No effect of weight loss on LINE-1 methylation levels in peripheral blood leukocytes from postmenopausal overweight women

Abstract

Objectives

Obesity and weight-loss are associated with methylation patterns in specific genes, but their effect on Long Interspersed Nuclear Elements (LINE-1) methylation, a measure of global methylation is largely unknown.

Methods

Three hundred overweight/obese post-menopausal women (50–75 years) were part of a completed, 1-year randomized controlled trial, comparing independent and combined effects of a reduced-calorie weight-loss diet, and exercise program, vs. control. DNA was extracted from peripheral blood leukocytes collected at baseline and 12-months, and LINE-1 methylation analyzed by pyrosequencing. We compared mean changes between groups using generalized estimating equations and examined effects of weight-loss on LINE-1 methylation using stratified analyses (gained weight/no weight-loss (N=84); <5% (N=45); 5–10% (N=45); >10% of baseline weight-loss (N=126)) within each arm, adjusted by blood cell counts. We also examined associations between LINE-1 methylation and previously measured biomarkers, and anthropometrics.

Results

No significant difference in LINE-1 methylation levels was detected in any intervention group vs. controls. The magnitude of weight-loss was not associated with LINE-1 methylation at 12-months. There were no associations between baseline characteristics of participants, or previously measured biomarkers, and LINE-1 methylation.

Conclusions

Our results suggest that lifestyle changes sufficient to significantly reduce weight over 12-months may not change LINE-1 DNA methylation levels.

Introduction

A number of chronic diseases such as cancer, diabetes and cardiovascular disease are associated with obesity. However, the underlying mechanisms linking obesity and increased risk for these conditions are not established DNA methylation is a heritable, reversible chemical/structural change that regulates gene activity in the absence of underlying changes to DNA sequence, involving the addition of methyl groups to cytosine to form 5-methyl-cytosine (5mC). Approximately 50% of the human genome is composed of repetitive sequences such as LINE (Long Interspersed Nuclear Elements), and Short Interspersed Nuclear Elements, including Alu.¹ Methylation levels of LINE-1 have been used as surrogate markers for global methylation status and play an important role in maintenance of genomic stability.^{2, 3} Disruption of the epigenetic profile is a feature of most cancers, and global hypomethylation is a risk factor for various cancers, including breast.^4–6 Studies in cardiovascular disease also demonstrated that global hypomethylation was associated with increased risk cardiovascular risk factors.⁷

Studies have examined the association between body mass index (BMI) and LINE-1 methylation levels with conflicting results,⁸ but there are few studies on the effect of weight loss on gene-specific or global methylation. To our knowledge, there are no studies examining the effect of weight loss on LINE-1 methylation levels in postmenopausal overweight/obese healthy women.

The primary aims of this study were to test the combined and independent effects over 12-months of a completed reduced calorie weight loss diet and an exercise program vs. control on LINE-1 methylation levels in peripheral blood leukocytes in post-menopausal overweight/obese women. We also stratified pre- and post-intervention LINE-1 methylation levels by weight loss (<5%, 5–10%, and >10% weight loss), compared to participants who did not change weight or who gained any weight.

Methods and procedures

This study is ancillary to the Nutrition and Exercise in Women (www.clinicaltrials.gov NCT00470119) study, a 12-month randomized controlled trial testing the effects of caloric restriction and/or exercise on circulating hormones and other outcomes. The study is described in detail elsewhere.⁹ Briefly, 438 postmenopausal, healthy overweight (BMI≥25 kg/m²), sedentary women aged 50–75 years, not taking hormonal therapy, were recruited through media and mass mailings. Exclusion criteria included: less than 100 min/week of moderate physical activity; diagnosed diabetes or other serious medical condition(s); postmenopausal hormone use; consumption of >2 alcoholic drinks/day; current smoking; participation in another structured weight loss program; contraindication to participation (e.g. abnormal exercise tolerance test, inability to attend sessions). Eligible women were randomized to one of: i) reduced-calorie dietary modification (N=118); ii) moderate-to-vigorous intensity aerobic exercise (N=117); iii) combined diet and exercise (N=117); or iv) control (no intervention) (N=87) (Figure 1). The dietary intervention was a modification of the Diabetes Prevention Program (DPP) and Look AHEAD (Action for Health in Diabetes) lifestyle behavior change programs with goals of: 1200–2000 kcal/day, <30% daily calories from fat, and 10% weight loss. The exercise intervention goal was 45 minutes of moderate-to-vigorous (≥4 metabolic equivalents [METs]) intensity exercise at a target heart rate of 70–85% observed maximum, 5 days/week. Participants attended three facility-based supervised sessions/week, and exercised 2 days/week at home.

Randomization was stratified by BMI (≥ or <30kg/m²) and race/ethnicity. A subsample of 300 participants with complete data from both time-points was chosen for this ancillary study: all women who lost ≥10% of starting weight (N=126); all women who gained some weight or had no weight loss (N=84); plus a random 30% sample of the remainder. The study was approved by the Institutional Review Board of the Fred Hutchinson Cancer Research Center, Seattle, WA and all participants signed informed consent.

Covariates

All study measures were obtained and analyzed by trained personnel who were blinded to participants’ randomization status. Fasting blood samples, anthropometrics and questionnaire data were collected at baseline (pre-randomization) and at 12-months, when the study was completed. Data included demographic information, height, weight, medical history, dietary intake, supplement use and physical activity data were collected. Body composition was measured by DXA (Dual-energy X-ray absorptiometry) whole-body scanner (GE Lunar, Madison, WI). Cardiorespiratory fitness (VO₂max) was assessed using a maximal graded treadmill test according to a modified branching protocol.¹⁰ Biomarkers including c-reactive protein insulin, glucose, adiponectin and leptin were analyzed, homeostatic model assessment (HOMA) score, a method used to quantify insulin resistance calculated, and complete blood counts were performed as previously described.^11–16

LINE-1 methylation

Fasting venous blood (50 mL) was collected during clinic visits (no exercise or non-steroidal anti-inflammatory medications within 24 hours, no alcohol within 48 hours). Blood was processed within 1 hour and samples were stored at −80°C. DNA was extracted from buffy coat preparations using the Qiagen Midi Kit (Qiagen, Valencia, CA).

DNA was extracted using Qiagen Midi prep kits (Qiagen). Unmethylated cytosine residues were converted to uracil using CpGnome DNA Modification Kit (Millipore, Billerica, MA, USA). LINE-1 was amplified using the Pyromark Q24 LINE-1 assay, and pyrosequenced on a Pyromark Q24 (Qiagen, Valencia, CA) at the Santella laboratory (Herbert Irving Comprehensive Cancer Center NY). Samples were analyzed in batches such that each participant’s samples were assayed simultaneously, the number of samples from each study arm was approximately equal, participant randomization dates were similar, and sample order was random. Laboratory personnel were blinded to sample identity. Duplicate DNA samples (N=40) were used as internal quality control samples. Intra-assay coefficient of variation was 1.7%. The assay of two samples failed, leaving 298 samples for analysis.

Statistical analyses

LINE-1 data was non-normally distributed and was log-transformed. Partial Pearson correlation coefficients were calculated between baseline LINE-1 methylation levels and previously measured biomarker measures, and anthropometrics. Mean 12-month change in LINE-1 methylation in each intervention group was compared to controls using generalized estimating equations modification of linear regression to account for intra-individual correlation over time. Effect of weight-loss on LINE-1 methylation was also examined using stratified analyses (gained weight/no weight loss (N=84); <5% (N=45); 5–10% (N=45); >10% of baseline weight loss (N=126)) performed within each arm. For 2 groups in this sample (diet; diet+exercise) all participants lost weight; reference groups were therefore those participants who lost <5% of baseline weight. Assuming 80% power, the minimum detectable absolute difference in mean change between 3 weight-loss groups and the group that gained any weight was 0.9% (%5mC), (type I error=0.05). Analyses were adjusted by a categorical variable representing percentages of different blood count cell populations. Descriptive data are presented as geometric means (95% confidence intervals (CI)). All statistical analyses were performed using SAS software version 9.1 (SAS Institute, Cary, NC, USA). Statistical tests are 2-sided.

Results

Participant characteristics, intervention adherence, weight loss and body composition changes over the course of the study have been previously described.⁹ The parent study demonstrated significant reductions in BMI, waist circumference, total and percent fat mass⁹ and HOMA scores, and circulating levels of insulin, glucose,¹¹ c-reactive protein, IL-6 ¹⁷, and leptin¹⁴ and a significant increase in circulating adiponectin levels¹⁴ in the intervention arms 12-months post-randomization, compared to the control arm. Characteristics of the participants in the subsample of all randomized participants (N=298), and effects of the intervention on circulating biomarkers or on anthropometrics (see supplementary table 1), did not differ significantly from all randomized participants (N=438; data not shown). Mean age and BMI of participants were 58.0 years and 31.1 kg/m² (Table 1). The majority were college graduates (71.5%) and non-Hispanic white (86.9%). Mean weight changes were −2.5% (p=0.03) in the exercise group, −11.0% (p<0.001) in the diet group, and −12.9% (p<0.001) in the diet + exercise group, compared to −0.18% among controls. Women randomized to the exercise and exercise+diet groups participated in moderate-to-vigorous activity for a mean (SD) 163.3 (70.6) mins/wk (exercise), and 171.5 (62.9) mins/wk (diet+exercise) and both groups significantly increased average pedometer steps/day compared to baseline. In both diet groups, women attended an average of 27 diet counseling sessions (86%).

Table 1.

Baseline Characteristics by Study Arm

	Subsample N=298					Parent trial
Variable	Control N=59	Diet N=82	Exercise N=70	Diet+ Exercise N=87	All N=298	All Participants N=438
	Mean (s.d.)	Mean (s.d.)	Mean (s.d.)	Mean (s.d.)	Mean (s.d.)	Mean (s.d.)
LINE-1 (% 5mC)	77.3 (3.1)	77 (4.2)	77.1 (3.3)	77.3 (3.4)	77.2 (3.6)	-
Age (years)	57.2 (4.1)	57.9 (6.3)	58.0 (4.9)	58.4 (4.5)	58.0 (5.1)	57.9 (4.9)
BMI (kg/m2)	30.9 (3.5)	31.0 (3.9)	31.3 (3.8)	31.1 (4.3)	31.1 (3.9)	30.9 (4.0)
Body fat (kg)	40.1 (7.1)	39.2 (7.7)	41.2 (8.1)	39.3 (8.2)	40.0 (7.8)	39.8 (8.2)
VO2max (L/min)	1.96 (0.38)	1.89 (0.31)	1.88 (0.26)	1.95 (0.30)	1.92 (0.31)	1.90 (0.33)
Race/Ethnicity
Non-Hispanic White	51 (86.4%)	71 (87.6%)	60 (85.7%)	77 (88.5%)	259 (86.9%)	372 (84.9%)
Hispanic	2 (3.4%)	2 (2.4%)	2 (2.9%)	5 (5.8%)	11 (3.7%)	12 (2.7%)
African American	5 (8.5%)	4 (4.9%)	8 (11.4%)	2 (2.3%)	19 (6.4%)	35 (8.0%)
Asian/Pacific Islander	0	2 (2.4%)	0	1 (1.2%)	3 (1.0%)	8 (1.8%)
Other	1 (1.7%)	3 (3.7%)	0	2 (2.3%)	6 (2.0%)	11 (2.5%)
Education
College degree	43 (72.9%)	51 (62.2%)	42 (60.0%)	63 (72.4%)	199 (71.5%)	287 (65.5%)

There were no statistically significant baseline associations between LINE-1 methylation levels and age, anthropometrics (BMI, DXA measures of total fat and lean mass, percent body fat, or waist circumference); circulating biomarkers (adiponectin, leptin, insulin, and glucose), MET hours/week, VO_2max or percent calorie intake from protein, carbohydrate or fat, adjusted by complete blood counts (Table 2).

Table 2.

Baseline associations between LINE-1 and anthropometric measurements, calorie intake, physical activity, and previously measured biomarkers^#

Baseline Covariates	LINE-1 (% 5mC)
	Pearson Partial Correlation Coefficients*
Age	0.08
BMI (kg/m2)	0.00
Waist circumference (cm)	−0.01
Body fat (%)	−0.06
Lean mass (%)	0.05
Total Lean Mass (kg)	0.05
Total Fat (kg)	−0.01
MET** (mins/wk)	0.01
VO2max (ml/kg/min)	0.01
Calories from fats (%)N=290	−0.06
Calories from proteins (%) N=290	−0.10
Calories from carbohydrates (%) N=290	−0.11
Insulin (μU/mL)	−0.01
Glucose (mg/dL)	−0.01
HOMA	−0.01
C-reactive protein (mg/L)	−0.04
Leptin (ng/mL)	−0.03
Adiponectin (μg/mL)	−0.00

^#All N=298 except where indicated.

^*Adjusted for cell count. All P>0.05

^**Metabolic Equivalent of Task (MET)

Compared to controls, there were no significant changes in LINE-1 methylation levels over 12 months in any intervention group (Table 3). Similarly, there was no statistically significant effect on LINE-1 methylation levels when intervention arms were stratified by weight change (Table 4). Adjusting the models for baseline BMI or race/ethnicity did not meaningfully change any of the results (data not shown).

Table 3.

Geometric mean (95% CI) of LINE1 methylation levels (% 5mC) at baseline and 12-months by intervention arm

	Baseline		12 Months		Change	P*
Intervention	N	Mean (95% CI)	N	Mean	Difference (%)
Control	59	77.3 (76.5–78.1)	57	77.4 (76.5–78.1)	0.1 (0.1)	Ref.
Diet	82	76.9 (76.0–77.8)	81	77.0 (76.1–77.8)	0.1 (0.1)	.72
Exercise	70	77.1 (76.3–77.9)	70	76.9 (76.0–77.7)	−0.2 (−0.3)	.52
Diet+Exercise	87	77.3 (76.5–78.0)	87	77.1 (76.4–77.9)	−0.1 (−0.2)	.64

^*P value: GEE model, comparing the change from baseline to 12 month follow-up between control and intervention group, adjusted for cell count (percent of neutrophils, lymphocytes, monocytes, eosinophils, and basophils)

Table 4.

Geometric mean (95% CI) of LINE1 methylation levels at baseline and 12-months, by intervention arm, and by intervention arm and stratified by percentage weight loss

		BASELINE		12 MONTH		CHANGE	P†	P‡	Ptrend§
	Stratification	N	Mean (95% CI)	N	Mean	Difference (%)
Controls	-	59	77.2 (76.5–78.1)	57	77.3 (76.5–78.1	0.1 (0.1)	Ref.
Exercise	No change/gained weight	29	78.0 (77.0–79.1)	29	78.1 (77.0–79.2)	0.1 (0.1)	0.69	Ref.	0.60
	Lost <5% baseline weight	18	75.94 (74.2–77.7)	19	75.5 (73.6–77.4)	−0.4 (−0.6)	0.67	0.68
	Lost >5–10% baseline weight*	23	76.7 (75.4–78.1)	22	76.5 (75.2–77.8)	−0.2 (−0.3)	0.98	0.62
Diet	No change/gained weight	0	-		-				0.17
	Lost <5% baseline weight	17	76.4 (74.5–78.3)	17	76.6 (74.9–78.2)	0.2 (0.3)	0.46	Ref.
	Lost 5–10% baseline weight	16	77.7(75.1–80.5)	16	78.2 (76.4–80.1)	0.5 (0.6)	0.32	0.81
	Lost ≥10% baseline weight	49	76.8 (75.7–77.9	48	76.6 (75.5–77.8)	−0.2 (−0.2)	0.51	0.23
Diet+Exercise	No change/gained weight	0							0.37
	Lost <5% baseline weight	12	77.0 (75.5–78.5)	12	77.2 (75.3–79.0)	0.2 (0.2)	0.77	Ref.
	Lost 5–10% baseline weight	7	78.1 (76.9–79.3)	7	78.9 (77.6–80.1)	0.8 (1.0)	0.27	0.46
	Lost ≥10% baseline weight	68	77.2 (76.3–78.1)	68	77.0 (76.1–77.8)	−0.2 (−0.3)	0.50	0.66
All participants	No change/gained weight	43	77.7 (76.7–78.6)	43	77.6 (76.6–78.6)	−0.6 (−0.1)	0.94	Ref.	0.30
	Lost <5% baseline weight	33	76.1 (74.9–77.4)	34	76.1 (74.9–77.4)	0.0 (0.0)	0.66	0.71
	Lost 5–10% baseline weight	42	77.5 (76.3–78.7)	41	77.7 (76.7–78.7)	0.2 (0.3)	0.41	0.54
	Lost ≥10% baseline weight	121	77.0 (76.3–77.7)	120	76.8 (76.1–77.5)	−0.2 (−0.3)	0.41	0.43

^†GEE model, P value testing the difference in change from baseline to 12 months in LINE-1 methylation compared to Controls, adjusted for cell count

^‡GEE model, P value testing the difference in change from baseline to 12 months in LINE-1 methylation compared to No Change or Gained some weight group, excluding Controls, adjusting for cell count

^§P_trend: testing for a trend in change from baseline to 12 months in LINE-1 methylation from No change or gained some weight group through most weight loss group, adjusting for cell count

^*None of the participants in this group lost more than 10% of their baseline weight

Discussion

Few studies have examined the association between BMI, weight-loss, physical activity and global methylation, and to our knowledge, no previous randomized controlled trial has tested the effect of weight loss on LINE-1 methylation levels in postmenopausal overweight/obese women. We observed no statistically significant changes in LINE-1 methylation levels from peripheral blood leukocytes in a 12-month weight loss intervention, vs. controls, in a sample of post-menopausal overweight/obese women, adjusted by blood cell type. Furthermore, we found that there were no statistically significant baseline associations between LINE-1 methylation and anthropomorphic measures, circulating biomarkers, or measures of physical activity.

Both weight loss and the obese state appear to cause alterations in methylation patterns in specific genes.¹⁸ A recent study reported that increased BMI in adults of European origin was associated with increased methylation at the HIF3A locus both in peripheral blood and in adipose tissue.¹⁹ A small study of 46 obese and 46 lean adolescents, aged 14–18, demonstrated obesity-related methylation changes in specific genes of peripheral blood leukocytes.²⁰ Weight-loss is associated with methylation at specific genes, though in one study, these associations appear to be sex-specific. A study of 25 overweight/obese men who participated in an 8-week caloric restriction intervention, reported that DNA (isolated from peripheral blood leukocytes) methylation levels in several CpGs located in the ATP10A and CD44 genes showed statistically significant baseline differences depending on the degree of weight-loss. Post-intervention, DNA methylation levels of several CpGs on the WT1 promoter were more methylated in the high than low responders.²¹ Twenty-four obese men who lost ≥5% of their initial body weight after 8 weeks of a hypocaloric diet had lower levels of total TNF-alpha promoter methylation from DNA from peripheral blood; women in the same study with similar weight loss patterns had no such reductions.²² Finally, 14 overweight/obese postmenopausal women were recruited for a 6-month caloric restriction intervention. Samples of subcutaneous adipose tissue were biopsied before and after the intervention, and DNA extracted from these samples. There were significant differences in DNA methylation at 35 loci between the high and low responders before dieting, and 3 regions showed differential methylation post-intervention.²³ Although these small studies of specific loci are intriguing, our study suggests that weight loss over a one year period may not be sufficient to change global measures of DNA methylation and specifically LINE-1 methylation levels. Despite these findings few studies examined the association between LINE-1 methylation and either obesity or weight loss. In one study, participants with BMI≥40 kg/m² had higher levels of LINE-1 methylation than participants with a BMI≤25.²⁴ As our study only examined women with an initial BMI≥25 kg/m², there was insufficient variation to make a comparable analysis. A large study (n= 465) of a combined analysis of 5 separate studies reported no association between LINE-1 methylation and BMI; mean %5mC of the 5 studies was 76.2 (s.d. 1.8), compared to 77.2 (s.d. 3.6) in our study.²⁵ Similarly, a study of healthy Koreans (N=86), with a mean %5mC of 76.6 (s.d. 2.9),²⁶ and a U.S. study (n=161, median LINE-1 % mC 73.9%)²⁷ showed no association between BMI, and LINE-1 methylation. In contrast, 470 premenopausal women, with abnormal Pap tests reported a statistically significant inverse association between LINE-1 methylation levels and BMI, and percent body-fat; participants in that study had lower levels of LINE-1 methylation, and more variability (mean (s.d.) 63.6 (7.6)), compared to our study.²⁸ However the study had a high proportion of Black participants (54%) and participants were younger, compared to ours.

It appears that compared to methylation of other repetitive elements, or of CpG islands in specific genes, LINE-1 methylation is likely to be genetically determined or to be determined by early life exposures.^29,30 For example, an analysis of 51 girls aged 6–17 years reported that LINE-1 methylation levels in peripheral blood were statistically significantly lower in girls with a family history of breast cancer compared to those with no such history.³⁰ Another study in families with a history of testicular cancer reported that global methylation at LINE-1 in peripheral blood of offspring were significantly positively correlated with parental levels, particularly between mother-daughter, father-daughter and affected father-affected son pairs.³¹ Finally, a longitudinal study of global methylation changes over time in two separate populations also suggested that methylation maintenance may be under genetic control.³² Finally, while some methylation markers changed significantly in a longitudinal study over a 10 year period, LINE-1 was relatively stable later in life.³³ These data suggest that alterations in lifestyle, even those resulting in significant weight-loss for example, may have less overall impact on methylation levels at LINE-1 elements, compared to those at specific genes.

Finally, while some studies have reported a statistically significant association between LINE-1 methylation and dietary blood biomarkers including blood glucose and lipid profiles,^34–36 we found no association between LINE-1 methylation and percent calories from fat, protein or carbohydrate, or with either glucose or insulin. In contrast to our results, a cross-sectional study reported a statistically non-significant trend of higher levels of LINE-1 methylation with higher levels of physical activity.²⁷

Limitations of our study include the relatively homogenous population: these results may not be generalizable to other populations. We did not measure methylation of LINE-1 in DNA from adipose tissue, and it is possible that fat tissue during weight-loss would display tissue-specific methylation changes that are not reflected in DNA from peripheral blood leukocytes. However, previous studies have shown that DNA methylation patterns are largely conserved across tissues, and that leukocyte DNA methylation levels of selected genes may serve as surrogate markers of DNA methylation in specific tissues.³⁷ Strengths of our study include the randomized controlled trial design, a well-characterized population, an intervention of 12-months’ duration, adequate power to examine changes in LINE-1 methylation patterns in three separate lifestyle interventions compared to controls, and our ability to adjust methylation patterns by specific blood-cell populations.

Continued research in humans is needed to characterize the degree, to which methylation, both global and gene-specific, can be altered by lifestyle weight loss and physical activity interventions, and if so, what intensity, dose, and duration are needed.

What is already known about this subject?

• Obesity is associated with methylation at specific genes

• Long Interspersed Nuclear Elements (LINE-1) methylation is a measure of an individual’s global methylation levels, itself associated with disease risk.

• It is unclear whether weight-loss can affect LINE-1 methylation levels

What this study adds

• This is the first study to examine the independent and combined effects of a reduced-calorie weight-loss diet, and exercise program, vs. control on LINE-1 methylation levels in the context of a randomized controlled trial

• None of the intervention arms statistically significantly altered LINE-1 methylation levels

• Weight loss (gained weight/no weight-loss; <5%; >10% of baseline weight-loss) was not associated with changes in LINE-1 methylation levels