Levels of 5-Hydroxymethyl-2′-deoxyuridine in DNA from Blood of Women Scheduled for Breast Biopsy¹

Systemic oxidative stress levels, as evidenced by DNA damage and lipid oxidation products in blood and urine, appear to be associated with increased risk of various cancers, including that of the breast (1, 2). DNA repair ability in blood lymphocytes also has been shown to be related to breast cancer risk (3, 4, 5, 6). Interestingly, individuals with BRCA1 or BRCA2 mutations may be at increased cancer risk because of deficiencies in repair of DNA lesions caused by oxygen free radicals (7). Oxidative DNA damage is therefore an attractive marker of risk because it takes into account not only exposure to and production of oxidants, but also DNA repair ability.

Our previous data indicated that women with breast cancer have higher levels of oxidative DNA damage in the blood than do control women (1). That study, however, was performed using blood obtained in a retrospective design with a convenience sample of controls. We now present data from a prospective study: blood was obtained from subjects prior to breast biopsy. This design resulted in the enrollment of women with a variety of breast lesions, including women who were at increased risk of breast cancer because of the presence of atypical hyperplasia or carcinoma in situ. It has been hypothesized that atypical hyperplasia and carcinoma in situ are precursor lesions to invasive breast cancer (8, 9). These types of lesions are known to increase a woman’s risk of breast cancer substantially (10). It was therefore of particular interest to include this group of women to help us investigate 5-OHmdU³as a marker of increased breast cancer risk.

Women were recruited from the Karmanos Cancer Institute Comprehensive Breast Clinic during the period October 1994 through January 1996. A research assistant asked women who were scheduled for a biopsy whether they would fill out a questionnaire and provide a blood sample for a biomarker study. This occurred at a clinic visit that was typically several days before the biopsy. Eligibility criteria were that subjects be at least 18 years of age and had never had a previous diagnosis of cancer. During this time period, a total of 252 subjects were enrolled.

Heparinized blood (10-ml) was used to prepare nuclei by the method of Ciulla et al. (11). Nuclei were frozen at−70°C in 50 mm mannitol, 1 mm EDTA, and 1% SDS until DNA could be extracted. DNA was isolated from nuclei by a modification of the procedure of Miller et al. (12). Briefly, the nuclei were treated with RNases A and proteinase K. The proteins were precipitated by the addition a one-third volume of 6 m sodium chloride, shaking, and centrifugation. One extraction with chloroform-isoamyl alcohol (48:2, v/v) and one extraction with n-butyl alcohol were used to remove residual protein. The DNA was precipitated twice and dissolved in 200μl of water, and a UV scan was obtained.

The isolated DNA (100–150 μg) was hydrolyzed enzymatically and derivatized, and 5-OHmdU levels were determined by gas chromatography-mass spectrometry using isotopically labeled internal standards as described previously (13). The samples were derivatized with N,O-bis(trimethylsilyl)trifluoroacetamide containing 1% trimethylchlorosilane and acetonitrile (2:1,v/v) by heating at 120°C for 20 min. Gas chromatographic separations were performed with a 25-m Hewlett-Packard SE54 Ultra 2 column using helium as the carrier gas.

Simple descriptive statistics were used to summarize the baseline demographic variables and levels of 5-OHmdU, by diagnosis category. Spearman’s rank correlation was used to assess the relationship of continuous covariables and 5-OHmdU levels. Comparison of pre- and postbiopsy 5-OHmdU levels (in a small subset of the women) was performed using the nonparametric Wilcoxon signed-rank test for paired data. Simultaneous comparison of 5-OHmdU distributions by diagnosis group (and by other factors) relied on the k-sample Kruskal-Wallis rank-sum test. All pairs of diagnosis groups were then contrasted using a nonparametric multiple-comparisons procedure (14). Despite age-matching of the subjects with benign breast lesions, a significant correlation of age with levels of 5-OHmdU remained. Thus, comparison of 5-OHmdU levels by diagnosis group, after adjustment for age and smoking status, was performed via a two-way analysis of covariance model (15), with age as a continuous covariate. For this statistical modeling work, a double natural log [ln(ln)] transformation was necessary to normalize the 5-OHmdU data. The adjusted means (and confidence limits) were transformed back to the original scale of measurement to facilitate results interpretation.

A total of 252 subjects were enrolled. Of those, 225 were considered analysis-eligible based on whether the blood was in fact drawn, whether the biopsy was subsequently performed at our institution, and whether a definitive pathology report was available. The samples included 131 biopsies that were considered benign (adenosis, apocrine or squamous metaplasia, cysts, mild hyperplasia, duct ectasia, fibroadenoma,mastitis) and 35 biopsies that are associated with a slightly increased risk of breast cancer (moderate or florid hyperplasia, ductal papilloma with fibrovascular core and sclerosing adenosis). The high-risk biopsies included 15 with atypical hyperplasia, 9 with ductal carcinoma in situ, and 3 with lobular carcinoma in situ. The invasive cancers were 27 infiltrating ductal carcinomas, 2 infiltrating lobular carcinomas, and 3 other cancers(medullary, phyllodes, and a poorly differentiated carcinoma).

The blood samples analyzed were all those available for this study from the high-risk and invasive groups: 22 of 27 subjects with biopsies in the high-risk category (2 with lobular carcinoma in situ, 7 with ductal carcinoma in situ, and 13 with atypical hyperplasia), and 23 of 29 in the invasive category (22 ductal and 1 lobular infiltrating cancer). Of the samples that were not available for this study, three did not have pathology reports completed at the time of sample analysis and the others did not have a tube of blood available for this assay. There were also 28 age-matched subjects from the benign category selected for analysis, but 1 of those samples yielded insufficient DNA, which left 27 for analysis. The age matching was done by first determining the age distribution, by decades, of the subjects in the high-risk and invasive categories. This turned out to be similar in each of those two groups. Subjects in the benign category were then randomly selected to match this distribution (with one extra subject for each age group), at the same time ensuring that samples were evenly distributed from the early, mid, and late enrollment periods.

We examined the association of 5-OHmdU with the following covariates:age, race (Caucasian, African-American, Asian), body weight, smoking status (current, past, never), menopausal status (pre-, peri-, or postmenopause), first-degree family history of breast cancer or any other nonaerodigestive cancer (yes/no), and other health problems(cardiovascular/hypertension problem, diabetes, arthritis, asthma versus none of these). Some of these factors are summarized in Table 1. The only significant associations with 5-OHmdU were with age (rank correlation = 0.25; P = 0.033), and smoking status(Kruskal-Wallis test, P = 0.031). Information on use of medications, such as hormone replacement therapy, was not available and is a potential limitation of this study.

The mean levels of 5-OHmdU in blood of women with either invasive cancer or high-risk lesions were significantly higher than in women with benign lesions (Table 2). After adjustment for age and smoking status, the results were similar to the unadjusted results (see Table 2). The adjusted mean 5-OHmdU of women with benign lesions was significantly lower than that of women in either of the other two diagnosis groups. The high-risk and invasive cancer groups again were not statistically significantly different. Interestingly, the distribution of 5-OHmdU levels across quintiles was such that the women with high-risk lesions exhibited levels only in the upper three quintiles, whereas women in the other two groups were distributed across all five quintiles. The numbers of women in each quintile were as follows: quintile 1 (0–60 fg of 5-OHmdU/ng of dThd), 3 with invasive cancer and 11 with benign lesions;quintile 2 (61–91 fg/ng of dThd), 5 with invasive cancer and 10 with benign lesions; quintile 3 (92–145 fg/ng of dThd), 5 with invasive cancer, 6 with high-risk lesions, and 3 with benign lesions; quintile 4(146–308 fg/ng of dThd), 4 with invasive cancer, 9 with high-risk lesions, and 2 with benign lesions; and quintile 5 (>308 fg/ng of dThd), 6 with invasive cancer, 7 with high-risk lesions, and 2 with benign lesions.

In 13 women, a second blood sample was obtained at a follow-up clinic visit, after the results of the biopsy were known. This included 2 women with high-risk biopsies and 11 women with invasive cancer who had not begun treatment. Average time post biopsy or surgery (if surgery was also done) was 10.2 days (SD = 5.5 days; range, 3–21 days). There was no significant difference between the levels of 5-OHmdU pre and post biopsy by the Wilcoxon signed-rank test(P = 0.414; for samples obtained before biopsy,mean ± SD, 303 ± 257 fg/ng of dThd; median, 218 fg/ng of dThd; for samples obtained after biopsy, 248 ± 245 fg/ng of dThd,mean ± SD; median, 191 fg/ng of dThd). This would suggest that differential psychological factors associated with the clinic visits(i.e., anxiety that a biopsy is needed) or any effects of having had surgery or biopsy in the recent past did not appreciably affect 5-OHmdU levels.

Oxidative DNA damage has been suggested previously to be associated with increased cancer risk (16). Levels of such damage can be envisioned to represent the net result of oxidant exposures via environmental or endogenous sources and DNA repair ability. For example, family history of breast cancer in primary relatives has uniformly been shown to be associated with decreased DNA repair ability of lymphocytes (3, 4, 5, 6).

Some studies, but not all, suggest that levels of oxidative DNA damage are higher in human breast tumor tissue than in normal breast tissue (17, 18, 19, 20). This is also evident in blood, with higher levels of oxidative DNA damage having been indicated in breast cancer cases versus controls (1, 2). In the study of Frenkel et al. (2), blood was obtained 0.5–6 years before the diagnosis of breast cancer was made, which helps to implicate oxidative damage in cancer etiology. In that study,women with self-reported benign breast disease (chiefly fibrocystic disease) or a family history of breast cancer with no reported breast problems also exhibited higher serum antibodies to 5-OHmdU versus controls, which is a measure of 5-OHmdU levels in the DNA of WBCs (2).

There is sparse literature available on the association of high-risk breast lesions with biomarkers of oxidative stress and cancer risk. In one study, lymphocytes of eight women with preinvasive lesions of the breast exhibited decreased repair of X-ray-induced DNA damage (4). In another study, women with dense mammogram patterns, which are associated with increased breast cancer risk, were shown to have relatively increased levels of malondialdehyde in their urine, indicating increased levels of systemic lipid peroxidation (21). These studies are consistent with our findings of relatively increased oxidative DNA damage levels in blood of women with high risk breast lesions

The low-risk control women in our study may not be true controls because a biopsy was performed. There is evidence that a history of breast biopsy places a woman at increased breast cancer risk (22). Thus, if we had been able to enroll women with clear mammograms with no reason for biopsy, levels of oxidative DNA damage may have been even lower. The lesions we categorized as benign,however, have typically been associated with very low or no risk for subsequent invasive breast cancer (23).

To the best of our knowledge, this is the first report of oxidative DNA damage levels in blood of women with atypical hyperplasia or carcinoma in situ, both of which are associated with a substantial increase in risk of invasive breast cancer (10). The observations that these women have mean levels of 5-OHmdU similar to those of women with invasive breast cancer indicates that oxidative DNA damage may play a role in placing these women at increased risk for breast cancer.