Long-Term Air Pollution Exposure and Mitochondrial DNA Copy Number: An Analysis of UK Biobank Data

Introduction

The adverse health effects of air pollution are believed to be mediated in part through oxidative stress.¹ Mitochondrial DNA (mtDNA) is susceptible to oxidative stress and damaged mtDNA produces excessive reactive oxygen species (ROS), further aggravating mtDNA damage. mtDNA copy number (mtDNA-CN) is a marker of mitochondrial dysfunction that can be measured in peripheral blood, and low mtDNA-CN has been associated with adverse health effects.²

Although air pollution is a pervasive source of oxidative stress, the association between long-term exposure to air pollution and mtDNA-CN has been inconclusive,^3–5 limited by small sample sizes,^4,5 wide variations in exposure concentration and composition, and heterogeneity in exposure and mtDNA-CN measurement. We therefore evaluated the association between long-term exposure to air pollution [particulate matter $\le 10\mathrm{\mu }\mathrm{m}$ (${\mathrm{PM}}_{10}$) and $\le 2.5\mathrm{\mu }\mathrm{m}$ in aerodynamic diameter (${\mathrm{PM}}_{2.5}$), black carbon, and nitrogen dioxide (${\mathrm{NO}}_2$)] with mtDNA-CN in over 45,000 adults from the UK Biobank study.

Methods

We used data on 45,665 participants from the UK Biobank who were enrolled in 2010 with available whole genome sequencing (WGS) data.⁶ Study participants provided demographic information, medical history, blood samples, and underwent a physical examination at enrollment.

Air pollutant concentrations were modeled using a land-use regression model developed for the European Study of Cohorts for Air Pollution Effects (ESCAPE) project.^7,8 The leave-one-out cross-validation $R^2$ for ${\mathrm{PM}}_{10}$, ${\mathrm{PM}}_{2.5}$, black carbon, and ${\mathrm{NO}}_2$ in the UK were 0.75–0.88, 0.21–0.77, 0.81–0.92, and 0.75–0.87, respectively. The annual average concentrations of air pollutants in 2010 were estimated for each participant’s residential address.

To estimate mtDNA-CN, we ran the MitoHPC pipeline⁹ on WGS data, given that WGS performs substantially better than quantitative real-time polymerase chain reaction or whole exome sequencing techniques for mtDNA-CN. mtDNA-CN was calculated as mitochondrial coverage relative to nuclear genome coverage, log-transformed, and standardized so that the mtDNA-CN metric represents standard deviation (SD) units ($z$-scores).

We used linear regression models to estimate the association between a $10{-\mathrm{\mu }\mathrm{g}/\mathrm{m}}^3$ increase in air pollutant exposure ($1{\mathrm{\mu }\mathrm{g}/\mathrm{m}}^3$ for black carbon) and mtDNA-CN. Standard errors were estimated using sandwich covariance matrix estimation (sandwich package in R) to account for the dependence among genetically related participants. In addition, we compared the differences in mtDNA-CN by quintiles of air pollutant concentrations using the lowest quintile as the reference. We evaluated linear trends across quintiles of exposure by including the quintile indicator as a continuous variable. We further modeled the dose–response relationship using restricted cubic splines (3 knots at the 10th, 50th, and 90th percentiles of the distribution of each air pollutant). All analyses were performed using R (version 4.1.3; R Development Core Team).

All participants provided written informed consent. The protocol and procedures of the UK Biobank study were approved by the UK North West–Haydock Research Ethics Committee. Ethical approval was provided by the institutional review board of the Johns Hopkins School of Medicine.

Results

The mean age $\pm \text{SD}$ of the 45,665 study participants (20,771 men and 24,894 women) was $56.8\pm 8.1$ years. The median (interquartile range) annual average concentrations of ${\mathrm{PM}}_{10}$, ${\mathrm{PM}}_{2.5}$, black carbon, and ${\mathrm{NO}}_2$ were 16.2 (15.5–17.1), 9.9 (9.3–10.4), 1.2 (1.0–1.4), and $27.8\left(22.7–32.4\right){\mathrm{\mu }\mathrm{g}/\mathrm{m}}^3$, respectively. mtDNA-CN ranged from $-4.57$ to 6.92. Compared with participants in the lowest quintile of mtDNA-CN, those in the highest quintile were younger (mean age $=57.9\pm 8.1$ vs. $55.6\pm 8.1$), more likely to be never/former smokers (89.1% vs. 90.8%), and less likely to have prevalent hypertension, diabetes, hyperlipidemia, or cardiovascular disease.

In the fully adjusted model (Table 1), a $10{-\mathrm{\mu }\mathrm{g}/\mathrm{m}}^3$ increase in ${\mathrm{PM}}_{10}$, ${\mathrm{PM}}_{2.5}$, and ${\mathrm{NO}}_2$ was associated with a difference in mtDNA-CN of $-0.058$ units [95% confidence interval (CI): $-0.108$ units, $-0.007$], $-0.043$ units ($-0.146$, 0.060), and $-0.008$ units ($-0.021$, 0.005), respectively. For black carbon, the corresponding difference in mtDNA-CN associated with a $1{-\mathrm{\mu }\mathrm{g}/\mathrm{m}}^3$ increase in exposure was $-0.036$ units ($-0.068$, $-0.004$). There was a linear inverse association of ${\mathrm{PM}}_{10}$ with mtDNA-CN and an inverse association with mtDNA-CN at higher concentrations of black carbon and ${\mathrm{NO}}_2$ (Figure 1).

Table 1.

Average difference in mtDNA-CN (95% CI) associated with a $10{-\mathrm{\mu }\mathrm{g}/\mathrm{m}}^3$ increase in each air pollutant and by quintile of air pollutant in 2010, among 45,665 participants from the UK Biobank.

	Linear (per ${10-\mathrm{\mu }\mathrm{g}/\mathrm{m}}^3$ increase)	Quintile					$p_{\text{Trend}}$
		1	2	3	4	5
${\mathrm{PM}}_{10}$	—	11.78, 15.22 ($n=\mathrm{9,098}$)	15.23, 15.94 ($n=\mathrm{8,978}$)	15.95, 16.52 ($n=\mathrm{8,982}$)	16.53, 17.41 ($n=\mathrm{9,006}$)	17.42, 26.16 ($n=\mathrm{8,982}$)	—
Model 1	$-0.083$ ($-0.134$, $-0.031$)	Ref	$-0.022$ ($-0.051$, 0.007)	$-0.013$ ($-0.041$, 0.016)	$-0.055$ ($-0.084$, $-0.027$)	$-0.031$ ($-0.060$, $-0.002$)	0.004
Model 2	$-0.084$ ($-0.139$, $-0.029$)	Ref	$-0.006$ ($-0.037$, 0.025)	$-0.007$ ($-0.038$, 0.024)	$-0.057$ ($-0.088$, $-0.026$)	$-0.033$ ($-0.064$, $-0.002$)	0.001
Model 3	$-0.058$ ($-0.108$, $-0.007$)	Ref	0.004 ($-0.025$, 0.032)	0.010 ($-0.028$, 0.038)	$-0.029$ ($-0.057$, $-0.000$)	$-0.019$ ($-0.048$, 0.009)	0.03
${\mathrm{PM}}_{2.5}$	—	8.17, 9.16 ($n=\mathrm{9,129}$)	9.17, 9.65 ($n=\mathrm{9,029}$)	9.66, 10.04 ($n=\mathrm{8,973}$)	10.05, 10.54 ($n=\mathrm{8,937}$)	10.55, 18.41 ($n=\mathrm{8,978}$)	—
Model 1	$-0.079$ ($-0.181$, 0.024)	Ref	$-0.029$ ($-0.057$, $-0.000$)	0.006 ($-0.022$, 0.035)	$-0.014$ ($-0.043$, 0.015)	$-0.031$ ($-0.060$, $-0.02$)	0.16
Model 2	$-0.057$ ($-0.168$, 0.055)	Ref	$-0.022$ ($-0.052$, 0.009)	0.015 ($-0.016$, 0.046)	$-0.001$ ($-0.032$, 0.030)	$-0.026$ ($-0.058$, 0.005)	0.39
Model 3	$-0.043$ ($-0.146$, 0.060)	Ref	$-0.005$ ($-0.034$, 0.023)	0.021 ($-0.007$, 0.049)	0.011 ($-0.017$, 0.040)	$-0.017$ ($-0.046$, 0.012)	0.64
Black carbona	—	0.83, 1.01 ($n=\mathrm{9,358}$)	1.02, 1.14 ($n=\mathrm{9,031}$)	1.15, 1.28 ($n=\mathrm{8,735}$)	1.29, 1.44 ($n=\mathrm{8,977}$)	1.45, 3.71 ($n=\mathrm{8,945}$)	—
Model 1	$-0.038$ ($-0.071$, $-0.006$)	Ref	$-0.037$ ($-0.066$, $-0.009$)	0.005 ($-0.024$, 0.034)	$-0.025$ ($-0.054$, 0.004)	$-0.039$ ($-0.068$, $-0.009$)	0.05
Model 2	$-0.051$ ($-0.833$, $-0.185$)	Ref	$-0.038$ ($-0.069$, $-0.008$)	0.008 ($-0.023$, 0.040)	$-0.034$ ($-0.065$, $-0.002$)	$-0.047$ ($-0.079$, $-0.016$)	0.01
Model 3	$-0.036$ ($-0.068$, $-0.004$)	Ref	$-0.004$ ($-0.033$, 0.024)	0.035 (0.007, 0.064)	$-0.010$ ($-0.039$, 0.018)	$-0.027$ ($-0.056$, $-0.002$)	0.08
${\mathrm{NO}}_2$	—	12.93, 21.61 ($n=\mathrm{9,023}$)	21.62, 25.66 ($n=\mathrm{9,029}$)	25.67, 29.72 ($n=\mathrm{8,992}$)	29.73, 33.38 ($n=\mathrm{9,027}$)	33.39, 97.74 ($n=\mathrm{8,995}$)	—
Model 1	$-0.012$ ($-0.028$, 0.001)	Ref	$-0.007$ ($-0.036$, 0.022)	$-0.012$ ($-0.041$, 0.017)	$-0.002$ ($-0.031$, 0.027)	$-0.028$ ($-0.057$, 0.002)	0.14
Model 2	$-0.012$ ($-0.025$, 0.002)	Ref	$-0.001$ ($-0.031$, 0.030)	$-0.009$ ($-0.040$, 0.022)	0.004 ($-0.027$, 0.035)	$-0.028$ ($-0.059$, 0.004)	0.18
Model 3	$-0.008$ ($-0.021$, 0.005)	Ref	0.017 ($-0.012$, 0.045)	0.021 ($-0.008$, 0.049)	0.025 ($-0.003$, 0.054)	$-0.015$ ($-0.045$, 0.014)	0.57

Note: Model 1: Adjusted for age (continuous), sex (men and women), self-reported ethnic background (White, Black, Asian, and Other), and study center; Model 2: $\text{Model\hspace{0.17em}}1+\text{average\hspace{0.17em}annual\hspace{0.17em}income}$ ($<\pounds \mathrm{18,000}$, £18,000–£30,999, £31,000–£51,999, £52,000–£100,000, and $>\pounds \mathrm{100,000}$), education level (less than college, college or university degree, professional degree, and other), smoking (never, former, and current), alcohol intake (never, former, and current), physical activity (walking, moderate, and vigorous physical activity), body mass index (continuous), and history of hypertension, diabetes, hyperlipidemia, and cardiovascular disease; Model 3: $\text{Model\hspace{0.17em}}2+\text{cell\hspace{0.17em}counts}$ (red blood cells, neutrophils, lymphocytes, basophils, eosinophils, monocytes, and platelets as continuous variables). Hypertension was defined as a self-reported physician’s diagnosis of hypertension, a self-reported use of antihypertensive medication, or a measured systolic blood pressure $\ge 140\text{\hspace{0.17em}mmHg}$ or diastolic blood pressure $\ge 90\text{\hspace{0.17em}mmHg}$. Diabetes was defined as a self-reported physician’s diagnosis of diabetes, a self-reported use of antidiabetic medication, or a measured HbA1c $\ge 6.5\%$. Hyperlipidemia was defined as a self-reported use of lipid-lowering medication, or a measured total cholesterol $\ge 200\mathrm{mg}/\text{dL}$ or triglycerides $\ge 150\mathrm{mg}/\text{dL}$. Cardiovascular disease was defined as the presence of either myocardial infarction or stroke based on the algorithm developed by the UK Biobank. —, not applicable; CI, confidence interval; HbA1c, hemoglobin A1c; mtDNA-CN, mitochondrial DNA copy number; ${\mathrm{NO}}_2$, nitrogen dioxide; ${\mathrm{PM}}_{2.5}$, particulate matter $\le 2.5\mathrm{\mu }\mathrm{g}$ in aerodynamic diameter; ${\mathrm{PM}}_{10}$, particulate matter $\le 10\mathrm{\mu }\mathrm{g}$ in aerodynamic diameter; Ref, reference.

^a The linear estimates for black carbon are per $1{-\mathrm{\mu }\mathrm{g}/\mathrm{m}}^3$ increase.

Figure 1.

Average mtDNA-CN levels by 2010 residential ${\mathrm{PM}}_{10}$, ${\mathrm{PM}}_{2.5}$, black carbon, and ${\mathrm{NO}}_2$ concentrations, in 45,665 participants from the UK Biobank. The curves represent estimated mtDNA-CN levels (solid line) and their 95% CIs (gray area) by ${\mathrm{PM}}_{10}$, ${\mathrm{PM}}_{2.5}$, black carbon, and ${\mathrm{NO}}_2$ concentrations based on fully adjusted regression models using restricted cubic splines to model air pollutants with knots at the 10th, 50th, and 90th percentiles of its distribution. The spline regression model was adjusted for age (continuous), sex (men and women), self-reported ethnic background (White, Black, Asian, and Other), average annual income ($<\pounds \mathrm{18,000}$, £18,000–£30,999, £31,000–£51,999, £52,000–£100,000, and $>\pounds \mathrm{100,000}$), education level (less than college, college or university degree, professional degree, and other), smoking (never, former, and current), alcohol intake (never, former, and current), physical activity (walking, moderate, and vigorous physical activity), body mass index (continuous), history of hypertension, diabetes, hyperlipidemia, and cardiovascular disease, and blood cell counts (red blood cells, neutrophils, lymphocytes, monocytes, eosinophils, basophils, and platelets as continuous variables). The histograms show the distribution of the concentrations of each air pollutant in the overall study population. Corresponding numeric data for the spline figure are available in our GitHub repository (https://github.com/ArkingLab/Air-pollution-and-mtDNA_CN). Note: CI, confidence interval; mtDNA-CN, mitochondrial DNA copy number; ${\mathrm{NO}}_2$, nitrogen dioxide; ${\mathrm{PM}}_{2.5}$, particulate matter $\le 2.5\mathrm{\mu }\mathrm{g}$ in aerodynamic diameter; ${\mathrm{PM}}_{10}$, particulate matter $\le 10\mathrm{\mu }\mathrm{g}$ in aerodynamic diameter.

Discussion

In this large-scale population study, higher concentrations of air pollutants were inversely associated with mtDNA-CN. Exposure to air pollutants impact various mitochondrial functions, including oxidative phosphorylation, calcium regulation, and mitochondrial membrane potential, in both in vitro and in vivo studies.¹ In animal studies, particulate matter also accelerated the production of ROS and disturbed the fission and fusion of mitochondria, resulting in mitochondrial dysfunction.¹

Long-term exposure to ${\mathrm{PM}}_{2.5}$ was inversely associated with mtDNA-CN in 2,758 healthy women from the Nurses’ Health Study.³ In addition, ${\mathrm{PM}}_{2.5}$ exposure in the past year was inversely associated with mtDNA-CN in two populations of older adults.^4,5 In our study, ${\mathrm{NO}}_2$ was also inversely associated with mtDNA-CN at higher concentrations. On the other hand, 3-y average exposure to ${\mathrm{PM}}_{10}$, ${\mathrm{PM}}_{2.5}$, and ${\mathrm{NO}}_2$ measured in $10{\text{-km}}^2$ grids was positively associated with mtDNA-CN in a rural Chinese population ($N=\mathrm{2,707}$).¹⁰ The different results of this study compared with our study may be due to differences in the duration and concentration of exposure, in the methods of mtDNA-CN assessment, in the population characteristics, or to increased variability in small samples.

The limitations of our study include the cross-sectional design, the measurement of mtDNA-CN on a single occasion, and the use of air pollution measurements derived from prediction models for the participants’ residential addresses, which may be subject to misclassification. Moreover, the UK Biobank comprises mostly White individuals, and the generalizability of the findings to other populations is unknown.

In conclusion, long-term exposure to air pollutants was inversely associated with mtDNA-CN. These findings suggest that oxidative stress-induced mitochondrial dysfunction, reflected by reduced mtDNA-CN, may be a potential mechanism mediating the adverse health effects of air pollution.