Abstract

Global decreases in DNA methylation, particularly in repetitive elements, have been associated with genomic instability and human cancer. Emerging, though limited, data suggest that in white blood cell (WBC) DNA levels of methylation, overall or in repetitive elements, may be associated with cancer risk. We measured methylation levels of three repetitive elements [Satellite 2 (Sat2)], long interspersed nuclear element-1 (LINE-1) and Alu) by MethyLight, and LINE-1 by pyrosequencing in a total of 282 breast cancer cases and 347 unaffected sisters from the New York site of the Breast Cancer Family Registry (BCFR) using DNA from both granulocytes and total WBC. We found that methylation levels in all markers were correlated between sisters (Spearman correlation coefficients ranged from 0.17 to 0.55). Sat2 methylation was statistically significantly associated with increased breast cancer risk [odds ratio (OR) = 2.09, 95% confidence interval (CI) = 1.09–4.03; for each unit decrease in the natural log of the methylation level, OR = 2.12, 95% CI = 0.88–5.11 for the lowest quartile compared with the highest quartile]. These associations were only observed in total WBC but not granulocyte DNA. There was no association between breast cancer and LINE-1 and Alu methylation. If replicated in larger prospective studies, these findings support that selected markers of epigenetic changes measured in WBC, such as Sat2, may be potential biomarkers of breast cancer risk.

Introduction

Extensive epidemiological evidence has linked many risk factors to breast cancer, including reproductive history, exogenous hormone use and alcohol intake ( 1 ). One plausible mechanism by which these exogenous factors alter breast cancer susceptibility is through epigenetic effects on somatic cells, leading to activation or silencing of key genes in critical pathways ( 2 ). DNA methylation is a type of epigenetic change that plays an important role in cancer etiology by silencing tumor suppressor genes through hypermethylation or activating oncogenes through hypomethylation ( 3–5 ). Three mechanisms by which genomic hypomethylation contributes to malignancy have been proposed, including activation of oncogenes and/or transposable elements, loss of imprinting, and promotion of chromosomal instability ( 6 ).

DNA methylation typically occurs in CpG dinucleotides. Up to 80% of CpG dinucleotides occur in repetitive sequences. There are different types of repetitive sequences scattered throughout the genome (e.g. satellite repeat, short interspersed nuclear element, and LINE) ( 7 ). These sequences comprise approximately half the DNA genomic content and some of these elements have the ability to integrate themselves in alternate genomic regions disrupting expression of important genes, eventually leading to genomic instability. One of the mechanisms used by the cellular genome to control the potentially deleterious presence of these sequences is to keep these sequences highly methylated ( 8 , 9 ). In cancer cells, loss of methylation has been observed for elements Sat2, LINE-1 and Alu ( 10 , 11 ).

There is increasing evidence that DNA methylation patterns are altered not only within tumors but in some cases in precursor lesions or throughout the body ( 5 , 12 , 13 ). Hypomethylation across the genome and/or in selected repetitive elements was observed in the peripheral blood of bladder ( 14 , 15 ), colorectal ( 16 ), breast ( 17 ) and head and neck squamous cell carcinoma cases compared with controls ( 18 ) (reviewed in ref. 19). We previously found significant correlations in methylation levels between tumor tissue and DNA from WBC of three repetitive elements, Sat2, LINE-1 and Alu, in 40 breast cancer cases ( 20 ). In addition, methylation of Sat2 in WBC was significantly lower in these same cases compared with controls ( 20 ).

Using information from 282 breast cancer cases in families participating in the New York site of the BCFR and 347 sister controls unaffected by breast cancer, we examined whether repetitive element DNA methylation in peripheral blood cell DNA is associated with breast cancer risk using both MethyLight (for Sat2, LINE-1 and Alu) and pyrosequencing methods (LINE-1). MethyLight and pyrosequencing are both common methods used for measuring region-specific methylation in epidemiological studies, but the sequences of LINE-1 measured in both assays are different ( 21 , 22 ). It is unclear whether the different methods for measuring LINE-1 will affect the association of DNA methylation and breast cancer risk. Because we previously have found that WBC methylation may differ by source of DNA ( 23 ), we only include families where both sisters have WBC DNA from the same source.

Materials and methods

Study participants

The New York site of the BCFR is part of a six-site international registry (California, New York, Philladelphia and Utah in the USA; Ontario in Canada; and Melbourne and Sydney in Australia) (for details see ref. 24). At the New York site, we recruited high-risk breast and/or ovarian cancer families from clinical and community settings. All family members who participated in the Registry completed an epidemiological questionnaire, to provide information on demographics, ethnicity, smoking, alcohol consumption, reproductive history, hormone use, weight, height and physical activity, and a self-administered food frequency questionnaire. We collected blood from participants at the time of recruitment to permit the isolation of plasma and WBC fractions. For this study, we used a family-based study design with sister sets discordant for breast cancer ( 25–27 ). Specifically, we examined 282 breast cancer cases and 347 unaffected sisters, with a total of 260 sibling sets who had the same cell type of peripheral blood DNA available for methylation analysis. The study was approved by Columbia University’s Institutional Review Board.

DNA extraction and bisulfite treatment

Genomic DNA was extracted from total WBC or from granulocytes by a salting out procedure. DNA was extracted from isolated WBCs and granulocytes isolated after centrifugation over ficoll. There were 173 sister sets with WBC and 87 pairs with granulocytes. As blood processing protocols changed over time, either granulocyte or total WBC DNA was available for a given individual. We only included sister sets that were recruited within the same time period so that identical blood processing was used for both sisters. Differential blood counts were not available. Cells were lysed with SDS in a nuclei lysis buffer and treated with RNase A (final 133 µg/ml) and RNase T1 (final 20U/ml) to remove RNA. Proteins were coprecipitated with NaCl (330 µl of saturated NaCl added per 1ml solution) by centrifugation. Genomic DNA was recovered from the supernatant by precipitation with 100% ethanol, washed in 70% ethanol and dissolved in the Tris–EDTA buffer.

Aliquots of DNA (500ng) were bisulfite-treated with the EZ DNA methylation kit (Zymo Research, Orange, CA) to convert unmethylated cytosines to uracils while leaving methylated cytosines unmodified. The DNA was resuspended in 20 µl of distilled water and stored at −20°C until use.

MethyLight assay

We used the sequences of probes and forward and reverse primers of Sat2-M1, LINE-1-M1 and Alu-M2 as described in Weisenberger et al. ( 21 ). These regions were selected because their level of methylation measured in the MethyLight assay has been shown to be significantly correlated with HPLC-based global DNA methylation measurements ( 21 ). PCR was performed in a 10 ul reaction volume with 0.3 uM forward and reverse PCR primers, 0.1 ul probe, 3.5 uM MgCl 2 , using the following PCR program: 95°C for 10min, then 55 cycles of 95°C for 15 s followed by 60°C for 1min. Assays were run on an ABI Prism 7900 Sequence Detection System (Perkin-Elmer, Foster City, CA).

Universal methylated DNA served as a methylated reference, and the Alu-based control reaction (AluC4) was used to measure the levels of input DNA to normalize the signal for each methylation reaction. We checked the Ct values of AluC4 by case-control status and found no difference by case status within the same source of DNA. In addition, age did not affect Ct values for AluC4 within the same source of DNA. MethyLight data specific for the repetitive elements were expressed as percent of methylated reference (PMR) values.

PMR = 100% * 2 exp–[Delta Ct (target gene in sample − control gene in sample) − Delta Ct (100% methylated target in reference sample − control gene in reference sample)]

Each MethyLight reaction was performed in duplicate, and the PMR values represent the mean. The inter-assay coefficients of variation based on duplicate threshold cycle measures were 0.4, 0.4 and 0.6 for Sat-2M1, LINE-1 and Alu-M2, respectively.

Pyrosequencing assay

Pyrosequencing for LINE-1 methylation levels was carried out using PCR and sequencing primers as previously described, with minor modifications to the original protocol ( 22 ). Briefly, PCR was carried out in a 25 µl reaction mix containing 50ng bisulfite-converted DNA, 1× Pyromark PCR Master Mix (Qiagen, Valencia, CA), 1× Coral Load Concentrate (Qiagen) and 0.2 uM forward and reverse primers, using the following PCR program: 95°C for 15min, then 44 cycles of 95°C for 30 s followed by 56°C for 30 s and 72°C for 30 s, with a final extension at 72°C for 10min. Each set of amplifications included bisulfite-converted CpG universal methylated, unmethylated and non-template controls.

Following amplification, the biotinylated PCR products were purified and incubated with the sequencing primer designed to bind adjacent to the CpG sites of interest. Pyrosequencing was conducted using a PyroMark Q24 instrument (Qiagen), with subsequent quantitation of methylation levels determined with the PyroMark Q24 1.010 software. Relative peak height differences were used to calculate the percentage of methylated cytosines at each given site. Percent methylation within a sample was subsequently determined by averaging across all three interrogated CpG sites in the analysis. Non-CpG cytosine residues were used as internal controls to verify efficient sodium bisulfite DNA conversion. The inter-assay coefficient of variation was 1.0.

Statistical methods

We examined DNA methylation levels for each marker both as a continuous and categorical variable. We took the natural logarithm of the continuous methylation levels of the three repetitive elements by MethyLight because these were not normally distributed and we wanted to first evaluate the association between risk factors and these continuous measures as outcomes. We categorized each DNA methylation marker based on quartiles of the distribution of each marker. However, when we categorize the continuous values, whether or not we take the natural logarithm, the categorization of individuals into categories will be preserved. When we modeled methylation levels as continuous measures, higher values indicate more methylation, lower values are indicative of hypomethylation.

We assessed the Spearman correlation coefficients between methylation levels for each of the markers to determine the correlation of each marker between sisters discordant on breast cancer status. We used conditional logistic regression stratified on the family identification number to model the association between DNA methylation and breast cancer risk within sisters. We first modeled the association adjusted for age and then assessed confounding separately for each association of Sat2, LINE-1 and Alu on breast cancer risk. Potential confounders included age at blood draw, BMI (<25kg/m 2 versus ≥ 25kg/m 2 ), race/ethnicity (coded as White non-Hispanic, Hispanic and Other), age at menarche, parity, age at first pregnancy, oral contraceptive use (ever use and never), hormone replacement therapy use (ever use and never)), smoking status (never, former and current), alcohol intake, BRCA1/2 mutation status and menopausal status. We considered these priori confounders in the multivariable model but then simplified each model by excluding each construct if it did not result in at least a 10% change in the association between the DNA methylation variables and breast cancer risk ( 28 ).

In addition to the conditional logistic regression models that examined within-family differences in methylation markers and breast cancer risk, we also assessed population average effects of the methylation markers on breast cancer risk through the use of generalized estimating equations (GEE) and assuming a logit link to model the ORs ( 29 ). We were able to include women who did not have sibling controls in the GEE analyses. All analyses were performed with SAS software 9.2 (SAS Institute, Cary, NC).

Results

Table 1 presents the distributions of selected characteristics and methylation markers by cases and unaffected sisters. The distributions of race and age at blood draw were similar between cases and their unaffected sister controls. As expected, BRCA mutation frequencies were higher in cases than in controls (8.2 versus 3.7% for BRCA1 and 4.3 versus 2.6% for BRCA2 ). Mean levels of methylation in the three repetitive elements, measured by the MethyLight assay in both granulocytes and total WBC DNA, were similar between cases and unaffected sisters. The mean level of Sat2 methylation was 49.7, 50.7, 36.9 and 39.5 for granulocyte DNA of cases and unaffected sisters and for total WBC of cases and unaffected sisters, respectively. The mean level of LINE-1 was 138.6 and 133.4 for cases and unaffected sisters, respectively, in granulocytes, and 91.2 and 94.8 in WBC DNA. The mean levels of Alu were 87.2 and 86.6 for granulocyte DNA of cases and unaffected sisters. The corresponding mean level of Alu in WBC DNA was 99.8 and 105.4 for cases and unaffected sisters, respectively. The mean level of LINE-1 pyrosequencing in granulocytes was 75.5 and 75.0 for cases and unaffected sisters, respectively. The corresponding level in total WBC was 74.0 and 74.1, respectively. The mean levels of DNA methylation did not differ by BRCA1 and 2 mutation status between affected and unaffected sisters (data not shown). The distributions of oral contraceptive use and hormone replacement use were similar between cases and their unaffected sister controls. There was no difference in methylation by oral contraceptive use, but individuals who have ever used hormone replacement therapy have lower methylation in LINE-1 (98.1 versus 110.4 for ever and never use, respectively) and Sat2 (38.6 versus 43.4 for ever and never use, respectively) measured by MethyLight.

Table 1

Demographic characteristics and DNA methylation of repetitive elements of breast cancer cases and unaffected sisters, the New York site of the BCFR

  Cases, N = 282   Unaffected sisters, N = 347  
N Mean (SD) or % N Mean (SD) or % 
Age 282 49.5 (11.4) 347 48.0 (11.1) 
Race 
 White 171 61.3 195 56.2 
 Hispanic 74 26.5 107 30.8 
 Other 34 12.2 45 13.0 
BRCA1 Mutation  
 + 23 8.2 13 3.7 
 − 259 91.8 334 96.3 
BRCA2 Mutation  
 + 12 4.3 2.6 
 − 270 95.7 338 97.4 
Sat2 266 41.3 (24.4) 333 43.5 (32.9) 
LINE-1 (MethyLight) 265 107.4 (63.6) 333 108.5 (59.1) 
Alu 266 95.5 (36.6) 334 98.7 (51.5) 
LINE-1 Pyrosequencing 279 74.5 (3.0) 340 74.5 (2.6) 
  Cases, N = 282   Unaffected sisters, N = 347  
N Mean (SD) or % N Mean (SD) or % 
Age 282 49.5 (11.4) 347 48.0 (11.1) 
Race 
 White 171 61.3 195 56.2 
 Hispanic 74 26.5 107 30.8 
 Other 34 12.2 45 13.0 
BRCA1 Mutation  
 + 23 8.2 13 3.7 
 − 259 91.8 334 96.3 
BRCA2 Mutation  
 + 12 4.3 2.6 
 − 270 95.7 338 97.4 
Sat2 266 41.3 (24.4) 333 43.5 (32.9) 
LINE-1 (MethyLight) 265 107.4 (63.6) 333 108.5 (59.1) 
Alu 266 95.5 (36.6) 334 98.7 (51.5) 
LINE-1 Pyrosequencing 279 74.5 (3.0) 340 74.5 (2.6) 
Table 2

Within-family correlation between affected sisters and unaffected sisters of repetitive element DNA methylation markers, the New York site of the BCFR

  Granulocyte, N = 87   Total WBC, N = 173   Pooled source, N = 260  
N r P N r P N r P 
Sat2 82 0.42 <0.0001 168 0.55  250 0.50 <0.0001 
LINE-1 82 0.27 0.01 167 0.48 <0.0001 249 0.43 <0.0001 
Alu 82 0.49 <0.0001 168 0.46 <0.0001 250 0.49 <0.0001 
LINE-1 Pyrosequencing 87 0.17 0.12 173 0.40 <0.0001 260 0.37 <0.0001 
  Granulocyte, N = 87   Total WBC, N = 173   Pooled source, N = 260  
N r P N r P N r P 
Sat2 82 0.42 <0.0001 168 0.55  250 0.50 <0.0001 
LINE-1 82 0.27 0.01 167 0.48 <0.0001 249 0.43 <0.0001 
Alu 82 0.49 <0.0001 168 0.46 <0.0001 250 0.49 <0.0001 
LINE-1 Pyrosequencing 87 0.17 0.12 173 0.40 <0.0001 260 0.37 <0.0001 

The Spearman correlations of methylation markers between discordant sisters are shown in Table 2 . Methylation levels were modestly and significantly correlated between sisters for all measures and by source of DNA, with the exception of methylation in granulocyte DNA measured by LINE-1 pyrosequencing. The Spearman correlation coefficients ranged from 0.17 to 0.55. Coefficients in granulocytes, total WBC and pooled sources of DNAs were 0.42, 0.55 and 0.50 for Sat 2; 0.27, 0.48 and 0.43 for LINE-1 by MethyLight; and 0.49, 0.46 and 0.49 for Alu. Coefficients in granulocytes, total WBC and pooled sources of DNAs were 0.17, 0.40 and 0.37, respectively, for LINE-1 pyrosequencing. Levels of LINE-1 methylation measured by MethyLight and pyrosequencing were only very modestly correlated with a correlation of 0.16.

The ORs from conditional logistic regression models of the association between genomic methylation and breast cancer risk are presented in Table 3 . We observed evidence of an association between Sat2 methylation and breast cancer risk. When Sat2 methylation was modeled in quartiles, we found a non-statistically significant positive dose–response relationship (indicating higher breast cancer risk with lower levels of DNA methylation) for total WBC DNA and the pooled sources of DNA but not separately for the granulocyte DNA (OR = 2.12, 95% CI = 0.88–5.11; OR = 1.62, 95% CI = 0.86–3.05; and OR 0.92, 95% CI = 0.34–2.47, respectively, for the lowest quartile compared with the highest quartile). In total WBC DNA, the association between Sat2 methylation as a continuous measure and breast cancer risk was statistically significant (OR per 1 unit decrease in the natural log of Sat2 methylation = 2.09, 95% CI = 1.09–4.03). Overall, we did not observe an association between breast cancer and methylation of LINE-1 and Alu, both measured by MethyLight, and LINE-1 measured by pyrosequencing modeled as categorical or continuous variables.

Table III.

Repetitive element DNA methylation and breast cancer risk, overall and stratified by source of DNA, the New York site of the BCFR

Conditional Logistic (Discordant Case/Control) OR (95% CI) 
  OR c (95%CI)   OR d (95%CI)   OR d (95%CI)  
Sat2  Pooled ( N = 593)   Granulocyte ( N = 206)   Total WBC ( N = 387)  
Q1 (<=25.8) 1.62 (0.86–3.05) 0.92 (0.34–2.47) 2.12 (0.88–5.11) 
Q2 (>25.8–37.7) 1.44 (0.80–2.61) 1.73 (0.67–4.44) 1.44 (0.64–3.23) 
Q3 (>37.7–52.5) 1.34 (0.79–2.25) 1.44 (0.65–3.23) 1.37 (0.67–2.80) 
Q4 (>52.5) 1.00 1.00 1.00 
P trend  0.15 0.84 0.11 
OR per 1 unit decrease in log(Sat2) 1.44 (0.94–2.20) 1.04 (0.59–1.85) 2.09 (1.09–4.03) 
LINE-1   Pooled ( N = 590)   Granulocyte ( N = 205)   Total WBC ( N = 385)  
Q1 (<=72.9) 0.94 (0.51–1.72) 1.19 (0.48– 2.96) 0.97 (0.41–2.32) 
Q2 (>72.9–99.3) 1.02 (0.58–1.79) 0.69 (0.25–1.91) 1.17 (0.54–2.55) 
Q3 (>99.3–125.7) 0.90 (0.53–1.51) 0.64 (0.28–1.46) 1.11 (0.54–2.28) 
Q4 (>125.7) 1.00 1.00 1.00 
P trend  0.95 0.92 0.92 
OR per 1 unit decrease in log (LINE-1) 0.99 (0.63–1.56) 0.94 (0.50–1.77) 1.11 (0.57–2.14) 
Alu  Pooled ( N = 593)   Granulocyte ( N = 206)   Total WBC ( N = 387)  
Q1 (<=73.0) 1.16 (0.63–2.13) 0.90 (0.35–2.30) 1.42 (0.62–3.27) 
Q2 (>73.0–92.9) 0.76 (0.40–1.42) 0.33 (0.09–1.19) 1.06 (0.49–2.26) 
Q3 (>92.9–117.8) 1.21 (0.72–2.02) 1.07 (0.33–3.45) 1.35 (0.75–2.43) 
Q4 (>117.8) 1.00 1.00 1.00 
P trend  0.73 0.86 0.46 
OR per 1 unit decrease in log(Alu) 1.18 (0.72–1.93) 1.13 (0.61–2.09) 1.48 (0.61–3.61) 
LINE-1 Pyrosequencing  Pooled ( N = 613)   Granulocyte ( N = 217)   Total WBC ( N = 396)  
Q1 (<=73.4) 0.81 (0.46–1.45) 0.45 (0.14–1.38) 0.93 (0.46–1.89) 
Q2 (>73.4–75.0) 0.58 (0.34–0.97) 0.88 (0.39–1.97) 0.44 (0.22–0.88) 
Q3 (>75.0–76.1) 0.60 (0.37–1.00) 0.73 (0.34–1.56) 0.52 (0.26–1.04) 
Q4 (>76.1) 1.00 1.00 1.00 
P trend  0.33 0.30 0.73 
OR per 1 unit decrease 0.97 (0.90–1.05) 0.89 (0.75–1.05) 1.01 (0.92-1.11) 
Conditional Logistic (Discordant Case/Control) OR (95% CI) 
  OR c (95%CI)   OR d (95%CI)   OR d (95%CI)  
Sat2  Pooled ( N = 593)   Granulocyte ( N = 206)   Total WBC ( N = 387)  
Q1 (<=25.8) 1.62 (0.86–3.05) 0.92 (0.34–2.47) 2.12 (0.88–5.11) 
Q2 (>25.8–37.7) 1.44 (0.80–2.61) 1.73 (0.67–4.44) 1.44 (0.64–3.23) 
Q3 (>37.7–52.5) 1.34 (0.79–2.25) 1.44 (0.65–3.23) 1.37 (0.67–2.80) 
Q4 (>52.5) 1.00 1.00 1.00 
P trend  0.15 0.84 0.11 
OR per 1 unit decrease in log(Sat2) 1.44 (0.94–2.20) 1.04 (0.59–1.85) 2.09 (1.09–4.03) 
LINE-1   Pooled ( N = 590)   Granulocyte ( N = 205)   Total WBC ( N = 385)  
Q1 (<=72.9) 0.94 (0.51–1.72) 1.19 (0.48– 2.96) 0.97 (0.41–2.32) 
Q2 (>72.9–99.3) 1.02 (0.58–1.79) 0.69 (0.25–1.91) 1.17 (0.54–2.55) 
Q3 (>99.3–125.7) 0.90 (0.53–1.51) 0.64 (0.28–1.46) 1.11 (0.54–2.28) 
Q4 (>125.7) 1.00 1.00 1.00 
P trend  0.95 0.92 0.92 
OR per 1 unit decrease in log (LINE-1) 0.99 (0.63–1.56) 0.94 (0.50–1.77) 1.11 (0.57–2.14) 
Alu  Pooled ( N = 593)   Granulocyte ( N = 206)   Total WBC ( N = 387)  
Q1 (<=73.0) 1.16 (0.63–2.13) 0.90 (0.35–2.30) 1.42 (0.62–3.27) 
Q2 (>73.0–92.9) 0.76 (0.40–1.42) 0.33 (0.09–1.19) 1.06 (0.49–2.26) 
Q3 (>92.9–117.8) 1.21 (0.72–2.02) 1.07 (0.33–3.45) 1.35 (0.75–2.43) 
Q4 (>117.8) 1.00 1.00 1.00 
P trend  0.73 0.86 0.46 
OR per 1 unit decrease in log(Alu) 1.18 (0.72–1.93) 1.13 (0.61–2.09) 1.48 (0.61–3.61) 
LINE-1 Pyrosequencing  Pooled ( N = 613)   Granulocyte ( N = 217)   Total WBC ( N = 396)  
Q1 (<=73.4) 0.81 (0.46–1.45) 0.45 (0.14–1.38) 0.93 (0.46–1.89) 
Q2 (>73.4–75.0) 0.58 (0.34–0.97) 0.88 (0.39–1.97) 0.44 (0.22–0.88) 
Q3 (>75.0–76.1) 0.60 (0.37–1.00) 0.73 (0.34–1.56) 0.52 (0.26–1.04) 
Q4 (>76.1) 1.00 1.00 1.00 
P trend  0.33 0.30 0.73 
OR per 1 unit decrease 0.97 (0.90–1.05) 0.89 (0.75–1.05) 1.01 (0.92-1.11) 

a Adjusted for age at blood draw, smoking status.

b Stratified by source, adjusted for age at blood draw, smoking status.

c Adjusted for age at blood draw.

d Stratified by source, adjusted for age at blood draw.

The results from the GEE models were consistent with those from the conditional logistic models. For example, with each 1 unit decrease in the natural log of Sat2 methylation level, the OR for breast cancer increased by 1.39 (95% CI = 1.04–1.85) for total WBC DNA. Decreased methylation of LINE-1 and Alu, modeled as continuous and categorical variables, was not associated with breast cancer risk in GEE models (data not shown).

Given that the blood samples were collected after breast cancer diagnosis, we further stratified our final conditional logistic regression models to see whether time since diagnosis had an impact on the overall findings (age at diagnosis <2 years versus ≥2 years) ( Figure 1 ). In these stratified models, methylation levels in Sat2 were associated with breast cancer risk ( Figure 1A ). For cases with blood collected within 2 years of diagnosis, lower levels of Sat2 methylation measured in both granulocytes and total WBC were associated with increased breast cancer risk, but these associations were not statistically significant (OR = 1.89 per one unit change, 95% CI = 0.81–4.42; OR = 1.41 per one unit change, 95% CI = 0.43–4.58, respectively). For cases with more than 2 years between blood draw and diagnosis, lower Sat2 DNA methylation in total WBC DNA was significantly associated with increased breast cancer risk. With each 1 unit decrease in Sat2 methylation level, the OR increased by 2.41 (95% CI = 1.09–5.37). There were no associations observed for LINE-1 ( Figure 1B and 1D ) and Alu ( Figure 1C ).

Fig. 1.

Repetitive element DNA methylation and breast cancer risk (pooled and by separate source) using multivariable conditional logistic regression models, the New York site of the BCFR by date of blood draw after diagnosis. ( A ) Sat2 and breast cancer risk (OR per 1 unit change in the log(Sat2)). ( B ) LINE-1 and breast cancer risk (OR per 1 unit change in the log(LINE-1)). ( C ) Alu and breast cancer risk (OR per 1 unit change in the log(Alu)). ( D ) LINE-1 measured by pyrosequencing and breast cancer risk (OR per 1 unit change in LINE-1).

Fig. 1.

Repetitive element DNA methylation and breast cancer risk (pooled and by separate source) using multivariable conditional logistic regression models, the New York site of the BCFR by date of blood draw after diagnosis. ( A ) Sat2 and breast cancer risk (OR per 1 unit change in the log(Sat2)). ( B ) LINE-1 and breast cancer risk (OR per 1 unit change in the log(LINE-1)). ( C ) Alu and breast cancer risk (OR per 1 unit change in the log(Alu)). ( D ) LINE-1 measured by pyrosequencing and breast cancer risk (OR per 1 unit change in LINE-1).

Discussion

We found that decreased Sat2 methylation was associated with an increased breast cancer risk. The association was observed in total WBC and not granulocyte DNA. This association was observed when considering Sat 2 methylation both as a continuous variable and as a categorical variable, but was only statistically significant for the former. There was no association of LINE-1 or Alu methylation levels with breast cancer risk, although there was an inverse association observed for the second quartile as measured by LINE-1 pyrosequencing. Given the lack of difference between the highest and lowest quartile and the lack of association between LINE-1 and the inconsistent pattern with LINE-1 measured from the MethyLight assay, this association with the second quartile may be spurious. Our finding that Sat 2 methylation may be associated with cancer risk while LINE-1 methylation is consistent with another case–control study showing lower 5-methylcytosine content but not LINE-1 hypomethylation in leukocyte DNA from breast cancer patients compared with cancer-free controls (17).

It is well known that the 5-methylcytosine content of DNA is tissue-specific, which may explain differences in studies that use DNA from different sources ( 30 ). However, the relevance of variability in the levels of DNA methylation in an individual’s different blood cell types is still unclear. Genome-wide DNA methylation profiling of ovarian cancer cases and unrelated controls demonstrated that there is altered gene expression patterns in granulocytes and lymphocytes that corresponds with differential methylation in these cells ( 31 ). In a previous study, we measured the global methylation levels by the [ 3 H]-methyl acceptor assay and repetitive element (LINE-1, Sat2 and Alu) DNA methylation by the MethyLight assay and found that global methylation of granulocyte DNA measured by these assays was not correlated with that of total WBC DNA, except for methylation in Sat2 ( 23 ). Other studies have similarly found that LINE-1 methylation levels correlate differently with the percent of neutrophils and lymphocytes, suggesting that differential DNA methylation of some repetitive elements might be associated with blood cell type counts ( 32 ). There is ample evidence to support that disease states can lead to alterations in the percent of different blood cell fractions, particularly in cancer. In this study, the effect of Sat2 hypomethylation on breast cancer risk was mainly observed in total WBC DNA and not in the granulocyte fraction, highlighting the importance of considering blood cell types when studying global DNA methylation as a predictor of cancer risk.

Evidence suggests that both genetic and environmental factors affect the patterns of methylation. Familial clustering of methylation changes ( 33 ) suggests that methylation stability might be directly related to genetic variation, such as in genes controlling one-carbon metabolism or DNA methyltransferase activity. We also observed a modest correlation in genomic DNA methylation within sisters that was consistent across measures of methylation and source of DNA. These modest correlations suggest that most of the within-sister differences in methylation levels may be explained by endogenous and exogenous factors that differ between sisters across the life course. A study by Fraga et al. provides evidence that monozygotic twins are epigenetically indistinguishable during the early years of life, but older monozygotic twins exhibit remarkable differences in overall content and genomic distribution of 5-methylcytosine ( 34 ). WBC DNA methylation has been associated with several breast cancer risk factors that change for women across the life course ( 19 , 35 ).

Most epidemiological studies of genomic methylation have focused on LINE-1 and Alu methylation and have reported that decreased LINE-1 and/or Alu methylation in WBC has been associated with selected environmental exposures ( 36–41 ). Most studies that measure the methylation levels of these two elements use pyrosequencing, a method that relies on DNA sequencing to accurately quantify the methylation level of specific sites. In this study, we used pyrosequencing as well as MethyLight to determine LINE-1 methylation levels and found little suggestion of an association with breast cancer risk within sisters independent of the methodology.

Studies of DNA methylation and disease endpoints differ by marker type. For example, Choi et al. found that lower levels of LINE-1 methylation were not associated with breast cancer risk, while overall genomic hypomethylation was associated with breast cancer risk ( 17 ). The distinct behavior of the different markers may have important implications for the use of DNA methylation of repetitive elements as a surrogate marker of global DNA methylation ( 42 ). The same satellite sequence might be demethylated in some but not all tumor types ( 43 ). Sat2 DNA sequences are small tandem repeats located in the pericentromeric and juxtacentromeric regions of Chr1. Therefore, they represent a different set of CpG sites than the LINE-1 and Alu elements, and this may partly explain the different results found for the different elements. About 15–40% of human breast cancers exhibit pericentromeric rearrangements in Chr1, i.e. rearrangements of the heterochromatin in or adjacent (juxtacentromeric) to the centromere of this chromosome ( 44 , 45 ). Sat2 hypomethylation has been postulated as the cause for Chr1 instability, a hallmark of cancer. Extensive cancer-associated hypomethylation of juxtacentromeric satellite DNA and global DNA hypomethylation is common even in grade-1 or stage-1 breast adenocarcinomas, which suggests that hypomethylation of these sequences is an early event in breast carcinogenesis ( 46 , 47 ).

One of the biggest challenges in large epidemiological studies is identifying an assay that is both cost effective and valid. This is particularly important for measures of genomic methylation, which all have some degree of bias unless they are whole genome-based ( 42 ). To minimize this limitation, we selected two assays to measure DNA methylation levels of LINE-1. The assays we used in the present study had high reliability and all laboratory analyses were conducted blinded to outcome and exposure information. Therefore, we expect any misclassification of methylation to be non-differential with respect to the outcome. Some of the percentages of methylation levels in LINE-1 were above 100%. It is well known that populations are polymorphic for the presence of LINE-1 elements and the human genome contains more than 10 5 truncated LINE-1 elements ( 48 ). Since values in the MethyLight assay are calculated versus a fully methylated control sample, differences in copy number between sample and control can lead to values >100%.

As with most epidemiological studies, we were limited by only having collected blood specimens at a single time point for the overall study. Because WBC progenitor cells are rapidly dividing, they may respond more quickly to factors that influence DNA methylation than cells that turn over more slowly. For this reason, WBC DNA methylation may not identify DNA methylation errors directly indicative of cancer, a process that occurs over decades, but rather may serve as an early biomarker for biological processes that systemically influence DNA methylation. Retrospective studies are limited for inference because treatment such as chemotherapy may have affected the blood levels. However, studies that have comparing levels of DNA methylation in LRE1, a LINE region, in whole blood of patients with head and neck squamous cell carcinoma in blood drawn before and after surgery have not observed differences in methylation levels before and after ( 18 ). We observed a stronger association between Sat2 in total WBC and breast cancer risk among women whose breast cancer diagnosis was more than 2 years before blood collection. The longer time interval since chemotherapy may reduce its effect on blood methylation levels

Despite these limitations, this study has numerous strengths. We were able to compare within the same study population methylation levels from the most commonly used assays in epidemiological studies. We were also able to compare associations from assays using different sources of DNA, which enabled us to investigate the role of DNA methylation patterns in particular blood cell types as a biomarker of breast cancer. The sister set design allowed for control of fixed family-level effects that do not vary between sisters, such as race/ethnicity; it also allowed for tighter control of factors, such as family socioeconomic status, that vary less between sisters than unrelated individuals. We were also able to compare our results using both conditional logistic regression approaches and generalized estimating equations. The overall inferences from these two approaches were very similar, suggesting that there was minimal unmeasured fixed family-level confounding of these associations.

In summary, our study further highlights the importance of using the same source of blood DNA within a study when measuring methylation and of reporting the type of blood DNA used in publications. We found that Sat2 methylation, but not other DNA methylation markers, measured in peripheral blood was associated with breast cancer risk. If replicated in larger prospective studies, this study adds to the growing literature supporting the potential of using selected markers of methylation in WBC as biomarkers of cancer risk.

Acknowledgements

This work was supported by an award from the Breast Cancer Research Foundation and NIH grants U01 CA69398, P30 CA13696, P30 ES009089 and K07 CA131094. This work was also supported by the National Cancer Institute, National Institutes of Health, under RFA #CA-06-503 and through cooperative agreements with members of the BCFR and Principal Investigators, including Columbia University (U01 CA69398). The content of this manuscript does not necessarily reflect the views or policies of the National Cancer Institute or any of the collaborating centers in the BCFR, nor does mention of trade names, commercial products or organizations imply endorsement by the United States Government or the BCFR.

Conflicts of Interest: None declared.

References

1.
Dumitrescu
R.G.
et al
2005
Understanding breast cancer risk – where do we stand in 2005?
J. Cell. Mol. Med.
  ,
9
,
208
221
.
2.
Herceg
Z.
2007
Epigenetics and cancer: towards an evaluation of the impact of environmental and dietary factors.
Mutagenesis
  ,
22
,
91
103
.
3.
Jones
P.A.
et al
2002
The fundamental role of epigenetic events in cancer.
Nat. Rev. Genet.
  ,
3
,
415
428
.
4.
Ehrlich
M.
2002
DNA methylation in cancer: too much, but also too little.
Oncogene
  ,
21
,
5400
5413
.
5.
Tycko
B.
2003
Genetic and epigenetic mosaicism in cancer precursor tissues.
Ann. N. Y. Acad. Sci.
  ,
983
,
43
54
.
6.
Frühwald
M.C.
et al
2002
Global and gene-specific methylation patterns in cancer: aspects of tumor biology and clinical potential.
Mol. Genet. Metab.
  ,
75
,
1
16
.
7.
Lander
E.S.
et al
;
International Human Genome Sequencing Consortium
.
2001
Initial sequencing and analysis of the human genome.
Nature
  ,
409
,
860
921
.
8.
Yu
F.
et al
2001
Methyl-CpG-binding protein 2 represses LINE-1 expression and retrotransposition but not Alu transcription.
Nucleic Acids Res.
  ,
29
,
4493
4501
.
9.
Steinhoff
C.
et al
2004
Transcriptional regulation of the human LINE-1 retrotransposon L1.2B.
Mol. Genet. Genom.
  ,
270
,
394
402
.
10.
Yoder
J.A.
et al
1997
Cytosine methylation and the ecology of intragenomic parasites.
Trends Genet.
  ,
13
,
335
340
.
11.
Hoffmann
M.J.
et al
2005
Causes and consequences of DNA hypomethylation in human cancer.
Biochem. Cell Biol.
  ,
83
,
296
321
.
12.
Laird
P.W.
2003
The power and the promise of DNA methylation markers.
Nat. Rev. Cancer
  ,
3
,
253
266
.
13.
Patel
A.
et al
2003
DNA methylation as a cancer-specific biomarker: from molecules to populations.
Ann. N. Y. Acad. Sci.
  ,
983
,
286
297
.
14.
Moore
L.E.
et al
2008
Genomic DNA hypomethylation as a biomarker for bladder cancer susceptibility in the Spanish Bladder Cancer Study: sa case-control study.
Lancet Oncol.
  ,
9
,
359
366
.
15.
Cash
H.L.
et al
2012
LINE-1 hypomethylation is associated with bladder cancer risk among nonsmoking Chinese.
Int. J. Cancer
  ,
130
,
1151
1159
.
16.
Lim
U.
et al
2008
Genomic methylation of leukocyte DNA in relation to colorectal adenoma among asymptomatic women.
Gastroenterology
  ,
134
,
47
55
.
17.
Choi
J.Y.
et al
2009
Association between global DNA hypomethylation in leukocytes and risk of breast cancer.
Carcinogenesis
  ,
30
,
1889
1897
.
18.
Hsiung
D.T.
et al
2007
Global DNA methylation level in whole blood as a biomarker in head and neck squamous cell carcinoma.
Cancer Epidemiol. Biomarkers Prev.
  ,
16
,
108
114
.
19.
Terry
M.B.
et al
2011
DNA methylation in white blood cells: association with risk factors in epidemiologic studies.
Epigenetics
  ,
6
,
828
837
.
20.
Cho
Y.H.
et al
2010
Aberrant promoter hypermethylation and genomic hypomethylation in tumor, adjacent normal tissues and blood from breast cancer patients.
Anticancer Res.
  ,
30
,
2489
2496
.
21.
Weisenberger
D.J.
et al
2005
Analysis of repetitive element DNA methylation by MethyLight.
Nucleic Acids Res.
  ,
33
,
6823
6836
.
22.
Yang
A.S.
et al
2004
A simple method for estimating global DNA methylation using bisulfite PCR of repetitive DNA elements.
Nucleic Acids Res.
  ,
32
,
e38
.
23.
Wu
H.C.
et al
2011
Global methylation profiles in DNA from different blood cell types.
Epigenetics
  ,
6
,
76
85
.
24.
John
E.M.
et al
;
Breast Cancer Family Registry
.
2004
The Breast Cancer Family Registry: an infrastructure for cooperative multinational, interdisciplinary and translational studies of the genetic epidemiology of breast cancer.
Breast Cancer Res.
  ,
6
,
R375
R389
.
25.
Kennedy
D.O.
et al
2005
DNA repair capacity of lymphoblastoid cell lines from sisters discordant for breast cancer.
J. Natl. Cancer Inst.
  ,
97
,
127
132
.
26.
Machella
N.
et al
2008
Double-strand breaks repair in lymphoblastoid cell lines from sisters discordant for breast cancer from the New York site of the BCFR.
Carcinogenesis
  ,
29
,
1367
1372
.
27.
Zipprich
J.
et al
2009
Plasma protein carbonyls and breast cancer risk in sisters discordant for breast cancer from the New York site of the Breast Cancer Family Registry.
Cancer Res.
  ,
69
,
2966
2972
.
28.
Greenland
S.
1989
Modeling and variable selection in epidemiologic analysis.
Am. J. Public Health
  ,
79
,
340
349
.
29.
Zeger
S.L.
et al
1988
Models for longitudinal data: a generalized estimating equation approach.
Biometrics
  ,
44
,
1049
1060
.
30.
Ehrlich
M.
et al
1982
Amount and distribution of 5-methylcytosine in human DNA from different types of tissues of cells.
Nucleic Acids Res.
  ,
10
,
2709
2721
.
31.
Teschendorff
A.E.
et al
2009
An epigenetic signature in peripheral blood predicts active ovarian cancer.
PLoS ONE
  ,
4
,
e8274
.
32.
Zhu
Z.Z.
et al
2011
Repetitive element hypomethylation in blood leukocyte DNA and cancer incidence, prevalence, and mortality in elderly individuals: the Normative Aging Study.
Cancer Causes Control
  ,
22
,
437
447
.
33.
Bjornsson
H.T.
et al
2008
Intra-individual change over time in DNA methylation with familial clustering.
JAMA
  ,
299
,
2877
2883
.
34.
Fraga
M.F.
et al
2005
Epigenetic differences arise during the lifetime of monozygotic twins.
Proc. Natl. Acad. Sci. U.S.A.
  ,
102
,
10604
10609
.
35.
Terry
M.B.
et al
2008
Genomic DNA methylation among women in a multiethnic New York City birth cohort.
Cancer Epidemiol. Biomarkers Prev.
  ,
17
,
2306
2310
.
36.
Kim
K.Y.
et al
2010
Association of low-dose exposure to persistent organic pollutants with global DNA hypomethylation in healthy Koreans.
Environ. Health Perspect.
  ,
118
,
370
374
.
37.
Rusiecki
J.A.
et al
2009
Cancer incidence among pesticide applicators exposed to permethrin in the Agricultural Health Study.
Environ. Health Perspect.
  ,
117
,
581
586
.
38.
Pilsner
J.R.
et al
2009
Influence of prenatal lead exposure on genomic methylation of cord blood DNA.
Environ. Health Perspect.
  ,
117
,
1466
1471
.
39.
Wright
R.O.
et al
2010
Biomarkers of lead exposure and DNA methylation within retrotransposons.
Environ. Health Perspect.
  ,
118
,
790
795
.
40.
Bollati
V.
et al
2007
Changes in DNA methylation patterns in subjects exposed to low-dose benzene.
Cancer Res.
  ,
67
,
876
880
.
41.
Pavanello
S.
et al
2009
Global and gene-specific promoter methylation changes are related to anti-B[a]PDE-DNA adduct levels and influence micronuclei levels in polycyclic aromatic hydrocarbon-exposed individuals.
Int. J. Cancer
  ,
125
,
1692
1697
.
42.
Nelson
H.H.
et al
2011
Global methylation in exposure biology and translational medical science.
Environ. Health Perspect.
  ,
119
,
1528
1533
.
43.
Choi
S.H.
et al
2009
Changes in DNA methylation of tandem DNA repeats are different from interspersed repeats in cancer.
Int. J. Cancer
  ,
125
,
723
729
.
44.
Lu
Y.J.
et al
1993
Direct chromosome analysis of 50 primary breast carcinomas.
Cancer Genet. Cytogenet.
  ,
69
,
91
99
.
45.
Pandis
N.
et al
1995
Chromosome analysis of 97 primary breast carcinomas: identification of eight karyotypic subgroups.
Genes. Chromosomes Cancer
  ,
12
,
173
185
.
46.
Jackson
K.
et al
2004
DNA hypomethylation is prevalent even in low-grade breast cancers.
Cancer Biol. Ther.
  ,
3
,
1225
1231
.
47.
Narayan
A.
et al
1998
Hypomethylation of pericentromeric DNA in breast adenocarcinomas.
Int. J. Cancer
  ,
77
,
833
838
.
48.
Ewing
A.D.
et al
2010
High-throughput sequencing reveals extensive variation in human-specific L1 content in individual human genomes.
Genome Res.
  ,
20
,
1262
1270
.

Abbreviations

    Abbreviations
  • BCFR

    Breast Cancer Family Registry

  • CI

    confidence interval

  • LINE-1

    long interspersed nucleotide element-1

  • OR

    odds ratio

  • Satellite 2

    Sat2

  • WBC

    white blood cell