Exposure to Traffic-Related Air Pollution is Associated with Greater Bacterial Diversity in the Lower Respiratory Tract of Children

Background Exposure to particulate matter has been shown to increase the adhesion of bacteria to human airway epithelial cells. However, the impact of traffic-related air pollution (TRAP) on the respiratory microbiome is unknown. Methods Forty children were recruited through the Cincinnati Childhood Allergy and Air Pollution Study, a longitudinal cohort followed from birth through early adolescence. Saliva and induced sputum were collected at age 14 years. Exposure to TRAP was characterized from birth through the time of sample collection using a previously validated land-use regression model. Sequencing of the bacterial 16S and ITS fungal rRNA genes was performed on sputum and saliva samples. The relative abundance of bacterial taxa and diversity indices were compared in children with exposure to high and low TRAP. We also used multiple linear regression to assess the effect of TRAP exposure, gender, asthma status, and socioeconomic status on the alpha diversity of bacteria in sputum. Results We observed higher bacterial alpha diversity indices in sputum than in saliva. The diversity indices for bacteria were greater in the high TRAP exposure group than the low exposure group. These differences remained after adjusting for asthma status, gender, and mother’s education. No differences were observed in the fungal microbiome between TRAP exposure groups. Conclusion Our findings indicate that exposure to TRAP in early childhood and adolescence may be associated with greater bacterial diversity in the lower respiratory tract. Asthma status does not appear to confound the observed differences in diversity. It is still unknown whether the development of asthma changes the lower respiratory tract microbiome or if an altered microbiome mediates a change in disease status. However, these results demonstrate that there may be a TRAP-exposure related change in the lower respiratory microbiota that is independent of asthma status.

it is plausible that the gut and respiratory system are impacted simultaneously by TRAP exposure. 93 Previously, we studied the association between respiratory end environmental microbiome. We 94 found that the environmental microbiome affected the fungi found in saliva but not bacteria found 95 in saliva or sputum [43]. 96 In this study, we examined the association between childhood exposure to TRAP and the 97 microflora in the lower respiratory tract of children. The participants were recruited from a 98 longitudinal cohort, followed from birth through early adolescence, with a well-characterized 99 TRAP exposure history. 6 131 they spent more than eight hours per week, up until age 12, where only the home address was used. 132 For this study, we categorized participants as exposed to high or low levels of TRAP if their 133 average exposure from birth through age 12 was above or below the median exposure of all 134 participants who completed the age 12 study visit.  Immediately prior to sputum induction, participants rinsed their mouths with nuclease-free water, 139 and then 2 mL of saliva was collected using the Norgen Saliva DNA Collection Kit (Norgen 140 BioTek Corp., Thorold, ON, Canada) according to the manufacturer's instructions. Saliva was 141 collected first so that we could compare the bacterial communities in saliva to those in sputum to 142 ensure that the bacterial communities were distinct and that the sputum samples were not entirely 143 contaminated by the oral microbiome. Next, participants underwent a spirometry assessment and 144 received a dose of albuterol. To induce sputum production, participants breathed in a nebulized 145 hypertonic saline solution for five minutes, then coughed the sputum into a Norgen Sputum DNA 146 Collection Kit (Norgen BioTek Corp., Thorold, ON, Canada). This cycle was repeated up to five 147 times, or until 2 mL of sputum had been collected. After adding the Norgen Collection Kit a 100 µg/mL solution of dithiothreitol and incubated at 37ºC for 60 minutes. The manufacturer's 155 protocol was used for DNA isolation of sputum and saliva samples, with one modification: after 156 adding the proteinkinase K and lysozyme, the sample was incubated in an ultrasonic water bath at  213 We accounted for differences in sequencing depth by multiplying the relative abundance 214 values by the qPCR values for each sample to calculate absolute abundance [71,72]. Two sputum 215 samples were removed from the bacterial dataset due to a low number of reads (<5000 total reads).

216
Four other sputum samples and four saliva samples were not included in the bacterial dataset 217 because the rDNA did not amplify either during qPCR or during PCR amplification prior to 16s 218 sequencing. Four saliva samples were not included in the bacterial dataset because the rDNA did 219 not amplify during qPCR or during PCR amplification prior to 16s sequencing. For fungi, saliva 220 samples were not included if they did not amplify or had <5000 reads after sequencing. Due to 221 low fungal abundance in sputum, sputum samples were included if they amplified and had >400 222 reads. These samples were treated as pilot samples. The number of observed ASVs and Shannon alpha diversity were calculated using phyloseq::estimate_richness and Faith's phylogenetic 224 diversity was calculated using the function pd in the picante R package (v1.8.0) [73]. 225 Alpha diversity indices between sample types, between exposure groups, between genders, 226 and between asthma status groups were first univariately compared using the Wilcoxon rank sum 227 test. We used multiple linear regression to model the effect of TRAP, adjusted for gender, asthma 228 status, and mother's education as a measure of socioeconomic status, on the alpha diversity in 229 sputum. TRAP exposure was modeled both as a categorical (high/low) and continuous (ECAT) 230 variable (high/low) in separate models. The overall bacterial abundance from qPCR was compared 231 between TRAP exposure groups, asthma status groups, and genders using the Wilcoxon rank sum 232 test.

233
The Adonis function in the vegan package (2.5-6) was used to implement a permutational 234 multivariate analysis of variance to test for differences in beta diversity between sample types. A 235 dispersion test for the homogeneity of variance across sample type was also conducted using the

246
Sputum samples from 34 of the participants were included in the bacterial analyses. There 247 were 17 participants in each TRAP exposure group for bacteria. The median ECAT in the high 248 exposure group was 0.46 µg/m 3 , and the median ECAT in the low exposure group was 0.29 µg/m 3 .

249
The cut-off between the low and high exposure groups was ECAT = 0.33 µg/m 3 . In the high 250 exposure group, 42% of the participants were female, 35% were asthmatic, and 82% of the mothers 251 had education beyond high school. In the low exposure group, 53% were female, 18% were 252 asthmatic, and 100% of the mothers had education beyond high school. were female, 50% were asthmatic, and 90% had mothers with education beyond high school. were female, 37.5% were asthmatic, and 100% had mothers with education beyond high school.
As so few samples returned amplified fungal DNA sequence reads, the fungal results discussed 267 are considered pilot samples.  We used the Bray-Curtis dissimilarity metric to compare the microbial community 284 composition between the two sample types for bacteria (Fig. 3). The Adonis test indicated that 285 6% of the variance in the distance matrix between sputum and saliva could be attributed to the 286 sample type (p=0.001). The homogeneity of dispersion test failed to reject the null hypothesis that   There was greater diversity in the sputum of the high TRAP exposure group when 298 compared to the low exposure group (Fig. 4). Univariate analysis showed that the number of 299 observed ASVs, Shannon diversity, and Faith's phylogenetic diversity were all greater in the high 300 TRAP exposure group for bacteria (Table 1). For phylogenetic diversity, there was also a 301 statistically significant difference between genders (Table 1,    phylogenetic diversity after adjusting for asthma status, gender and mother's education (Table 2).

314
TRAP as a categorical variable was positively associated also with Shannon diversity. Female The Bray-Curtis dissimilarity measure figures did not show a distinction in bacterial 322 microbiota in sputum between TRAP exposure group, asthma status group, or gender (Fig. 5).

323
However, the Adonis test indicated that 6% of the variance in the distance matrix between the 324 sputum of asthmatics and non-asthmatics could be attributed to asthma status (p=0.04). The Adonis   According to the xdc.sevsample test, the distribution of major phyla did not differ in 336 sputum between each TRAP exposure group nor each gender (p=0.43 and p=1.0, respectively).

337
However, a significant difference in major phyla was found between asthma status groups 338 (p≤0.001). These results are in agreement with the relative abundance bar plots ( Fig. 6 and S5 Fig).

339
The relative abundance of phyla in sputum does not appear to differ between TRAP exposure 340 group nor gender, but asthmatics appear to have more Bacteroidetes and fewer Proteobacteria.  While the overall microbial community composition did not appear to differ between 346 TRAP exposure groups, negative binomial regression was able to identify several individual ASVs 347 with a log2 fold-change greater than 2 and FDR p≤0.05 according to TRAP exposure groups (high 348 vs. low) (Fig. 7). Fusobacterium nucleatum had a log2 fold change of 25 (FDR p≤0.001). Two   Overall Bacterial Load 368 We compared the overall bacterial load in sputum, as measured by qPCR with universal 369 bacterial primers, between TRAP exposure groups, between asthmatics and non-asthmatics, and 370 between genders using Wilcoxon rank sum test. No significant difference in the overall bacterial 371 load was observed between high and low TRAP exposure groups (median high TRAP= 6.9 x10 5 372 bacterial genome copies per mL of sputum, median low TRAP=3.6 x 10 5 bacterial genome copies 373 per mL of sputum; p=0.43) (Fig. 8). However, there was a higher bacterial load in the sputum of asthmatic participants than in non-asthmatic participants (median asthmatic=1.3 x 10 6 bacterial 375 genome copies per mL of sputum, median non-asthmatic=3.6 x 10 5 bacterial genome copies per 376 mL of sputum; p=0.07) and in the sputum of male than in female participants (median male=9.4 x 377 10 5 bacterial genome copies per mL of sputum, median female=1.7 x 10 5 bacterial genome copies 378 per mL of sputum; p=0.006) (Fig. 8). We also found a higher bacterial load in the saliva of 379 asthmatics (median asthmatic=4.0 x 10 6 10 5 bacterial genome copies per mL of saliva, median   388 We compared the sputum mycobiome of high and low TRAP-exposed participants, as well as 389 by gender and asthma status. The median ECAT for the high exposure group was 0.40 mg/m 3 and 390 the median for the low exposure group was 0.29 mg/m 3 . There was an overlap between confidence 391 intervals for all three alpha diversity measures when comparing TRAP exposure groups, genders, 392 and asthma status groups (S1 Table). There also were no significant differences in beta diversity 393 (Bray-Curtis) between asthma status groups (p=0.35), between TRAP exposure groups (p=0.94), 394 nor between genders (p=0.23), using the Adonis test. The xdc.sevsample test showed a difference 395 in the distribution of major fungal classes between asthma status groups (p=0.009), and between 396 gender (p<0.001), but not between TRAP exposure groups (p=1.0). It should be noted, however, 397 that due to overall low fungal abundance in sputum, we had a very small sample size.

400
The most abundant bacterial phyla in both sputum and saliva samples were Actinobacteria, 401 Firmicutes, Bacteroidetes, and Proteobacteria. The major phyla identified in the sputum and saliva  As previous studies have shown that asthmatics tend to have greater alpha diversity in their 418 respiratory tracts than non-asthmatics, and exposure to TRAP is associated with a higher incidence 419 of asthma, asthma status was examined as a potential confounder [33]. We compared the diversity 420 indices between asthmatics and non-asthmatics. While the mean number of observed ASVs, asthmatics, the confidence intervals of asthmatics and non-asthmatics overlapped in each alpha 423 diversity measure. In contrast, the confidence intervals of the high and low TRAP-exposure groups 424 did not overlap in the mean number of ASVs nor the mean phylogenetic diversity. These results 425 are consistent with the Wilcoxon rank sum test p-values and the diversity measure box plots. Thus, 426 asthma status does not appear to have a significant impact on our results of overall diversity 427 comparisons between high and low TRAP exposure groups. However, as there were far fewer 428 asthmatics than non-asthmatics included in the study, it is difficult to make a clear determination 429 of differences in relative abundance between the asthma status groups. We also observed greater with higher levels of specific taxa found in the gut of males, and overall higher bacterial diversity 435 found on the skin of males [78,79]. In contrast to sputum, we did not observe any significant 436 differences in the alpha diversity indices of saliva between TRAP exposure groups, gender, or 437 asthma status groups. This further supports the notion that the bacterial microbiome in the sputum 438 samples is distinct from that in the saliva samples.

439
The results of the multiple regression model were consistent with the results of the 440 univariate analysis (Table 2). Neither asthma status nor mother's education were significant 441 predictors of alpha diversity indices, further supporting that asthma status did not have a significant  [88]. As TRAP has been shown to increase the adhesion of bacteria to respiratory 464 tract epithelial cells, it is possible that the alteration in growth conditions in the TRAP-exposed 465 respiratory tract could promote the development of a more diverse bacterial community. 466 Additionally, this change in microbial communities may elicit a local immune response, or even impact immune system development in children. Future studies should examine the relationship 468 between bacterial diversity in the lungs and respiratory health.

469
There was not a significant difference in beta diversity between the TRAP exposure groups.  While we cannot meaningfully remark on the biological significance of these findings, these may 491 be taxa of interest in future investigations regarding TRAP exposure and the respiratory 492 microbiome. We also identified differentially abundant ASVs across gender and asthma status. Our pilot study results did not indicate a difference between the mycobiomes of sputum in 501 the high and low TRAP-exposed participants. This could be because we had a small sample size 502 and the samples that did amplify had very low abundance. Additionally, fungi do not proliferate 503 to the same extent as bacteria in the lower respiratory tract. When comparing bacteriomes, we had 504 enough samples to conduct a linear regression to determine the association between alpha diversity 505 and TRAP exposure, both as a categorical variable (high vs. low) and continuous variable (ECAT), 506 while adjusting for gender, asthma status, and socioeconomic status. Unfortunately, because of the 507 small sample size we could not do the same for the mycobiome. Other sequencing methods besides 508 marker gene analysis, such as shotgun metagenomics, may be better choices for these types of 509 samples and should be explored further in future studies.

511
A major strength of this study is the well-characterized TRAP exposure history of the 512 CCAAPS cohort. Due to this characterization, we were able to examine the effect of TRAP 24 513 exposure on bacterial diversity in the lower respiratory tract as both a categorical and continuous 514 variable. However, we were limited by the small sample size. Additionally, while the data support 515 that asthma status did not significantly impact our bacterial results, it must be noted that we had 516 very few asthmatics compared to non-asthmatics in this study. We also did not have the specific 517 endotypes of the asthmatic participants. Previous studies have shown that specific asthma 518 endotypes may impact the respiratory microbiome in different ways [42].

519
Another strength of this study is that we conducted both 16s and ITS metagenomic sequencing 520 and qPCR with universal bacterial and fungal primers on all samples. Therefore, we were able to 521 normalize our sequencing results with the qPCR data instead of relying on statistical methods, 522 such as rarefaction, to account for sequencing depth. One limitation of the qPCR method is that 523 for both bacteria and fungi, species contain variable numbers of the amplified genomic region [56]. 524 Therefore, the measurement of the total number of bacterial and fungal genome copies per mL of 525 sputum may be affected by the species present in the sample. 526 We selected the induced sputum method over bronchoalveolar lavage because it is less 527 invasive. Therefore, oral contamination of the sputum samples was a major concern in this study.

528
However, one strength of this study is that our results demonstrated that the saliva and sputum 529 samples had distinct bacterial communities, indicating that the sputum samples were not entirely 530 contaminated by the oral microbiome. It should also be noted that the lungs have a wide range of 531 microgeographic conditions, with a temperature gradient from ambient air temperature to body 532 temperature in the short distance from the point of inhalation to the alveoli, so future studies may 533 want to focus on specific locations within the lower respiratory tract.

535
These findings indicate that exposure to TRAP in early childhood and adolescence is 536 associated with greater diversity in the lower respiratory tract in our sample of participants. It is 537 still unknown whether the development of asthma changes the lower respiratory tract microbiome 538 or if an altered microbiome mediates a change in disease status. However, these results 539 demonstrate that there may be a TRAP-exposure-related change in the lower-respiratory 540 microbiome that is independent of asthma status. We also identified several taxa of interest for 541 future studies, including Fusobacterium, Atopobium, and Prevotella. A major limitation of this 542 study was the small sample size, so a larger pool of participants is needed to confirm our findings.

543
Additionally, a study with a larger sample size could use model-based approaches for a more robust