Flavonoid intake and telomere length attrition among middle-aged women: a cross-sectional analysis of the Nurses’ Health Study

Abstract

Although mechanistic studies support a beneficial effect of several dietary flavonoids on telomere length (TL), to our knowledge no studies have examined associations between habitual flavonoid intakes and TL in population-based studies. We examined the associations between habitual intake of major flavonoid subclasses (flavonols, flavones, flavanones, flavan-3-ols and their polymers, and anthocyanins) and TL in a cross-sectional analysis of 4,944 disease-free females from the Nurses’ Health Study (NHS). Flavonoid intakes were collected using food frequency questionnaire data, and TL was measured in peripheral blood leukocytes using quantitative real-time polymerase chain reaction. Multivariable-adjusted least squares mean leukocyte TL (z scores and corresponding standard error [SE]) for total and all flavonoid subclasses were calculated using generalized linear models. Although no individual flavonoid subclass was significantly associated with TL in the overall population, when we restricted analyses to younger and middle-aged participants (aged < 55 y), a higher anthocyanin intake associated with longer TL (least squares means ± SE: 0.06 ± 0.06 for the highest versus 0.24 ± 0.07 for the lowest quintile; P-trend = 0.042), corresponding to 6.6 (95% CI, 1.7–14.5) years of aging. In food-based analyses, participants aged < 55 y who consumed more berries had longer TL (Mean TL: 0.10 ± 0.04 for never/rarely; 0.21 ± 0.04 for ≤ 1 serving/week; 0.44 ± 0.26 for ≥ 2 servings/week; P = 0.033 for trend). Similarly, a higher anthocyanin intake was associated with longer TL among pre-menopausal females (0.46 ± 0.11 for the highest versus 0.02 ± 0.08 for the lowest quintile; P-trend = 0.001). Overall, habitual flavonoid intake was not associated with TL attrition in this cross-sectional analysis. However, in females aged < 55 y a higher anthocyanin intake was associated with longer TL, which raises the possibility that anthocyanin-rich foods may promote healthy aging and warrants further investigation with respect to age.

Introduction

Telomeres are the dynamic nucleoprotein structures at the ends of linear chromosomes that maintain the genomic integrity of a cell. Telomere length (TL) shortens with age as a result of incomplete replication of DNA ends with each cell division, oxidative stress, and inflammation [1]. TL dynamically changes across the lifespan with a gradual decrease in shortening rate until around age 50 and remains at a relatively stable rate towards the elderly period [2]. Shorter telomeres have been associated with lower life expectancy and higher risks of age-related chronic diseases [3].

Prior evidence suggested that longer TL was associated with greater adherence to a Mediterranean-style dietary pattern, a plant-based diet, an anti-inflammatory diet, or higher intakes of dietary fiber, vegetables, fruit, or coffee [4–11]. For example, a recent study found that a 100 g/d increment in fruit and vegetable intake corresponded to 1.9 years of lower biological aging as measured by telomere length [12]. Notably, a common characteristic of these plant-based foods and dietary patterns is their high dietary flavonoid content. Flavonoids are bioactive compounds as part of a plant-rich dietary pattern that act as a potent regulator of telomerase by enhancing endogenous antioxidant systems and reducing oxidative stress and systemic inflammation, thereby increasing telomerase activity and providing subsequent protection for telomeres and increasing their length [13]. Prospective cohort studies have also found that higher flavonoid intake was inversely associated with inflammatory biomarkers and the risks of mortality and several age-related diseases [14–18]. However, to our knowledge, no population-based studies have examined the association between dietary flavonoids and major subclasses with TL.

In this cross-sectional study in the Nurses’ Health Study (NHS), we examined whether greater intakes of dietary flavonoids and major subclasses in the habitual diet were associated with longer TL, exploring if the previously observed associations between a plant-rich diet and TL were related to its flavonoid constituents.

Methods

Study population

The NHS began in 1976, when 121,700 female registered nurses from 11 US states and aged 30 to 55 years responded to a questionnaire of health-related information. Subsequently, each participant received follow-up questionnaires biennially to update lifestyle factors and identify new cases of disease [19, 20]. Dietary information was obtained every 4 years through semiquantitative food frequency questionnaires (FFQs) [21, 22]. In addition, blood samples were provided by 32,826 female participants between 1989 and 1990, and all samples were immediately processed upon arrival and aliquoted into cryotubes as plasma, buffy coat, and red blood cells. Details of the blood collection and archival methods have previously been described, and women who provided blood samples were similar to those who did not [23].

Multiple nested case–control studies have been conducted within the NHS blood sub-cohort to investigate the association between leukocyte TL and cancer, cardiovascular disease, and cognitive function, among others [24–30]. All of these case–control studies employed the same design: for each incident case of a disease, a risk-set sampling scheme was used to select one to three controls from the remaining participants in the blood sub-cohort who were disease-free at the time the case was diagnosed. All cases and controls were free of the specific disease of interest at the time blood was drawn in each individual study. In addition, cases and controls were matched for menopausal status and postmenopausal hormone use (except in the myocardial infarction case–control study), age, time of blood draw, and other factors carefully selected for each individual study. In addition to these case–control studies, telomere length data were available from a study of cognitive function [27]. In total, the current analysis included data on 5,515 female participants selected as healthy controls from nested case–control studies with previously measured leukocyte TL. After excluding those with incomplete FFQ data, having a history of cancer or cardiovascular disease, or with missing covariates, a total of 4,944 participants were included in the current analysis (Fig. 1).

Fig. 1.

Participant enrollment flowchart

The study protocol was approved by the institutional review boards (IRBs) of the Brigham and Women’s Hospital (IRB Protocol Number: 1999P011114), and participants provided consent by returning their questionnaires and providing bloods.

Assessment of dietary flavonoid intake

Dietary flavonoid intake was determined using our database which has previously been described [31]. We calculated intakes from the 1990 validated FFQ, the closest to the blood draw (1989–90) that was used for quantifying telomere length [31, 32]. Dietary information was judged as unreliable and excluded from study if reported energy intakes were < 600 or > 4000 kcal/d or if > 12 food items were left blank.

Intakes of individual compounds were calculated as the sum of the consumption frequency of each food multiplied by the specific flavonoid content for the stated portion size. We derived the intake of the six flavonoid subclasses commonly consumed in the US diet: flavonols, flavones, flavanones, flavan-3-ols, proanthocyanidins, anthocyanins, as well as total flavonoids. We did not include isoflavones in our analysis because the intake is very low (< 3 mg/d) in the habitual US diet, and we know that an isoflavone intake of more than 25 mg/d is needed for any biological/clinical effect [33, 34]. The validity and reproducibility of the semi-quantitative FFQ have been reported previously, and correlations between major dietary sources of flavonoids (including fruits, vegetables, tea and wine) measured by diet records and the FFQ were 0.70, 0.50, 0.77, and 0.83, respectively [35, 36], which were correlations similar to those reported for an FFQ in a previous urinary flavonoid biomarker study [37].

Telomere length measurement

All buffy-coat cryotubes were stored in the vapor phase of liquid nitrogen freezers at ≤  − 130 °C. Genomic DNA was extracted from peripheral blood leukocytes using the QiAmp 96-spin protocol (Qiagen, Chatsworth, CA, USA). Pico-Green quantification of genomic DNA was performed using a Molecular Devices 96 well spectrophotometer (Sunnyvale, CA, USA). Subsequent standardization by drying down the genomic DNA and re-suspending ensured accurate and uniform DNA concentrations. DNA was extracted at the same lab and then stored in freezers at ≤  − 80 °C for all studies. During 2007–2010, relative TL was determined using a modified, high-throughput version of the real-time quantitative polymerase chain reaction-based telomere assay that was run on the Applied Biosystems 7900HT Sequence Detection System (Foster City, CA) [38, 39]. Each sample was assayed in triplicate by laboratory technicians who were blind to participant characteristics; variability was assessed with the inclusion of quality control samples on each plate. The average relative leukocyte TL was determined as the copy-number ratio between telomere repeats and a single-copy (36B4) reference gene (T/S Ratio, -ΔCt). Leukocyte TL is reported as the exponentiated T/S ratio corrected for a reference sample, and only samples that were successfully assayed were included [24–27]. The coefficients of variation were below 4% for the telomere and single-gene assays and below 18% for the exponentiated T/S. In the NHS, the three-year and ten-year interclass correlations for reliability were determined to be 0.80 and 0.60. Although this assay measures relative TL, the T/S ratio correlates highly with absolute TL provided by Southern Blot (r = 0.68–0.85; P < 0.001) [40]. Because TL was assayed separately for these included studies in various batches, to minimize the impact of potential batch effect on leukocyte TL measurements across different studies, we calculated z scores of log-transformed leukocyte TL by standardizing the leukocyte TL in comparison with the mean within each individual study [5, 41].

Assessment of covariates

Information on covariates was obtained from self-reports from the 1990 follow-up questionnaire, closest in time to the blood collection used to measure TL. All these covariates were selected a priori based on potential confounders previously associated with TL in the NHS, [42–44] including: age; body mass index (BMI, kg/m²); cigarette smoking (never, past, and current); menopausal status and postmenopausal hormone use (premenopausal or postmenopausal; never, past, or current postmenopausal hormone user); multivitamin user; total caloric intake (kcal/day); physical activity (metabolic equivalents [MET] per week); protein (g/day); cereal fiber intake (g/day); alcohol consumption (g/day); saturated, trans, polyunsaturated, and monounsaturated fat intake (g/day).

Statistical analyses

Relative leukocyte TL values were z-transformed to improve normality within batches. Energy-adjusted flavonoids were calculated using the residual method and presented as quintile categories of intake. We used multivariable linear regression models to determine the cross-sectional association between intake of flavonoid subclasses and leukocyte TL. Model 1 adjusted for age. Model 2 further adjusted for smoking, BMI, physical activity, menopausal status and postmenopausal hormone use, and total energy intake. Model 3 further adjusted for dietary factors, including alcohol intake, multivitamin use, intakes of polyunsaturated fatty acids, saturated fatty acids, trans fatty acids, monounsaturated fatty acids, cereal fiber, and protein. We calculated multivariable adjusted least squares mean leukocyte TL z scores (and their corresponding 95% confidence intervals) across all flavonoid subclasses using generalized linear models. The α level of 0.05/7 was used to determine statistical significance for flavonoids and TL (Bonferroni correction).

Since TL differences tend to be more pronounced in younger and middle-aged adults [2], we also explored whether chronological age modified the association between flavonoid intakes and TL stratified by ≤ 55, 55–64, and ≥ 65 years. Moreover, associations stratified by smoking status, BMI, and menopausal status were also explored as prior evidence suggested that telomere dynamics are more strongly influenced by smoking, adiposity, and estrogen-related hormonal milieu [45–47]. Subgroup analyses were not adjusted for multiple testing and were hypothesis-generating; however, all analyses were planned a priori based on the wealth of existing data on flavonoids and healthy ageing and the supportive mechanistic data.

In addition, we conducted secondary food-based analyses based on previously published work and evaluated the association for foods known to contribute to the flavonoid subclasses associated with telomere length in the primary analyses [31].  Specifically, we evaluated the association of berry intake (blueberries and strawberries; categorized as never/rarely, ≤ 1 serving/week, and ≥ 2 servings/week; one serving equals 0.5 cup) with leukocyte TL to further explore observed findings for anthocyanin intake. To help interpretation, associations were summarized as the number of equivalent years of age-related change in TL by dividing the β coefficients for flavonoids by the absolute value of the β coefficient for the age-related change in TL, and confidence intervals were estimated using 1,000 bootstrap resamples. As only very few participants (1%) had missing data on covariates, we only included complete cases in the current analysis. The SAS version 9.4 software (SAS Institute) was used for analyses.

Results

Characteristics of the study population

A total of 4944 female participants were included in the current analysis, and the participants ranged in age from 42–70 years. Baseline characteristics of the study population by total flavonoid intake are presented in Table 1. Participants with higher flavonoid intake smoked less and consumed less alcohol, but participants with higher intakes of total flavonoids did not have a more favorable dietary pattern aside from lower saturated fat intakes. Participants with moderate flavonoid intake consumed more fruits and vegetables. The total flavonoid intake ranged from 6.8 to 3191.4 mg/d, with the most contribution on average from the proanthocyanidin subclass (2.6–3061.4 mg/d), and anthocyanin intakes ranged from 0 to 251 mg/d (Table 2). The correlation between flavonoid subclasses and major food sources is presented in Fig. 2, with high correlations observed for apples/pears with proanthocyanidins, citrus with flavones and flavanones, berries with anthocyanins, and tea with flavan-3-ols (all r > 0.7).

Table 1.

Baseline characteristics of the cohort (n = 4944) at blood raw, by quintiles of flavonoid intake^1–3

	Quintiles of total flavonoid intake
Baseline Characteristics	Q1	Q2	Q3	Q4	Q5
	(≤ 142.8; n = 989)	(142.9–205.6, n = 989)	(205.7–301.3; n = 989)	(301,4–491.0; n = 989)	(≥ 491.1; n = 988)
Age (years)	58.3 ± 6.6	59.1 ± 6.7	58.7 ± 6.7	59.5 ± 6.3	58.7 ± 6.7
BMI (kg/m2)	25.3 ± 4.6	25.0 ± 4.3	25.1 ± 4.0	25.3 ± 4.1	25.1 ± 4.3
Smoking status (%)
Never smoker	39	44	46	50	50
Past smoker	38	42	43	38	37
Current smoker	22	15	12	11	13
Physical activity (METs/wk)	14.5 ± 18.5	17.1 ± 20.2	17.5 ± 20.6	18.2 ± 27.2	15.3 ± 16.4
Caloric intake (kcal/day)	1755 ± 500	1801 ± 483	1795 ± 502	1790 ± 442	1711 ± 478
Protein (g/day)	76.81 ± 14.24	76.22 ± 11.97	75.46 ± 12.99	76.13 ± 11.81	75.05 ± 11.93
Cereal Fiber (g/day)	5.08 ± 3.24	5.47 ± 3.39	5.51 ± 3.13	5.44 ± 3.14	5.33 ± 3.31
Alcohol (g/day)	7.37 ± 12.38	6.40 ± 9.89	5.10 ± 8.07	4.83 ± 7.61	4.53 ± 9.20
Total Saturated Fat (g/day)	20.2 ± 4.67	18.65 ± 3.97	17.67 ± 3.81	17.82 ± 4.08	18.2 ± 4.27
Total Polyunsaturated Fat (g/day)	10.97 ± 3.22	10.63 ± 3.09	10.47 ± 2.76	10.39 ± 2.56	10.57 ± 2.87
Total Monounsaturated Fat (g/day)	22.72 ± 5.08	21.2 ± 4.42	20.46 ± 4.28	20.46 ± 4.19	20.89 ± 4.60
Total Trans Fat (g/day)	2.88 ± 1.10	2.56 ± 0.90	2.45 ± 0.88	2.47 ± 0.91	2.55 ± 0.97
Fruit and vegetable intake (servings/wk)	4.40 ± 1.90	5.82 ± 2.10	6.52 ± 2.45	6.50 ± 2.502	6.11 ± 2.63

BMI body mass index, METs, metabolic equivalent tasks

¹All values (except age) are age standardized

²All dietary variables (except alcohol) are calorie adjusted

³Mean ± SD (all such values)

Table 2.

Total flavonoid and class intakes in 4944 participants aged 42 to 70 years in the Nurses’ Health Study ¹

	Q1	Q2	Q3	Q4	Q5
Total Flavonoids, mg/d	108.1 (6.8–142.8)	174.1 (142.9–205.6)	245.2 (205.7–301.3)	374.7 (301.4–491)	797.15 (491.1–3191.4)
Flavonols, mg/d	7.45 (1.34–9.39)	11.22 (9.4–12.87)	14.91 (12.88–17.5)	20.94 (17.51–26.73)	35.04 (26.75–108.89)
Flavones, mg/d	0.56 (0.03–0.81)	1.05 (0.82–1.3)	1.565 (1.31–1.88)	2.22 (1.89–2.58)	3.195 (2.59–25.09)
Flavanones, mg/d	5.72 (0–10.77)	17.135 (10.78–24.81)	33.12 (24.83–41.59)	50.69 (41.6–61.88)	80.7 (61.91–364.03)
Flavan-3-ols, mg/d	7.9 (0.2–11.2)	14.4 (11.3–17.6)	23.4 (17.7–35.8)	56.4 (35.9–80.6)	153.7 (80.8–738.5)
Proanthocyanidins, mg/d	46.4 (2.6–67.5)	85.8 (67.6–104.3)	134.65 (104.4–184)	267.15 (184.1–372.4)	682.7 (372.5–3061.4)
Anthocyanins, mg/d	1.55 (0.01–2.72)	3.6 (2.73–4.7)	6.5 (4.71–9.66)	12.375 (9.67–15.62)	21.1 (15.63–251.06)

¹All values are median (minimum–maximum)

Fig. 2.

Correlations between flavonoid subclasses and major food groups in the Nurses’ Health Study

Association of flavonoid intakes with telomere length

Table 3 presents the multivariate-adjusted associations between quintiles of different flavonoid subclass intakes and TL. Overall, we found no significant association between total flavonoid intake and TL across all models following multivariate adjustment (all p for trend > 0.05). Similarly, we found no significant associations for all subclasses (flavonols, flavones, flavanones, flavan-3-ols, proanthocyanidins, or anthocyanins).

Table 3.

Associations between quintiles of different flavonoid subclass intakes and leukocyte telomere length in 4944 participants aged 42 to 70 years in the Nurses’ Health Study

	Quintiles (Q) 1–5: Least square mean ± SE for telomere length					P-value (Q5 vs Q1)	P-trend
	Q1	Q2	Q3	Q4	Q5
Total Flavonoids
Model 11	0 ± 0.03	0 ± 0.03	0.05 ± 0.03	0.02 ± 0.03	−0.03 ± 0.03	0.49	0.33
Model 22	0.01 ± 0.03	0 ± 0.03	0.05 ± 0.03	0.02 ± 0.03	−0.03 ± 0.03	0.33	0.24
Model 33	0.02 ± 0.03	0 ± 0.03	0.05 ± 0.03	0.02 ± 0.03	−0.03 ± 0.03	0.31	0.24
Flavonols
Model 11	−0.02 ± 0.03	−0.02 ± 0.03	0.07 ± 0.03	0.04 ± 0.03	−0.03 ± 0.03	0.78	0.66
Model 22	−0.01 ± 0.03	−0.02 ± 0.03	0.07 ± 0.03	0.04 ± 0.03	−0.03 ± 0.03	0.67	0.57
Model 33	0 ± 0.03	−0.01 ± 0.03	0.06 ± 0.03	0.03 ± 0.03	−0.03 ± 0.03	0.49	0.40
Flavones
Model 11	−0.02 ± 0.03	−0.01 ± 0.03	0.01 ± 0.03	0.04 ± 0.03	0.02 ± 0.03	0.37	0.22
Model 22	−0.01 ± 0.03	−0.01 ± 0.03	0.01 ± 0.03	0.04 ± 0.03	0.02 ± 0.03	0.55	0.34
Model 33	0 ± 0.03	−0.01 ± 0.03	0 ± 0.03	0.04 ± 0.03	0.02 ± 0.03	0.69	0.43
Flavanones
Model 11	−0.04 ± 0.03	0.08 ± 0.03	−0.06 ± 0.03	0.08 ± 0.03	−0.01 ± 0.03	0.59	0.89
Model 22	−0.03 ± 0.03	0.08 ± 0.03	−0.06 ± 0.03	0.08 ± 0.03	−0.02 ± 0.03	0.81	0.87
Model 33	−0.02 ± 0.03	0.08 ± 0.03	−0.07 ± 0.03	0.08 ± 0.03	−0.02 ± 0.03	0.95	0.77
Flavan-3-ols
Model 11	−0.02 ± 0.03	0.02 ± 0.03	0.08 ± 0.03	−0.03 ± 0.03	−0.01 ± 0.03	0.94	0.32
Model 22	−0.01 ± 0.03	0.02 ± 0.03	0.08 ± 0.03	−0.03 ± 0.03	−0.02 ± 0.03	0.9	0.26
Model 33	−0.01 ± 0.03	0.02 ± 0.03	0.08 ± 0.03	−0.03 ± 0.03	−0.01 ± 0.03	0.89	0.29
Proanthocyanidins
Model 11	−0.01 ± 0.03	0.01 ± 0.03	−0.01 ± 0.03	0.04 ± 0.03	0.02 ± 0.03	0.53	0.39
Model 22	0 ± 0.03	0.01 ± 0.03	−0.02 ± 0.03	0.04 ± 0.03	0.01 ± 0.03	0.83	0.64
Model 33	0.01 ± 0.03	0.01 ± 0.03	−0.02 ± 0.03	0.04 ± 0.03	0.01 ± 0.03	0.95	0.88
Anthocyanins
Model 11	−0.02 ± 0.03	0.01 ± 0.03	−0.01 ± 0.03	0.03 ± 0.03	0.03 ± 0.03	0.25	0.26
Model 22	−0.01 ± 0.03	0.01 ± 0.03	−0.01 ± 0.03	0.02 ± 0.03	0.03 ± 0.03	0.37	0.36
Model 33	−0.01 ± 0.03	0.02 ± 0.03	−0.01 ± 0.03	0.02 ± 0.03	0.02 ± 0.03	0.57	0.55

¹Adjusted for age

²Adjusted for smoking, BMI, physical activity, menopausal status and postmenopausal hormone use, total energy intake

³Adjusted for alcohol intake, multivitamin use, intakes of polyunsaturated fatty acids, saturated fatty acids, trans fatty acids, monounsaturated fatty acids, cereal fiber, and protein

Age-stratified analyses

Since TL shortening tends to be more pronounced in younger and middle-aged adults [2], we next stratified our analyses by age (≤ 55, 55 to < 65, and ≥ 65 y) to examine any potential differential associations between intakes of different flavonoid subclasses and TL (Table 4). Although there was no evidence of effect modification by age for total flavonoids and several subclasses (flavonols, flavones, flavanones, flavan-3-ols, proanthocyanidins, anthocyanins; all p > 0.05 for interaction) we did observe that a higher anthocyanin intake was significantly associated with longer TL in females aged ≤ 55 y (Mean TL in Q5: 0.24 ± 0.07 versus Q1: 0.06 ± 0.06; P-trend = 0.042). The mean difference in telomere length between the highest and lowest quintiles corresponds with 6.6 (95% CI, 1.7–14.5) years of aging. Given these findings, we conducted food-based analyses of berries, which were the primary source of anthocyanins in our cohort [31]. We found that females aged ≤ 55 y with higher berry intake had longer TL (Mean TL: 0.10 ± 0.04 for never/rarely; 0.21 ± 0.04 for ≤ 1 serving/week; 0.44 ± 0.26 for ≥ 2 servings/week; P = 0.033 for trend); however, the associations were not significant for total berries or strawberries.

Table 4.

Associations between different flavonoid subclass intakes and telomere length in 4944 participants from the Nurses’ Health Study, stratified by < 55, 55–64, and ≥ 65 years of age

	< 55 years (N = 1,423)				55–64 years (N = 2,469)				≥ 65 years (N = 1,053)				Pinteraction
	Q1	Q3	Q5	Ptrend	Q1	Q3	Q5	Ptrend	Q1	Q3	Q5	Ptrend
Total Flavonoids
Model 12	0.22 ± 0.06	0.16 ± 0.06	0.09 ± 0.06	0.27	−0.05 ± 0.04	0.04 ± 0.04	−0.10 ± 0.04	0.10	−0.17 ± 0.07	−0.08 ± 0.07	−0.02 ± 0.07	0.10	0.063
Model 23	0.23 ± 0.06	0.17 ± 0.06	0.08 ± 0.06	0.19	−0.04 ± 0.04	0.04 ± 0.04	−0.10 ± 0.04	0.10	−0.17 ± 0.07	−0.06 ± 0.07	−0.02 ± 0.07	0.13	0.063
Model 34	0.23 ± 0.06	0.17 ± 0.06	0.07 ± 0.06	0.16	−0.02 ± 0.05	0.03 ± 0.05	−0.09 ± 0.05	0.09	−0.18 ± 0.08	−0.05 ± 0.07	−0.01 ± 0.07	0.09	0.059
Flavonols
Model 12	0.14 ± 0.06	0.22 ± 0.06	0.07 ± 0.06	0.43	−0.08 ± 0.04	0.04 ± 0.04	−0.07 ± 0.04	0.58	−0.08 ± 0.07	−0.08 ± 0.07	−0.08 ± 0.07	0.42	0.50
Model 23	0.14 ± 0.06	0.22 ± 0.06	0.06 ± 0.06	0.36	−0.06 ± 0.05	0.04 ± 0.04	−0.06 ± 0.04	0.54	−0.09 ± 0.07	−0.08 ± 0.07	−0.08 ± 0.07	0.44	0.49
Model 34	0.15 ± 0.06	0.22 ± 0.06	0.05 ± 0.07	0.26	−0.05 ± 0.05	0.03 ± 0.04	−0.07 ± 0.04	0.39	−0.09 ± 0.08	−0.09 ± 0.07	−0.07 ± 0.07	0.36	0.48
Flavones
Model 12	0.17 ± 0.06	0.10 ± 0.06	0.08 ± 0.07	0.59	−0.06 ± 0.05	0 ± 0.04	0.02 ± 0.04	0.12	−0.19 ± 0.07	−0.11 ± 0.07	−0.08 ± 0.06	0.34	0.33
Model 23	0.17 ± 0.06	0.11 ± 0.06	0.08 ± 0.07	0.54	−0.05 ± 0.05	−0.01 ± 0.04	0.02 ± 0.04	0.19	−0.19 ± 0.07	−0.11 ± 0.07	−0.09 ± 0.06	0.40	0.35
Model 34	0.18 ± 0.06	0.11 ± 0.06	0.07 ± 0.07	0.42	−0.03 ± 0.05	−0.02 ± 0.04	0.02 ± 0.05	0.27	−0.20 ± 0.08	−0.12 ± 0.07	−0.08 ± 0.06	0.33	0.34
Flavanones
Model 12	0.12 ± 0.06	0.03 ± 0.06	0.09 ± 0.07	0.36	−0.08 ± 0.04	−0.04 ± 0.04	−0.04 ± 0.04	0.77	−0.17 ± 0.07	−0.23 ± 0.07	−0.11 ± 0.06	0.32	0.39
Model 23	0.13 ± 0.06	0.04 ± 0.06	0.08 ± 0.07	0.26	−0.07 ± 0.04	−0.05 ± 0.05	−0.04 ± 0.05	0.92	−0.17 ± 0.07	−0.23 ± 0.07	−0.11 ± 0.06	0.39	0.37
Model 34	0.13 ± 0.06	0.05 ± 0.06	0.07 ± 0.07	0.2	−0.05 ± 0.05	−0.05 ± 0.05	−0.04 ± 0.05	0.89	−0.18 ± 0.07	−0.24 ± 0.07	−0.11 ± 0.06	0.33	0.34
Flavan-3-ols
Model 12	0.10 ± 0.06	0.22 ± 0.06	0.10 ± 0.06	0.34	−0.01 ± 0.04	0.04 ± 0.04	−0.07 ± 0.05	0.12	−0.19 ± 0.07	−0.02 ± 0.07	−0.03 ± 0.07	0.19	0.14
Model 23	0.10 ± 0.06	0.23 ± 0.06	0.09 ± 0.06	0.25	0.01 ± 0.04	0.04 ± 0.04	−0.06 ± 0.05	0.15	−0.20 ± 0.07	−0.02 ± 0.07	−0.04 ± 0.07	0.23	0.14
Model 34	0.10 ± 0.06	0.23 ± 0.06	0.08 ± 0.06	0.21	0.02 ± 0.05	0.03 ± 0.04	−0.06 ± 0.05	0.16	−0.20 ± 0.07	−0.02 ± 0.07	−0.03 ± 0.07	0.19	0.13
Proanthocyanidins
Model 12	0.11 ± 0.06	0.12 ± 0.06	0.13 ± 0.07	0.83	−0.01 ± 0.04	−0.04 ± 0.05	−0.01 ± 0.04	0.91	−0.17 ± 0.07	−0.14 ± 0.06	−0.05 ± 0.07	0.13	0.47
Model 23	0.11 ± 0.06	0.11 ± 0.06	0.12 ± 0.07	0.91	0.01 ± 0.05	−0.05 ± 0.05	−0.02 ± 0.04	0.88	−0.16 ± 0.07	−0.13 ± 0.07	−0.06 ± 0.07	0.21	0.47
Model 34	0.12 ± 0.06	0.11 ± 0.06	0.12 ± 0.07	0.99	0.02 ± 0.05	−0.05 ± 0.05	−0.03 ± 0.04	0.57	−0.17 ± 0.07	−0.13 ± 0.07	−0.04 ± 0.07	0.13	0.47
Anthocyanins
Model 12	0.06 ± 0.06	0.16 ± 0.06	0.24 ± 0.06	0.03	−0.01 ± 0.05	−0.10 ± 0.04	−0.02 ± 0.04	0.68	−0.14 ± 0.07	0 ± 0.07	−0.12 ± 0.07	0.54	0.15
Model 23	0.06 ± 0.06	0.17 ± 0.06	0.24 ± 0.07	0.04	0.01 ± 0.05	−0.10 ± 0.04	−0.01 ± 0.04	0.80	−0.13 ± 0.07	0 ± 0.07	−0.12 ± 0.07	0.53	0.15
Model 34	0.06 ± 0.06	0.17 ± 0.06	0.24 ± 0.07	0.04	0.02 ± 0.05	−0.10 ± 0.04	−0.03 ± 0.04	0.87	−0.14 ± 0.07	0 ± 0.07	−0.11 ± 0.07	0.60	0.15

¹Values are least square mean ± SE for telomere length

²Adjusted for age

³Adjusted for smoking, BMI, physical activity, menopausal status and postmenopausal hormone use, total energy intake

⁴Adjusted for alcohol intake, multivitamin use, intakes of polyunsaturated fatty acids, saturated fatty acids, trans fatty acids, monounsaturated fatty acids, cereal fiber, and protein

Other subgroup analyses

Similarly, when we stratified by menopausal status, a higher anthocyanin intake was associated with longer TL among pre-menopausal participants (Mean TL in Q5: 0.46 ± 0.11 versus Q1: 0.02 ± 0.08; P-trend = 0.001) but not among post-menopausal participants (Mean TL in Q5: −0.03 ± 0.03 versus Q1: −0.01 ± 0.04; P-trend = 0.61; P-interaction = 0.005). In other analyses, we found no evidence for effect modification by either smoking status or BMI for the association of total flavonoid intake or individual subclasses with TL (all p > 0.05 for interactions), except that higher anthocyanin intake was associated with shorter TL among current smokers (n = 732; Mean TL in Q5: −0.22 ± 0.08 versus Q1: −0.05 ± 0.06; P-trend = 0.032; P-interaction = 0.044). Compared with never smokers, current smokers with higher dietary anthocyanin intake were more likely to be post-menopausal (6% in the current-smoker group versus 10% in the never-smoker group) although age was slightly younger (mean age, 58.92 versus 59.54); and they also have lower physical activity levels, higher alcohol intake, and lower diet quality (including lower fiber, protein, fruit, and vegetable intake). Further stratified analyses showed the inverse association was only observed among current smokers who were ≥ 65 years or post-menopausal.

Discussion

To our knowledge, this is the first large population-based study to examine associations between dietary flavonoids, major flavonoid subclasses and flavonoid food sources, and telomere length. In this cross-sectional study of well-characterized females, we observed no overall associations for total flavonoids and subclasses. However, we observed that a higher intake of anthocyanins was independently associated with longer telomere length in females aged < 55 y after comprehensive adjustment for established lifestyle and dietary risk factors. In food-based analyses, 2 servings/week of blueberries (equal to one cup) were also associated with significantly longer telomere length among females aged < 55 y. These findings provide important preliminary data that simple dietary changes, including several servings/week of anthocyanin-rich berries as part of a habitual diet, could promote healthy aging among middle-aged females.

We only observed an association between anthocyanins and TL when we restricted our analyses to participants under 55 y. These data support previous findings from NHS where we observed that female participants consuming the highest quintiles of intakes of several flavonoids (including anthocyanins) at midlife had greater odds of healthy aging [17, 48]. In previous work we have also shown that a higher habitual anthocyanin intake is associated with the greatest benefit in middle-aged females for cardiovascular outcomes such as myocardial infarction and hypertension [42, 49]. Our current data adds further weight to the importance of TL as a potential mechanism to explain these previously observed associations. However, given the null association in the overall population, the subgroup-specific association of anthocyanins and TL in middle-aged women should be further validated using longitudinal repeated measures in a more diverse population.

Our understanding of telomere dynamics supported the potential for diet to more likely exert a protective impact on TL in midlife rather than older ages [50]. Prior meta-analysis has revealed a non-linear shortening trend of TL across the human lifespan, with the shortening rate remaining at a relatively stable rate after age 50 [2]. Longitudinal analyses also suggested that the rate of age-dependent leukocyte TL attrition in women is faster during the premenopausal period than the postmenopausal period [51], which could be explained by the fact that estrogen can potentially stimulate telomerase, as an estrogen-response element is present in the promoter of the catalytic subunit of telomerase [52]. Consistently, we also observed that the association of anthocyanin intake was even more pronounced when limited to premenopausal women. Together, these patterns may point to a window of greater telomere “plasticity” at midlife, where telomeres change dynamically and are more modifiable. At older ages, cumulative TL attrition, reduced residual variability, and survivor selection likely dilute the associations, which may explain why we only observed the association of anthocyanins with TL in women aged < 55 y and those who were premenopausal. Further studies applying objective biomarkers are still needed to understand the dynamics between anthocyanin intake, hormonal status, and telomere attrition.

However, we did not find any evidence between other flavonoid subclasses, which was not unexpected given the differences in bioactivity, bioavailability, and health effects across flavonoid subclasses [53–55]. Among flavonoid subclasses, anthocyanins are most strongly linked to anti-inflammatory effects, consistent across biomarkers functionally related to acute inflammation, cytokines, and oxidative stress [14]. Another study found that anthocyanidins exhibited the strongest inverse associations with biological aging processes [56]. Mechanistic data also support the biological plausibility for TL preservation with anthocyanins through reduction in oxidative stress and inflammation that typically drive telomere attrition [50, 57–60]. The lack of association for other flavonoid subclasses might also be related to intake levels. For example, flavan-3-ol levels that have shown vascular benefits in intervention studies were 112 mg/d, [61] whereas the median intake in our population was only 23.4 mg/d, with only the top quintile of intake reaching this amount (median: 153.7 mg/d; range: 80.8–738.5 mg/d). The association of flavonoid subclasses with TL warrants further exploration across a wider range of intake levels.

Previous studies have examined the associations between dietary constituents and TL. Increased adherence to a Mediterranean-style dietary pattern was associated with slower TL attrition, and positive associations for other foods/constituents including dietary fiber, vegetables, fruit, coffee and longer TL have also been reported [5–9]. The magnitude of the association that we observed for anthocyanin intake at midlife was comparatively greater than that previously observed for the Mediterranean-style dietary pattern. The difference in least squares mean TL z scores between the highest and lowest quintiles were 0.11 for the Mediterranean-style dietary pattern and 0.18 for the anthocyanin intake [5]. When we extrapolate this, these data suggest that difference in telomere length between the highest and lowest quintiles corresponds on average to 6.6 years of aging, which was also more pronounced than the difference observed when comparing smokers with nonsmokers (4.6 years) and physically active participants with inactive participants (4.4 years) [43, 47].

An unexpected finding was observed among current smokers, with higher anthocyanin intake associated with shorter TL. This pattern contradicts prior studies that reported an association between higher anthocyanin intakes and a lower risk of incident COPD among current and former smokers, with indications that anthocyanin intake may partially mitigate the adverse effects of smoking intensity [62]. However, current smokers with greater anthocyanin intake may distinctly have poorer lifestyle patterns and be menopausal. Further, some residual unmeasured confounding within smokers may exist, such as smoking dose and duration and differences in correlated lifestyle characteristics and menopausal status. More broadly, assumed nutrient content levels ignore heterogeneity in growing conditions and manufacturing processes of anthocyanin content of plant-based foods. Additionally, the small number of current smokers introduce change as an explanation for our results. Accordingly, this subgroup-specific result warrants confirmation in larger samples.

The strengths of this study include the large sample size, detailed data on important risk factors and confounders, and a comprehensive assessment of flavonoids and subclasses present in the habitual diet. The limitations of our study also warrant discussion. First, the cross-sectional design limits causal inference. Second, despite detailed adjustments for many self-reported clinical, lifestyle and dietary factors, unmeasured or imperfectly measured confounding factors (e.g., smoking intensity and duration, hormonal status, flavonoid sources) cannot be fully controlled in the current study. Third, self-reported measurement error was inevitable. Our FFQ assessed the frequency of intake for each food up to ≥ 6 times per day; therefore, we can still differentiate between higher and lower intake. While the flavonoid content of plant-based foods varies by growing conditions and manufacturing processes, the ranking of intake and comparison of high versus low intake in a large cohort permits focus on the relative ranking of intake levels with TL. Additionally, only a subset of more commonly consumed anthocyanin-rich foods was measured by self-report; further studies incorporating more diverse anthocyanin food sources or objective measures of anthocyanin intake are warranted. Fourth, our study relies on a single measure of telomere length; therefore, the interindividual variability in telomere length over longer periods cannot be examined. The lack of serial measurements may limit our ability to detect associations, and further validation is needed through longitudinal repeated measures of both diet and TL. Fifth, DNA oxidation or degradation during long-term storage may result in measurement errors in the TL assay. In the NHS, buffy coat fractions were stored at ≤  − 130 °C, and extracted DNA samples were stored at ≤  − 80 °C before the TL assay. Evidence has suggested that buffy coats are a reliable source of good-quality DNA for genome-wide association analysis and other genetic testing after up to 9 years' storage in a − 80 °C frozen state, [63] although it still needs to be confirmed whether this is also true for TL assays. We used the TL z-score to minimize between-batch variation of the TL assay, but the non-differential measurement error might attenuate true associations. Sixth, we conducted a wide range of analyses, with only an inverse association observed between anthocyanin intake and TL in women aged < 55 y. While this finding is supported by potential mechanisms that explain the attenuation of TL, [50, 57–59, 64, 65] this inverse association should be interpreted with caution and viewed as hypothesis-generating as the possibility of a false positive finding cannot be fully excluded. Seventh, our estimates of age-related changes in TL are based on cross-sectional estimates derived within the NHS, which might not be directly comparable to other studies. Finally, our study included predominantly white middle-aged and older female participants. Our participants were healthy controls free of major diseases, selected from multiple nested case–control studies, and healthy user bias may exist (e.g., higher diet quality, physical activity, preventive health behaviors), which could influence both flavonoid intake and TL and thereby bias estimates. Therefore, residual confounding and limited generalizability should be considered when interpreting our findings, and replication in men and other population groups is needed.

In conclusion, our results did not support an association between total flavonoid intake and telomere length attrition in this cross-sectional analysis. However, our subgroup analysis provides evidence to suggest that a higher habitual intake of anthocyanins and anthocyanin-rich foods (i.e., blueberries and other red/blue colored fruits and vegetables) are associated with longer TL in middle-aged females, which may offer potential mechanistic insight into the previously observed associations for anthocyanins and reduction in risk in a range of health outcomes, including MI, hypertension, and cognitive decline. Prospective studies and randomized trials are now needed to examine whether anthocyanin-rich foods promote healthy aging, as assessed by repeated TL measurements.

Authors’ contributions

The authors’ responsibilities were as follows: IDV, HDS, AC: design and concept; MVD: conducted statistical analysis; SL, IDV, HDS, AC: interpreted data and drafted manuscript; SD, VS, GS: provided critical review of the manuscript; IDV, HDS, AC: guarantors of the work; and all authors: read and approved the final manuscript.

Funding

The Nurses’ Health Study reported in this manuscript was supported by National Institutes of Health grants UM1 CA186107, P01 CA87969, and R01 CA49449.

Data availability

A detailed plan for common data sharing procedures to be used in the Nurses’ Health Study has been reviewed by Nurses’ Health Study external advisory committee, which endorses the use of a data enclave approach to data sharing. Along with a general description of the cohort, the policies and guidelines for access to biospecimens and questionnaire data are published on the Nurses’ Health Study website (https://www.nurseshealthstudy.org/researchers) for outside investigators interested in accessing the resources of the study.

Declarations

Ethics and Consent to Participate

The study protocol was approved by the institutional review boards of the Brigham and Women’s Hospital (IRB Protocol Number: 1999P011114), and participants provided consent by returning their questionnaires and providing blood samples.

Conflicts of interest

AC has received funding from the US Highbush Blueberry Council (USHBC) with oversight from USDA and acts as an advisor to the USHBC grant committee. All other authors declare no relevant conflicts of interest. The funders of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the manuscript.