by grants (CA60100, CA21115, CA62924, and CA23318) from the National Institutes of Health.
We are indebted to Drs. Bert Vogelstein, Kenneth W. Kinzler, and James Eshleman for advice; to Dr. Asif Rashid for assistance; and to Nancy Folker and Cheryl Willis for secretarial support.