Polyunsaturated fatty acids modulation of smoking-related epigenetics and pulmonary outcomes

Abstract

Background

Cigarette smoking results in epigenetic alterations that persist after smoking cessation and is a major risk factor for chronic lung disease. Polyunsaturated fatty acids (PUFAs) may promote epigenetic recovery and support lung health. We aimed to characterize PUFA biomarker associations with lung phenotypes in high-risk populations with a smoking history and determine whether smoking-related DNA methylation (DNAm) mediates these associations.

Methods

In this observational study we analyzed blood-based omega-3 and omega-6 PUFA biomarkers, spirometry, chest computed tomography (CT) measures, and smoking-related DNAm in 3857 former and current smokers in the Genetic Epidemiology of COPD (COPDGene) study. PUFA associations with lung phenotypes and DNAm were modeled with robust linear regression and linear mixed models. Mediation analysis estimated PUFA effects mediated through DNAm. Models adjusted for demographics, smoking history, genotype principal components, and, where relevant, cell type proportions and CT scanner. Replication in blood and extension to lung tissue were tested in the Lung Tissue Resource Consortium (LTRC).

Results

Higher omega-3s were associated with higher lung function, less emphysema, and less airway wall thickening, while higher omega-6s and a higher omega 6:3 ratio were associated with worse lung phenotypes. Higher omega-3s and omega-6s were each associated with higher DNAm at cg05575921 in the aryl hydrocarbon receptor repressor gene (AHRR), lower epigenetic smoking scores, and a slower epigenetic pace of aging; the omega 6:3 ratio showed opposite associations. Smoking-related DNAm partially mediated some PUFA- lung phenotype asssociations. Omega-3s showed beneficial direct and mediated effects; some omega-6s showed detrimental direct effects but beneficial mediated effects; and the omega 6:3 ratio showed detrimental direct and mediated effects. For example, direct and AHRR DNAm-mediated effects on FEV1 were 0.0201 and 0.0049 for total omega-3s, -0.0049 and 0.0009 for total omega-6s, and − 0.0074 and − 0.0014 for the omega 6:3 ratio. Replication analysis in LTRC generally showed consistent directions of effects across blood and lung tissue.

Conclusions

PUFAs may mitigate smoking-related epigenetic alterations, with omega-3s particularly associated with better lung outcomes. These findings have implications for PUFA-focused precision nutrition strategies in high-risk populations.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12263-026-00812-0.

Background

Respiratory health is impacted by multiple environmental factors throughout the lifespan. Cigarette smoke exposure is a major contributor, leading to airway inflammation, emphysema, impaired lung function and increased risk of chronic obstructive pulmonary disease (COPD). While smoking cessation can improve prognosis, former smokers remain at higher risk for respiratory disease morbidity and mortality. Polyunsaturated fatty acids (PUFAs), including omega-3 (N3) and omega-6 (N6) PUFAs, represent modifiable dietary factors with immunomodulatory properties. N3 and N6 PUFAs are generally thought to have contrasting effects on inflammation, with the majority of their metabolites serving as potent anti- and pro-inflammatory signaling molecules, respectively. Accordingly, higher N3 PUFAs have been associated with better lung outcomes, with some evidence suggesting larger effects in those with a history of smoking [1–6]. However associations of higher N6 PUFAs and their interactions with smoking are less clear [7–9], and the ratio of N6 to N3 PUFAs (N6:N3 ratio), which has been implicated in inflammation-related chronic disease [10–12], remains understudied.

Inflammation is hypothesized to sit at the intersection of smoking, PUFAs and lung disease, but precise mechanisms are unknown. Cigarette smoking is associated with robust epigenetic alterations that can persist long after smoking cessation [13–22]. Notable among these is hypomethylation at cg05575921 in the aryl hydrocarbon receptor repressor gene (AHRR DNAm), which is associated with multiple smoking behaviors, including current, former and never smoking status, pack-years and duration of smoking, and time since smoking cessation [14, 15, 18, 21]. Smoking and AHRR DNAm also associate with epigenetic aging, which may play a mechanistic role in the adverse effects of smoking on chronic disease outcomes [23–27]. PUFAs have also been linked to DNA methylation (DNAm) and epigenetic aging [28–32], with some evidence suggesting N3 PUFAs specifically may protect against the harmful epigenetic effects of environmental toxins [33–37]. Characterizing PUFA interactions with smoking and DNAm in relation to respiratory disease traits could advance our understanding of the molecular mechanisms underlying COPD heterogeneity and inform precision nutrition approaches to promote lung health in populations with a higher inflammatory load.

With this study we evaluated PUFA biomarker associations with lung function, emphysema and airway-wall thickness in high-risk populations enriched for ever-smokers and investigated whether these associations were mediated through differences in DNAm marks of cigarette smoking. We hypothesized that associations of PUFA biomarkers with lung function may depend on smoking behaviors, and that PUFAs may help promote epigenetic recovery for smoking-related DNAm. Preliminary results informing this study have previously been reported in abstract form [38–41].

Methods

This study is reported in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines. Completed STROBE checklists, including the nutritional epidemiology (STROBE-nut) and molecular epidemiology (STROBE-ME) extensions, are provided as additional information.

Participants

Our primary analysis was conducted in the Genetic Epidemiology of COPD (COPDGene) study (NCT00608764, registered January 28, 2008). We included all COPDGene participants with a history of smoking who had PUFA, DNAm, and spirometry and/or CT imaging measurements at the 5-year follow-up visit (e-Figure 1). Replication analyses and extension to lung tissue DNAm were performed in the Lung Tissue Research Consortium (LTRC) study (NCT02988388, registered December 7, 2016). Both studies were approved by the Institutional Review Boards of respective research centers. All participants provided written informed consent. Please see e-Appendix 1 for additional details.

Measurements

Details on PUFA measurements are in e-Appendix 1. Briefly, PUFAs in COPDGene were measured in plasma samples by gas-chromatography mass-spectrometry (GCMS) at Cornell University. We focused on the PUFAs regarded as most important to human health, including the N3s ALA, EPA and DHA, and N6s LA and AA, and relevant summary metrics, including the N3 index (EPA + DHA), total N3s (ALA + DPA + EPA + DHA), total N6s (LA + GLA + DGLA + AA), and the N6:N3 ratio. PUFAs in LTRC were measured in plasma samples as part of the Broad Metabolomics targeted amide negative ion mode analysis of central metabolites at the TOPMed Metabolomics Core at BIDMC. PUFAs measured in LTRC included the N3s EPA and DHA and N6s LA and AA.

Peripheral blood leukocyte (COPDGene and LTRC) and lung tissue (LTRC only) DNAm was measured using the Illumina EPIC array. Raw DNAm data were imported and processed using the Bioconductor packages minfi and Enmix [42, 43]. Postprocessed DNAm beta values were used for all calculations. This study focused on DNAm at cg05575921 in the aryl hydrocarbon receptor repressor gene (referred to as AHRR DNAm), a robust epigenetic marker of smoking [14, 15, 18, 21], two multi CpG epigenetic smoking scores (smokingScore and methylationScore, estimated using the R package EpiSmokEr [44]) and the epigenetic pace of aging (DunedinPACE, estimated using the R package DunedinPACE [45]).

Lung phenotypes were measured as previously described [46, 47]. Briefly, lung function (FEV₁, FVC and FEV₁/FVC) was measured by spirometry using the ndd EasyOne Spirometer (Zurich, Switzerland). Emphysema traits (low-attenuation area at -950 Hounsfield units [LAA950] and adjusted lung density) and airway wall thickening (Pi10, defined as the predicted square root of airway wall thickness for an airway with a 10-mm internal perimeter) were measured by CT imaging using GE, Philips, or Siemens multi-detector CT scanners.

Statistical analyses

PUFA biomarker levels by smoking status were compared with two-sample t-tests. Differences in PUFAs between current and former smokers accounting for age, sex, genotype principal components (PCs), COPD status, income category, BMI, and exercise were evaluated with ordinary least squares regression. Cross-sectional associations of PUFA biomarkers with lung phenotypes and smoking-related DNAm were modeled with robust linear regression. Exploratory analysis evaluated effect modification by smoking behaviors associated with disease trajectories and epigenetic variability, including current vs. former smoking status, time since smoking cessation, and smoking pack-years, by including PUFA × effect modifier multiplicative interaction terms. Sensitivity analyses adjusting for BMI, income category, COPD status and recent exercise were performed. Longitudinal associations of PUFAs with changes in lung phenotypes across the baseline, 5-year and 10-year follow-up visits were evaluated with linear mixed effects models with a random intercept and a PUFA × age multiplicative interaction term. Mediation analyses evaluating if PUFA associations were mediated through smoking-related DNAm was performed with the R package mediation [48]. All models adjusted for age, age², height, height², sex, smoking status, smoking pack-years, genotype PCs 1–10 and, for DNAm models, cell type proportions. Models for LAA950, Pi10 and adjusted lung density were further adjusted for CT scanner, and LAA950 was log transformed to approximate a normal distribution. A Bonferroni-corrected significance threshold of 0.01 (0.05/the number of individual PUFAs evaluated [N3s ALA, EPA and DHA and N6s LA and AA]) was applied for each outcome for our primary analyses. Results significant at the nominal level of p < 0.05 are also indicated. Interaction terms significant at p < 0.05 were considered as suggestive evidence of effect modification.

Analyses were replicated in LTRC using identical models, except for models evaluating LAA950, which did not adjust for CT scanner model as this information was missing for most LTRC participants. Models of DNAm markers were evaluated separately in blood and lung tissues. Replication was declared for associations significant at p < 0.05 with the same direction of effect.

Results

Participant characteristics

Our primary analysis included 3857 COPDGene participants from the 5-year follow-up visit (Table 1, e-Figure 1). The mean (SD) age was 65 (8.6) years and the mean BMI was 29 (6.3) kg/m2. Participants were 51% (N = 1964) female, 71% (N = 2756) self-reported white and 29% (N = 1101) self-reported black, and 38% (N = 1479) currently smoking. The mean smoking history was 44 (24) pack-years. Two-thirds (N = 2548) of the participants completed at least some college or technical school, and half (N = 1942) of the participants reported an income of $35,000 or less.

Table 1.

Participant characteristics by smoking status

Characteristic	Overall	Former smokers	Current smokers	P-value
	Mean (SD) or N (%)
N	3857	2378	1479
Age, years	65.15 (8.58)	67.99 (8.21)	60.59 (7.07)	< 0.001
Sex
Female	1964 (51%)	1229 (52%)	735 (50%)	< 0.001
Male	1893 (49%)	1149 (48%)	744 (50%)
Height, cm	169.55 (9.58)	169.36 (9.60)	169.84 (9.56)	0.132
BMI, m/kg2	29.00 (6.31)	29.47 (6.21)	28.24 (6.40)	< 0.001
Race
White	2756 (71%)	1986 (84%)	770 (52%)	< 0.001
Black	1101 (29%)	392 (16%)	709 (48%)
School completed
8th grade or less	68 (1.8%)	32 (1.3%)	36 (2.4%)	< 0.001
Some high school	332 (8.6%)	123 (5.2%)	209 (14%)
High school graduate/GED	909 (24%)	487 (20%)	422 (29%)
Some college/technical school	1059 (27%)	642 (27%)	417 (28%)
College/technical school graduate	1061 (28%)	748 (31%)	313 (21%)
Master’s or Doctoral degree	428 (11%)	346 (15%)	82 (5.5%)
Income
<$15,000	1019 (26%)	415 (17%)	604 (41%)	< 0.001
$15,000–35,000	923 (24%)	587 (25%)	336 (23%)
$35,000–50,000	516 (13%)	382 (16%)	134 (9.1%)
$50,000–75,000	465 (12%)	360 (15%)	105 (7.1%)
>$75,000	431 (11%)	345 (15%)	86 (5.8%)
declined to answer	502 (13%)	288 (12%)	214 (14%)
Lung phenotypes
FEV1, liters	2.15 (0.84)	2.09 (0.87)	2.26 (0.78)	< 0.001
FVC, liters	3.14 (0.96)	3.10 (0.95)	3.22 (0.95)	< 0.001
FEV1/FVC	0.68 (0.15)	0.66 (0.16)	0.70 (0.13)	< 0.001
LAA950	5.35 (8.95)	6.89 (10.26)	2.94 (5.56)	< 0.001
Pi10	2.27 (0.58)	2.21 (0.55)	2.37 (0.61)	< 0.001
Adjusted Lung Density	86.23 (25.25)	79.61 (23.94)	96.65 (23.72)	< 0.001
PUFA biomarkers
N3 ALA, % of total FAs	0.76 (0.35)	0.78 (0.37)	0.71 (0.32)	< 0.001
N3 EPA, % of total FAs	0.67 (0.55)	0.75 (0.62)	0.54 (0.36)	< 0.001
N3 DHA, % of total FAs	1.60 (0.71)	1.71 (0.76)	1.43 (0.59)	< 0.001
N3 index, % of total FAs	2.27 (1.14)	2.45 (1.26)	1.97 (0.84)	< 0.001
N3 total, % of total FAs	3.47 (1.30)	3.69 (1.42)	3.12 (0.99)	0.008
N6 LA, % of total FAs	28.30 (5.06)	28.47 (5.01)	28.03 (5.13)	0.005
N6 AA, % of total FAs	9.41 (2.77)	9.31 (2.82)	9.56 (2.69)	0.457
N6 total, % of total FAs	39.69 (5.64)	39.74 (5.52)	39.60 (5.83)	< 0.001
N6:N3 ratio	12.72 (4.26)	12.09 (4.24)	13.74 (4.10)	< 0.001

Participant characteristics and PUFA biomarker descriptives in the overall sample and by smoking status. Values are presented as mean (standard deviation [SD]) or N (%). P-values are from two-sample t-tests or chi-squared tests for differences in means or proportions between former and current smokers. Additional variables along with participant characteristics by self-reported race are included in the online supplement (e-Table 1)

N3 and N6 PUFA levels in COPDGene differed by smoking status, with N3 PUFAs and N6 LA being higher in former smokers, and N6 AA and the N6:N3 ratio being higher in current smokers (Table 1). PUFA differences by smoking status persisted with adjustment for age, sex, genotype PCs, FEV1, COPD status, BMI, income and recent exercise for all PUFAs except for N6 AA (e-Figure 2a). For N6 AA, current smoking was associated with higher N6 AA in unadjusted models but with lower N6 AA in models accounting for covariates. Stepwise variable addition revealed that this change in direction was driven by the first genotype PC (e-Table 2), which correlated with self-reported race (e-Figure 3). Analyses stratified by self-reported race supported this (e-Figure 2b, c).

PUFA associations with lung phenotypes

Higher N3 PUFAs generally associated with higher FEV1, higher FEV1/FVC, and lower Pi10, while higher N6 PUFAs associated with lower FEV1/FVC, higher LAA950, and lower adjusted lung density (Fig. 1, e-Table 3), with most results robust to multiple testing correction for the total number of individual PUFAs evaluated. Sensitivity analyses adjusting for exercise, BMI, income, and/or COPD status yielded largely similar results (e-Figure 4).

Fig. 1.

PUFA associations with lung phenotypes across smoking strata. Bubble plot of PUFA biomarker associations with spirometry and CT imaging traits in the overall sample and by smoking status. Bubble color represents covariate-adjusted Z-values from robust linear regression of lung outcomes on PUFA biomarkers. Bubble size represents -log(p-values). P-values significant at p < 0.05 and p < 0.01 (0.05/total number of individual PUFAs tested) are indicated with a black border and star respectively. LAA950: low-attenuation area at -950 Hounsfield units; ALD: adjusted lung density; Pi10: airway wall thickening, defined as the predicted square root of airway wall thickness for an airway with a 10-mm internal perimeter

Suggestive evidence (p < 0.05) of interactions with smoking status was seen for some PUFA—lung phenotype associations (e-Table 4), with current smoking generally attenuating (driving toward the null) the associations. Stratified analyses similarly revealed that many N3 PUFA—lung phenotype associations were larger in magnitude and significant only in former smokers, though with consistent directions of effect in current smokers (Fig. 1, e-Table 5). In contrast, some N6 PUFAs associated with higher FEV1 (LA) and FVC (LA, total N6) in current smokers only. Evidence of effect modification by time since smoking cessation in former smokers (Fig. 2, e-Table 6), and smoking pack-years in current smokers (e-Figure 5, e-Table 7) was also observed, with PUFA main and interaction effects tending to go in opposite directions, suggesting PUFA effects may be largest around the time of smoking cessation, but may be overwhelmed at high cumulative smoking doses.

Fig. 2.

PUFA x time since smoking cessation interactions in former smokers. Bar plot illustrating effects of time since smoking cessation on PUFA biomarker associations with lung outcomes in former smokers. PUFA main effects are shown in red and PUFA x time since smoking cessation effects are shown in blue. Effect sizes are presented as covariate-adjusted Z values. Bars crossing the dashed red lines are significant at p < 0.05. LAA950: low-attenuation area at -950 Hounsfield units; ALD: adjusted lung density; Pi10: airway wall thickening, defined as the predicted square root of airway wall thickness for an airway with a 10-mm internal perimeter

In longitudinal models, higher N3 PUFAs were generally associated with slower decline in lung function and adjusted lung density, while higher N6 PUFAs and the N6:N3 ratio were associated with faster decline in lung function and adjusted lung density (e-Figure 6, e-Table 8). This trend largely persisted across analyses using lung phenotype data from all study visits, from study visits preceding PUFA measurements, and from study visits after PUFA measurements (e-Figure 6, e-Table 8). Analysis stratified by smoking status revealed that many longitudinal associations, especially for N3 PUFAs, were significant in current but not former smokers (e-Figure 6).

PUFA associations with smoking-related DNAm markers

As expected, smoking-related DNAm differed by smoking status (e-Table 9). Mean AHRR DNAm was lower in current smokers, while the mean epigenetic smoking scores and epigenetic pace of aging were higher in current smokers. In contrast, higher N3 and N6 PUFAs were generally associated with higher AHRR DNAm, lower epigenetic smoking scores, and slower paces of epigenetic aging, while a higher N6:N3 ratio showed the opposite associations (Fig. 3, e-Table 10). Sensitivity analyses adjusting for recent exercise, BMI, income and/or COPD status yielded similar results (e-Figure 7).

Fig. 3.

PUFA associations with smoking-related DNAm markers. Bubble plot of PUFA biomarker associations with smoking-related DNAm in the overall sample and by smoking status. Bubble color represents covariate-adjusted Z-values from robust linear regression of smoking-related DNAm marks on PUFA biomarkers. Bubble size represents -log(p-values). P-values significant at p < 0.05 and p < 0.01 (0.05/total number of individual PUFAs tested) are indicated with a black border and star respectively

Significant PUFA × smoking status interactions were observed for some DNAm markers (Table S11), with current smoking associated with greater magnitude effects for some PUFAs. Stratified analyses revealed that most PUFA—DNAm associations were consistent in direction across smoking strata, except for the association of the N6:N3 ratio with DunedinPACE, where a higher N6:N3 ratio was associated with a slower pace of aging in current smokers only (Fig. 3, table S12). PUFA × years since quit interactions in former smokers and PUFA × pack-years interactions in current smokers showed that PUFA main and interaction effects on smoking-related DNAm tended to go in opposite directions, again suggesting that PUFA effects may be largest around the time of smoking cessation, but may be overwhelmed at high cumulative smoking doses.

Mediation of PUFA effects on lung phenotypes through smoking-related DNAm

Mediation analysis revealed significant mediation of PUFA–lung phenotype associations through smoking-related DNAm (Fig. 4, e-Table 15, e-Figure 10) For both N3 and N6 PUFAs, effects mediated through AHRR DNAm and the epigenetic smoking scores were generally positive for FEV1, FVC and FEV1/FVC and negative for LAA950 and Pi10. In contrast, mediated effects for the N6:N3 ratio were negative for spirometry traits and positive for LAA950 and Pi10. Interestingly, for N3 PUFAs, both the direct and mediated effects tended to be beneficial (i.e. associated with higher lung function, less emphysema and lower airway wall thickness); for some N6 PUFAs, the direct effects were detrimental while the mediated effects remained beneficial; and for the N6:N3 ratio, both the direct and mediated effects tended to be detrimental. For example, for total N3s the AHRR DNAm-mediated and direct effects on FEV1 were both positive (proportion mediated 0.18); for total N6s, the AHRR DNAm-mediated effect on FEV1 was positive while the direct effect was negative (proportion mediated: -0.22); and for the N6:N3 ratio, the AHRR DNAm-mediated and direct effects on FEV1 were both negative (proportion mediated: 0.15)(Fig. 4). We note that the proportion mediated should be interpreted with caution when the mediated and direct effects are not in the same direction [49].

Fig. 4.

Mediation effects on FEV1 through AHRR DNAm. Mediation diagrams showing PUFA biomarker associations with AHRR DNAm (cg05575921) and FEV1 for (a) total N3s, (b) total N6s and (c) the N6:N3 ratio. a=PUFA associations with AHRR DNAm. b=AHRR DNAm associations with FEV1, accounting for PUFA levels, axb=average causal mediated effect (ACME) of PUFA on FEV1 through AHRR DNAm, c’=average direct effect (ADE) of PUFA on FEV1 and c=total effect of PUFA on FEV1. Positive associations are shown in blue and inverse associations are shown in red. Stars indicate significance, with *=p < 0.05, **=p < 0.01, and ***=p < 0.001

Replication in LTRC blood samples and extension to lung tissue

Similar to COPDGene, N3 DHA was significantly higher in former smokers than current smokers in LTRC. Similar trends were seen for N3 EPA and N6 LA, though these differences were not significant (e-Table 16). PUFA—lung phenotype associations in LTRC were consistent in direction to COPDGene for both N3 and N6 PUFAs, with the positive associations of N3 EPA with FEV1/FVC and N3 DHA with FEV1 and FVC, and the negative associations of N6 LA with FEV1/FVC replicating at a significance level of 0.05 (e-Figure 11).

PUFA associations with blood-based smoking-related DNAm also replicated for both N3 and N6 PUFAs (e-Figure 12). Specifically, higher N3 DHA and N6 LA were associated with higher blood AHRR DNAm, lower epigenetic smoking scores, and a slower pace of aging. PUFA associations with DNAm measured in lung tissue were generally consistent in direction to those measured in blood, with a significant association seen for N3 DHA with lung AHRR DNAm (e-Figure 12).

Mediation analyses using blood-based smoking-related DNAm showed replication of some causal mediated effects, specifically for N3 DHA and N6 LA (e-Figure 13). Mediation analyses using lung-based smoking-related DNAm were largely consistent in direction to those seen in blood but did not reach statistical significance for any PUFA—mediator—outcome relationship (e-Figure 13).

Discussion

This study evaluated PUFA biomarker associations with lung function, emphysema and airway wall thickness in a population of ever-smokers at high risk of chronic lung disease, and investigated whether these associations were mediated through differences in DNAm markers of smoking. We found that higher blood N3 PUFAs generally had beneficial associations with lung function and emphysema traits, while higher N6 and a higher ratio of N6:N3 PUFAs tended to have detrimental associations. In contrast, we found that both N3 and N6 PUFAs had associations with smoking-related DNAm consistent with epigenetic recovery, which mediated PUFA effects on lung phenotypes. These patterns replicated in an independent cohort, with some signals persisting across blood and lung tissues.

Our results for N3 PUFAs build on prior evidence supporting a role of N3 PUFAs in lung function and chronic lung disease [1–6, 9, 50]. Additionally, we showed the protective associations of N3 PUFAs extended to radiographic measurements of emphysema and airway wall thickness, findings which, we believe, are novel. Our study is, to our knowledge, the first large-scale evaluation of blood N6 PUFAs and the N6:N3 ratio in relation to lung function, emphysema and airway wall thickness. Prior studies of N6 PUFAs and lung outcomes focused on reports of dietary intake and yielded mixed results [7, 8, 9]. Our findings that higher blood levels of total N6 PUFAs, N6 AA, and the N6:N3 ratio associate with lower lung function, more emphysema and greater airway wall thickness align with the known pro-inflammatory effects of AA and the growing evidence of a role in the N6:N3 ratio for chronic inflammatory diseases. In contrast, we found that the effects of N6 LA varied. LA, which is the predominant N6 in the western diet, is converted to AA in the body. While physiological effects of AA are predominantly pro-inflammatory, intermediates in the conversion of LA to AA can have both pro- and anti-inflammatory properties. Further research aimed at understanding the effects of upstream vs. downstream N6 PUFAs is warranted.

Moving beyond associations with lung phenotypes, we demonstrated for the first time that both N3 and N6 PUFAs associate with higher AHRR methylation, lower epigenetic smoking scores, and a slower epigenetic pace of aging, suggesting that PUFAs may mitigate some epigenetic effects of smoking. DNAm is a plausible mechanism through which PUFAs impact health outcomes [28–32], and PUFA effects on DNAm have also been explored as mechanisms through which PUFAs help alleviate adverse environmental effects [33–37]. Here we found that some PUFA effects on lung outcomes were mediated through effects on smoking-related DNAm and epigenetic aging. Interestingly, for N3 PUFAs the mediated and direct effects were both beneficial, while for N6 PUFAs the mediated effects were beneficial while the direct effects were generally detrimental. This suggests that PUFAs may impact lung phenotypes through multiple pathways, and that the epigenetic effects of N6 PUFAs in particular may be independent from their inflammatory signaling effects.

Smoking cessation improves disease outcomes, yet there is clinical variability among former smokers in lung health trajectories. We investigated PUFA interactions with time since smoking cessation, and found suggestive evidence that PUFA effects on both lung phenotypes and smoking-related DNAm were greatest closest to the time of smoking cessation, a finding of clinical and translational relevance. Previous studies on smoking cessation have demonstrated curvilinear associations with lung phenotypes and AHRR methylation recovery, such that the greatest improvements occur immediately following smoking cessation [13, 21, 51]. Together these findings suggests there may be a window of opportunity around smoking cessation where increasing N3 PUFAs and decreasing the N6:N3 PUFA ratio could support smoking-related epigenetic recovery and optimize lung disease risk reduction. Further research characterizing the dynamic interactions of PUFAs with smoking epigenetics is needed.

This study has several strengths and novel aspects. Our primary analyses were done in COPDGene, a deeply phenotyped prospective cohort study ascertained for current and former smokers with rich epigenomic data, a panel of plasma N3 and N6 PUFAs quantified using gold standard GCMS methods, and 15 years of follow-up. We explored novel, hypothesis-oriented epigenetic mechanisms through which PUFAs influence lung health. We evaluated respiratory disease traits that have not previously been evaluated in relation to PUFAs and considered the impact of current smoking status and time since smoking cessation. Finally, we tested replication in LTRC, a tissue biobank resource ascertained from individuals undergoing lung surgery or biopsy, which provided the opportunity to explore tissue-specificity by extending our analyses to lung tissue. Given the high worldwide prevalence of suboptimal blood and dietary N3 levels and imbalanced N6:N3 ratios, and the global morbidity and mortality associated with chronic lung disease, our findings have clear nutritional, clinical, and public health relevance [12].

Key limitations of our study relate to replication of our findings. LTRC has fewer phenotypic measures, and PUFAs were measured through a targeted metabolomics panel, which may yield less precise quantification. The metabolomics panel did not capture all PUFAs, so we were unable to replicate our results for ALA, total N3s, total N6s, and the N6:N3 ratio. The sample size in LTRC was also smaller than in COPDGene, especially for current smokers, and therefore had lower statistical power. Despite this, multiple signals replicated across COPDGene and LTRC, with signals for DHA modulation of smoking-related DNAm extending across blood and lung tissue. This replication increases confidence in our findings and highlights the need for further research on the N6:N3 ratio in relation to cigarette smoking and lung health.

Conclusions

In this study we found contrasting effects of N3 and N6 PUFA biomarkers on lung function, emphysema, and airway wall thickness, with higher blood N3s being generally beneficial, and higher blood N6s and the blood N6:N3 ratio being detrimental, in line with their known anti- and pro-inflammatory effects. We demonstrated for the first time that PUFAs associate with smoking-related DNAm in patterns consistent with epigenetic recovery, and that these molecular shifts mediate the effects of PUFAs on clinically relevant lung outcomes. These findings highlight potential opportunities for nutritional interventions around smoking cessation and have implications for precision nutrition approaches for lung disease prevention.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

PUFAs

Polyunsaturated fatty acids

N3

Omega-3

N6

Omega-6

FEV1

Forced expiratory volume in the first second

FVC

Forced vital capacity

LAA950

Low-attenuation area at -950 Hounsfield units

Pi10

Predicted square root of airway wall thickness for an airway with a 10-mm internal perimeter

DNAm

DNA methylation

Author contributions

BKP and DLD designed research with input from EKS, DBH and PAC; BKP conducted research and analyzed data; BKP and DLD wrote the paper. DLD had primary responsibility for final content. All authors read and approved the final manuscript.

Funding

Research reported in this manuscript was supported by NHLBI R01 HL149352. BKP, AR and EKP are supported by NIH T32 HL007427. DLD is supported by NIH K24 HL171900, NIH P01HL114501, NIH R01 HL178032 and R01 HG011393. BDL is supported by R01 HL122531 and the Charles and Amelia Gould Trust. This work was supported by NHLBI grants U01 HL089897 and U01 HL089856 and by NIH contract 75N92023D00011. The COPDGene study (NCT00608764) has also been supported by the COPD Foundation through contributions made to an Industry Advisory Committee that has included AstraZeneca, Bayer Pharmaceuticals, Boehringer-Ingelheim, Genentech, GlaxoSmithKline, Novartis, Pfizer, and Sunovion. This study utilized biological specimens and data provided by the Lung Tissue Research Consortium (LTRC) supported by the National Heart, Lung, and Blood Institute (NHLBI).

Data availability

Data used in this study are controlled access data. Data requests should be directed to dbGaP (https://dbgap.ncbi.nlm.nih.gov) with study accession numbers phs000951.v6.p5 (COPDGene) and phs001662.v4.p2 (LTRC).

Declarations

Ethics approval and consent to participate

The Institutional Review Board at Brigham and Women’s Hospital approved this study (protocol numbers 2018P000186 and 2007P000554). All participants provided written informed consent. All data were anonymized prior to analysis.

Competing interests

In the past three years, EKS and DLD received institutional grant support from Bayer and Northpond Laboratories and DLD received a consulting fee from Astra Zeneca. DLD is also a member of the Board of Directors for the COPD foundation. No other authors declared conflicts of interest.