Association of household air pollution exposure with cellular immune responses among women in rural Bangladesh

Abstract

Household air pollution (HAP) arising from combustion of biomass fuel (BMF) is a leading cause of morbidity and mortality in low-income countries. Air pollution may stimulate pro-inflammatory responses by activating diverse immune cells and cyto/chemokine expression, thereby contributing to diseases. We aimed to study cellular immune responses among women chronically exposed to HAP through use of BMF for domestic cooking.Among 200 healthy, non-smoking women in rural Bangladesh, we assessed exposure to HAP by measuring particulate matter 2.5 (PM_2.5), black carbon (BC) and carbon monoxide (CO), through use of personal monitors RTI MicroPEM^™ and Lascar CO logger respectively, for 48 hours. Blood samples were collected following HAP exposure assessment and were analyzed for immunoprofiling by flow cytometry, plasma IgE by automated analyzer and cyto/chemokine response from monocyte-derived-macrophages (MDM) and -dendritic cells (MDDC) by multiplex immunoassay. In multivariate linear regression model, a doubling of PM_2.5 was associated with small increments in naïve/immature B cells (CD19⁺CD38⁺) and plasmablasts (CD19⁺CD38⁺CD27⁺). In contrast, a doubling of CO was associated with 1.20% reduction in CD19⁺ B lymphocytes (95% confidence interval (CI)=−2.38, −0.001). A doubling of PM_2.5 and BC each was associated with 3.06% (95%CI=−5.73, −0.39) and 4.17% (95%CI=−7.76, −0.58) decrements in memory B cells (CD19⁺CD27⁺), respectively. Exposure to CO was associated with increased plasma IgE levels (beta(β)=240.4, 95%CI=3.06, 477.8). PM_2.5 and CO exposure was associated with increased MDM production of CXCL10 (beta(β)=12080, 95%CI=1201, 22960) and CCL5 (β=754.6, 95%CI=4.16, 1505), respectively. Conversely, BC exposure was associated with reduction in MDDC-produced CCL5 (β=−2973, 95%CI=−5281, −664.6) and TNF-α (β=−15761, 95%CI=−26924, −4597). Our findings suggest that chronic HAP exposure adversely affects proportions of B lymphocytes, particularly memory B cells and functions of antigen presenting cells in women.

1. Introduction

According to the World Health Organization (WHO), globally an estimated 4.2 million people die from various diseases that are connected to air pollution, such as chronic obstructive pulmonary disease, respiratory infections, cardiovascular disease, stroke and lung cancer (Abrams et al. 2017; Falcon-Rodriguez et al. 2016; Leni et al. 2020; Lodovici and Bigagli 2011; WHO 2014). The common source of HAP is burning of solid BMF such as wood, saw dust, agricultural residues for cooking and heating in low- and middle-income countries (LMIC) including Bangladesh, Nepal, India, Pakistan as well as in China. The major particulate and gaseous pollutants of HAP include PM_2.5, CO, nitrogen dioxide (NO₂), sulfur dioxide (SO₂) (Seow et al. 2016; Weaver et al. 2019; WHO 2014; Zhang and Srinivasan 2020). According to the US Environmental Protection Agency (EPA), the levels of indoor air pollutants are two to five times higher (sometimes 100 times higher) than those of outdoor air pollutants, and have been ranked among the top 5 environmental risks to public health. The association of biomass fuel generated air pollution and respiratory diseases have been shown to be common in Nepal (Budhathoki et al. 2020), India (Mondal and Paul 2020), Pakistan (Naz and Ghimire 2020) and other LMICs (Al-Janabi et al. 2021; Odo et al. 2021; Van Vliet et al. 2019; Woolley et al. 2021).

The impact of air pollution on morbidity/mortality outcomes are likely mediated by alterations in immune competence. Experimental and ex vivo studies have shown that before entering into the blood stream, air pollutants accumulate in the lung alveoli and bronchioles to activate the epithelial barrier, trigger inflammatory cascade modulating production of anti- and pro-inflammatory cytokines, reactive oxygen species, altering expression of various cell surface markers and induce immune dysfunction leading to ultrastructural damage in the lungs (Guttenberg et al. 2021; Jantzen et al. 2016; Lee et al. 2021; Maestre-Batlle et al. 2018; Sigaux et al. 2019; Yang et al. 2019). Emerging evidence suggest a causative relationship between ambient air pollution and increased prevalence of allergic diseases including asthma and allergic rhinitis (Hajime Takizawa 2011; PMID: 22016586; PMID: 17000005). Immunoglobulin E (IgE) is known to play vital role in certain allergic conditions including allergic rhinitis and allergies to various substances (González-Díaz et al. 2017). Concomitant exposure to diesel exhaust particles and air borne pollutants increase the levels of allergen-specific IgE, asthma severity, and airway inflammation and hyper-reactivity (Acciani et al. 2013; Brandt et al. 2013). A great majority of studies have focused on ambient air pollution (Li et al. 2019; Prunicki et al. 2020; Tripathy et al. 2021; Wang et al. 2020). Only a handful of epidemiological studies have reported association between chronic exposure to HAP, particularly from burning BMF and immune function (Dutta et al. 2012; Lehmann et al. 2001; Saha et al. 2016; Yao et al. 2019).

It is important to learn how HAP disrupts different components of immune system to cause disease so that the knowledge gained would help to design strategies to mitigate the pathological conditions. We carried out a prospective study to evaluate the associations of indoor/household air pollutants (assessed through PM_2.5, BC and CO concentrations) with a number of preclinical markers of cardiopulmonary diseases and immune functions among rural Bangladeshi women as part of Bangladesh GEOHealth study (Shahriar. MH et al. 2021). In the present study, we used the baseline cross-sectional data in order to investigate the underlying effects of HAP-mediated immune dysfunction and inflammation. We aimed to assess lymphocyte immunoprofile, humoral IgE response and functional capacity of antigen-presenting cells that are involved in both the innate and adaptive immune responses in relation to HAP. Thus, production of chemokines in relation to HAP exposure by monocyte-derived macrophages (MDM) and monocyte-derived dendritic cells (MDDC) were studied.

2. Methods

2.1. Study population and site

The study was carried out in Araihazar and Matlab, two rural subdistricts in Bangladesh, about 40 to 55 km from Dhaka. We measured 48-hour exposure to air pollutants (PM_2.5, BC and CO) and markers of immune function for 200 women (Araihazar n=100; Matlab n=100) (Figure 1) who use solid BMF (wood, dried straw, dung, agricultural residue) in traditional stoves for cooking purpose. The inclusion criteria for women were, age between 25 to 65 years, non-smokers and living with non-smoking spouses, living in households with ground water (tubewell) arsenic levels <10 μg/L(since both these areas are known to have tubewell water widely contaminated with arsenic), not known to have any immune related illness, apparently healthy, not taking any prescription medication known to suppress or enhance immune function, and not known to have any clinical events of CVD or lung disease including stroke or coronary heart disease.

Fig. 1.

Flow chart of the study showing recruitment process, timing of exposure assessment, sampling and outcome analysis. status of participants.

The study was approved by the Ethical Review Committee of icddr,b.

2.2. Field data collection

Based on the available data on tubewell water arsenic levels from Matlab and Araihazar, only those households were selected for screening that had low water arsenic levels. After obtaining informed consents from participants, structured questionnaire was applied to collect data on sociodemographic features; biomass/air pollution exposure related data included number and types of stoves used, fuel usage pattern, type of kitchen used on the days of exposure assessment (enclosed, semi-enclosed, open), distance of cooking area to living quarter, ventilation in both kitchen and home, and other sources of exposures. The participants were invited to the field clinics for anthropometric measurements and blood sample collection. Medical history was collected by study physicians. Trained field staff organized wearing of the portable personal air sampling devices by each participant and retrieving it from their homes. The study supervisors were responsible for loading filters into the device, calibration of personal air samplers before every measurement, changing of air filters and downloading of data files after every use in the data repository.

2.3. Personal Air quality sensor/ exposure monitor and exposure data collection

RTI MicroPEM^™, developed by RTI International, is a wearable, personal exposure monitor or sensing device that measures real time PM_2.5 (range of detection of PM_2.5 1 to 10,000 mg/m³) and collects the air into a filter at a flow rate of 0.4L/min. The fine particles accumulate in an internal Teflon air filter (diameter: 25mm; pore size: 3.0 micron) which is later weighted for PM mass gravimetrically (described below). Exposure to CO was monitored using a Data Logger (Lascar Electronics Inc, Erie, PA) which uses electrochemical method to detect CO and has a capacity to detect up to 1000 ppm. Each woman wore a small pouch containing the microPEM and CO logger placing it around the breathing zone during the measurement period (48 hours) except for during taking a bath and at night when sleeping. The MicroPEM harbors an internal accelerometer that is sensitive to sense breathing movement even when an individual is sitting still. Data from the internal accelerometer was checked for wearing compliance of microPEM devices.

The MicroPEM also contains a light scattering nephelometer that measures real-time PM concentration every 10s. For the current analysis, we used the gravimetric PM data; nephelometric data was not used. The particle concentration was calculated by dividing the PM mass with total volume of air consumed by the PM filter during measurement period (time-weighted average). The real-time PM_2.5 was processed with MicroPEMs docking station software and used for quality assurance. If quality of the data was not good, the measurement was repeated. The MicroPEM air filters were sent to Atomic Energy Center Dhaka, Bangladesh for weighing of filters before and after use as well as for measuring PM_2.5. PM weighing were done using a microbalance (METTLER Model MT5) (accuracy of 1 μg) which was maintained at room temperature of ~22 °C and relative humidity at 50%. An electrostatic charge eliminator (STATICMASTER) was used to eliminate the static charge accumulated on the filters before each weighing. Appropriate laboratory and field blanks were used to ensure the quality control of filter weight. HAP was measured twice for each women with 6-months interval between the two measurements and the average was used to capture the long-term exposure as well as seasonal variation in air pollution (Shahriar. MH et al. 2021). In this study, November to April was considered dry season while May to October was considered as wet season.

Data from CO Logger was downloaded and analyzed separately. For the quality control of CO data, the device was checked to see whether it ran for the entire measurement period. If data logging was stopped before completion of measurement, the measurements were repeated. During installing the device, the CO logger was carried in an airtight bag. We used logged data during travel time to check functionally of the device. A log sheet was developed to record participant movement and checked for any abnormal value during data download.

BC was measured from the PM_2.5 filters by EEL-type Smoke Stain Reflectometer. The concentration of BC was determined based on the amount of reflected light that was absorbed by the filter by using standards of carbon with known areal density and an assumed mass absorption coefficient. For BC calculation, 10 m²/g was used as the extinction coefficient. Secondary standards of known black carbon concentrations were applied to calibrate the reflectometer.

2.4. Specimen collection

Whole blood (20 ml) was collected from adult women at the field office within 2 weeks of the second HAP exposure measurement, and transported to icddr,b Laboratory in Dhaka within 3-4 hours. Participants were asked about their general health condition before collection of blood. Peripheral blood mononuclear cells (PBMC) was separated from whole blood by Ficoll-paque (Amersham Pharmacia Biotech, Inc; 800 centennial avenue piscataway NJ, USA) density gradient centrifugation. After removal of plasma, PBMC were washed with culture media containing RPMI 1640, 10% fetal bovine serum (FBS) and antibiotics (Gibco, Life technology, Grand Island, NY, USA). The PBMC were counted and divided into 4 parts/aliquot. One aliquot of PBMC was used for cell phenotyping, second aliquot (10 million) for T cell proliferation assay, the third and fourth aliquots were cultured for developing monocyte-derived macrophage (MDM) (5 million) and monocyte-derived dendritic cell (MDDC) (5 million) respectively for functional assay.

2.5. Immunoprofiling of lymphocytes

Cell concentrations of PBMC were adjusted to 0.5x10⁶ PBMCs for T cells and 1x10⁶ PBMCs for B cells in separate tubes and re-suspended in FACS staining buffer (FSB) (BD Biosciences, San Jose, CA, USA). The percentage and mean fluorescence intensity for each surface marker were determined by flow cytometry in a 4-color BD Accuri C6 flow cytometer system (BD Biosciences) by labelling lymphocytes with monoclonal antibodies against specific markers (CD markers) in FSB and incubated for 20 min at 4°C in dark. After fixation, cells were washed and re-suspended in FSB. The following combinations were used in two separate flow tubes: (1) T lymphocytes: anti-CD4-FITC, anti-CD8-APC, anti-CD25-PE, anti CD127-PerCP-Cy5.5; (2) B lymphocytes: anti-CD19-PE, anti-CD27 FITC, anti-CD38-APC antibodies (all from BD Pharmingen, 10975 Torreyana Road San Diego, USA). CD4 and CD8 are expressed on CD4⁺ T lymphocytes, and CD8⁺ T lymphocytes respectively; CD4⁺CD25⁺CD127^low for identifying T regulatory cells; CD19 is a marker for CD19⁺B lymphocytes; CD19⁺CD27⁺ for memory B cells; CD19⁺CD38⁺ for early/naive B cells; and CD19⁺CD27⁺CD38⁺ for identifying plasmablasts. Figure S1 (A-E) shows the plots defining the gating process of subpopulations of B and T lymphocytes. In BD Accuri C6 flow cytometer system, a maximum of 4 fluorescent dyes can be used at a time. Thus, for the participants having adequate PBMC count (≥22 million PBMC), B lymphocytes (CD3⁻CD19⁺) gate had been defined by using additional flow tube with monoclonal antibody against CD3 cell surface marker along with T and B panel tubes. The percentages of CD3⁻CD19⁺ B lymphocytes and CD19⁺ B lymphocytes in total lymphocytes were similar (Fig S1F-G). At least 200,000 events per sample were collected in BD Accuri C6 flow cytometer and analyzed using FlowJo10 software (Tree Star, Inc., Ashland, OR, USA). Lymphocytes were identified by forward and side-angle light scatter characteristics. From lymphocyte gated population CD19⁺ B cells and CD4⁺ T cell and CD8⁺ T cells were defined. From CD19⁺ B cells gated population CD19⁺CD27⁺ memory B, CD19⁺CD38⁺ naive B cells and CD19⁺CD27⁺CD38⁺ plasmablasts were outlined. From CD4⁺ T cells CD4⁺CD25⁺CD127^low Treg cells were identified (Milward et al. 2019; Seddiki et al. 2006). At least 20,000 lymphocyte-gated cells were analyzed for each sample. Fluorescence-minus one (FMO) controls were performed for each cell surface marker.

2.6. Plasma Immunoglobulin E measurement

Plasma level of immunoglobulin E (IgE) was measured by electrochemiluminescence immunoassay (ECLIA) with Roche automated immunoassay analyzers Cobas e601 using Elecsys IgE II kit (Roche Diagnostics, GmbH, 68305 Mannheim, Germany) according to the manufacturer’s instruction. This method has been standardized against the 2nd WHO International Reference Reagent for serum IgE (75/502; 5000 IU/ampoule). Two control serums (PreciControl Universal, Roche Diagnostics) having high and low level of IgE were used to check both accuracy and precision.

2.7. Culture of macrophages and dendritic cells for functional assessment of cytokine secretion

For macrophage and dendritic cell culture assays, we stratified both PM_2.5 and BC by median split (134.8 and 6.1 μg/m3 for PM_2.5 and BC, respectively) and selected 40 samples from above the median (for both PM_2.5 and BC) and remaining 40 from below the median. In preparation for growing and isolating MDM, PBMCs (5x10⁶ PBMCs/well) were cultured in 4-well culture plate (Thermo Scientific/NUNC, Roskilde, Denmark) in culture media (RPMI media containing 10% autologous plasma and antibiotics) at 37°C in CO₂ incubator for 6 days. Thereafter, non-adherent cells were removed by washing and fresh culture media was added to the mature MDM adhered to culture plates (Nunc, NY, Roche ster, USA) were obtained. Similarly, for producing monocyte-derived dendritic cells (MDDCs), PBMCs (5x10⁶ PBMCs/well) were cultured in culture media at 37°C CO₂ incubator. After overnight incubation, the non-adherent cells were removed by washing and the adherent cells were further cultured with GM-CSF and IL-4 (Gibco, Life technology, Grand Island, NY, USA) for 6 days at 37°C to obtain mature MDDCs. Non-adherent lymphocytes were removed by washing and fresh culture media was added to the mature MDDC. Both MDM and MDDC were cultured with or without LPS (5.0 μg/ml) for 48 hours. Extracellular fluid was collected from both cell types, followed by treatment with 0.1% saponin in RPMI to release intracellular chemokines as performed earlier (Lore et al. 1998; Mily et al. 2013; Mily et al. 2015; Raqib et al. 2014) . Both extra-and intra-cellular fluid were combined and stored at −80°C for analyzing chemokines.

The Bio-Plex 200 system (BioRad Laboratories, Inc. Hercules, CA, USA) was used to measure chemokines. Bio-Plex Pro Human Cytokine 1-8 plex (MIP-1β (CCL4), MCP-1 (MCAF), CCL5 (RANTES), CXCL10 (IP-10), IFN-γ, IL-8, TNF-α, IL-1β) (Bio-Rad) kit was used for both MDM and MDDC culture supernatants to determine the concentration of 9 chemokines, while Bio-Plex Pro Human Chemokine 1plx EXP (x-Plex and Express Assay), was used as a single plex set to assess CCL-22 concentration in MDDC culture supernatant only.

Adhered MDMs and MDDCs (above) were harvested from the 4-well culture plate (NUNC) using cold EDTA in PBS by cell scraper (CELLTREAT Scientific Products, Pepperell, MA). The collected MDMs and MDDMs were washed and taken for phenotyping using fluorophores tagged cell surface markers (CD14⁺ for MDM and CD80⁺CD1a⁺CD83⁺ for MDDC) with BD Accuri C6 cell sorter; the recovered cells were generally >90% mature monocytes and dendritic cells respectively.

2.8. Anthropometric and Demographic information (Covariates)

A structural questionnaire was used to collect the socioeconomic and sociodemographic features of the study participants. Trained field staff performed anthropometric assessments of the women. Standard approach was used to collect the household biomass/air pollution exposure related data including number and types of stoves used, fuel usage pattern (natural gas, dried leaf, jute straw and pressed wood powder), duration of cooking per day, types of kitchen used on the days of exposure assessment (enclosed, semi-enclosed and open), and ventilation in kitchen The complete covariate data for all participants were available. In sensitivity analysis, a significant difference between the two study sites was found in PM_2.5 and BC exposure levels. To examine the possible interaction effects of HAP exposures and locality (Matlab and Araihazar) with respect to outcomes (subtypes of T and B cells and cytokines), multivariate regression analysis was performed (a value <0.10 was considered significant for p-for interaction). Except for T regulatory cells, no effect modification of HAP exposures and locality was found on the outcomes (B and T cell subsets, cyto/chemokines derived from macrophages and dendritic cells), thus locality-based stratification was not evaluated.

2.9. Statistical analyses

For data visualization, several statistical plots such as normal k-density curve, probability of skewness and kurtosis and q-q plots, scatter diagram were used. Data were presented as descriptive statistics as either mean, with standard deviation, median with inter-quartile range, frequencies, or percentages. Bivariate association between exposures (PM_2.5, BC, CO), potential covariates (age, BMI, overall duration of cooking (years), duration of cooking per day (hours), ventilation surface area of the kitchen) and outcomes (immune cells, , and chemokines) were assessed using either Spearman’s rank correlation, Mann-Whitney U-test, or Kruskal-Wallis, depending on the type of data (continuous or categorical). All associations and linearity between the HAP exposure variables (PM_2.5, BC and CO) and outcomes (B and T lymphocyte subpopulations, chemokines) were examined graphically using Lowess moving average. To check the multicollinearity of the exposures on outcomes in the model, Ridge regression and ordinary least square multicollinearity diagnostic test was performed. The variance influence factors (VIF) was below 2 and Farrar-Glauber multicollinearity Chi-square and F-test were below 1 showing no multicollinearity. For linear associations between exposure biomarkers and different outcomes, multivariate adjusted linear regression model was applied. In the multivariate regression analysis, all exposure markers (PM_2.5, BC, CO) were left skewed thus were log₂-transformed excluding 3 outliers from PM_2.5 and 2 from BC and CO each, in order to obtain normally distributed residuals in the models. The residual distribution in each model was checked using residual versus fitted plot before transformation of the exposure. Multivariate linear regression models were constructed to test the associations between air pollution (log₂ transformed PM_2.5, BC, CO) and immunological markers. In the regression analyses, log₂-transformation simplifies the interpretation of the beta-coefficients i.e. average changes in immunological outcome associated with each doubling of HAP exposure. Various covariates were assessed to select the confounding variables and in the final model, age, BMI, years of cooking, and duration of cooking per day were adjusted to prevent the confounding effects of these covariates. The statistical analyses were performed with Stata 15 (StataCorp, LP, College Station, Texas, USA) and SPSS (version 22.0).

3. Results

3.1. Study population and air pollution exposures

The age range of women enrolled in the study was 25-55 years. On average, women spent more than 2.5 hours per day for 17.42 years for cooking (including meal preparation) in the kitchen. About 92% households had cooking spaces outside the main house or living quarters, with 42% having closed kitchen rooms with windows and >50% having semi-open kitchen (with 2 or 3 sides open) in the courtyard. In 21% of the households, LPG based stove was available that were rarely/seldom used because the households could not afford the monthly cost of LPG use; it was limited to occasional warming of baby food or meal in the night (Supplemental Table S1).

The exposure levels assessed through MicroPEM and CO data logger were 48-hours’ measures of PM_2.5, BC and CO at individual level. HAP data was used for up to 48 hours for uniformity, as in most participants, HAP could not be collected up to 72 hours due to stopping of the air sampling device. The average time gap between HAP measurement and blood collection was 7.6±3.7 days. The mean of two HAP exposure levels measured at dry and wet seasons on each participant, were used for all assessments to minimize the seasonal effects Exposure data for individual sites showed that both PM_2.5 and BC levels were significantly higher in Araihazar than in Matlab; however CO levels were similar in the two study sites (Table 1).

Table 1.

Personal air pollution exposure data in Matlab and Araihazar study areas.

Air pollutants	Total participants (n=200)	Matlab (n=100)	Araihazar (n=100)	p-value
PM2.5, μg/m3	134.5 (103.4, 172.3)	124.1 (94.0, 160.9)	145.6 (113.8, 180.6)	0.030
BC, μg/m3	6.04 (4.84, 7.58)	5.08 (3.98, 5.91)	7.58 (6.28, 8.78)	<0.001
CO, ppm	0.97 (0.62, 1.35)	0.86 (0.58, 1.27)	1.02 (0.66, 1.42)	0.072

Data was presented as median with inter quartile range (IQR). Wilcoxon signed rank test was used to estimate the p-value. Particulate matter 2.5, PM_2.5; Black carbon, BC; Carbon monoxide, CO

3.2. Association of air pollution with immunoprofile of lymphocytes

The frequency distribution of different subpopulations of T and B lymphocytes from women are given in Supplemental Table S2. The distribution pattern shows that among the B cell subpopulations, the largest fraction was memory B cells (CD19⁺CD27⁺) in women We used scatter plots to visually inspect the associations between the HAP exposure markers (PM_2.5, BC and CO; log₂ transformed) and immune cell outcomes (T and B cells) (Fig 2). There were no indications of nonlinear associations. Thereafter, we applied multivariate-adjusted linear regression analysis by modelling the associations (Fig 2) (Table 2). A doubling of both PM_2.5 (β=−3.12, 95% CI=−5.85, −0.38) and BC (β=−4.07, 95% CI=−7.96, −0.17) were significantly associated with reduction in memory B cell fractions (CD19⁺CD27⁺) in women . Likewise, a doubling of CO exposure was significantly associated with 1.18% reduction in CD19⁺ B lymphocyte counts (95% CI=−2.36, −0.01) in women. Contrarily, a doubling of PM_2.5 exposures was significantly associated with small fractional increases in CD19⁺CD38⁺ immature B cells (0.83%) and CD19⁺CD27⁺CD38⁺ plasmablast (0.79%)2). No significant associations were noted between HAP exposure and different T cell subsets. To investigate possible interaction effects of HAP exposures and locality (Matlab and Araihazar) with respect to outcomes (subtypes of T and B cells and cyt/chemokines), multivariable regression analysis was performed (a p-value <0.10 was considered significant). A significant positive association was obtained between BC exposure in Matlab (β=1.30, 95%CI=- 0.38, 2.21; p=0.006; p for interaction=0.004) and Treg cells. The association between BC exposure and Treg was plotted using scatter plots with Lowess smoothing line to explore potential non-linear relationship (Fig S2). We found that up to BC concentration of 5.6 μg/m³ (median BC levels in Matlab was 5.08 μg/m³), significant positive association was found between BC and Treg cells. After that point, the association did not remain significant. No such effect modification of HAP exposures and locality was found on other outcome variables.

Fig. 2.

Scatter/regression plots showing the associations between exposure to air pollutants (PM_2.5, BC and CO) and lymphocyte subpopulations (A-L). The multivariate linear regression model was applied to estimate the p-value and the model was adjusted by age, BMI, years of cooking, duration of cooking per day, ventilation surface area of the kitchen, locality and seasonality. M, N, O plots represent the associations of PM_2.5, BC and CO exposure respectively with lymphocyte subpopulations in women.

Table 2.

Association between exposure to air pollutants (PM_2.5, BC and CO) and lymphocyte subpopulations

Lymphocytes	Women n=198
	β(95% CI)	p-value	1P for Interaction	2P for Interaction
CD8+ T cells
PM2.5 (n=197)	−0.84(−2.56, 0.87)	0.334	0.535	0.667
BC (n=198)	−0.12(−2.51, 2.27)	0.922	0.757	0.734
CO (n=198)	0.21(−0.94, 1.36)	0.720	0.562	0.846
CD4+ T cells
PM2.5 (n=198)	0.44(−1.06, 1.94)	0.565	0.919	0.520
BC (n=197)	0.81(−1.27, 2.90)	0.443	0.589	0.597
CO (n=197)	0.82(−0.18, 1.81)	0.107	0.922	0.609
T regulatory cells
PM2.5 (n=198)	0.02(−0.44, 0.49)	0.925	0.129	0.483
BC (n=197)	0.46(−0.18, 1.11)	0.156	0.004	0.898
CO (n=197)	0.01(−0.30, 0.32)	0.964	0.106	0.600
CD19+ B cells
PM2.5 (n=198)	0.46(−1.35, 2.27)	0.616	0.627	0.336
BC (n=197)	0.33(−2.24, 2.89)	0.801	0.614	0.731
CO (n=197)	−1.18(−2.36, −0.01)	0.048	0.200	0.213
Early plasma cells
PM2.5 (n=198)	0.83(0.02, 1.78)	0.047	0.533	0.900
BC(n=197)	0.06(−1.41, 1.30)	0.931	0.161	0.459
CO (n=197)	0.12(−0.51, 0.74)	0.716	0.757	0.556
Plasmablasts
PM2.5 (n=198)	1.63(1.03, 2.24)	0.044	0.131	0.971
BC (n=197)	0.45(−0.32, 1.22)	0.588	0.151	0.655
CO (n=197)	0.27(−0.17, 0.71)	0.709	0.205	0.123
Memory B cells
PM2.5 (n=198)	−3.12(−5.85, −0.38)	0.026	0.695	0.098
BC (n=197)	−4.07(−7.96, −0.17)	0.041	0.727	0.881
CO (n=197)	0.71(−1.12, 2.54)	0.447	0.105	0.514

Multivariate regression model was used to estimate the p-value. The multivariate regression model was adjusted by age, body mass index (BMI), period of cooking, ventilation surface space of the kitchen, and locality (Matlab and Araihazar).

¹Interaction effects of HAP and ¹locality and

²seasonality on lymphocyte subpopulations.

3.3. Association of air pollution with plasma immunoglobulin E

The multivariate regression associations of HAP exposure markers (PM2.5, BC & CO) with total plasma IgE concentrations are shown in Fig S3. A positive association between CO exposure and plasma IgE was found; a doubling of CO exposure was significantly associated with 240.4 unit increase in plasma total IgE levels (95% CI=3.06, 477.8; p=0.047). No associations were observed between the other HAP exposure markers and IgE levels.

3.3. Association of air pollution exposure with chemokines from macrophage and dendritic cells

Concentration of cyto/chemokines secreted by MDM and MDDC into the culture supernatants were measured in a sub-group of 80 Bangladeshi women. The concentrations of HAP exposure markers or other demographic features in the sub-group did not differ significantly with those of the total population of 200 women (Supplemental Table S4).

Ex vivo concentrations of chemokines secreted by MDM and MDDC in the cell culture supernatants are provided in Supplemental Table S3. To visually examine the associations between the HAP exposure markers and chemokines, we used scatter plots between HAP markers and chemokines secreted from MDM as well as from MDDC (Fig. 3.). Multivariate-adjusted linear regression model was performed to assess the effect of HAP markers on chemokine expression after adjusting with covariates (Table 3) (Fig. 3.). The analysis showed a positive association between PM_2.5 exposures and CXCL10 (IP-10) levels produced by macrophages, whereby a doubling of PM_2.5 was associated with 12287 pg/ml increase in CXCL10 levels (95% CI=1038, 23536). Again, a doubling of CO exposure was associated with 835.7 pg/ml increase in CCL5 (RANTES) levels (95% CI=95.5, 1576).

Fig. 3.

Scatter/regression plots showing the association of air pollutants (PM_2.5, BC and CO) exposure with chemokines secreted from macrophages (A-I). Multivariate linear regression model was applied to estimate the p-value and the model was adjusted by age, BMI, years of cooking, duration of cooking per day, ventilation surface area of the kitchen, locality and seasonality. 3 J-L represent the associations of air pollutants exposure and macrophage secreted chemokine levels.

Table 3.

Association of PM_2.5, BC and PM with chemokines secreted from macrophages and dendritic cells.

	Macrophages (n=80)			Dendritic cells (n=80)
Cyto/chemokines	1β (95% CI) a	p- value	1P for interaction	β (95% CI)	p-value	1P for interaction
CCL5
PM2.5	417.3(−671.4, 1506)	0.447	0.133	−806.3(−2901, 1288)	0.445	0.602
BC	−1201(−2699, 297.7)	0.115	0.320	−3583(−6358, −807.8)	0.012	0.928
CO	835.7(95.5, 1576)	0.027	0.237	167.2(−1301, 1636)	0.821	0.426
CXCL10
PM2.5	12287(1038, 23536)	0.033	0.415	−5321(−16327, 5686)	0.338	0.997
BC	2517(−13673, 18707)	0.757	0.367	−10177(−25443, 5090)	0.188	0.270
CO	4545(−3528, 12618)	0.265	0.925	4569(−2123, 12262)	0.240	0.137
IFN-γ
PM2.5	296.8(−100.7, 694.3)	0.141	0.676	112.1(−129.5, 353.6)	0.358	0.125
BC	250.4(−309.5, 810.2)	0.376	0.571	−78.3(−416.7, 260.1)	0.646	0.947
CO	209.2(−69.4, 487.9)	0.139	0.754	137.6(−29.7, 304.8)	0.105	0.867
IL-8
PM2.5	12775(−21311, 46862)	0.457	0.846	−5088(−11370, 1194)	0.111	0.208
BC	11558(−36097, 59214)	0.630	0.620	−1523(−10437, 7390)	0.734	0.528
CO	16213(−7476, 39903)	0.177	0.718	3915(−473.8, 8303)	0.080	0.404
TNF-α
PM2.5	−31925(−75707, 11857)	0150	0.465	−1363(−11346, 8620)	0.786	0.126
BC	−3610(−65568, 58347)	0.908	0.484	−15521(−28968, −2074)	0.024	0.105
CO	9994(−21064, 41052)	0.523	0.883	−4146(−11082, 2789)	0.237	0.191
CCL4 (MIP-1β)
PM2.5	−368.6(−3155, 2417)	0.793	0.583	−74.2(−1593, 1445)	0.923	0.511
BC	−2857(−6686, 973.3)	0.141	0.168	−2012(−4077, 54.0)	0.056	0.712
CO	−239.9(−2194, 1714)	0.807	0.473	131.5(−933.4, 1196)	0.806	0.389
IL-1β
PM2.5	−111.0(−259.4, 37.4)	0.140	0.140	−311.7(−711.4, 88.0)	0.124	0.779
BC	21.4(−188.8, 231.5)	0.840	0.155	−124.8(−690.8, 441.2)	0.662	0.265
CO	52.4(−52.6, 157.3)	0.323	0.192	−106.9(−390.7, 176.8)	0.455	0.461
MCP-1
PM2.5	9882(−16967, 36730)	0.466	0.334	−83.7(−754.9, 587.5)	0.804	0.181
BC	−227.0(−37819, 37365)	0.990	0.896	−83.4(−1020, 853.2)	0.860	0.494
CO	7457(−11358, 26272)	0.432	0.450	381.7(−80.5, 843.9)	0.104	0.267
CCL-22
PM2.5	-	-		226.0(−110.5, 562.5)	0.185	0.101
BC	-	-		247.5(−224.1, 719.2)	0.299	0.843
CO	-	-		82.8(−155.3, 320.9)	0.490	0.699

Multivariate regression model was used to estimate the p-value. The multivariate regression model was adjusted by age, BMI, period of cooking, ventilation surface space of the kitchen, seasonality and locality (Matlab and Araihazar).

¹Interaction effects of HAP and locality on chemokines.

Adjusted regression models of association between HAP exposure biomarkers and MDDC secreted chemokines showed that, a doubling of BC exposure (μg/m³) was associated with reductions of 3,583 pg/ml of CCL5 (95% CI=−6358, −807.8), 15,521 pg/ml of TNF-α (95% CI=−28968, −2074) and 2,012 pg/ml of CCL-4 (95% CI=−4077, 54.0) in the rural women. No associations were observed between the HAP exposure markers and concentrations of other chemokines. No effect modification of HAP exposures and locality was found on the chemokine concentrations.

4. Discussion

We have recently reported that the average exposure to PM_2.5 among rural Bangladeshi women was much higher than the WHO recommended levels (10 μg/m³) (Shahriar. MH et al. 2021). In this study we have shown that exposure to household air pollution in women was associated with increased fractions of naïve/immature B lymphocytes and plasma cells, reduced fractions of memory B cells without any notable changes in T lymphocyte subsets, increased concentration of plasma IgE, reduced concentration of cyto/chemokines (CCL5 and TNF-α) secreted by dendritic cells and elevated levels of inflammatory cyto/chemokines (CCL5 and CXCL10) produced by macrophages.

Reports on effects of HAP exposure on immune function, particularly in relation to BMF combustion-generated air pollution are scarce. The available studies show conflicting results, and are limited to particulate matter (PM_2.5 and PM₁₀) only. In a study in rural West Bengal in India, women cooking solely with BMF for >5 years exhibited association of PM_2.5 with reduction in CD4⁺T and CD19⁺B cell fractions, and increments in CD8⁺, Treg and CD16⁺CD56⁺NK cell fractions (Dutta et al. 2012). We found that Treg cells were positively associated with BC concentrations up to a threshold level of 5.6 μg/m³, beyond which the association was no longer significant. It is possible that at higher BC concentration, the regulatory control of Treg cells to maintain self-tolerance and suppress chronic inflammatory response wanes. In a similar setting in India, assessment of effects of indoor air pollution (PM₁₀ and PM_2.5) revealed increased fractions of CD4⁺ and CD19⁺ lymphocytes, granulocytes, monocytes and neutrophils in BMF-users compared with women using cleaner LPG for cooking (Saha et al. 2016), which supports our finding of positive association between PM_2.5 and immature B cells and plasmablasts. Additionally, we found substantial decrements in memory B cells that were affected by both PM_2.5 and BC exposure and CD19⁺B cells affected by CO exposure. Reduction in memory B cells and changes in other B cell subsets have not been reported previously in association with various air pollution markers. Memory B cells typically persist for years or a lifetime after the first contact with a pathogen, and can generate rapid antibody response to it in future encounters. Our findings are indicative of lower reserve of memory B cells due to chronic HAP exposure which may dampen adaptive immunity against recurrent and future infections. The non-specific increase in immature B cells and plasmablasts may be a counterbalancing strategy in response to decreased memory B cells and may be indicative of ongoing inflammation. Relationship of smoke or air pollution exposure and IgE-mediated allergy or atopic diseases have been shown in previous studies (Cuinica et al. 2015; Hou et al. 2021; Kaji et al. 2014). Our finding of increased levels of plasma IgE in women exposed to high levels of CO indicates is supported by a large population based study where short-term exposure to CO and PM_2.5 was associated with increased daily hospital visits for IgE-mediated allergy (Hou et al. 2021).

The innate immune system at the lung-environment interface is critical to maintain homeostasis. Air pollutants have been shown to affect the functions of macrophages, including reduced phagocytosis, increased production of pro-inflammatory cytokines and other innate factors, which may eventually result in cardiopulmonary /lung diseases (Bekki et al. 2016; Laskin et al. 2019; Raji et al. 2020). Cytokines and chemokines are crucial intermediaries of the innate and the adaptive immune responses that play major roles in signaling, activation and functioning of antigen presenting cells. Both CCL5 and CXCL10 are inflammatory chemokines that actively recruit leukocyte into inflammatory sites. Increased secretion of these chemokines by peripheral MDM in chronically HAP exposed Bangladeshi women may show indications of persistent activation of innate inflammatory responses. In support of our findings, chronic inhalation of biomass smoke (PM_2.5 and PM₁₀) by Indian women was shown to induce elevated serum levels of pro-inflammatory mediators (cytokines, reactive species/free radicals), as well as reduced anti-oxidants compared to women who cooked with cleaner LPG (Banerjee et al. 2012). In contrast to our findings, long term exposure to cooking smoke in healthy Malawian adults was associated with increased particulate content of alveolar macrophages with concomitant reduction in macrophage function (oxidative burst, phagocytosis of microbes, inflammatory cytokines), reflecting impaired innate immune defense against infections (Rylance et al. 2015). The discrepancy could be related to variations in concentration of pollutants and cellular source of chemokines.

Dendritic cells are professional antigen-presenting cells and key orchestrators of immune responses. The capacity of DCs to initiate immune responses depends on their expression of chemokines and chemokine mediated interaction between DCs and T cells. CCL5 functions as a chemoattractant for activated T cells, DC and other leukocytes and are important for homing and migration of effector and memory T lymphocytes to inflamed tissue; TNF-α is important for DC maturation (Crawford et al. 2011). DC-derived CCL5 and TNF-α regulate interactions between Th1/NK cells and DC, to initiate T cell response (Hirata et al. 2010; Palomino-Segura et al. 2019; Summers deLuca and Gommerman 2012). Disruption of DC function may contribute to development or exacerbation of allergic, autoimmune or other diseases. Recent evidence shows that through inhibition of MAPK and NF-κB signaling pathways, PM_2.5 significantly suppresses cytokine production by DC (Arooj et al. 2020; Glencross et al. 2020; Williams et al. 2007). DCs can also transfer unprocessed antigens to immature B cells to initiate antigen-specific antibody responses (Heath et al. 2019). B cells are dominant antigen-presenting cells that activate naive CD4⁺T cells upon infection (Heath et al. 2019; Hong et al. 2018). Altogether in the present study, the reduction of B cell fractions and hypo-responsiveness of DC functions linked to personal HAP exposure, suggest that impaired B cell- and DC-driven activation of T cell trafficking and migration may impede the T-dependent as well as T-independent humoral immunity.

The findings of our study are applicable for all LMIC settings especially where solid biomass fuel is used for domestic purpose such as cooking or heating rooms in winter season for prolonged periods and may significantly contribute to respiratory diseases in the vulnerable population (Budhathoki et al. 2020; Mondal and Paul 2020), Naz and Ghimire 2020); Al-Janabi et al. 2021; Odo et al. 2021; Van Vliet et al. 2019; Woolley et al. 2021).

The study has several strengths. We have applied 3 different exposure markers (PM_2.5, BC and CO) measured through personal exposure monitors to investigate their effects on B and T lymphocyte subpopulations that have not been demonstrated by previous studies. Another unique feature is the assessment of relationship of HAP with functional capacities of antigen-presenting cells (macrophages and dendritic cells). Instead of evaluating cyto/chemokines in serum showing pre-existing levels from diverse cell types, we assessed ex vivo concentration of chemokines actively secreted in culture supernatant from in vivo HAP-exposed antigen-presenting cells. The study has some limitations. We did not carry out a B cell-based functional assay or adaptive immune function to evaluate whether B cell capacity to produce antigen-specific antibodies is reduced in parallel to reduced frequencies of B cells. Recall responses to a vaccine might have revealed more striking differences; assessment of effector and memory T cells in interactions with PM activated-DC in co-cultures would have further given indications of impacts on the functions of subtypes of T cells (Matthews et al. 2016; Pfeffer et al. 2018; Rahmani et al. 2020; Shahbaz et al. 2021).

5. Conclusion

Our findings suggest that chronic HAP exposure adversely affects proportions of B lymphocytes, particularly memory B cells and functions of macrophages and dendritic cells in women. Future studies may investigate mechanisms by which HAP exerts its negative impact, including oxidative stress-mediated damage, the capacities of B cells/plasmablasts to produce antigen-specific antibody responses, functional capacities of cytolytic T cells. Our recently completed LPG based intervention study (NCT02824237 2016) in rural Bangladesh may help to better understand the association of oxidative stress, HAP and immune cells and the impact of intervention on the immune function.

Supplementary Material

Supplementary Materials

Fig. S1. Gating strategies for the characterization of B and T lymphocytes and identification of CD19⁺ B lymphocytes from lymphocyte population (1A shows the lymphocyte population identification based on side-scatter (SSC) and forward-scatter characteristics (FSC) of the cells. 1B represents CD19 expressing B cells in the lymphocyte population from B lymphocyte tube. 1C represents the selection of CD4⁺ T lymphocytes and CD8⁺ T lymphocytes in the lymphocyte population gate from T cell flow tube. In 1D, CD19⁺ B cell population gate is further selected for the expression of CD27, CD38. The following B cell sub-populations have been delineated: CD27⁺ (memory B cells), CD38⁺ (immature B cells) and CD27⁺CD38⁺ (plasmablasts). In 1E, CD4⁺ T lymphocyte population is further selected for the expression of CD25 and CD127 where CD4⁺CD25⁺CD127^low subpopulation is delineated as regulatory T cell (Treg cell). F. shows the selection of B lymphocytes (CD19⁺) based on side-scatter (SSC) characteristics and CD19 expression of lymphocytes. G. represents selection of B lymphocytes (CD3⁻CD19⁺) based on the expression of CD3 and CD19.

Fig. S2. Scatter plots showing the association of black carbone (BC) exposure with proportion of regulatory T cells.

Fig. S3. Association of air pollutants (PM_2.5, BC and CO) exposure with plasma total IgE concentrations. Multivariate linear regression model was applied to estimate the p-value and the model was adjusted by age, BMI, years of cooking, duration of cooking per day, ventilation surface area of the kitchen, locality and seasonality.

Fig. 4.

Scatter/regression plots showing the association of air pollutants (PM_2.5, BC and CO) exposure with chemokines secreted from dendritic cells (A-I). Multivariate linear regression model was applied to estimate the p-value and the model was adjusted by age, BMI, years of cooking, duration of cooking per day, ventilation surface area of the kitchen, locality and seasonality. 4. J-L represent the associations of air pollutants exposure and dendritic cells secreted chemokine levels.