Background
Esophageal adenocarcinoma (EAC) is one of the fastest rising cancers in the West, particularly among white men[
1]. Survival remains poor despite experimentation with numerous cytotoxic agents and therapeutic approaches, as well as improvements in diagnostic, surgical, and radiation technique[
2]. In the U.S., a widely used standard to attempt to cure locally advanced disease, which is a common stage at presentation, is to provide concurrent radiochemotherapy (RCT) followed by surgery. Cisplatin often forms the base of the chemotherapy regimen[
3].
This approach causes considerable toxicity in the vast majority of patients. Moreover, therapies are given with little foreknowledge of outcome. Therefore, one method to improve outcomes in EAC is to identify which patients will respond best to a particular therapeutic approach or agent - or, conversely, to identify patients who will most likely fail standard therapy so as to deliver alternative therapy at or near the outset[
4]. Such patient stratification holds the potential for maximizing efficacy and minimizing toxicity.
Current methods of patient stratification in the trimodality setting - e.g., tumor grade, lymph node status, and other clinicopathologic traits - inadequately forecast clinical outcome. Therefore, the identification of biologic or molecular predictors is a rational step to tailor therapy according to an individual clinical-molecular profile.
Importantly, EAC represents a useful human model in which to study radiochemoresistance in gastrointestinal carcinomas. RCT leads to complete obliteration of tumor (
i.e., complete pathologic response [pCR]) in approximately 25% of EAC patients [
5]. pCR has been repeatedly shown to be one of the strongest prognosticators of long-term outcome in EAC, associated with a 2- to 4-fold longer median overall survival[
6‐
8]. Its importance is underscored by its increasing use as a primary endpoint in clinical trials for esophageal cancer. RCT typically leads to a higher pCR rate in EAC than in other gastrointestinal carcinomas, including those of the rectum[
9] or pancreas[
10].
Recent attention has focused on genetic variations in DNA repair pathways as a strategy for predicting response to DNA-damaging agents. DNA lesions induced by radiotherapy or cisplatin are repaired by the base excision (BER) and double-stranded break repair (DSBR) or by the nucleotide excision repair (NER) pathways, respectively. Key enzymes in these multistep complexes are XRCC1 (BER), XRCC2 (DSBR), and ERCC1 and XPD (both NER)[
11]. It has been hypothesized that impaired BER/DSBR/NER activity in tumor cells leads to greater DNA damage after treatment with platins and/or ionizing radiation, thus causing greater tumor cell death[
12‐
14]. Alternatively, greater DNA damage and genetic instability may produce tumor heterogeneity, giving rise to malignant clones that resist apoptosis after platin or radiation exposure[
15].
Single nucleotide polymorphisms (SNP) in these pathways may alter DNA repair capacity. SNPs in the
XRCC1 gene (Arg399Gln) and
XPD gene (Lys751Gln) have been associated with increased DNA damage[
16,
17]. Studies of these SNPs in EAC are only recently emerging. Efforts to identify markers that predict radioresistance, which encompasses the radiosensitizing effects of chemotherapy, would focus naturally on BER or DSBR genes. A co-investigator (X.W.) found that the
XRCC1 Arg399Gln SNP (BER) was associated with pCR in esophageal cancer patients treated with cisplatin-based RCT and surgery[
18]. However, this cohort was retrospectively collected over a long period (1985-2003), received heterogeneous therapeutic approaches (
e.g., induction chemo prior to combined RCT were included), and combined squamous cell carcinoma (ESCC) and adenocarcinoma (EAC) histology. ESCC and EAC appear to be distinct from one another in epidemiology and biology[
1,
2,
19]. Our objective in the current study was to build on this recent finding, by assessing this SNP and 4 others (
ERCC1, XPD, XRCC2) in a cohort of only-EAC subjects, enrolled to a prospective, interventional clinical trial during a recent and short period and treated with cisplatin-based RCT in the first-line setting, without induction chemotherapy. We found that the relationship between the
XRCC1 SNP and pCR pointed in the same direction as found by our co-investigator. To our knowledge, this is the first assessment of this SNP in EAC patients receiving RCT in a prospective clinical trial.
In addition, we assessed whether germline genotype may serve as a surrogate for tumor genotype at this XRCC1 SNP by examining allelic imbalance (AI). AI is the loss or gain of a DNA region in tumor (as compared to germline) cells. This issue has relevance in both the research and clinical context. Studies evaluating the relation between SNPs and therapeutic efficacy in humans have typically studied germline DNA (peripheral blood lymphocytes or normal tissue) under the assumption that germline genotype usually reflects tumor genotype (i.e., that AI is rare at these loci). However, the rate of AI in resectable EAC at this XRCC1 SNP is unknown. A low rate of AI would provide biologic plausibility that germline genotype may be used as a surrogate for tumor genotype at these loci in predicting treatment response. Therefore, we examined tumor tissues with matched histologically normal tissues to assess AI. We found that AI at this XRCC1 SNP is uncommon; this finding supports the use of germline DNA, which is far more accessible clinically, when attempting to use this SNP to predict therapeutic tumor response in the tumor using this SNP to therapy.
Discussion
The ability to predict, in the pre-RCT setting, which human EACs are radiochemosensitive has enormous clinical relevance. This is because RCT carries substantial toxicity, and because preoperative chemotherapy alone, without radiotherapy, is a valid curative alternative to RCT[
28,
29]. The likelihood that a patient's tumor is radiochemosensitive may be incorporated into a risk/benefit model at the time of diagnosis which could be used as a clinical decision-making tool to select patients for preoperative RCT
vs chemotherapy alone. pCR is a clinically relevant marker of radiochemosensitivity, due to its consistent and strong association with overall survival in this disease[
4]. Because traditional clinical-pathologic factors in the pretreatment setting do not adequately predict radiochemosensitivity, it is reasonable to evaluate molecular or genetic factors[
4,
19].
In this study we found that 52% of EAC subjects had the variant 399Gln allele in XRCC1, and that subjects with the variant allele had five times higher odds of failing to achieve pCR after cisplatin-based RCT, compared to subjects without the variant allele. To our knowledge, our study is the first to assess this SNP in relation to radiochemoresistance in a clinical trial cohort of EAC. This association did not reach statistical significance at the 0.05 alpha level, which may reflect our relatively modest sample size. A number of observations support the probability that our observed findings are real, and would have reached greater statistical significance had more samples been available for analysis. One, the only other study we are aware of which assessed the relationship between this SNP in a cohort of mostly EAC patients receiving platin-based RCT and pCR (co-investigator X.W.) found the same association. Our cohort was completely independent from the first and was analyzed using a different genotyping platform.
Two, our study employed strong methodology. Therapeutic approach, staging, and disease characteristics were highly uniform, well-characterized, and prospectively collected. Our subjects were accrued over a recent, short time period across multiple centers, and staged by modern methods. Only adenocarcinoma histology was included. We focused on adenocarcinoma because of its greater relevance in the U.S., and the epidemiologic and potential biologic differences compared to squamous cell carcinoma[
19]. Lab investigators remained blinded to clinical outcomes.
Three, we found that tumor genotype reflected germline genotype at the
XRCC1 399 loci (19q13.2) in 90% of informative cases, suggesting germline DNA may be an appropriate surrogate of tumor genotype. To our knowledge, this is the first reported assessment of AI specifically in
XRCC1 in EAC. Comparative genomic hybridization (CGH) studies of EAC have not found substantial genetic alterations in this general region[
30,
31]. One study found >10% of 28 GEJ cancers (including 3 cell lines and 2 xenografts) had gene amplification at a neighboring site (19q13.1) [
31]; however, the density of CGH coverage is not clearly reported [
32] so it is unknown whether amplification occurred at
XRCC1. Together, these AI data, which require further evaluation and confirmation, provides biologic plausibility that assessing germline DNA is appropriate when attempting to predict response to RCT in EAC tumors.
In addition,
XRCC1 is a key player in BER, the major repair pathway for nonbulky damaged bases, abasic sites, and DNA single-stranded breaks after treatment with ionizing radiation[
33,
34]. Prior reports in human populations suggested the 399Gln variant of
XRCC1 was associated with greater DNA and chromosomal damage[
16,
35]. Worsened pCR and survival[
18] related to the variant may be due to increased genetic instability and the development of multiple clonal populations, given the substantial data linking chromosomal aberrations and poor prognosis[
15,
18]. One co-investigator (G.L.) assessed this
XRCC1 SNP in patients with esophageal cancer treated with cisplatin-based trimodality therapy and, similar to the current study, did not find an association with disease-free or overall survival[
36]. Also consistent with our findings, the 399Gln variant allele was associated with decreased tumor response (and worse survival) in patients with stage III-IV non-small cell lung cancer (NSCLC) and metastatic colorectal cancer, respectively, treated with platinum chemotherapy[
22,
37]. By contrast, the variant allele was associated with favorable OS in patients with stage IV squamous cell carcinoma of the head and neck (SCCHN) treated with cisplatin-based chemotherapy or RCT[
12]. Investigators of the latter study hypothesized their divergent result may have been due to biologic or tissue-specific factors underlying the greater chemosensitivity of SCCHN, compared to lung cancer, or to the differing types of platin compounds used in other studies (mostly carboplatin in NSCLC and oxaliplatin in colorectal cancer).
The
XRCC1 SNP deviated somewhat from Hardy-Weinberg equilibrium, with under-representation of the heterozygote genotype
vs over-representation of one of the homozygous genotypes. This raises the possibility that one of the homozygous genotypes increases EAC risk, as supported by some [
38] but not other [
39,
40] case-control data; however, our study design does not enable firm conclusions to be drawn.
Because this study focused on previously reported alleles in a few genes, we have not accounted for the potential influence of other SNPs on clinical outcome. In addition, because the two arms of the study were pooled, the study does not account for the potential differential effects of the second chemoagent of each arm. While the effect size noted in our study between the XRCC1 SNP and pCR was considerable, the results did not reach statistical significance at the alpha 0.05 level; therefore, our results should be viewed with caution pending validation in an independent cohort. Despite substantial effort to identify prognostic or predictive biomarkers in EAC, the process remains in its early stages. Because chemo- or radiotherapy deliver their effects through multigenic steps, it is unlikely that a single SNP or marker will robustly predict therapeutic efficacy. In the practical clinical context, a combination of markers will likely be required. Our finding supports the further evaluation of this XRCC1 SNP in larger cohorts.
Acknowlegdements
We thank M. Eileen Dolan, MD, for helpful discussion, and Candace Kostelec, Karen J. Hanson, and Carol Chami for administrative assistance.
FUNDING:
The parent study was conducted by the Eastern Cooperative Oncology Group (Robert L. Comis, M.D.). The parent and/or translational study were supported by an American Society of Clinical Oncology Young Investigator Award and Paul Calabresi Program in Clinical Translational Research CA90628-08U (both H.H.Y), and in part by Public Health Service Grants CA23318, CA66636, CA21115, CA13650, CA16116, CA49883, CA17145, CA39229 from the National Cancer Institute, National Institutes of Health, and the Department of Health and Human Services. Its contents are solely the responsibility of the authors and do not necessarily represent the official views of the National Cancer Institute.
PRIOR PRELIMINARY REPORTING
These results were presented in preliminary form at ASCO Annual Meeting, May 31, 2009, and at ECOG Gastrointestinal Cancers Meeting, June 13, 2009.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
The following authors were involved in study design: HHY, MKG, ABB, AAF, KMM, GL, XW, TCK, SJM, LRK, MP, and PJC. HHY, KMM, MJH, SMO, AAF, XW, GL, SJM, EAC, SP, and TCK contributed to execution and interpretation of lab assays and/or interpretation of study results. Statistical analysis was performed by PJC and MP. All authors critically reviewed and approved the final manuscript.