The Associations of Prenatal Exposure to Fine Particulate Matter and Its Chemical Components with Allergic Rhinitis in Children and the Modification Effect of Polyunsaturated Fatty Acids: A Birth Cohort Study

Abstract

Background:

Polyunsaturated fatty acids (PUFAs) have been shown to protect against fine particulate matter $<2.5\mathrm{\mu }\mathrm{m}$ in aerodynamic diameter (${\mathrm{PM}}_{2.5}$)-induced hazards. However, limited evidence is available for respiratory health, particularly in pregnant women and their offspring.

Objectives:

We aimed to investigate the association of prenatal exposure to ${\mathrm{PM}}_{2.5}$ and its chemical components with allergic rhinitis (AR) in children and explore effect modification by maternal erythrocyte PUFAs.

Methods:

This prospective birth cohort study involved 657 mother–child pairs from Guangzhou, China. Prenatal exposure to residential ${\mathrm{PM}}_{2.5}$ mass and its components [black carbon (BC), organic matter (OM), sulfate ($\mathrm{SO}_4{}^{2-}$), nitrate ($\mathrm{NO}_3{}^{-}$), and ammonium ($\mathrm{NH}_4{}^{+}$)] were estimated by an established spatiotemporal model. Maternal erythrocyte PUFAs during pregnancy were measured using gas chromatography. The diagnosis of AR and report of AR symptoms in children were assessed up to 2 years of age. We used Cox regression with the quantile-based g-computation approach to assess the individual and joint effects of ${\mathrm{PM}}_{2.5}$ components and examine the modification effects of maternal PUFA levels.

Results:

Approximately $5.33\%$ and 8.07% of children had AR and related symptoms, respectively. The average concentration of prenatal ${\mathrm{PM}}_{2.5}$ was $35.50\pm 5.31{\mathrm{\mu }\mathrm{g}/\mathrm{m}}^3$. ${\mathrm{PM}}_{2.5}$ was positively associated with the risk of developing AR [hazard ratio $\left(\text{HR}\right)=1.85$; 95% confidence interval (CI): 1.16, 2.96 per $5{\mathrm{\mu }\mathrm{g}/\mathrm{m}}^3$] and its symptoms ($\text{HR}=1.79$; 95% CI: 1.22, 2.62 per $5{\mathrm{\mu }\mathrm{g}/\mathrm{m}}^3$) after adjustment for confounders. Similar associations were observed between individual ${\mathrm{PM}}_{2.5}$ components and AR outcomes. Each quintile change in a mixture of components was associated with an adjusted HR of 3.73 (95% CI: 1.80, 7.73) and 2.69 (95% CI: 1.55, 4.67) for AR and AR symptoms, with BC accounting for the largest contribution. Higher levels of n-3 docosapentaenoic acid and lower levels of n-6 linoleic acid showed alleviating effects on AR symptoms risk associated with exposure to ${\mathrm{PM}}_{2.5}$ and its components.

Conclusion:

Prenatal exposure to ${\mathrm{PM}}_{2.5}$ and its chemical components, particularly BC, was associated with AR/symptoms in early childhood. We highlight that PUFA biomarkers could modify the adverse effects of ${\mathrm{PM}}_{2.5}$ on respiratory allergy. https://doi.org/10.1289/EHP13524

Introduction

Allergic rhinitis (AR) is an IgE-mediated inflammatory disease of the nasal mucosa, characterized by repetitive sneezing, runny nose, and nasal congestion.¹ According to a large population-based survey in China, AR affects 5.3% to 25.8% of preschool children.² Environmental irritants and allergens, including air pollutants, play a crucial role in inducing and exacerbating AR.^3,4 With significant pro-inflammatory capacity,⁵ atmospheric fine particulate matter $<2.5\mathrm{\mu }\mathrm{m}$ in aerodynamic diameter (${\mathrm{PM}}_{2.5}$) has been associated with increased prevalence of AR.^6–10 The vulnerability to ${\mathrm{PM}}_{2.5}$ may begin early in life when the airway structure and immune system are immature.¹¹ Several birth cohorts in Germany¹² and the US¹³ have suggested a positive association between prenatal ${\mathrm{PM}}_{2.5}$ exposure and the risk of allergies and AR in childhood. However, evidence from prospective cohorts is still scarce in China, where ${\mathrm{PM}}_{2.5}$ pollution is notably high.

The toxicity of ${\mathrm{PM}}_{2.5}$ is likely attributable to its chemical constituents.¹⁴ For example, black carbon (BC), a key ${\mathrm{PM}}_{2.5}$ component, has been detected in fetal circulation and lung tissues,¹⁵ contributing to the risk of respiratory allergies after birth.¹⁶ Exposures to other toxic ${\mathrm{PM}}_{2.5}$ constituents in early life have also been linked to respiratory problems,^17,18 although their effect on AR is much less clear. To date, we are only aware of the Chinese Children Health Home (CCHH) study, which has retrospectively demonstrated that exposure to multiple ${\mathrm{PM}}_{2.5}$ components during pregnancy was associated with a higher AR risk in children.⁸ However, this study might be unable to disentangle the effects of different components due to their complex synergy and high collinearity.^19,20 Additional studies employing advanced analytical methods are needed to address the effects of mixture exposures and identify the critical components responsible for AR risk in early life.

Dietary nutrients, particularly polyunsaturated fatty acids (PUFAs), have been shown by research to alleviate the inflammatory response and oxidative stress caused by ${\mathrm{PM}}_{2.5}$.^21–23 During pregnancy, PUFAs are important inflammation and immune modulators,²⁴ and we have shown beneficial associations of maternal PUFAs with allergic diseases in toddlers.²⁵ An animal study also suggested that n-3 PUFAs could combat ${\mathrm{PM}}_{2.5}$ toxicity by suppressing IgE production and improving respiratory outcomes,²⁶ but the results have not yet been verified in pregnant women. Further exploration of the moderating effect of PUFAs may provide insights into preventive measures against the adverse effects of air pollution for vulnerable populations.

This birth cohort study investigated the association between prenatal exposure to ${\mathrm{PM}}_{2.5}$ chemical components and the onset of AR and AR symptoms within 2 years of age. We used a novel approach—quantile-based g-computation—to bridge the research gap in making valid inferences on the individual and joint effects of ${\mathrm{PM}}_{2.5}$ components.²⁷ We further analyzed the modification effects of maternal PUFAs on the associations between ${\mathrm{PM}}_{2.5}$ and AR outcomes using the erythrocyte biomarkers, which reflect dietary PUFA intakes in the past few months.

Methods

Study Population

The study involved mother–child pairs enrolled in a prospective birth cohort from Guangzhou, China (registration number NCT03023293). Between 2017 and 2018, 1,035 pregnant women were recruited at 20–28 wk of gestation from Yuexiu District Maternal and Child Health Hospital. Eligible women were a) 20–45 years of age; b) without pregestational diabetes mellitus, cardiovascular disease, thyroid disease, hematopathy, polycystic ovary syndrome, pregnancy infection, or mental disorder; and c) singleton pregnancy. A total of 950 women provided blood samples during 20–28 wk of gestation, and 866 delivered babies at the study hospital. This study included 705 children whose parents completed the follow-up visit for allergic outcomes at 2 years of age. We further excluded those who changed residence at or after birth ($n=37$) and those with missing information on exposure assessment ($n=6$) and AR outcomes ($n=5$), resulting in a final analysis of 657 mother–child pairs (Figure S1). Compared with all women enrolled in this cohort ($n=\mathrm{1,035}$), those included in the final analysis tended to have a higher level of education (Table S1). The study protocol was approved by the ethics committee of the School of Public Health at Sun Yat-sen University. All participants signed written informed consent at the study enrollment.

Environmental Exposure Assessment

We assigned gridded ${\mathrm{PM}}_{2.5}$ and its chemical composition data to the geocoded home addresses for each participant. The daily concentrations were aggregated into average exposure in pregnancy, which spanned from the last menstrual period to the delivery date. Average exposure was also estimated for the first and second years of life (i.e., during 1 and 2 years of age) based on the birth date. The pollution data were derived from the established Tracking Air Pollution in China (TAP) framework.^28–30 A two-stage machine learning algorithm was used to estimate ${\mathrm{PM}}_{2.5}$ at a $10\text{-km}$ spatial resolution by integrating ground observations, satellite aerosol optical depth (AOD), operational chemical transport model simulations, and other ancillary data. We primarily utilized a high-resolution ($1\mathrm{km}\times 1\mathrm{km}$) ${\mathrm{PM}}_{2.5}$ product, further developed by fusing $10\text{-km}$ TAP ${\mathrm{PM}}_{2.5}$ predictions, ground observations, $1\text{-km}$ satellite-retrieved AOD, and improved land use parameters. Temporally continuous land use predictors were constructed through statistical and spatial modeling techniques, which incorporated road maps, population distribution, artificial impervious area, and vegetation index.

In addition, TAP estimated five major chemical components of ${\mathrm{PM}}_{2.5}$ at a $10\text{-km}$ spatial resolution in China, including sulfate ($\mathrm{SO}_4{}^{2-}$), nitrate ($\mathrm{NO}_3{}^{-}$), ammonium ($\mathrm{NH}_4{}^{+}$), organic matter (OM), and BC.²⁹ The Weather Research and Forecasting-Community Multiscale Air Quality (WRF-CMAQ) modeling system was conducted to simulate component-specific conversion factors. To further enhance accuracy and reduce biases, these conversion factors underwent training and refinement using ground observations. The daily ${\mathrm{PM}}_{2.5}$ component concentrations were derived by applying the revised conversion factors to partition the total mass concentrations of $10\text{-km}$ ${\mathrm{PM}}_{2.5}$. The estimated concentrations for all species are in good agreement with the available ground observations, with correlation coefficients ranging from 0.67 to 0.80 from 2013 to 2020.²⁹

Outcomes Measurements

At 2 years of age, we identified AR diagnoses and the onset age of children by reviewing the hard copies of the outpatient medical records provided by the parent. To gather further information on AR manifestations, we asked the parent the standardized questions adapted from the International Studies on Asthma and Allergies in Childhood (ISAAC) questionnaire during the interview.³¹ Specifically, parent-reported AR symptoms were assessed using the question: “Has your child ever experienced repeated sneezing, a runny or blocked nose when not suffering from a cold or flu, persisting for at least 1 hour per day and 2 days per week?” If the answer was affirmative, we then inquired about the time when these symptoms first occurred.

Lab Analysis of PUFA Biomarkers

During 20–28 wk of gestation, maternal blood samples were collected by professional nurses after overnight fasting. We measured erythrocyte PUFAs using gas chromatography, following the methods previously described.²⁵ In short, we thawed the red cell samples and added Tris-HCl buffer, which hemolyzed the cells. The resulting red blood cells were then centrifuged at $\mathrm{3,000}\text{\hspace{0.17em}rpm}$ for 10 min to obtain the bottom layer of milky red blood cell fragments. Next, we extracted the lipid component of erythrocyte fragments, where fatty acid methyl esters were derived through transesterification with ${\mathrm{H}}_2{\mathrm{SO}}_4$ in methanol (5%, vol/vol), together with toluene, at 70°C for 2 h in sealed tubes. After that, these fatty acid methyl esters were analyzed with an Agilent 7820 gas chromatograph (Agilent Corporation), equipped with a flame ionization detector and a $60\text{-m}$ by $0.25\text{-mm}$ [inside diameter (i.d.)] by $0.25\text{-}\mathrm{\mu }\mathrm{m}$ (film thickness) fused silica bonded phase column (DB-23; Agilent Corporation). The carrier gas was nitrogen, maintained at a pressure of 300 kPa. The injector and the detector temperatures were both set at 270°C. The initial temperature of the column was programmed to increase from 150 to 180°C at a rate of 10°C/min, holding for 2 min. The temperature was then raised to 215°C at a rate of 2.5°C/min and maintained for 6 min. Finally, it was increased to 230°C at a rate of 10°C/min and held for 5 min. PUFAs were identified by comparing retention time with standard mixtures of fatty acid methyl ester (Nu-Chek Prep, Inc.). To quantify the PUFA compositions, the peak areas were compared with the internal standard, which was added to the samples ($1\mathrm{mg}$ of internal standard in $500\mathrm{mg}$ sample) prior to extraction. PUFAs were expressed as fractions (%) of their peak area relative to the total peak area of all fatty acids.

This study examined important n-3 and n-6 PUFAs, which included a-linolenic acid (ALA) (C18:3 n-3), eicosapentaenoic acid (EPA) (C20:5 n-3), docosapentaenoic acid (DPA) (C22:5 n-3), docosahexaenoic acid (DHA) (C22:6 n-3), linoleic acid (LA) (C18:2 n-6), and arachidonic acid (AA) (C20:4 n-6).

Covariates

Maternal sociodemographic characteristics and lifestyle factors were collected at baseline interviews, including maternal age ($\ge 35$ years), the highest education level (high school and below, or college and above), employment (yes or no), monthly household income per capita [$<\mathrm{4,000}$, $\mathrm{4,000}\text{\hspace{0.17em}to}<\mathrm{8,000}$, or $\ge \mathrm{8,000}$ Chinese Renminbi (RMB)], and maternal smoking and/or passive smoking (yes or no). Birth information was obtained from the hospital birth registry system, including the birth season (winter–spring or summer–autumn), child’s sex, parity, low birth weight ($<\mathrm{2,500}\mathrm{g}$), and preterm birth (gestational age $<37$ weeks). We also collected children’s postnatal information via parent-reported questionnaires, including exclusive breastfeeding duration ($\ge 6$ months, yes or no) at age 6 mo and family history of allergies (yes or no), household mold and dampness (yes or no), and secondhand smoke exposure (yes or no) at age 2 years.

As ${\mathrm{NO}}_2$ contributes to secondary particulate formation³² and airway inflammation,³³ we estimated ${\mathrm{NO}}_2$ with daily concentration at a $10\text{-km}$ spatial resolution based on the China High Air Pollutants dataset.³⁴ The daily temperature was obtained from the National Oceanic and Atmospheric Administration (https://www.ncdc.noaa.gov/cdo-web/). Residential greenness was estimated using the normalized difference vegetation index (NDVI) derived from the MODIS satellite (https://modis.ornl.gov/documentation.html). We calculated mean levels of ${\mathrm{NO}}_2$, temperature, and NDVI over the time interval of interest (i.e., pregnancy and the first and second postnatal years) for each participant.

Statistical Analysis

Characteristics of participants were summarized as means and standard deviations (SDs) or frequencies and percentages. Differences between AR and nonAR groups were tested using tests of variance or ${\mathrm{\chi }}^2$ tests.

We employed the Cox proportional hazards model to fit the association between exposure to ${\mathrm{PM}}_{2.5}$ and its components and AR or AR symptoms. The follow-up duration was determined as the time elapsed from birth until the occurrence of the outcome or the end of follow-up (2 years of age). The proportional hazards assumption was confirmed by the Schoenfeld residuals test. We estimated hazard ratios (HRs) with 95% confidence intervals (CIs) per $5{\mathrm{\mu }\mathrm{g}/\mathrm{m}}^3$ change in ${\mathrm{PM}}_{2.5}$ and $1{\mathrm{\mu }\mathrm{g}/\mathrm{m}}^3$ increase in its components, except for BC [per interquartile range (IQR) increase in BC due to its narrow distribution]. The joint effect of five ${\mathrm{PM}}_{2.5}$ components was evaluated by a quantile-based g-computation (qg-computation) approach. The qg-computation method is a generalization and extension of weighted quantile sum (WQS) regression, with the strengths of being simple to implement and computationally tractable.²⁷ It can avoid the effects of potential collinearity among exposures and allow the directional homogeneity assumption. The weight index of each component with direction (positive or negative) can indicate their relative contributions. We used quintiles for constructing the quantile indicator variables representing the exposures. Estimated HRs were associated with one quintile increase in exposure to a mixture of five ${\mathrm{PM}}_{2.5}$ components. All main analytic models were adjusted for potential confounding factors according to the directed acyclic graphic (DAG) (Figure S2). These factors included maternal age, maternal employment, maternal education level, monthly household income, parity, average temperature, residential NDVI, ${\mathrm{NO}}_2$ during pregnancy, as well as child’s sex. Other important risk factors for AR were additionally considered in the sensitivity analysis. The nonlinear association between ${\mathrm{PM}}_{2.5}$ and AR outcomes was tested by the restricted cubic splines ($\text{knot}=3$).

We conducted stratified analyses to examine the modification effects of erythrocyte PUFAs on the associations between exposure and AR outcomes. For each ${\mathrm{PM}}_{2.5}$ component and their mixtures, two-sample z-tests³⁵ were conducted using the following formula (Equations 1). It compared the difference between the stratum-specific point estimates ($\mathrm{\beta }=\text{ln HR}$) and their standard errors (SEs) of the exposure in the lower and higher PUFA groups, stratified by median levels of the study population (i.e., below or above median).

$$z=\frac{\beta {}_{\text{lower\hspace{0.17em}PUFA}}-\beta {}_{\text{higher\hspace{0.17em}PUFA}}}{\sqrt{\mathrm{SE}_{\text{lower\hspace{0.17em}PUFA}}{}^2+\mathrm{SE}_{\text{higher\hspace{0.17em}PUFA}}{}^2}}$$ (1)

We performed several sensitivity analyses to verify the robustness of our results: a) additional adjustment for other risk factors of AR, including family history of allergies, exclusive breastfeeding, birth season, maternal smoking and/or passive smoking, childhood secondhand smoke, and household mold and dampness; b) adjustment for ${\mathrm{PM}}_{2.5}$ or its component exposures after birth (0–2 years) to control the potential effect of postnatal exposure; c) subsample analysis excluding children with preterm birth (gestational age $<37$ weeks) and low birth weight ($<\mathrm{2,500}\mathrm{g}$) ($n=36$), as they might act as potential mediators which should not be included in the main model; d) reanalysis of the association using covariates data imputed by the chained equations with the “mice” package (missing rate 1.7% to 5.6%, imputation $\text{number}=5$), enabling comparisons with the results of complete cases analysis in the main regression model; e) tests for stability of association between $10\mathrm{km}$ ${\mathrm{PM}}_{2.5}$ and the outcomes; and f) exploration for the association of postnatal ${\mathrm{PM}}_{2.5}$ components with AR and AR symptoms while examining the potential moderating effect of prenatal PUFAs.

All analyses were performed in the statistical software R 4.1.1 (R Core Team, 2021). A $p$-value $<0.05$ for a two-sided test was considered statistically significant.

Results

Characteristics of the Study Population

Table 1 summarizes the general characteristics of 657 mother–child pairs. Of all children, 49.47% were female and 34.09% had a family history of allergies. Up to 2 years of age, 5.33% (35/657) of children had been diagnosed with AR and 8.07% (53/657) had related symptoms, with a mean onset age of $14.55\pm 6.21$ and $15.54\pm 5.47$ months, respectively. AR symptoms were more common in males than in females, and children who developed AR and its symptoms were more likely to have mothers with higher education and who smoked during pregnancy compared with those without (Table 1; Table S2). No significant differences were observed in other characteristics between AR (or AR symptoms) and non-AR (or AR symptoms) groups.

Table 1.

General characteristics of study population recruited in 2017 and 2018 ($n=657$ Chinese children with and without diagnosed allergic rhinitis by age two).

Characteristic	Total	Allergic rhinitis
		No ($n=622$)	Yes ($n=35$)	$p$
Maternal characteristic at pregnancy
Maternal age [$n$ (%)]	—	—	—	0.878
$<35$ years	538 (81.89)	509 (81.83)	29 (82.86)	—
$\ge 35$ years	119 (18.11)	113 (18.17)	6 (17.14)	—
Maternal employmenta [$n$ (%)]	—	—	—	0.755
No	168 (27.10)	158 (26.96)	10 (29.41)	—
Yes	452 (72.90)	428 (73.04)	24 (70.59)	—
Education level [$n$ (%)]	—	—	—	0.010
High school and below	227 (34.55)	222 (35.69)	5 (14.29)	—
College and above	430 (65.45)	400 (64.31)	30 (85.71)	—
Monthly household income per capita [$n$ (%)]	—	—	—	0.734
$<\mathrm{4,000}$ RMB	121 (18.42)	113 (18.17)	8 (22.86)	—
4,000–8,000 RMB	259 (39.42)	245 (39.39)	14 (40.00)	—
$\ge \mathrm{8,000}$ RMB	277 (42.16)	264 (42.44)	13 (37.14)	—
Maternal smoking and/or passive smokinga [$n$ (%)]	—	—	—	0.041
No	444 (69.70)	425 (70.60)	19 (54.29)	—
Yes	193 (30.30)	177 (29.40)	16 (45.71)	—
Parity [$n$ (%)]	—	—	—	0.846
1	365 (55.56)	345 (55.47)	20 (57.14)	—
$\ge 2$	292 (44.44)	277 (44.53)	15 (42.86)	—
Child’s characteristic by 2 years of age
Birth season [$n$ (%)]	—	—	—	0.576
Summer–Autumn	540 (82.19)	510 (81.99)	30 (85.71)	—
Winter–Spring	117 (17.81)	112 (18.01)	5 (14.29)	—
Child’s sex [$n$ (%)]	—	—	—	0.250
Male	332 (50.53)	311 (50.00)	21 (60.00)	—
Female	325 (49.47)	311 (50.00)	14 (40.00)	—
Preterm birth [$n$ (%)]	—	—	—	0.701
No	630 (95.89)	596 (95.82)	34 (97.14)	—
Yes	27 (4.11)	26 (4.18)	1 (2.86)	—
Low birth weight [$n$ (%)]	—	—	—	0.918
No	640 (97.41)	606 (97.43)	34 (97.14)	—
Yes	17 (2.59)	16 (2.57)	1 (2.86)	—
Family history of allergies [$n$ (%)]	—	—	—	0.261
No	433 (65.91)	413 (66.40)	20 (57.14)	—
Yes	224 (34.09)	209 (33.60)	15 (42.86)	—
Secondhand smoke at 2 years of age [$n$ (%)]	—	—	—	0.889
No	577 (87.82)	546 (87.78)	31 (88.57)	—
Yes	80 (12.18)	76 (12.22)	4 (11.43)	—
Exclusive breastfeedinga [$n$ (%)]	—	—	—	0.159
$<6$ months	369 (57.12)	345 (56.46)	24 (68.57)	—
$\ge 6$ months	277 (42.88)	266 (43.54)	11 (31.43)	—
Household mold and dampness at age 2 years [$n$ (%)]	—	—	—	0.233
No	471 (71.69)	449 (72.19)	22 (62.86)	—
Yes	186 (28.31)	173 (27.81)	13 (37.14)	—
Environmental exposure during pregnancy
T [°C ($\text{mean}\pm \text{SD}$)]	$23.82\pm 1.60$	$23.83\pm 1.59$	$23.62\pm 1.70$	0.340
NDVI ($\text{mean}\pm \text{SD}$)	$0.33\pm 0.10$	$0.33\pm 0.10$	$0.33\pm 0.10$	0.761
${\mathrm{NO}}_2$ [${\mathrm{\mu }\mathrm{g}/\mathrm{m}}^3$ ($\text{mean}\pm \text{SD}$)]	$57.54\pm 6.50$	$57.57\pm 6.50$	$57.03\pm 6.61$	0.574

Note: Data were presented as $\text{mean}\pm \text{SD}$ or $n$ (%). $p$-Value from chi-square test or analysis of variance. —, no data; NDVI, normalized difference vegetation index; ${\mathrm{NO}}_2$, nitrogen dioxide; RMB, Renminbi; SD, standardized deviation; T, temperature.

^a Maternal employment had missing data for 5.6% ($n=37$), maternal smoking and/or passive smoking for 3.0% ($n=20$), and exclusive breastfeeding for 1.7% ($n=11$).

Distribution of $P{M}_{2.5}$ and Its Chemical Components

As shown in Table 2, the average mass concentration of ${\mathrm{PM}}_{2.5}$ during pregnancy was $35.50\pm 5.31{\mathrm{\mu }\mathrm{g}/\mathrm{m}}^3$_. Among five ${\mathrm{PM}}_{2.5}$ components, OM had the highest $\text{mean}\pm \text{SD}$ concentration of $8.86\pm 1.09{\mathrm{\mu }\mathrm{g}/\mathrm{m}}^3$, followed by $\mathrm{SO}_4{}^{2-}$ ($6.31\pm 0.90{\mathrm{\mu }\mathrm{g}/\mathrm{m}}^3$), $\mathrm{NO}_3{}^{-}$ ($4.26\pm 0.99{\mathrm{\mu }\mathrm{g}/\mathrm{m}}^3$), $\mathrm{NH}_4{}^{+}$ ($3.10\pm 0.64{\mathrm{\mu }\mathrm{g}/\mathrm{m}}^3$), and BC ($1.88\pm 0.25{\mathrm{\mu }\mathrm{g}/\mathrm{m}}^3$). ${\mathrm{PM}}_{2.5}$ chemical components were highly positively correlated with each other and negatively correlated with temperature as presented in Table 2. The distribution of ${\mathrm{PM}}_{2.5}$ composition in the first 2 years of life was summarized in Table S3.

Table 2.

Summary distribution of maternal exposure to ${\mathrm{PM}}_{2.5}$ and its chemical components and their correlations during pregnancy among 657 Chinese women in 2017 and 2018.

Prenatal exposure (${\mathrm{\mu }\mathrm{g}/\mathrm{m}}^3$)	$\text{Mean}\pm \text{SD}$	Median (IQR)	Spearman correlation coefficient
			${\mathrm{PM}}_{2.5}$	$\mathrm{SO}_4{}^{2-}$	$\mathrm{NO}_3{}^{-}$	$\mathrm{NH}_4{}^{+}$	OM	BC	${\mathrm{NO}}_2$	T
${\mathrm{PM}}_{2.5}$	$35.50\pm 5.31$	35.35 (7.71)	1	0.86	0.84	0.86	0.89	0.88	0.23	$-0.75$
$\mathrm{SO}_4{}^{2-}$	$6.31\pm 0.90$	6.09 (1.27)	—	1	0.88	0.92	0.95	0.99	0.14	$-0.71$
$\mathrm{NO}_3{}^{-}$	$4.26\pm 0.99$	4.24 (1.67)	—	—	1	0.99	0.95	0.91	0.33	$-0.93$
$\mathrm{NH}_4{}^{+}$	$3.10\pm 0.64$	3.08 (1.02)	—	—	—	1	0.96	0.94	0.28	$-0.88$
OM	$8.86\pm 1.09$	8.73 (1.69)	—	—	—	—	1	0.98	0.26	$-0.85$
BC	$1.88\pm 0.25$	1.83 (0.31)	—	—	—	—	—	1	0.21	$-0.77$

Note: —, no data; BC, black carbon; IQR, interquartile range; $\mathrm{NH}_4{}^{+}$, ammonium; ${\mathrm{NO}}_2$, nitrogen dioxide; $\mathrm{NO}_3{}^{-}$, nitrate; OM, organic matter; ${\mathrm{PM}}_{2.5}$, particulate matter with an aerodynamic diameter $\le 2.5\mathrm{\mu }\mathrm{m}$; SD, standardized deviation; $\mathrm{SO}_4{}^{2-}$, sulfate; T, temperature.

Association of $P{M}_{2.5}$ and Its Components with AR or AR Symptoms

In the adjusted model, prenatal ${\mathrm{PM}}_{2.5}$ and its chemical components showed a positive association with AR and AR symptoms in early childhood (Table 3). A $5{\mathrm{\mu }\mathrm{g}/\mathrm{m}}^3$ increase in ${\mathrm{PM}}_{2.5}$ was associated with an HR of 1.85 (95% CI: 1.16, 2.96) and 1.79 (95% CI: 1.22, 2.62) for developing AR and AR symptoms, respectively. Exposure to $1{\mathrm{\mu }\mathrm{g}/\mathrm{m}}^3$ in $\mathrm{SO}_4{}^{2-}$, $\mathrm{NO}_3{}^{-}$, $\mathrm{NH}_4{}^{+}$, OM, and BC was linked to an HR of 2.07 (95% CI: 1.20, 3.55), 2.22 (95% CI: 1.05, 4.68), 2.67 (95% CI: 1.03, 6.90), 1.91 (95% CI: 1.10, 3.27), and 2.23 (95% CI: 1.20, 4.18) in AR, respectively. Similar associations were observed between ${\mathrm{PM}}_{2.5}$ components and AR symptoms. The prediction data for the concentration–response relationship between ${\mathrm{PM}}_{2.5}$ and its components and AR or AR symptoms are presented in Table S4. We did not observe nonlinear associations between the exposure and AR outcomes (Figure S3 and S4).

Table 3.

Association of prenatal exposure to ${\mathrm{PM}}_{2.5}$ and its chemical components in 2017 and 2018 with allergic rhinitis (AR) or AR symptoms by age two in 657 Chinese children.

		AR ($n=35$)		AR symptomsa ($n=53$)
	Exposure increment (${\mathrm{\mu }\mathrm{g}/\mathrm{m}}^3$)	HR (95% CI)	$p$	HR (95% CI)	$p$
Crude model
${\mathrm{PM}}_{2.5}$	5.00	1.40 (1.03, 1.90)	0.032	1.32 (1.03, 1.69)	0.028
$\mathrm{SO}_4{}^{2-}$	1.00	1.52 (1.10, 2.10)	0.011	1.34 (1.02, 1.75)	0.037
$\mathrm{NO}_3{}^{-}$	1.00	1.26 (0.90, 1.77)	0.170	1.16 (0.88, 1.52)	0.294
$\mathrm{NH}_4{}^{+}$	1.00	1.49 (0.90, 2.45)	0.117	1.29 (0.86, 1.95)	0.218
OM	1.00	1.33 (1.00, 1.78)	0.054	1.22 (0.96, 1.55)	0.103
BCb	0.30	1.55 (1.07, 2.23)	0.020	1.37 (1.01, 1.86)	0.042
Main model
${\mathrm{PM}}_{2.5}$	5.00	1.85 (1.16, 2.96)	0.010	1.79 (1.22, 2.62)	0.003
$\mathrm{SO}_4{}^{2-}$	1.00	2.07 (1.20, 3.55)	0.009	1.74 (1.12, 2.71)	0.014
$\mathrm{NO}_3{}^{-}$	1.00	2.22 (1.05, 4.68)	0.037	2.16 (1.19, 3.94)	0.011
$\mathrm{NH}_4{}^{+}$	1.00	2.67 (1.03, 6.90)	0.043	2.47 (1.15, 5.31)	0.020
OM	1.00	1.91 (1.12, 3.27)	0.018	1.81 (1.16, 2.80)	0.008
BCb	0.30	2.23 (1.20, 4.18)	0.012	1.99 (1.20, 3.32)	0.008

Note: The crude Cox regression model was unadjusted. The main Cox regression model was adjusted for maternal age, maternal employment, maternal education, monthly household income, parity, child’s sex, temperature, residential greenness, and ${\mathrm{NO}}_2$. Maternal employment had a missing rate of 5.6% ($n=37$). —, no data; AR, allergic rhinitis; BC, black carbon; CI, confidential interval; HR, hazard ratio; $\mathrm{NH}_4{}^{+}$, ammonium; $\mathrm{NO}_3{}^{-}$, nitrate; OM, organic matter; ${\mathrm{PM}}_{2.5}$, particulate matter with an aerodynamic diameter $\le 2.5\mathrm{\mu }\mathrm{m}$; $\mathrm{SO}_4{}^{2-}$, sulfate.

^a AR symptoms were defined as having experienced repeated sneezing and a runny or blocked nose when not suffering from a cold or flu, persisting for at least 1 hour per day and 2 days per week within 2 years of age.

^b Estimated effects associated with per interquartile range increase ($0.30{\mathrm{\mu }\mathrm{g}/\mathrm{m}}^3$) in BC.

Table 4 illustrates the joint effect of five ${\mathrm{PM}}_{2.5}$ components on AR or AR symptoms from the qg-computation model. Each quintile increase in the prenatal mixture exposure was associated with an increased HR of 3.73 (95% CI: 1.80, 7.73) and 2.69 (95% CI: 1.55, 4.67) in AR and AR symptoms in the main model, respectively. The weight index indicated that BC positively contributed to the adverse effect of mixture exposure and had the largest weight of 0.436 and 0.454; whereas the negative index of $\mathrm{SO}_4{}^{2-}$ might suggest that it influenced the overall association by diminishing the contribution of other toxic components.

Table 4.

Quantile based g-computation models for the association of prenatal exposure to a mixture of five ${\mathrm{PM}}_{2.5}$ components in 2017 and 2018 with allergic rhinitis (AR) or AR symptoms by age two in 657 Chinese children.

	Joint effecta		Weights for mixture components
	HR (95% CI)	$p$	$\mathrm{SO}_4{}^{2-}$	$\mathrm{NO}_3{}^{-}$	$\mathrm{NH}_4{}^{+}$	OM	BC
Crude model
AR ($n=35$)	1.55 (1.11, 2.16)	0.010	0.394b	0.371b	$-1.000$	0.059b	0.176b
AR symptoms ($n=53$)c	1.37 (1.05, 1.78)	0.019	0.645b	0.309b	$-0.798$	0.046b	$-0.202$
Main model
AR ($n=35$)	3.73 (1.80, 7.73)	$<0.001$	$-0.614$	0.356b	$-0.386$	0.208b	0.436b
AR symptoms ($n=53$)c	2.69 (1.55, 4.67)	$<0.001$	$-1.000$	0.418b	0.002b	0.126b	0.454b

Note: The Cox regression model was adjusted for maternal age, maternal employment, maternal education, monthly household income, parity, child’s sex, temperature, residential greenness, and ${\mathrm{NO}}_2$. Maternal employment had a missing rate of 5.6% ($n=37$). —, no data; AR, allergic rhinitis; BC, black carbon; CI, confidential interval; HR, hazard ratio; $\mathrm{NH}_4{}^{+}$, ammonium; $\mathrm{NO}_3{}^{-}$, nitrate; OM, organic matter; ${\mathrm{PM}}_{2.5}$, particulate matter with an aerodynamic diameter $\le 2.5\mathrm{\mu }\mathrm{m}$; $\mathrm{SO}_4{}^{2-}$, sulfate.

^a Estimated effects associated with one quintile increase in exposures to five ${\mathrm{PM}}_{2.5}$ components.

^b Positive index weight.

^c AR symptoms were defined as having experienced repeated sneezing and a runny or blocked nose when not suffering from a cold or flu, persisting for at least 1 hour per day and 2 days per week within 2 years of age.

Modification Effects of PUFAs

Table S5 displays the average percentage (%) of maternal erythrocyte n-3 and n-6 PUFAs of total fatty acids. The median (IQR) levels of ALA, EPA, DPA, and DHA were 0.24% (0.16), 0.88% (0.73), 1.29% (0.55), and 7.20% (3.65), respectively; while LA and AA had median (IQR) levels of 15.28% (3.12) and 17.34% (4.35), respectively. Their modification effects are presented in Table 5. ${\mathrm{PM}}_{2.5}$ showed stronger adverse associations with AR outcomes in the group with DPA levels lower than 1.29% (AR, ${\text{HR}}_{\text{lower}\text{\hspace{0.17em}DPA}}=3.02$, 95% CI: 1.41, 6.47; AR symptoms, ${\text{HR}}_{\text{lower}\text{\hspace{0.17em}DPA}}=3.42$, 95% CI: 1.73, 6.78). We also observed that higher LA levels (above 15.28%) aggravated PM-associated risks of AR outcomes (AR, ${\text{HR}}_{\text{higher}\text{\hspace{0.17em}LA}}=6.48$, 95% CI: 2.27, 18.49; AR symptoms, ${\text{HR}}_{\text{higher}\text{\hspace{0.17em}LA}}=3.68$, 95% CI: 1.78, 7.59). when examining ${\mathrm{PM}}_{2.5}$ components, similar modification effects of DPA and LA were observed in single-component and mixture exposure models (Table 6). In the higher LA group, for instance, each quintile increase in prenatal mixture exposure was associated with an HR of 5.70 (95% CI: 2.01, 16.61) for AR symptoms.

Table 5.

The modification of maternal erythrocyte PUFAs on the association between prenatal exposure to ${\mathrm{PM}}_{2.5}$ (per $5{\mathrm{\mu }\mathrm{g}/\mathrm{m}}^3$) in 2017 and 2018 and AR or AR symptoms by age two in 657 Chinese children.

	AR ($n=35$) [HR (95% CI)]	AR symptomsa ($n=53$) [HR (95% CI)]
n-3 PUFAsb
ALA
Lower level ($\le 0.24$, $n=333$)	1.27 (0.50, 3.21)	1.37 (0.66, 2.83)
Higher level ($>0.24$, $n=324$)	2.30 (1.10, 4.78)	2.38 (1.29, 4.37)
$p_{\text{for}\text{\hspace{0.17em}difference}}$	0.324	0.244
EPA
Lower level ($\le 0.88$, $n=329$)	2.11 (0.94, 4.76)	2.26 (1.13, 4.53)
Higher level ($>0.88$, $n=328$)	2.15 (0.91, 5.06)	1.90 (1.02, 3.56)
$p_{\text{for}\text{\hspace{0.17em}difference}}$	0.977	0.720
DPA
Lower level ($\le 1.29$, $n=332$)	3.02 (1.41, 6.47)	3.42 (1.73, 6.78)
Higher level ($>1.29$, $n=325$)	1.05 (0.42, 2.60)	1.05 (0.56, 1.99)
$p_{\text{for}\text{\hspace{0.17em}difference}}$	0.081	0.033
DHA
Lower level ($\le 7.20$, $n=329$)	1.69 (0.74, 3.85)	2.10 (1.02, 4.33)
Higher level ($>7.20$, $n=328$)	2.40 (0.98, 5.87)	1.71 (0.93, 3.17)
$p_{\text{for}\text{\hspace{0.17em}difference}}$	0.576	0.671
n-6 PUFAsb
LA
Lower level ($\le 15.28$, $n=329$)	0.74 (0.32, 1.73)	0.96 (0.50, 1.87)
Higher level ($>15.28$, $n=328$)	6.48 (2.27, 18.49)	3.68 (1.78, 7.59)
$p_{\text{for}\text{\hspace{0.17em}difference}}$	0.002	0.007
AA
Lower level ($\le 17.34$, $n=329$)	1.61 (0.70, 3.68)	1.95 (0.94, 4.04)
Higher level ($>17.34$, $n=328$)	2.23 (1.01, 4.91)	1.85 (1.03, 3.32)
$p_{\text{for}\text{\hspace{0.17em}difference}}$	0.578	0.907

Note: The Cox regression model was adjusted for maternal age, maternal employment, maternal education, monthly household income, parity, child’s sex, temperature, residential greenness, and ${\mathrm{NO}}_2$. Maternal employment had a missing rate of 5.6% ($n=37$). —, no data; AA, arachidonic acid; ALA, a-linolenic acid; AR, allergic rhinitis; CI, confidential interval; DHA, docosahexaenoic acid; DPA, docosapentaenoic acid; EPA, eicosapentaenoic acid; HR, hazard ratio; LA, linoleic acid; ${\mathrm{PM}}_{2.5}$, particulate matter with an aerodynamic diameter $\le 2.5\mathrm{\mu }\mathrm{m}$; PUFAs, polyunsaturated fatty acids.

^a AR symptoms were defined as having experienced repeated sneezing and a runny or blocked nose when not suffering from a cold or flu, persisting for at least 1 hour per day and 2 days per week within 2 years of age.

^b Each of the n-3 and n-6 PUFAs was stratified into higher and lower levels based on the median levels (%), respectively.

Table 6.

The modification of maternal erythrocyte DPA and LA on the association between prenatal exposure to ${\mathrm{PM}}_{2.5}$ chemical components in 2017 and 2018 and allergic rhinitis (AR) or AR symptoms by age two in 657 Chinese children.

	Stratified by DPAa [HR (95% CI)]			Stratified by LAa [HR (95% CI)]
	Lower level ($\le 1.29$, $n=332$)	Higher level ($>1.29$, $n=325$)	$p_{\text{for}\text{\hspace{0.17em}difference}}$	Lower level ($\le 15.28$, $n=329$)	Higher level ($>15.28$, $n=328$)	$p_{\text{for}\text{\hspace{0.17em}difference}}$
Single componentb
$\mathrm{SO}_4{}^{2+}$
AR ($n=35$)	3.10 (1.34, 8.01)	1.66 (0.59, 4.51)	0.357	1.20 (0.49, 2.97)	4.88 (1.66, 14.38)	0.051
AR symptoms ($n=53$)c	4.19 (1.93, 9.89)	0.94 (0.46, 1.89)	0.007	0.87 (0.39, 1.91)	3.94 (1.72, 9.00)	0.009
$\mathrm{NO}_3{}^{-}$
AR ($n=35$)	4.03 (1.27, 12.73)	2.05 (0.42, 9.56)	0.497	1.32 (0.42, 4.14)	3.49 (1.06, 11.43)	0.248
AR symptoms ($n=53$)c	4.98 (1.84, 13.48)	1.25 (0.48, 3.25)	0.049	1.19 (0.44, 3.22)	4.12 (1.52, 11.13)	0.084
$\mathrm{NH}_4{}^{+}$
AR ($n=35$)	4.24 (1.53, 11.76)	0.84 (0.31, 2.26)	0.026	1.29 (0.54, 3.12)	1.97 (0.87, 4.47)	0.491
AR symptoms ($n=53$)c	2.28 (1.02, 5.09)	1.00 (0.50, 1.98)	0.126	1.06 (0.32, 3.53)	4.69 (1.28, 17.17)	0.100
OM
AR ($n=35$)	3.32 (1.32, 8.35)	1.30 (0.45, 3.70)	0.188	0.99 (0.36, 2.71)	4.01 (1.45, 11.11)	0.055
AR symptoms ($n=53$)c	4.10 (1.78, 9.45)	1.05 (0.52, 2.11)	0.014	0.89 (0.39, 2.02)	3.63 (1.64, 8.05)	0.016
BC
AR ($n=35$)	4.42 (1.46, 13.39)	1.51 (0.45, 5.04)	0.198	1.08 (0.36, 3.27)	3.79 (1.43, 10.05)	0.096
AR symptoms ($n=53$)c	5.66 (2.06, 15.53)	0.84 (0.35, 1.98)	0.005	0.84 (0.34, 2.07)	3.38 (1.51, 7.57)	0.024
Mixture exposured
AR ($n=35$)	4.56 (1.13, 11.18)	1.64 (0.49, 5.52)	0.974	1.64 (0.49, 5.52)	6.80 (1.73, 26.77)	0.128
AR symptoms ($n=53$)c	3.30 (1.21, 8.97)	1.42 (0.52, 3.91)	0.356	1.42 (0.52, 3.91)	5.70 (2.01, 16.61)	0.060

Note: The Cox regression model was adjusted for maternal age, maternal employment, maternal education, monthly household income, parity, child’s sex, temperature, residential greenness, and ${\mathrm{NO}}_2$; Maternal employment had a missing rate of 5.6% ($n=37$). —, no data; AR, allergic rhinitis; BC, black carbon; CI, confidential interval; DPA, docosapentaenoic acid (n-3); HR, hazard ratio; LA, linoleic acid; $\mathrm{NO}_3{}^{-}$, nitrate; OM, organic matter; ${\mathrm{PM}}_{2.5}$, particulate matter with an aerodynamic diameter $\le 2.5\mathrm{\mu }\mathrm{m}$.

^a DPA and LA were stratified into higher and lower levels based on the median levels (%), respectively.

^b Single-component model: estimated effects associated with one unit (${\mathrm{\mu }\mathrm{g}/\mathrm{m}}^3$) increase in exposures to individual ${\mathrm{PM}}_{2.5}$ components (an interquartile range of $0.30{\mathrm{\mu }\mathrm{g}/\mathrm{m}}^3$ in BC).

^c AR symptoms were defined as having experienced repeated sneezing and a runny or blocked nose when not suffering from a cold or flu, persisting for at least 1 hour per day and 2 days per week within 2 years of age.

^d Quantile-based g-computation model: estimated effects associated with one quintile increase in exposures to all five ${\mathrm{PM}}_{2.5}$ components.

Sensitivity Analysis

The associations between prenatal exposure to ${\mathrm{PM}}_{2.5}$ and its components and AR or AR symptoms were similar in the models additionally adjusted for family history of allergies, exclusive breastfeeding, birth season, maternal smoking and/or passive smoking, childhood secondhand smoke, and household mold and dampness (Table S6 and S7). These associations were also consistent in the sample excluding children with preterm birth or low birth weight (Table S8). Furthermore, the association of prenatal ${\mathrm{PM}}_{2.5}$ and BC with AR outcomes remained consistently positive after adjustment of postnatal exposure (Table S9). The results were comparable using the covariates computed for missing data (Table S10) and $10\mathrm{km}$ ${\mathrm{PM}}_{2.5}$ (Table S11).

When further examining the effect of postnatal exposure (Table S12 and S13), the first-year exposure to ${\mathrm{PM}}_{2.5}$ and its components were adversely associated with AR or AR symptoms. Prenatal erythrocyte DPA had a moderating effect on the observed association (Table S14).

Discussion

This birth cohort study demonstrated a positive association between prenatal exposure to ${\mathrm{PM}}_{2.5}$ and its chemical components with AR or AR symptoms in children under age 2 years. BC contributed most to the risk of AR or AR symptoms when assessing the mixture exposure. The adverse effect of ${\mathrm{PM}}_{2.5}$ was alleviated in the higher DPA group and lower LA group. Additionally, maternal erythrocyte LA consistently modified the risk of AR symptoms in both individual and mixed exposures to ${\mathrm{PM}}_{2.5}$ composition.

${\mathrm{PM}}_{2.5}$ exposure during pregnancy has garnered great interest as a part of AR etiology in prior studies.^{7,8,10,12,36–38} Our study supported the detrimental association between prenatal exposure to ${\mathrm{PM}}_{2.5}$ and AR while controlling the potential effect of postnatal exposure. Despite some discrepant results observed in European birth cohorts (median ${\mathrm{PM}}_{2.5}$ $8–17{\mathrm{\mu }\mathrm{g}/\mathrm{m}}^3$),^11,35 this finding aligned with results reported in other highly polluted cities across mainland China and Taiwan (mean ${\mathrm{PM}}_{2.5}$ $33–72{\mathrm{\mu }\mathrm{g}/\mathrm{m}}^3$).^7,8,10,36,37 The ${\mathrm{PM}}_{2.5}\text{-associated}$ AR risk may involve chemical components reflecting the sources and toxicity.¹⁹ These components have been suggested as better indicators of particulate substances harmful to health than undifferentiated particle mass.^39,40 We also found an association between a mixture of various organic and inorganic compounds of ${\mathrm{PM}}_{2.5}$ and an elevated risk of AR. Because of the high collinearity and complex interplay,^19,20 these components exhibited different effects when measured individually and as a whole.⁴¹ Nonetheless, BC was consistently associated with AR risk and was the largest contributor to mixture exposure in our study. The results were aligned with previous studies that showed BC significantly contributed to airway inflammation in asthmatic children from California,⁴² respiratory diseases (including rhinitis) of children living in Brazilian urban areas,⁴³ and AR risk in preschool children from China.⁸ Generally, BC was emitted by incomplete combustion of fuels in vehicles and industrial activities.⁴⁴ In the urban area of Guangzhou, traffic emission was the most important source of carbonaceous aerosols,⁴⁵ indicating that promoting clean energy vehicles and relieving traffic congestion might be a key strategy for reducing BC mission and related AR risk.

Mechanistically, ambient carbonaceous particles could penetrate the placental barrier and translocate into human fetal tissues in the first and second trimesters.¹⁵ The development of nasal mucosa begins in early pregnancy and experiences progressive increases in thickness after that.⁴⁶ The damage to the nasal epithelial cells can affect their role in innate and adaptive immune responses.⁴⁷ ${\mathrm{PM}}_{2.5}$ and BC might disrupt the early life development of the nasal mucosa and immune system through several possible pathways. First, they could act synergistically with allergic pollen by increasing the allergenicity and bioavailability.^48,49 BC has been linked to an $\sim 30\%$ increased risk for allergen sensitization,¹³ potentially through inducing or enhancing nasal-specific IgE and Th2 cytokine responses to allergens.³⁹ Second, BC exhibits a higher potential to inflict oxidative stress than other ${\mathrm{PM}}_{2.5}$ compounds such as soluble salts,⁴⁰ which can amplify the inflammatory response in nasal epithelial cells.⁵⁰ In an in vivo experiment,⁴⁷ BC induced reactive oxygen species production to activate the inflammasome leading to maturation and secretion of interleukin $1\mathrm{\beta }$ $\left(\text{IL-}1\mathrm{\beta }\right)$ in pollen-sensitized human nasal epithelial cells, predisposing the development of AR. Third, exposure to sources of oxidative stress also exhibits the potential to alter epigenetic patterns with downstream consequences for gene expression related to airway inflammation and immune regulation.⁵¹ Transplacental exposure to ${\mathrm{PM}}_{2.5}$ and BC has been associated with key changes to DNA methylation patterns in fetuses and neonates.⁵² The resulting long-term consequences might be supported by the hypothesis “Developmental Origins of Health and Diseases,” where adverse in utero exposure can alter the epigenome and influence immune programming after birth.⁵³

Nutrients could serve as antioxidants to alleviate the effect of ${\mathrm{PM}}_{2.5}$,^13,22 and our findings support the benefits of n-3 PUFAs against ${\mathrm{PM}}_{2.5}$ for pregnant women and their offspring. We found that the adverse association between prenatal ${\mathrm{PM}}_{2.5}$ and AR symptoms was attenuated for children with maternal erythrocyte DPA levels above the median. DPA is an elongation metabolite of eicosapentaenoic acid (EPA) and a precursor of docosahexaenoic acid (DHA), which can function as a reservoir or buffer of n-3 long-chain (LC) PUFAs.⁵⁴ Previous studies and experiments have provided evidence on dietary n-3 LC PUFAs for their potential protective effect against the pro-inflammation and pro-sensitization effect of ${\mathrm{PM}}_{2.5}$.^21,55,56 For example, in a trial of Mexico residents, supplementation with fish oil (contained $52.4\%\text{\hspace{0.17em}DHA}+25.0\%\text{\hspace{0.17em}EPA}+5.8\%\text{\hspace{0.17em}DPA}$) was found to modulate oxidative response to ${\mathrm{PM}}_{2.5}$ exposure.²¹ We specifically identified the moderating effect of DPA by measuring PUFA levels in erythrocytes, which were accurate and stable biomarkers that objectively reflect both dietary intake and biological processes.⁵⁷ The protective levels of erythrocyte DPA in our study (above 1.29%) were comparable to those reported in previous studies (median levels of 1.13% and 1.56%).^58,59 The biological effect of DPA might be further explained by its derivatives, a group of specialized pro-resolution lipid mediators (protectins, maresins, and resolvins) that orchestrate key signaling processes in the resolution of inflammation and the regulation of the immune.⁶⁰ Mid-pregnancy is a critical window for the expansion of T cells and development of the immune system.⁶¹ We also observed that the moderating effect of maternal DPA persisted in the first year after birth, indicating DPA during mid-pregnancy might have lasting effects on the immune response.

Meanwhile, erythrocyte n-6 LA showed more robust modification effects on the ${\mathrm{PM}}_{2.5}$ components-induced risk of AR or AR symptoms in our study. High n-6 LA might increase allergic inflammation by competing for the same enzymes and thus suppressing the biosynthesis of n-3 LC-PUFAs.⁶² Moreover, LA in the human body serves as the precursor of n-6 arachidonic acid (AA), which can further activate lipid mediators that promote the production of inflammatory eicosanoids and impact immune cells.²⁴ As LA is an essential fatty acid that must be absorbed from food, we also explore the moderating effect of dietary LA intake ($p_{\text{for}\text{\hspace{0.17em}difference}}=0.018$ regarding mixed components exposure for AR), which has been suggested to modify allergic disease risk in our previous study.⁶³ There has been a noticeable shift in the fatty acid composition of the Westernized diet toward increased intake of LA in recent decades.⁶⁴ Consistent evidence from dietary LA and its blood biomarker suggested that reducing LA intake in the dietary pattern might be a potential approach to combat the effect of ${\mathrm{PM}}_{2.5}$ components and reduce the risk of AR.

To our knowledge, this is the first birth cohort study assessing the prenatal association between mixture exposure to ${\mathrm{PM}}_{2.5}$ components and AR while identifying PUFA biomarkers as potential modifiers. We adjusted for various environmental factors (e.g., residential greenness and ${\mathrm{NO}}_2$) during pregnancy to minimize the possible impacts of these factors on the associations and ensure the robustness of our results. However, several limitations are noted. First, we did not collect immunological indicators of clinical tests for allergic diseases, which may compromise causal inference on the biological mechanism linking ${\mathrm{PM}}_{2.5}$ and AR. Nevertheless, the observed association should not be biased as the AR diagnosis was confirmed by medical records and a validated questionnaire. Second, the study could not demonstrate the contributing effects of other ${\mathrm{PM}}_{2.5}$ elements (e.g., Fe, K, Si).¹⁸ However, we estimated the components that were more prevalent in China, which accounted for $\sim 70\%$ of the total ${\mathrm{PM}}_{2.5}$ concentration. Third, we used the $1\text{-km}\times 1\text{-km}$ resolution of ${\mathrm{PM}}_{2.5}$ for primary analysis, while its components were only available at a $10\text{-km}$ resolution. Nevertheless, the concentration of $10\mathrm{km}$ ${\mathrm{PM}}_{2.5}$ was similar to that measured at $1\mathrm{km}$ ($35.08\pm 4.25$ vs. $35.50\pm 5.31{\mathrm{\mu }\mathrm{g}/\mathrm{m}}^3$), and its association with AR was consistent in the sensitivity analysis. Fourth, we did not measure erythrocyte fatty acids at multiple time points during pregnancy and after birth. However, erythrocyte PUFAs at mid-pregnancy are considered stable and representative. Furthermore, the strengths of correlations between PUFAs in maternal erythrocytes and PUFAs in umbilical cord erythrocytes were almost the same throughout the second and third trimesters.⁶⁵ Given that postnatal exposure to ${\mathrm{PM}}_{2.5}$ was also associated with AR risk in children, further studies are warranted to confirm the interaction between ${\mathrm{PM}}_{2.5}$ and PUFAs across different exposure windows in early life. Fifth, we have considered the potential influence of household mold and dampness in our sensitivity analysis, but the unmeasured confounding effect of other indoor environmental factors (e.g., air pollution from renovations⁶⁶) should be better addressed in future research. Sixth, this study investigated early onset AR outcomes in children up to 2 years of age. Further follow-up visits should be conducted to explore whether prenatal exposure to ${\mathrm{PM}}_{2.5}$ continues to influence the development of AR in later childhood. Finally, our study participants were from a central district of Guangzhou, and those who remained in the follow-up visit were more likely highly educated. Although we found no effect modification by maternal education level on the observed associations ($p_{\text{for}\text{\hspace{0.17em}interaction}}>0.100$), the generalizability of our findings should be approached with caution.

Conclusion

Prenatal exposure to ${\mathrm{PM}}_{2.5}$ and chemical components, particularly BC, was associated with developing AR in early childhood. The detrimental associations between ${\mathrm{PM}}_{2.5}$ and AR symptoms were attenuated with the higher level of DPA and lower level of LA in maternal erythrocytes. These findings have implications for air quality control programs to prioritize regulating BC emission sources and provide clues for nutrient intervention for pregnant women to mitigate the adverse effect of ${\mathrm{PM}}_{2.5}$ by optimizing PUFA intakes.

Conclusions and opinions are those of the individual authors and do not necessarily reflect the policies or views of EHP Publishing or the National Institute of Environmental Health Sciences.