Clinical and inflammatory features of traffic-related diesel exposure in children with asthma

Abstract

Background:

Epidemiologic studies have shown associations between traffic-related pollutants such as diesel particulate matter (PM) and asthma outcomes in children, but the inflammatory features associated with diesel PM exposure in children with asthma are not understood.

Objective:

This study examined symptoms, exacerbations, and lung function measures in children with uncontrolled asthma and their associations with residential proximity to major roadways. Biomarker studies were performed to determine associations between diesel PM exposure and systemic inflammatory cytokines, circulating markers of T-cell activation and exhaustion, and metabolomic features.

Methods:

Children 5 through 17 years of age with physician-diagnosed, uncontrolled asthma despite treatment with an asthma controller medication completed a research visit involving questionnaires, lung function testing, and venipuncture for biomarker studies. Geocoding was performed to quantify residential proximity to major roadways and pollutant exposure.

Results:

Four hundred forty-seven children with uncontrolled asthma were enrolled. Children living closer to highly trafficked roadways were more disadvantaged and had more exposure to diesel PM, more exacerbations prompting an emergency department visit, and lower lung function measures. Children with the highest diesel PM exposure, compared to children with the lowest diesel PM exposure, also had blunted cytokine secretion and evidence of T cell exhaustion, as well as disturbances in several metabolites associated with glutathione formation and oxidative stress.

Conclusion:

Traffic-related diesel PM exposure in children with poorly controlled asthma is associated with poorer clinical outcomes and unique patterns of inflammation and oxidative stress. These findings argue for continued mitigation efforts to improve traffic-related air quality and health equity in children with asthma.

Introduction

Asthma currently affects 8.1% of all school-age children in the United States¹ and is a significant public health issue. Many children with asthma have symptoms that are not well controlled and nearly half of all children with asthma have an exacerbation (attack) each year.¹ Although the factors associated with asthma symptom control and exacerbations are complex, recent data suggest that perturbations in traffic-related air quality may aggravate asthma in children, possibly through alteration of airway function and host defenses.^2–4 Indeed, an ecological study conducted during the 1996 Summer Olympic games (which necessitated alternative transportation to downtown Atlanta, Georgia), observed improved traffic-related air quality and a substantial reduction in asthma acute care events in children.⁵ However, most prior studies of traffic exposure and asthma are epidemiologic in nature, without attention to individual patients, or alternatively were performed in highly controlled laboratory settings which do not generalize to real-world exposures. There are very few studies involving children and even fewer studies of inflammatory biomarkers in children living in high exposure areas. This is a significant gap since children may be more sensitive to airborne pollutants given developmental differences in their respiratory physiology, more time spent outdoors, and developing immune systems.⁶

To address this gap, we examined symptoms, exacerbations, and lung function measures in a well characterized sample of school-age children with uncontrolled asthma and their associations with residential proximity to major roadways. We hypothesized that closer residential proximity to highly trafficked roadways would be associated with poorer asthma control, more exacerbations, and lower lung function measures in children with uncontrolled asthma despite treatment with asthma controller medications. We then performed biomarker studies in these same children to determine associations between diesel particulate matter (PM) exposure and systemic inflammatory cytokines, markers of T-cell exhaustion and activation, and metabolomic features to identify potential biomarkers for future study.

Methods

Children 5 through 17 years of age with physician-diagnosed, uncontrolled asthma despite treatment with an asthma controller medication who were attending a specialty clinic for asthma at Children’s Healthcare of Atlanta in Atlanta, Georgia, were eligible for the study. Children’s Healthcare of Atlanta is a large, urban pediatric healthcare system which manages more than 414,000 children each year, including more than 22,500 unique patients with asthma. Families were approached for participation after a review of asthma-related outpatient encounters in the electronic medical records system. Children were eligible for the study if they: 1) prescribed an asthma controller medication for at least 6 months, 2) had asthma symptoms that were not well controlled, defined by daytime symptoms more than twice weekly, any night awakening due to asthma, or any activity limitation over the past month,⁷ and 3) had confirmed asthma with historical evidence of ≥ 12% reversibility in their forced expiratory volume in one second (FEV₁) relative to baseline after administration of a bronchodilator or systemic corticosteroid. Exclusion criteria included residence outside of metropolitan Atlanta, premature birth before 35 weeks of gestation, and other chronic airway disorders that could mimic asthma such as pulmonary aspiration disorders, congenital airway anomalies, immune deficiency, cystic fibrosis, or vocal cord dysfunction. Permission to proceed with this study was granted by the Emory University and Children’s Healthcare of Atlanta Institutional Review Boards. Informed written consent was obtained from legal guardians. Verbal assent was obtained from children 6–10 years and written assent was obtained from children and adolescents 11 to 17 years.

Study design and residential geocoding for exposure quantification.

Participants completed a single outpatient research visit that was postponed if an asthma exacerbation treated with systemic corticosteroids was reported within the preceding two weeks. Residential geocoding was performed as described previously⁸ by mapping participant residential addresses to census tracts using the R package tidygeocoder (R version 4.0.2).⁹ U.S. Census 2020 Geographic Identifiers (GEOIDs) were mapped to the 2010 GEOIDs for each census tract in Georgia using the R package tigris.¹⁰ Air pollution indices (ozone, PM_2.5, and diesel PM) were determined for each participant’s census tract from the Center for Disease Control and Prevention’s Environmental Justice Index Environmental Burden Module,¹¹ which relies on historical data with varying time scales to characterize cumulative impacts across communities.¹² In this module, ozone is defined as the mean annual number of days with maximum 8-hour average ozone concentration of the National Ambient Air Quality Standard (NAAQS) (averaged over three years), PM_2.5 is defined as the mean annual percent of days with daily 24-hour PM_2.5 concentrations over the NAAQS (averaged over three years), and diesel PM is defined as the diesel concentration in air in μg/m³ from the U.S. Environmental Protection Agency National Air Toxics Assessment, modeled over one year.¹² Residential proximity to a major roadway was estimated with the 2019 Topologically Integrated Geographic Encoding and Referencing/Line shapefile for the primary and secondary roads in Georgia, which were obtained using the R package tigris.¹⁰ The Euclidian distance, or the shortest distance between a point (residence) and a line segment (roadway) “as the crow flies” was measured in meters using the R package sf.¹³ Distances were verified by mapping the geographical coordinates of the residential address and the Topologically Integrated Geographic Encoding and Referencing/Line road shapefile using the R package mapview.¹⁴

Clinical characterization procedures.

Participants and their caregivers completed questionnaires pertaining to symptoms, medical history, and demographics. Asthma symptom control was assessed with the 6-item Asthma Control Questionnaire (ACQ) validated in children¹⁵ that was administered by a trained interviewer.¹⁶ Spirometry (KoKo^® PDS, Ferraris, Louisville, Colorado) was performed according to technical standards¹⁷ and the best of three forced vital capacity (FVC) maneuvers was interpreted for FEV₁, FEV₁/FVC, and forced expiratory flow at 25–75% of vital capacity (FEF_25–75) according to race/ethnicity-corrected Global Lung Function Initiative prediction equations.¹⁸ Lung function data were verified with race/ethnicity neutral equations which have been shown to increase the prevalence and severity of respiratory impairment in Black children¹⁹ and were interpreted as percentages of predicted values. Venipuncture was also performed for biomarker quantification. Sensitization to eight aeroallergens (tree mix, grass mix, weed mix, mold mix, dog dander, cat dander, Blatella germanica, and dust mite mix (Dermatophagoides farinae and Dermatophagoides pteronyssinus) was assessed with specific immunoglobulin E (IgE) testing by the ImmunoCap method (Children’s Healthcare of Atlanta, Atlanta, GA). Tests with specific IgE values >0.35 kU/L were considered positive. Venipuncture was also performed for quantification of blood eosinophils and total serum IgE, which were measured in a hospital laboratory (Children’s Healthcare of Atlanta, Atlanta, Georgia).

Clinical outcome methods.

Outcome measures of interest included asthma symptom control measured by the ACQ-6 instrument on the day of enrollment, the occurrence of any asthma exacerbations prompting emergency department (ED) utilization or requiring hospitalization in the previous 12 months, and lung function measures (FEV₁, FEV₁/FVC, and FEF_25–75 percent) on the day of enrollment. Asthma exacerbations were verified by a review of electronic medical records. Only ED visits in which a course of systemic corticosteroids was initiated were counted as an exacerbation in accordance with consensus recommendations.²⁰

Laboratory methods for biomarker quantification.

For biomarker studies, whole blood was collected into EDTA tubes and immediately centrifuged for isolation of plasma. Plasma samples were stored at −80°C and analyzed in a single batch for each experiment. A 21-plex human T-cell Panel array (HSTCMAG 28SK, T cell Panel, Millipore, Burlington, Massachusetts) was used to quantify plasma concentrations of C-X-C motif chemokine ligand 11 (CXCL11), granulocyte–macrophage colony-stimulating factor (GMCSF), C-X3-C motif chemokine ligand 1 (CX3CL1), C-C motif chemokine ligand (CCL)-3, CCL4, CCL20, interferon gamma (IFNγ), interleukin (IL)-1β, IL-2, IL-4, IL-5, IL-6, IL-7, IL-8, IL-10, IL-12 (p70), IL-13, IL-17, IL-21, IL-23, and tumor necrosis factor alpha (TNFα) with a Luminex MAGPIX system (Millipore) according to technical standards. T-cell exhaustion and circulating markers of T cell activation were measured in plasma with the 10-plex LEGENDplex Human Immune Checkpoint Panel 1 kit (BioLegend, San Diego, California) on a FACSAria flow cytometer (Becton Dickinson, Franklin Lakes, New Jersey). Analytes were measured according to the manufacturer’s instructions and included soluble CD25 (i.e., IL-2Rα), CD86, CD137, CD223, CD274 (i.e., Programmed Cell Death 1 Ligand-1), CD279 (i.e., Programmed Cell Death-1), free active transforming growth factor beta 1 (TGFβ1), cytotoxic T-lymphocyte associated protein 4 (CTLA4), T-cell membrane protein 3 (Tim3), and galectin-9. Samples were analyzed in duplicate with positive and negative controls.

Amino acid metabolites (52 in total) were measured in plasma by solid phase extraction as described previously²¹ followed by derivatization and liquid/liquid extraction (EZ:faast Kit, Phenomenex, Torrance, California) according to the manufacturer’s instructions. Samples were mixed with internal standards (homoarginine, methionine-d3 and homophenylalanine), extracted, and derivatized with propyl chloroformate. The organic phase was evaporated at room temperature under a stream of nitrogen and re-dissolved in mobile phase. Samples were analyzed using a Thermo Vanquish ultra-high performance liquid chromatograph coupled to a Thermo TSQ Quantis triple quadrupole mass spectrometer (Thermo Scientific, Waltham, Massachusetts). Using an autosampler at 4°C, a volume of 1 μL was injected onto a 250 × 2.0 mm × 4 μ AAA-MS column (Phenomenex) at a flow rate of 0.25 mL/min. The column was held at 35°C. Mobile phase A was 10 mM ammonium formate in water, and mobile phase B was 10 mM ammonium formate in methanol. Samples were separated using an 18-minute gradient, from 68 to 83% of mobile phase B, with a 7-minute re-equilibration between samples. The ion transfer tube and vaporizer were maintained at 275°C and 225°C respectively. Positive electrospray ionization mode at 5000 V was used to monitor selected reaction transitions as outlined in the EZ:faast manual. Transitions were optimized for the mass spectrometer using derivatized standards. Quantitation of amino acids was performed using TraceFinder software (Thermo Scientific).

Statistical analyses.

Data were analyzed with SPSS^® Statistics (Version 29, IBM, Armonk, NY). Associations between residential distance to a major roadway and clinical outcomes were assessed with linear regression models that were adjusted for age, sex, race, ethnicity, and obesity. Diesel exposure groups were defined by tertiles of diesel PM concentrations. Cytokine concentrations and T cell exhaustion markers were log-normalized prior to analysis. Cytokines and T cell exhaustion markers were compared between diesel exposure tertile groups with analysis of variance and Tukey’s Least Significant Different tests for pairwise comparisons.

Metabolites were log-normalized prior to analysis and were first visualized with linear discriminant analysis, which was performed using the Fisher method.²¹ This procedure yields a set of discriminant functions based on the linear combination of all the measured metabolites that provide the best discrimination between groups. Linear discriminant analysis is related to principal component analysis and explicitly attempts to model the differences between the groups of interest. The resulting linear discriminant functions were plotted to visualize the proximity of each participant to others in the same group. To determine the individual metabolites that differed significantly between groups, significant metabolites were first identified with analysis of variance and those significant metabolites underwent Tukey’s Least Significant Different tests for pairwise comparisons. Pathway analysis was then performed on the metabolite data to identify the biologic pathways that were altered between groups. Pathway analysis integrated pathway enrichment analysis and pathway topology analysis and was performed with MetaboAnalyst 5.0²² using global test enrichment methods, and relative-betweenness centrality.

All analyses utilized a p-value of 0.05 as the threshold for statistical significance. Biomarker analyses including metabolomics analyses were not adjusted for multiple comparisons given the exploratory nature of those experiments.

Results

Four hundred forty-seven children with poorly controlled asthma despite treatment with asthma controller medications participated in the study. The children enrolled were disproportionately Black and were predominantly atopic, with elevated blood eosinophil counts and total IgE concentrations (Table 1). Eighty nine percent of the participants had either aeroallergen sensitization or blood eosinophil counts above 300 cells/microliter, consistent with Type 2 inflammation,²³ and 94% of the participants had either aeroallergen sensitization, blood eosinophil counts above 300 cells/microliter, or eczema. Other features of the children are shown in Table 1 and eTable 1.

Table 1.

Features of the participants. Data represent the number of participants (%) or the median (25^th, 75^th percentile).

Feature	Participants (N=447)
Age (years)	10.8 (8.3, 13.5)
Asthma duration (years)	8.0 (5.0, 11.4)
Males	259 (57.9)
Hispanic or Latino ethnicity	7 (1.6)
Primary race (self-reported)
White	92 (20.6)
Black	332 (74.3)
Asian	4 (0.9)
More than one race	19 (4.3)
Highest household education
Unknown/not reported	114 (25.5)
Did not complete high school	13 (2.9)
High school degree	52 (11.6)
Technical training/some college	120 (26.8)
Bachelor’s degree	148 (33.1)
Obese (body mass index >95th percentile)	112 (25.1)
Asthma controller medications (prescribed)
Inhaled corticosteroid	354 (79.2)
Long-acting beta agonist	215 (48.1)
Leukotriene receptor antagonist	251 (56.2)
Omalizumab	11 (2.6)
Number of controller medications	2 (1, 3)
Atopic features
Eczema	246 (55.0)
Aeroallergen sensitization1	304 (84.4)
Blood eosinophil count (cells/microliter)2	261 (143, 434)
Blood eosinophils >300 cells/microliter2	162 (41.6)
Total IgE (kU/L)3	355 (115, 813)
Residential ZIP code features
Unemployment (%)	13.4 (9.5, 18.6)
Median household income (thousand $)	46.9 (39.5, 57.9)
Private insurance (%)	60.5 (51.9, 69.5)
Bachelor’s degree completion (%)	27.4 (20.2, 37.4)
Families in poverty (%)	15.7 (10.4, 21.4)
Rental homes (%)	40.2 (26.6, 51.0)

¹Participants with specific IgE results available, N=360

²Participants with blood eosinophil counts available, N=389

³Participants with total IgE results available, N=368

Associations between roadway proximity and asthma clinical outcomes.

We first examined associations between residential proximity to a major roadway and asthma clinical outcomes of interest, including symptom severity, exacerbations necessitating healthcare utilization, and lung function measures. There were no associations between roadway proximity and the magnitude of asthma symptoms measured by ACQ-6 scores (univariate p=0.425; adjusted p=0.494). However, children with an asthma exacerbation prompting an ED visit in the previous 12 months lived closer to major roadways. (Figure 1A). Compared to children living ≥3000 meters, children living <1000 meters and children living 1000–2999 meters of a major roadway had increased odds of exacerbation in both univariate (<1000 meters, odds ratio and 95% confidence interval: 2.04 (1.14–3.66); 1000–2999 meters: 1.94 (1.04–3.63)) and adjusted models (1000 meters: 2.24 (1.12–4.48); 1000–2999 meters: 2.35 (1.12–4.94). Hospitalizations were less frequent and were not associated with roadway proximity in this sample (Figure 1A). Closer residential proximity to major roadways was also associated with very slightly yet significantly lower FEV_1, FEV₁/FVC, and FEF_25–75 percent predicted values with race/ethnicity adjusted equations (eFigure 1A–C). There were no differences in FVC percent predicted values (adjusted p=0.396). Children living within 1000 meters of a major roadway, compared to children living 3000 meters or beyond, had mean reductions of approximately 5.5% for FEV_1, 4.1% for FEV₁/FVC, and 8.3% for FEF_25–75 percent predicted values, respectively (eFigure 1D–F). However, with application of the race-neutral prediction equations (which do not analyze FEF_25–75), the associations between roadway proximity and FEV₁ and FEV₁/FVC were more pronounced (Figure 1B–C). Children living within 1000 meters of a major roadway, compared to children living 3000 meters or beyond, had mean reductions of approximately 6.9% and 3.9% for for FEV₁ and FEV₁/FVC percent predicted values, respectively (Figure 1D–E).

Figure 1.

Associations between residential proximity to major roadways and (A) healthcare utilization for acute exacerbations, (B-C) race/ethnicity-neutral FEV₁ % predicted values and (D-E) FEV₁/FVC % predicted values. P-values are adjusted for age, sex, race, ethnicity, and obesity. ED = Emergency department, m = meters.

Airborne constituents associated with roadway proximity.

To further understand the airborne constituents associated with roadway proximity, we then examined residential indices of ozone, PM_2.5, and diesel PM. Roadway proximity was not associated with ozone measured as days above the regulatory standard (adjusted p=0.358), but was associated with PM_2.5 days above the regulatory standard (adjusted p<0.001) and diesel PM concentrations (adjusted p<0.001) (eFigure 2A–B). PM_2.5 days and diesel PM concentrations were significantly higher in children residing within 1000 meters of a major roadway (eFigure 2C–D).

Diesel particles are highly respirable and constitute a significant fraction of PM_2.5.²⁴ However, because there is no threshold for defining low, moderate or high aggregate residential diesel exposure, children were then grouped into tertiles of diesel exposure for biomarker studies using the following cut points: 0–0.45 μg/m³ diesel PM (defined as “low” exposure), 0.46–0.59 μg/m³ diesel PM (defined as “moderate” exposure), and 0.60–1.47 μg/m³ diesel PM (defined as “high” exposure). Features of the children in each diesel exposure tertile are shown in Table 2. Children with the highest diesel exposure, who also had the greatest residential proximity to major roadways and more exposure to PM_2.5 and ozone, were more likely to identify as Black and had families with lesser educational attainment (Table 2). Although the percentage of children with indoor and outdoor aeroallergen sensitization was not different between the diesel exposure tertiles, children with the highest diesel exposure had significantly less sensitization to dog dander and to molds (eTable 1). Children with the highest diesel exposure also had the highest exposure to PM_2.5 and ozone (eFigure 3A–B).

Table 2.

Features of the participants, grouped by diesel exposure tertiles. Data represent the number of participants (%) or the median (25^th, 75^th percentile).

Feature	Low diesel exposure N=132	Moderate diesel exposure N=142	High diesel exposure N=173	Overall p-value
Age (years)	10.7 (8.1, 13.1)	11.1 (8.3, 14.1)	10.7 (8.0, 13.2)	0.183
Asthma duration (years)	7.2 (4.8, 11.0)	8.0 (5.2, 11.3)	8.1 (5.4, 11.6)	0.441
Males	70 (53.0)	83 (58.5)	106 (61.3)	0.348
Hispanic ethnicity	3 (2.3)	2 (1.4)	2 (1.2)	0.726
Race				<0.001
White	48 (36.4)	23 (16.2)	21 (12.1)
Black	79 (59.8)	111 (78.2)	142 (82.1)
Asian	0	1 (0.7)	3 (1.7)
More than one race	5 (3.8)	7 (4.9)	7 (4.0)
Household education				0.002
Unknown/Not reported	45 (34.1)	28 (19.7)	41 (23.7)
Did not complete high school	3 (2.3)	3 (2.1)	7 (4.0)
High school degree	6 (4.5)	16 (11.3)	30 (17.3)
Technical training/some college	27 (20.5)	46 (32.4)	47 (27.2)
Bachelor’s degree	51 (38.6)	49 (34.5)	48 (27.7)
Obese (body mass index >95th percentile)	34 (26.0)	29 (20.7)	49 (28.3)	0.297
Asthma medications (prescribed)
Inhaled corticosteroid	106 (80.3)	114 (80.3)	134 (77.5)	0.772
Number of medications	2 (1, 3)	2 (1, 3)	2 (1, 3)	0.309
Atopic features
Blood eosinophils (per microliter)	292 (176, 444)	240 (141, 412)	255 (126, 440)	0.191
IgE (kU/L)	335 (78,823)	391 (114, 852)	349 (127, 797)	0.487

Biomarkers of inflammation in exposure tertiles.

Biomarker studies were then performed in each diesel exposure group. There were 331 plasma samples available for cytokine analyses (low exposure, n=87; moderate exposure, n=110; high exposure, n=134) and 230 samples available for T-cell exhaustion experiments (low exposure, n=46; moderate exposure, n=65; high exposure, n=94). Children with samples available for biomarker studies were not different from the parent sample (data not shown). Children with the highest diesel exposure had slightly yet significantly lower concentrations of IFNγ (median for low vs. moderate vs. high exposure: 6.6 vs. 4.7 vs. 4.9 pg/mL); lower concentrations of CXCL11, which is induced by IFNγ (median: 14.4 vs. 13.5 vs. 12.1 pg/mL); lower concentrations of IL-12p70, which drives production of IFNγ (median: 1.4 vs. 1.3 vs. 1.3 pg/mL); lower concentrations of IL-2, which also drives production of IFNγ and controls the survival and proliferation of regulatory T-cells (median: 1.2 vs. 0.9 vs. 0.9 pg/mL); and lower concentrations of IL-7, which promotes T cell development and regulates activated T cell proliferation (median: 4.5 vs. 3.8 vs. 3.9 pg/mL) (Figure 2A–E). There were no differences in GMCSF, CX3CL1, CCL3, CCL4, CCL20, IL-1β, IL-4, IL-5, IL-6, IL-8, IL-10, IL-13, IL-17, IL-21, IL-23, or TNFα (eTable 2). Children with the highest diesel exposure also had higher expression of circulating sCD25, which is released from T cells during chronic T-cell activation, and higher expression of the circulating T cell exhaustion markers, CD86 and Tim3 (Figure 2F–H). Higher expression of sCD25 was associated with lower concentrations of IFNγ (r= −0.191, p=0.004), CXCL11 (r= −0.152, p=0.023), IL-12p70 (r= −0.152, p=0.023), IL-2 (r= −0.179, p=0.007), and IL-7 (r= −0.276, p<0.001) (eFigure 4A–E). Higher expression of CD86 was also associated with lower concentrations of IFNγ (r= −0.252, p<0.001) and IL-7 (r= −0.218, p<0.001) (eFigure 4F–G). There were no differences in CD137, CD223, CD274, CD279, active TGFβ1, CTLA4, or galectin-9 between diesel exposure tertiles (eTable 3).

Figure 2.

Plasma concentrations of inflammatory cytokines (A) IFNγ, (B) CXCL11, (C) IL-12p70, (D) IL-2 and (E) IL-7; and T-cell exhaustion markers (F) sCD25, (G) CD86, and (H) Tim3 in tertiles of children based on diesel particulate matter exposure.

Metabolomic features in exposure tertiles.

Fifty-eight plasma metabolites were quantified in plasma from 109 children (low exposure, n=33; moderate exposure, n=34; high exposure, n=42) who were representative of the parent sample (data not shown). Linear discriminant analysis of the combination of all 58 plasma metabolites demonstrated significant differences between the diesel exposure tertiles (Figure 3A). The resulting model yielded the greatest separation between the low and high diesel exposure groups (discriminant function 1 Eigenvalue=1.63, 52.7% of variance explained, Wilks’ lambda=0.155, p=0.022), whereas more overlap was noted between children with moderate exposure and the other groups (discriminant function 2 Eigenvalue=1.461, 47.3% of variance explained, Wilks’ lambda=0.406, p=0.089). However, the discriminant model yielded classification probabilities >0.80 (Figure 3B). Twelve metabolites differed in children with the highest diesel exposure, including ethanolamine, which stimulates cellular proliferation and affects lipid metabolism and short-chain fatty acid biosynthesis; pyroglutamic acid, methionine, glutamic acid and proline, which are involved in synthesis of the antioxidant, glutathione; methionine sulfoxide, a marker of oxidative stress; 4-aminobenzoic acid, an intermediate in the synthesis of folate; α-aminobutyric acid, a metabolite in isoleucine biosynthesis; α-aminoadipic acid, which is a precursor in the biosynthesis of lysine; and the amino acids valine, tyrosine, and aspartic acid (Figure 4).

Figure 3.

Discriminant analysis of the linear combination of all 58 plasma metabolites demonstrating (A) separation tertiles of children based on diesel particulate matter exposure and (B) a high probability (>0.80) of correct classification of each tertile group.

Figure 4.

Concentrations of (A) 4-aminobenzoic acid, (B) α-aminoadipic acid, (C) α-aminobutyric acid, (D) aspartic acid, (E) ethanolamine, (F) glutamic acid, (G) methionine, (H) methionine sulfoxide, (I) pyroglutamic acid, (J) proline, (K) tyrosine, and (L) valine in tertiles of children based on diesel particulate matter exposure. p-values are for the post-hoc pairwise comparisons obtained from Tukey testing.

To understand the pathways associated with these metabolomic differences, exploratory metabolic pathway analyses were then performed using the data from the high versus the low diesel exposure groups. With an unadjusted p-value of 0.05 as the threshold for significance, eight pathways were significantly mapped, including ubiquinone and other terpenoid-quinone biosynthesis (1/18 pathway metabolites matched, p=0.005), tyrosine metabolism (3/42 matched, p=0.005), phenylalanine, tyrosine and tryptophan metabolism (2/4 matched, p=0.006), phenylalanine metabolism (2/8 matched, p=0.006), pantothenate and CoA biosynthesis (2/20 matched, p=0.019), valine, leucine and isoleucine degradation (3/40 matched, p=0.019), and arginine and proline metabolism (3/36 matched, p=0.043). The pathway impact, a reflection of the relative importance of the metabolites in the pathway, was highest for phenylalanine, tyrosine and tryptophan biosynthesis, followed by phenylalanine metabolism, arginine and proline metabolism, and tyrosine metabolism (Figure 5A). The two phenylalanine pathways and the tyrosine metabolism pathway were also significant at a false discovery rate of 0.05 (adjusted p=0.048 for each).

Figure 5.

Exploratory metabolic pathway analyses for the high versus the low diesel exposure groups. Pathway impact reflects the relative importance of the identified metabolites in the pathway. Larger circles reflect more important pathway impact. Significant pathways at an unadjusted p-value of 0.05 are shown in red. Non-significant pathways are shown in orange, yellow and white.

Discussion

Children living near heavily trafficked roadways are exposed to complex mixtures of pollutants, which not only disrupt air quality, but also worsen respiratory outcomes. In the present study, we found that closer residential proximity to highly trafficked roadways was associated with more exposure to diesel PM, more historical exacerbations prompting an ED visit, and lower lung function measures in children with poorly controlled asthma despite prescription of asthma controller medications. In children with the highest diesel PM exposure, we also observed blunted cytokine secretion with chronic T cell activation and T cell exhaustion, as well as disturbances in several metabolites associated with oxidative stress. Together, these observations highlight the potential negative health consequences of traffic-related diesel exposure in children with poorly controlled asthma and argue for continued mitigation efforts to improve health equity in this population.

Our results are similar to those of others who have shown associations between roadway proximity and asthma outcomes in children. In a recent analysis of inner-city, school-age children with asthma, proximity to heavily trafficked roadways was associated with significantly more asthma symptoms and healthcare utilization for asthma.²⁵ Other studies have also observed similar associations between roadway pollutant exposure and asthma symptoms.²⁶ Although we did not observe associations between asthma symptoms measured by ACQ-6 scores and residential roadway proximity in the present study, our inclusion criteria were limited to children with poorly controlled asthma, who were, by definition, all symptomatic at enrollment. However, our observations of more ED visits for asthma exacerbations and lower lung function measures in children residing closer heavily trafficked roads are in agreement with other studies.^{25, 27–30} Furthermore, in the present study, FVC was not lower in children residing closest to major roadways, suggesting that the traffic-related airflow limitation was likely inflammatory in nature and not related to impairment of lung growth.

Prior studies have shown inconsistent associations between traffic-related pollutants and inflammatory cytokines in children. Higher plasma pro-inflammatory cytokines have previously been reported in children with asthma exposed to higher residential air pollutants compared to healthy controls.^{28, 31, 32} However, in one study of whole blood from children with asthma exposed to traffic-related PM ex vivo, cytokine responses (namely IL-8) were lower in children with poorly controlled asthma compared to children whose asthma was well controlled.³³ In the present study, we also observed lower baseline cytokine expression in children with the highest residential exposure to diesel PM. This observation of lower cytokine expression was also associated with higher expression of sCD25 and CD86. Although the concept of T cell exhaustion is broad,³⁴ one interpretation of our findings is that the systemic T cells of patients with poorly controlled asthma exposed to traffic-related diesel PM are hypofunctional in response to chronic antigen stimulation. Indeed, nearly all the patients enrolled in the present study had evidence of allergic sensitization with T2 inflammation. Other studies have shown that chronic allergen exposure drives Th2 cells to exhaustion in sensitized mice with allergic asthma³⁵ and patients with allergic rhinitis.³⁶ T cell exhaustion has also been observed in settings of chronic pro-inflammatory cytokine stimulation, in which T cells display a progressive loss of cytokine production beginning with IL-2, then finally IFNγ.³⁷ In response, exhausted T cells also express a variety of inhibitory receptors (including sCD25, CD86, and Tim3) and exhibit dysregulated metabolism to regulate T cell proliferation and effector functions.³⁷ However, the present study was limited to children with poorly controlled asthma and did not include a control group of children with well controlled asthma in the same geocode, so we cannot definitely conclude that the T cell marker data are related to worse asthma outcomes in this population.

In the present study, we also observed several metabolic derangements in children with the highest traffic-related diesel PM exposure. Most notably, we observed increases in glutathione intermediates and increases in methionine sulfoxide. Glutathione plays a critical role in cellular defense against unregulated oxidative stress by acting directly as an antioxidant against free radicals and pro-oxidants and as a co-factor for enzymes such as glutathione peroxidases.³⁸ Methionine sulfoxide is formed from oxidation of the sulfur of methionine and can cause tissue proteins to be dysfunctional.³⁹ Higher concentrations of glutathione-related biomarkers and methionine sulfoxide have previously been observed in adults with higher versus lower traffic exposure.⁴⁰ These metabolites were also recognized as important traffic-related biomarkers in adults in a systematic pathway and network analysis of 139 publications focused on traffic-related air pollution.⁴¹ This present study extends those findings to a vulnerable group of children with poorly controlled asthma, who were not represented in many previous studies.

The exploratory metabolic pathway analyses performed in this study also identified disturbances in phenylalanine, tyrosine, and arginine pathways in children with high versus low diesel exposure, which have biologic plausibility. Phenylalanine is associated with increased pro-inflammatory cytokine release and lung inflammation⁴² and is also increased in children with acute asthma exacerbations.^{43, 44} Tyrosine is a non-essential amino acid that is synthesized from phenylalanine and perturbations in tyrosine metabolism have been associated with corticosteroid-refractory asthma in children⁴⁵ and neutrophilic asthma phenotypes in adults.⁴⁶ Similarly, arginine metabolism promotes airway inflammation and hyperresponsiveness⁴⁷ and lower concentrations of arginine have been reported in children with uncontrolled^{21, 48} and exacerbating⁴⁴ asthma.

It is well recognized that the negative effects of traffic-related diesel PM exposure are not equally distributed across the population but are instead disproportionately increased in socially disadvantaged populations. In the present study, we observed that disadvantaged Black children were exposed to higher traffic-related diesel PM. Others have similarly shown that Black and Hispanic children are more likely to live in lower opportunity neighborhoods with less recreational green space due to systemic structural racism and historic red lining practices, are exposed to more pollutants, and have more severe asthma exacerbations attributable to air pollution.^{27, 49–55} These same children are also projected to have a 1.3-fold risk of living in a neighborhood with the highest projected increases in asthma due to increased PM concentrations.⁵⁶

This study does have limitations. First, it is recognized that traffic-related pollutants such as PM_2.5 are heterogeneous, and the toxicity can vary depending on the chemical composition.⁵⁷ Our analyses involving stratification by diesel PM exposure are therefore oversimplifications, but were pursued given reports of the clinical relevance of diesel PM. Indeed, others have shown an increase in 0.2% above the mean rate of pediatric asthma for every unit increase in diesel PM.²⁷ Furthermore, other studies have shown that the PM generated from diesel-fueled vehicle sources is associated with acute asthma exacerbations in school aged children.⁵⁸ The clinical effects from a single day of diesel PM exposure may also be sustained for up to a week.⁵⁸ Second, we cannot accurately quantify the dose, duration or timing of the traffic-related exposures in relation to the clinical data that we obtained. Our data source, which provided census tract level estimates, did not permit estimation of day-by-day exposure quantification for estimation of lag effects, but instead yielded an average of the exposures over one to three years. We also did not capture information on roadway proximity and traffic-related exposures in the children’s schools, where they spend a significant amount of time. Third, since traffic densities and exposure mitigation strategies can vary widely across regions, the findings from this study may not be generalizable to other regions outside of metropolitan Atlanta, Georgia. Finally, since this study is associative in nature, we cannot rule out residual confounding since the factors contributing to asthma exacerbations in children are complex. For example, although the children enrolled were prescribed asthma controller medications, adherence measures were not quantified. We also did not enroll children with well controlled asthma in the same geocode as a control group and we did not measure indoor allergen exposure and housing quality. Therefore, the differences in the clinical and molecular features that we observed between the diesel exposure groups could be related to these and other unmeasured factors.

Conclusion.

In conclusion, our findings provide additional evidence that roadway proximity and traffic-related diesel PM may contribute to clinical outcomes in children with asthma through alteration in systemic inflammatory host defenses and oxidative stress. Since the children with the highest roadway exposures in this study were also more disadvantaged, the findings from this study also highlight the need for ongoing public policies and focused interventions to reduce the respiratory impacts of traffic-related exposures in children with asthma in metropolitan Atlanta, Georgia.

Abbreviations:

ACQ

Asthma Control Questionnaire

CCL

C-C motif chemokine ligand

CTLA4

Cytotoxic T-lymphocyte associated protein 4

CX3CL1

C-X3-C motif chemokine ligand 1

CXCL11

C-X-C motif chemokine ligand 11

ED

Emergency Department

FEF25–75

Forced expiratory flow at 25–75% of vital capacity

FEV1

Forced expiratory volume in one second

FVC

Forced vital capacity

GEOID

Geographic identifier

IFNγ

Interferon gamma

IgE

Immunoglobulin E

IL

Interleukin

NAAQS

National Ambient Air Quality Standard

PM

Particulate matter

TGFβ1

Transforming growth factor beta 1

Tim3

T-cell membrane protein 3

TNFα

Tumor necrosis factor alpha