Association of single nucleotide polymorphisms in XRCC1 and ATM with recurrence in patients with cervical cancer

Design of study

The null hypothesis for sample size calculation was “the 2-year RFS of patients with CC genotype of XRCC1 580C>T is equal to that of patients with CT or TT genotype.” The ratio between CC vs. CT or TT genotype in my previous study was 52:48.⁸ Therefore, the ratio between CC vs. CT or TT genotype was assumed as 1:1. To estimate 2-year RFS according to genotypes, the RFS of 27 patients who were included in my previous study⁸ and received adjuvant CRT after radical surgery were estimated. The estimated 2-year RFS of patients with CC genotype was 85.9% and that of patients with CT or TT genotype was 63.6%. A difference in RFS of 22.3% (85.9% minus 63.6%) is considered a clinically important difference to detect. This difference require 63 patients in each genotype group to afford statistical power of at least 0.80 when keeping the probability of type 1 error at significance level, α=0.05. Considering the 10% failure rate to extract DNA from paraffin block or to determine genotypes, 70 patients in each genotype group is required.

Through searching the patient registry, I identified 159 patients who underwent radical hysterectomy followed by adjuvant CRT between 1999 and 2004. Because the number of patients identified (n=159) was similar to the statistically required number of patients (n=140), I decided to include all patients identified. Of 159 patients, six patients were excluded because both cancer and normal tissue blocks were unavailable. The remaining 153 patients were included in this study. Of 153 patients, 104 patients had both cancer and normal tissue blocks. Six patients had only normal tissue blocks and 43 patients had only cancer tissue blocks. After approval by the Institutional Review Board (K-1101-002-058), I abstracted clinicopathologic variables from medical records. Patients with parametrial invasion, positive resection margin or lymph node involvement were determined as high-risk, and those with intermediate-risk factors were determined as intermediate-risk based on the criteria used in a previous clinical trial. ⁵ RFS was defined as the length of time from the date of surgery to the date of recurrence. Likewise, OS was defined as the length of time from the date of surgery to the date of death from any cause.

Selection of genes and SNPs

Because most adjuvant CRT comprised of platinum-based chemotherapy and radiation, I decided to choose one gene related to response to platinum-based chemotherapy and another gene related to the response to radiation. In my previous study, SNPs at XRCC1 and gamma-glutamyl hydrolase (GGH) were associated with response to platinum-based chemotherapy. Because I thought that XRCC1 is more consistently reported to be associated with response to platinum-based chemotherapy than GGH in literature, I selected XRCC1 for the current study.

A recent review summarized genes reported to be associated with response to radiation in various cancers. According to the review, SNPs in nine genes were associated with response to radiation: ATM, checkpoint kinase 1 (CHK1), excision repair cross-complementing rodent repair deficiency complementation group 1 (ERCC1), ERCC2, ERCC5, Mdm2 p53 binding protein mouse homolog (MDM2), tumor protein p53 (TP53), xeroderma pigmentosum complementation group A (XPA), XRCC1.⁷ Because patients in the current study underwent CRT, I excluded genes (ERCC5, XPA) associated with response to radiation alone.⁹ Because the endpoint of the current study was RFS, I excluded a gene (ERCC1) which was not associated with RFS but with response rate.¹⁰ Genes (ERCC2, MDM2, TP53) associated with response to radiation only when they were combined with SNPs in other genes were excluded.^11,12 XRCC1 which is already selected for the current study was excluded. Between the remaining pair ATM and CHK1, I selected ATM because I thought that ATM is more consistently reported to be associated with response to radiation than CHK1 in literature.

A gene function could be affected by multiple SNPs and would be better predicted by a combination of multiple SNPs than single SNP.⁶ Therefore, I aimed to select SNPs which could accurately determine haplotypes in >80% of cases. For XRCC1, based on a study including 229 Korean patients with lung cancer,¹³ 580C>T (rs 1799782), and 1196A>G (rs 25487) were chosen. For an ATM gene, there is no study investigating the frequency of each genotype of SNPs in Korean population. Therefore, I used the results of a study¹⁴ including 556 French women and selected 3078-77C>T (rs 664677), 5557G>A (rs 1801516) and 6807+238G>C (rs 609429).

DNA extraction and SNP typing

In each patient, a formalin-fixed, paraffin embedded tissue block containing tumor tissue and another block containing uninvolved lymph node were chosen by a single pathologist (Cho SY). Three to five sections (10μm thick) were cut from each block and cells were scraped from the representative tumor portion or uninvolved lymph node. Scraped cells were stored at 4 °C until DNA extraction. QIAamp DNA Micro kit (Qiagen, Hilden, Germany) was used for DNA extraction according to the manufacturer’s instructions. The amount of DNA was measured by absorbance at 260 nm using an ultraviolet spectrometer, and the purity of the DNA was evaluated by measuring the 260/280 absorbance ratios.

Polymerase chain reaction (PCR) amplification of extracted DNA was performed with 50mM MgCl2 0.75μl, 25mM dNTP 0.2μl, 20pM forward and biotinylated reverse primers 0.5μl each, 5U/μl Taq polymerase 0.15μl, and 2.5μl DNA in a total volume of 25μl. The used primers for each SNP were summarized in Table 1. PCR conditions were 95°C for 10 min followed by 40 cycles (95°C for 30 seconds, 55°C for 30 seconds, 72°C for 30 seconds) and a final elongation step at 72°C for 5 minutes.

Each biotinylated PCR product (20μl) was denatured and immobilized on Streptavidin Sepharose High Performance beads (Amersham Bioscience, Piscataway, NJ, USA) using the PSQ Vacuum Prep Tool (Qiagen, Hilden, Germany) and Pyromark Q96 Vacuum Prep Worktable (Qiagen, Hilden, Germany). These were then incubated at 72-80°C for 3 minutes with 10pM of sequencing primer and placed in the room temperature. Pyrosequencing was performed using PyroGold Q96 Reagents (Qiagen, Hilden, Germany) on the Pyromark Q96 ID instrument (Qiagen, Hilden, Germany) according to the manufacturer’s instructions. Pyrogram was analyzed by the Pyromark software version 1.0 (Qiagen, Hilden, Germany)(Figure 1). When the result of pyrogram analysis was indeterminate, pyrosequencing and subsequent analyses were repeated two times. If the result of analysis was still indeterminate, the genotype was determined as genotyping failure.

• Download figure
• Open in new tab

Figure 1.

Example of the pyrogram analysis. Arrows indicate the existence of minor alleles.

Statistical analysis

Clinicopathologic variables and genotype of SNPs were summarized. Hardy-Weinberg equilibrium was tested using chi-square test. The concordance rate of genotype from cancer tissue and that from normal tissue was calculated according to SNP. To evaluate the association of clinicopathologic variables with RFS and OS, all clinicopathologic variables except risk group were dichotomized using the median or appropriate value. The association of clinicopathologic variables with RFS and OS was evaluated via Cox regression analysis. Forward stepwise selection method was used to select clinicopathologic variables independently associated with RFS or OS. The association of genotype of SNPs with RFS and OS was evaluated via Cox regression analysis under the adjustment for clinicopathologic variables independently associated with RFS or OS. The association of genotype of SNPs with completion of CRT was evaluated via chi-square or Fisher’s exact test. Haplotype estimation was performed using haplo.stats in R version 2.15.0 (R Development Core Team (2012). R: A language and environment for statistical computing, reference index version 2.15.0. R Foundation for Statistical Computing, Vienna, Austria. ISBN 3-900051-07-0, URL http://www.R-project.org.). A P-value<0.05 was regarded as statistically significant and all statistical analyses except haplotype estimation were performed using SPSS (version 18.0) (SPSS Inc., Chicago, IL, USA).

Baseline characteristics

The baseline characteristics of 153 patients are summarized in Table 2. The median age was 47 years and median clinical tumor size was 3.0 cm. Stage 1B1 accounted for half of population and squamous cell carcinoma is the most common histologic type. The median pathologic tumor size was 3.3 cm and invasion to outer 1/3 of cervical wall was present in 69% of patients. Lymphovascular space invasion (LVSI), parametrial invasion, positive resection margin and lymph node involvement were present in 56%, 25%, 7% and 44% of patients, respectively. Over half of patients were classified as high-risk and 16 of 153 patients could not complete the adjuvant CRT. During follow-up, 29 recurrences and deaths were observed (Table 2). Specifically, one vaginal, three pelvic and 25 extra-pelvic recurrences have occurred.

Genotype of SNP

All SNPs except ATM 6807+23G>C from cancer tissue satisfied the Hardy-Weinberg equilibrium. The frequency of minor genotypes was considerable except ATM 5557G>A. Genotyping failure occurred more frequently in normal tissue (7%) than cancer tissue (1%). In addition, genotyping failure occurred more frequently in ATM 6807+238G>C (14%) than other SNPs (1%)(Table 3).

Concordance between genotype from cancer tissue and that from normal tissue

Concordance rate varied according to SNP. Specifically, concordance rates of XRCC1 580C>T, XRCC1 1196A>G, ATM 3078-77C>T, ATM 5557G>A and ATM 6807+238G>C were 72% (Table 4), 88% (Table 5), 73% (Table 6), 100% and 54% (Table 7), respectively.

Association of clinicopathologic variables with RFS and OS

In univariate analysis, age, clinical tumor size, stage, histologic type, pathologic tumor size, LVSI, parametrial invasion, resection margin, lymph node involvement and risk group were associated with RFS. In multivariate analysis, age, stage, histologic type, LVSI and parametrial invasion were associated with RFS (Table 8).

In univariate analysis, clinical tumor size, histologic type, LVSI, parametrial invasion, resection margin, lymph node involvement, risk group and completion of CRT were associated with OS. In multivariate analysis, clinical tumor size, resection margin, lymph node involvement and completion of CRT were associated with OS (Table 9).

Adjusted RFS and OS according to genotype

Genotype of XRCC1 580C>T was not associated with RFS in patients with cervical cancer who underwent radical surgery followed by adjuvant CRT (Figure 2). Of all SNPs, genotype from cancer tissue was associated with neither RFS nor OS adjusted for clinicopathologic variables (Table 10). Similarly, genotype from normal tissue was associated with neither RFS nor OS adjusted for clinicopathologic variables (Table 11).

• Download figure
• Open in new tab

Figure 2.

Recurrence-free survival according to genotype of XRCC1 580C>T

Association of haplotype with RFS and OS

The distribution of haplotypes among XRCC1 and ATM SNPs was summarized in Table 12 and Table 13, respectively. Haplotype pairs in XRCC1 were associated with neither RFS nor OS (Table 14). Similarly, haplotype pairs in ATM were associated with neither RFS nor OS (Table 15).

Association of SNPs with completion of CRT

In the SNPs, genotype from cancer tissue was not associated with completion of CRT (Table 16). Similarly, genotype from normal tissue was not associated with completion of CRT (Table 17).

RFS according to XRCC1 580C>T

This study failed to show different RFS according to genotype of XRCC1 580C>T in patients with cervical cancer who underwent radical surgery followed by adjuvant CRT. There was a trend that patients with CT or TT genotype had a worse RFS than those with CC genotype but the difference was not significant.

I believe that inadequate sample size as a result of lower than expected effect is the main reason for the negative result. Specifically, I calculated the sample size using estimated 2-year RFS of 27 patients included in my previous study.⁸ The estimated 2-year RFS of patients with CC genotype was 85.9% and that of patients with CT or TT genotype was 63.6%. However, in this study, the estimated 2-year RFS of patients with CC genotype was 87.4% and that of patients with CT or TT genotype was 80.9%. The incorrect estimation of 2-year RFS in patients with CT or TT genotype resulted in small sample size and lack of power to detect the difference in RFS. I recalculated the sample size using the observed 2-year RFS in the current study. To achieve statistical power of at least 0.80 when keeping the probability of type 1 error at 0.05, 550 patients per arm is necessary considering the 10% genotyping failure rate, which is about eight-fold larger than initially calculated sample size (n=70 per arm).

Concordance between genotype from cancer tissue and that from normal tissue

Various small studies of colon and breast cancer reported high concordance rate between genotype from cancer tissue and that from normal tissue.^15-19 For example, in a study including 44 colorectal cancer samples and adjacent normal tissue identified strong concordance between paired samples for 28 SNPs in 13 genes and the concordance rate ranged between 92% and 100%.¹⁶ However, the current study showed quite a low concordance rate ranging between 54% and 100%.

The current study suggested that the variability between cancer and germline genotype at the SNP level could be significant in cervical cancer and the germline genome could not be used as a surrogate to assess SNP status in the cancer.

Association of SNPs in XRCC1 and ATM with RFS and OS

Whether DNA is extracted from cancer tissue or normal tissue, none of SNPs were associated with RFS or OS adjusted for clinicopathologic variables. In addition, none of haplotypes of SNPs was associated with RFS or OS. I believe that lack of power due to small sample size is one of the reasons for negative results.

Association of clinicopathologic variables with RFS and OS

Patients with high-risk factors such as parametrial invasion, positive resection margin and lymph node involvement were known to be at high-risk for recurrence and adjuvant treatment was indicated.2 Similarly, patients with intermediate-risk factors were also known to have modestly increased risk for recurrence and usually received adjuvant treatment.^3,5 However, there is scant literature regarding prognostic factors in patients who already received postoperative adjuvant treatment. Specifically, it is unclear whether the increased risk of patients with high- or intermediate-risk factors is negated by adjuvant treatment.

The current study showed that the recurrence risk of patients with age≤47, stage>1B1, non-squamous histologic type, LVSI or parametrial invasion is still elevated despite of adjuvant treatment and demonstrated that death risk of patients with large tumor, positive resection margin, lymph node involvement or failure to complete CRT is incompletely negated by adjuvant CRT. Patients with such risk factors could be the target population for additional treatment after radical surgery followed by adjuvant CRT.

Genotyping failure

If high quality DNA from lymphocyte or fresh frozen tissues is used, the genotype of SNP can be accurately determined in most cases. However, the quality of DNA extracted from paraffin blocks is known to be inferior to DNA from lymphocyte or fresh frozen tissues.²⁰ Therefore, SNP genotyping using DNA from paraffin blocks has lower call rates and accuracy compared to genotyping using fresh frozen tissues.²⁰ Reduced DNA extraction efficiency, reduced quality due to cross-linking and degradation, fragmentation of DNA and selective non-amplification of large fragments, inaccurate SNP calls in GC rich regions were suggested as the mechanisms of low performance of genotyping using DNA from paraffin blocks.²⁰

In the current study, genotyping failure occurred more frequently in normal tissue (7%) than cancer tissue (1%). The mechanism for more genotyping failure in normal tissue than in cancer tissue is unclear. In addition, genotyping failure occurred more frequently in ATM 6807+238G>C (14%) than other SNPs (1%). The large size of the amplicon of ATM 6807+238G>C could be the reason for frequent genotyping failure. The size and GC contents (%) of amplicon of each SNP were summarized in Table 18.

Summary

Genotype of XRCC1 580C>T was not associated with RFS and OS in patients with cervical cancer who underwent radical surgery followed by adjuvant CRT. Similarly, none of genotype of XRCC1 1196A>G, ATM 3078-77C>T, ATM 5557G>A and ATM 6807+238G>C was associated with RFS or OS. In addition, none of haplotype of SNPs was associated with RFS or OS. None of genotype of SNPs was associated with completion of CRT. Low concordance rate between cancer and germline genotype was identified in cervical cancer. Age, stage, histologic type, LVSI and parametrial invasion were independently associated with RFS. Clinical tumor size, resection margin, lymph node involvement and completion of CRT were associated with OS in patients who underwent radical surgery followed by adjuvant CRT.