There is debate in the literature over the relative importance of genetic instability and clonal expansion during progression to cancer. Barrett’s esophagus is a uniquely suited model to investigate neoplastic progression prospectively because periodic endoscopic biopsy surveillance is recommended for early detection of esophageal adenocarcinoma. We hypothesized that expansion of clones with genetic instability would predict progression to esophageal adenocarcinoma. We measured p16 (CDKN2A/INK4A) lesions (loss of heterozygosity, mutations, and CpG island methylation), p53 (TP53) lesions (loss of heterozygosity, mutation) and ploidy abnormalities (aneuploidy, tetraploidy) within each Barrett’s esophagus segment of a cohort of 267 research participants, who were followed prospectively with cancer as an outcome. We defined the size of a lesion as the fraction of cells with the lesion multiplied by the length of the Barrett’s esophagus segment. A Cox proportional hazards regression indicates that the sizes of clones with p53 loss of heterozygosity (relative risk = 1.27x for an x cm clone; 95% confidence interval, 1.07–1.50) or ploidy abnormalities (relative risk = 1.31x for an x cm clone; 95% confidence interval, 1.07–1.60) predict progression to esophageal adenocarcinoma better than the mere presence of such clones (likelihood ratio test, P < 0.01). Controlling for length of the Barrett’s esophagus segment had little effect. The size of a clone with a p16 lesion is not a significant predictor of esophageal adenocarcinoma when we controlled for p53 loss of heterozygosity status. The combination of clonal expansion and genetic instability is a better predictor of cancer outcome than either alone. This implies that interventions that limit expansion of genetically unstable clones may reduce risk of progression to cancer.
Multistage carcinogenesis is an evolutionary process in which genetic instability generates new clones and natural selection among such variants drives clonal expansions (1, 2, 3, 4, 5, 6) . The number of genetic lesions necessary to produce a cancer is unknown for most tissues but has been variously estimated to be 2 to 12 (7, 8, 9) . Normal mutation rates in the absence of clonal expansion cannot account for the accumulation of so many lesions (10) . There has been much debate in the literature as to the relative importance of a mutator phenotype (10 , 11) and genetic instability (3 , 12) versus clonal expansion (4 , 5) in neoplastic progression.
Both genetic instability and clonal expansions are commonly observed in neoplastic progression, and it is likely that the two factors cooperate in clonal evolution (11 , 13) . If this is true, then expansion of a genetically unstable clone in a premalignant neoplasm will associate with risk of progression to cancer.
The debate over the relative importance of genetic instability and clonal expansion has evolved largely around colon cancer, for which elegant studies can assess intermediate stages of adenoma progression (14) . However, abnormalities cannot be studied prospectively for progression to cancer because colonic polyps are typically removed when detected.
Barrett’s esophagus is a uniquely suited model to investigate in vivo clonal expansion and genetic instability prospectively as predictors of progression to cancer in humans (15, 16, 17, 18) . Periodic endoscopic biopsy surveillance of Barrett’s esophagus is recommended for early detection of esophageal adenocarcinoma (19 , 20) . Barrett’s esophagus is a neoplastic condition that is associated with a 30- to 40-fold increase in the risk of developing esophageal adenocarcinoma (21, 22, 23) . Barrett’s epithelium meets the criteria for a neoplasm (24) in that it is hyperproliferative relative to normal squamous epithelium (25) and generally clonal (18) , and progression to esophageal adenocarcinoma is associated with clonal evolution (2) . Two of the most commonly lost tumor suppressor genes in human cancers, p53 (TP53) and p16 (CDKN2A/INK4A), are also lost in Barrett’s esophagus (2) . Clonal expansions of cells with p16 lesions are observed in more than 85% of Barrett’s esophagus segments (18) and tend to fill the entire Barrett’s segment (24) . There is evidence for genetic instability in the form of loss of heterozygosity (LOH) even in the earliest stages of the disease (18) . LOH on chromosome 17p at the p53locus is associated with a 16-fold increase in the risk of progression to esophageal adenocarcinoma (16) . The presence of aneuploidy [relative risk (RR) = 9.5; 95% confidence interval (CI), 4.9–18] and tetraploidy (i.e., 4N fraction >6%, RR = 11.7; 95% CI, 6.2–22) are also associated with progression to esophageal adenocarcinoma (17) .
On the basis of the evolutionary theory of neoplastic progression, the size of the Barrett’s segment, as a surrogate for the evolving cell population size, should associate with progression to esophageal adenocarcinoma. One retrospective, hospital-based, case–control study in Rotterdam excluded segments shorter than 3 cm and found that a doubling in length of the Barrett’s segment was associated with a 1.7 odds ratio for esophageal adenocarcinoma (26) . Another retrospective, case–control study from the Hines Veterans Affairs Barrett’s Esophagus Cohort (27) , including all segment lengths, reported an odds ratio of 2.48 per cm. However, an earlier prospective study of the Seattle Barrett’s Esophagus Cohort reported only a nonsignificant trend with a 5-cm difference in segment length associated with a 1.7-fold increase in cancer risk (28) . The relative weakness of the evidence for an effect of segment length on cancer risk suggests the size of the relevant population of cells, i.e., those that progress to cancer, may not have been assessed. It may be that the number of cells that are genetically unstable is more important for progression than the number of proliferating cells. Therefore, we hypothesized that the size of the clone with p53lesions, or aneuploidy, or tetraploidy would predict progression to esophageal adenocarcinoma. Because clones with p53lesions seem to require a p16lesion before they can expand (24) , we hypothesized in addition that the size of clones with p16lesions would also predict progression to esophageal adenocarcinoma.
Carcinogenesis is often separated into processes of initiation and promotion. The initiating event in Barrett’s esophagus is currently unknown but is associated with chronic gastro-esophageal reflux (29 , 30) . By measuring the sizes of clones with genetic lesions in Barrett’s esophagus and following the patients to cancer outcome, we may determine the relative importance of genetic instability and clonal expansion in promotion. Although understanding initiation will be important for cancer prevention efforts, measuring lesions during progression is likely to be clinically relevant for both prognosis and cancer prevention.
In this study, we measured the size of clones with p53lesions, p16lesions, aneuploidy, and tetraploidy in the neoplasms of participants with Barrett’s esophagus at baseline and followed the research participants with serial endoscopies for up to 8 years with progression or lack of progression to esophageal adenocarcinoma as end points. Below, we present the first prospective data, with cancer as an end point, that addresses the roles of genetic instability and clonal expansion in a human, sporadic, premalignant neoplasm.
MATERIALS AND METHODS
Research participants were members of the Seattle Barrett’s Esophagus Cohort with intestinalized metaplastic epithelium documented in baseline endoscopic biopsies between January 5, 1995, and November 19, 2003. Fifty participants were excluded because of a lack of any follow-up data, which left 267 eligible participants. The Seattle Barrett’s Esophagus Study was approved by the Human Subjects Division of the University of Washington in 1983 and was renewed annually thereafter with reciprocity from the institutional review board (IRB) of the Fred Hutchinson Cancer Research Center (FHCRC) from 1993 to 2001. Since 2001, the study has been approved annually by the IRB of the FHCRC with reciprocity from the Human Subjects Division of the University of Washington.
Endoscopic procedures and the collection of biopsies have been described previously (31) . One biopsy every 2 cm was analyzed for molecular data in most cases. In 23 participants, one biopsy was analyzed per 1 cm. Excluding these participants from the analysis does not change the significance of the results.
Biopsies were flow-sorted based on Ki67 status and DNA content as described previously (17 , 31, 32, 33) . Each biopsy was separated into diploid proliferating cells (Ki67+) and either aneuploid cells, if they were present, or cells with 4N DNA content. Tetraploid clones were defined as 4N fractions >6% (17) .
DNA Extraction and Amplification.
DNA was extracted from both of the flow-purified cell populations with either standard phenol/chloroform or the Puregene DNA Isolation Kit as recommended by the manufacturer (Gentra Systems, Inc. Minneapolis, MN). Whole genome amplification with primer extension preamplification was performed as described previously (32) for each sorted fraction and three constitutive controls per participant.
Loss of Heterozygosity.
Genomic DNA was evaluated for p16promoter methylation in 317 flow-purified fractions from 121 participants with methods for bisulfite treatment and methylation-specific PCR described previously (18) . Human genomic DNA, treated in vitro with Sss I methyltransferase (New England Biolabs, Beverly, MA), was used as the methylated control. In a subset of cases, promoter methylation was determined and/or verified by directly sequencing PCR products of bisulfite-treated genomic DNA, as described previously (18) .
Genomic or primer extension preamplification DNA was sequenced with either BigDye or BigDyeV3 Terminator cycle sequencing (Applied Biosystems, Foster City, CA) on an ABI 377, 3730, or 3700 DNA sequencer. Wild-type sequences for each participant were confirmed with constitutive samples. All of the mutations were confirmed by at least two independent PCR and sequencing reactions, and, in cases of ambiguity, by direct sequencing of genomic DNA. Evaluation of mutation of exons 5 to 9 of the p53 gene was performed on 839 flow-purified fractions from 236 participants with conditions described previously (35) . Mutation analysis of exon 2 of the p16 gene was performed on 1,195 flow-purified fractions from 239 participants with an aliquot of genomic DNA that had undergone whole genome amplification (primer extension preamplification), as described previously (18) .
Follow-up time was calculated as the time between the baseline endoscopy and either the last endoscopy before the end of data collection (November 19, 2003) or the first endoscopy that resulted in a diagnosis of esophageal adenocarcinoma. The length of the evaluated segment was measured as the level (distance in cm from the incisors) of the most distal biopsy evaluated for molecular markers minus the level of the most proximal biopsy evaluated for molecular markers plus 1. We adjusted the calculated fraction of flow-sorted cells that carry a lesion within a Barrett’s segment by excluding nonproliferating diploid cells as described previously (24) . We define the size of a lesion to be the product of the segment length and the fraction of cells in the biopsies from the participant that carry the lesion. Thus the size is an estimate of the area covered by the clone measured in cm. The size of a clone is different from the number of cm over which a clone has expanded because a clone may only comprise a portion of the cells sampled from each level.
The Cox proportional hazards regression method was used to measure the association of lesion expanses and sizes with cancer outcome. Likelihood ratio tests (LRTs) were used to assess the benefit of adding additional predictor variables to the Cox models. Backward and forward stepwise selection based on the Akaike information criterion (36) in the R statistics package 4 was used to derive a unified multivariate Cox model starting from a model that included all of the univariate predictors found to be statistically significant predictors of cancer outcome. A logistic regression was used to measure the association of p16-deficient clone sizes with the presence of a p53lesion.
The 267 participants in the cohort were followed for an average of 4.4 years (range, 0.1–8.4 years). The age and sex of participants with p53LOH, aneuploidy, or tetraploidy were statistically indistinguishable from participants without those lesions (Table 1) ⇓ . The Barrett’s segment lengths of participants with p53LOH, aneuploidy, or tetraploidy were longer than participants without those lesions (2-sided t test, P < 0.001). Neither age nor sex had a statistically significant association with cancer outcome (Cox proportional hazards: age RR = 1.01, 95% CI: 0.98–1.04; sex RR = 1.15, 95% CI: 0.50–2.67). Patients who were lost to follow-up did not have a statistically significant difference compared with the cohort of study in the frequency or clone size of p53LOH, p16lesions, aneuploidy/tetraploidy, or segment length (Wilcoxon rank-sum and χ2 tests, P > 0.05).
When the cohort is restricted to the 53 participants that have a clone with p53LOH, the size of the p53LOH clone predicts cancer outcome (Table 2 ⇓ ; Fig. 1 ⇓ , univariate Cox model RR = 3.3 for 5 cm of the clone, 95% CI: 1.4–7.6). The size of clones with either p53LOH or p53mutation also predicts cancer outcome, but the size of clones with a p53mutation is not a significant predictor of progression to cancer when analyzed alone (RR = 1.84 for 5 cm of the clone, 95% CI: 0.56–5.98). Among participants with an aneuploid or tetraploid clone, clone size also predicts progression to cancer (RR = 3.9 for 5 cm of the clone, 95% CI: 1.4–10.5), although neither aneuploidy alone nor tetraploidy alone reached statistical significance as predictors. Although many of the predictors are significantly associated with progression to cancer, point estimates for their risks vary over large ranges, as indicated by the width of the confidence intervals. The size of a clone with a p16lesion (LOH, mutation, or methylation) predicts cancer outcome (Table 2) ⇓ , but this effect disappears when participants with a p16lesion are stratified by the presence or absence of p53LOH (Tables 2 ⇓ and 3) ⇓ . Participants with large clones with a p16lesion also tend to have a p53LOH lesion (logistic regression odds ratio = 1.16x for an x cm clone, 95% CI: 1.06–1.26).
We evaluated whether previously known predictors of cancer outcome were significantly improved by the addition of clone size measurements. Adding the size of the p53LOH clone to segment length alone significantly improved prediction of cancer outcome (Table 3 ⇓ , row 1; LRT P < 0.01). Accounting for the size of p53LOH clones along with the presence of p53LOH predicted cancer better than the presence of a p53LOH lesion alone (Table 3 ⇓ , row 3; LRT P < 0.01). Similarly, accounting for the size of aneuploid/tetraploid clones along with the presence of an aneuploid/tetraploid clone predicted cancer better than the presence of an aneuploid/tetraploid clone alone (Table 3 ⇓ , row 4; LRT P < 0.01). Combining p53LOH clone size and aneuploidy/tetraploidy clone size predicted cancer better than either clone size alone (Table 3 ⇓ , row 5; LRT P < 0.001). Thus, information on clone sizes significantly improves our estimates of risk for progressing to cancer in all analyses. The multivariate Cox model with the optimal Akaike information criterion included only the size and presence of p53LOH clones and the size and presence of aneuploid/tetraploid clones (Table 3 ⇓ , last row).
The cohort only included 12 cases of participants with p53LOH and no flow abnormality at baseline that developed aneuploidy or tetraploidy in follow-up. We were, thus, unable to determine whether the size of clones with a p53lesion predicts the future development of aneuploidy or tetraploidy. p16LOH clone size did not significantly predict progression to aneuploidy or tetraploidy (RR = 1.10 for 5 cm of the clone, 95% CI: 0.62–1.93; n = 100).
If the six participants with less than 6 months of follow-up are excluded, then the effect of p53LOH clone size on cancer outcome was slightly weakened (previously RR = 3.30, with six participants excluded RR = 2.70 for 5 cm of the clone, 95% CI; 1.00–7.10). However, the effect of aneuploid/tetraploid clone size was slightly strengthened (previously RR = 3.86, with six participants excluded RR = 4.32 for 5 cm of the clone, 95% CI: 1.40–12.99).
In contrast to the results for the absolute sizes of clones, the fraction of the Barrett’s segment covered by a clone with a p53LOH lesion does not predict cancer (RR = 1.08, 95% CI: 0.30–3.82). Nor does the fraction of the segment covered by an aneuploid or tetraploid clone predict cancer outcome (RR = 1.02, 95% CI: 0.25–4.23).
Evolution is driven by both mutation and natural selection (36) . The debates over genetic instability and clonal expansion in neoplastic progression have really been debates over the relative importance of mutation and natural selection in somatic evolution. We have shown evidence in a human neoplasm that both are important. The combination of clonal expansion and genetic instability predicts progression to esophageal adenocarcinoma in Barrett’s neoplasms better than either factor alone.
Previous studies have shown that p53LOH, aneuploidy, and tetraploidy predict progression to esophageal adenocarcinoma (16 , 17) . Our results suggest that not just the presence but the size of clones with these lesions predicts progression to esophageal adenocarcinoma. Like aneuploidy (37, 38, 39, 40) and tetraploidy (41) , p53lesions are associated with genetic instability (42, 43, 44) . It seems that an increasing number of genetically unstable cells favors the evolution of a malignant clone. The presence of instability may increase the rate of generation of genetic variants and subsequent selection of clones with an increased predisposition to cancer. This is consistent with data showing greater diversity of chromosomal variants in biopsies with p53LOH or aneuploidy than in diploid biopsies from Barrett’s esophagus. 5
The size of a clone with a p16lesion appears to predict cancer, but the effect disappeared with stratification on p53lesions and was not observed in Barrett’s segments that did not have a p53lesion. In view of the width of the confidence intervals and the weakness of the effects of p16-deficient clones (Table 2) ⇓ , stratification by p53LOH may result in an apparently nonsignificant effect of p16deficient clones on cancer outcome simply through a reduction in the power to distinguish an effect. The lack of statistical significance in the interactions between predictors in our data, thus, does not preclude the presence of such interactions. Sample size limitations make it difficult to accurately quantify risks and to detect interactions of the genetic lesions. Loss of p16is associated with large clonal expansions to the point that the p16-deficient clone often fills the entire Barrett’s segment (18 , 24) . In addition, large clones with p16lesions tend to include a p53lesion. Taken together, this suggests that p16-deficient clones are risk factors, in part, because they may carry p53lesions as hitchhikers (24) and thus increase the size of the genetically unstable clone. Alternatively, the loss of p16may be necessary to allow a clone to spread laterally in the Barrett’s epithelium; in this case, p16loss would be a prerequisite for subsequent expansion of genetically unstable clones with a selective advantage.
The debate over the relative importance of genetic instability and clonal expansion has concerned both neoplastic progression and initiation (3 , 4) . Our present results address factors that determine whether or not a neoplasm will progress to cancer, and, thus, they do not speak to the initiation of the premalignant neoplasm. Future studies of early neoplasms in Barrett’s esophagus may also be able to provide evidence of the roles of genetic instability and clonal expansion in neoplastic initiation.
The size of genetically unstable clones may be used for prognoses in other neoplasms in which clone sizes may be estimated through the analysis of multiple biopsies or flow cytometry. The fact that large, genetically unstable clones are more likely to progress to cancer than smaller clones implies that cancer prevention efforts might be focused on reducing the size or constraining the growth of genetically unstable clones. To modulate clonal expansion, we will need a better understanding of clonal competition in neoplasms (24) and methods to manipulate that competition (45 , 46) .
Genetic instability and clonal expansion are clearly important factors in neoplastic progression. Previous studies have focused on the role of one or the other of these factors. We have shown in a human premalignant condition that the combination of both genetic instability and clone size predict progression to cancer in a prospective cohort study.
We would like to thank Tom Vaughan, Peter Rabinovitch, David Cowan, Laura Prevo, Jessica Arnaudo, Heather Kissel, Christine Karlsen, Valerie Carera, Julie Ryan, and Michael Barrett for their contributions to the Seattle Barrett’s Esophagus Program. This work could not have been done without them.
Grant support: NIH grants NIH P01 CA91955, NIH K01 CA89267-02, and NIH K07 CA89147-03 (C. Maley).
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Requests for reprints: Carlo Maley, Fred Hutchinson Cancer Research Center, P.O. Box 19024, Seattle, WA 98109. Phone: (206) 667-4615; Fax: (206) 667-6132; E-mail:
↵4 Internet address: http://www.r-project.org.
↵5 V. J. Wongsurawat, J. C. Finley, P. C. Galipeau, C. A. Sanchez, Carlo Maley, P. L. Blount, R. D. Odze, P. S. Rabinovitch, and B. J. Reid. Genetic mechanisms of p53 loss in Barrett’s esophagus: implications for FISH as an alternative test for LOH, submitted for publication.
- Received May 19, 2004.
- Revision received August 2, 2004.
- Accepted August 18, 2004.
- ©2004 American Association for Cancer Research.