Multiomic characterization of DNA released after airway exposure to allergens and nanoparticles reveals selective export of repetitive and fragile site DNA

Abstract

Allergen and nanoparticle exposure generates reactive oxygen species (ROS) that endanger genome integrity, yet airway epithelial cells survive repeated exposures. Here we show that environmental allergens from Alternaria alternata and polyethyleneimine (PEI) nanoparticles induce oxidative stress, triggering rapid release of double-stranded DNA fragments enriched for repetitive elements (SINEs, LINEs, and LTRs), centromeric sequences and late-replicating, origin-poor fragile site containing genes. Nanopore sequencing along with a multi-omics approach were employed to analyze fragments of extracellular DNA (eDNA) in human and mouse airway epithelial cells and bronchoalveolar lavage fluid from mice. We show that eDNA release following Alternaria or PEI exposure produces similar genomic sequence profiles suggesting that a mechanism exists to selectively export these fragments out of the nucleus and into the airway lumen. However, epigenomic modifications localized around transcriptionally active regions differed between Alternaria and PEI exposures, with potentially lasting implications for gene expression and function in airway epithelia. We propose that export of DNA fragments is selective to reduce repetitive DNA fragment accumulation and secondary structure formation in the nucleus, thus ensuring airway epithelial cell survival.

New & Noteworthy

Exposure of airway epithelia to environmental allergens or nanoparticles invokes a novel genome-protective physiological response to oxidative stress. Fragments of double stranded genomic DNA released into the airway lumen originate from repetitive, transcriptionally active, and fragile sites within the genome. Exporting DNA with high repetitive content potentially aids cell survival through the preservation of genome integrity and fidelity during the repair process.

Graphical Abstract

Introduction

Alternaria alternata is a clinically important driver of allergic airway inflammation and asthma worldwide (1, 2). Sensitized individuals experience more severe asthma exacerbations, more frequent emergency room visits, and a greater need for asthma therapeutics. Alternaria is among the most consequential outdoor fungi for asthma, with sensitization linked to hospitalization and potentially fatal respiratory distress (1). Similarly, nanoparticle exposure triggers oxidative stress and inflammation in airways, disrupting epithelial barrier function, and reducing mucociliary clearance. In allergen exposed individuals, oxidative stress augments sensitization, strengthens Th2/Th17 polarization, and contributes to asthma exacerbation (3, 4). Consequently, while increases in global temperature, wildfires, atmospheric CO₂ and extreme weather events continue to intensify, the collective impacts of allergenic spores and oxidative stress-inducing nanoparticles will likely increase the occurrence, severity and economic burden of allergic airway diseases.

Previously we showed that airway exposure to Alternaria induced oxidative stress, ATP release, and a sustained increase in intracellular calcium concentration ([Ca²⁺]_i) in human airway epithelial cells (5). PEI exposure produced a similar ATP release and Ca²⁺ mobilization response (3). The increase in [Ca²⁺]_i after Alternaria treatment stimulated furin activity, a proprotein convertase enzyme that produced noncanonical cleavage and activation of procaspase 3, which subsequently induced DNA fragmentation. A portion of these DNA fragments were shown to be released from the nucleus, across the apical membrane and into the airway lumen (6), thus preventing activation of intracellular DNA sensors. Termination of DNA fragmentation and release occurred within a few hours following calpain activation, which effectively preserved cell viability. These findings demonstrated that Alternaria-induced mobilization and release of DNA fragments is regulated and that epithelial cells are able to engage genome repair mechanisms to ensure cell survival.

The initial characterization of eDNA fragments revealed a range of sizes from ~1000 - ~5000 bp indicating that eDNA did not form long continuous filaments capable of functioning as extracellular traps. These fragments strongly deviated from the classical small fragments associated with apoptosis, and the export of eDNA is not associated with cell death but instead cell viability. Furthermore, our initial study suggested that eDNA was not composed of genomic sequences that were randomly exported out of the cell. In the present study we tested the hypothesis that Alternaria and PEI exposure activated a mechanism capable of selecting specific dsDNA sequences that would facilitate DNA repair and contribute to sustained cell viability during exposure. Here, we show that Alternaria and PEI induce a remarkably similar pattern of eDNA composition from human bronchial epithelial cells (hBECs), mouse airway epithelial cells (mAECs) and in vivo mouse airways, comprised of repetitive elements (LINES, SINES and TLRs), centromeric sequences and late-replicating, origin-poor fragile-site genes that are not essential for epithelial cell viability.

Materials and Methods

Alternaria alternata extract was purchased from Greer Laboratories (Lenoir, NC). Polyethyleneimine (PEI), L-glutathione (GSH) and N-acetyl-L-cysteine (NAC) were purchased from Sigma-Aldrich (St. Louis, MO, USA). Hanks balanced salt solution (HBSS), with or without calcium, was purchased from Thermo Fisher Scientific (Waltham, Massachusetts, USA). Caspase-3 was purchased from Santa Cruz Biotechnology (Santa Cruz, California, USA), and Caspase-3 dye was purchased from Sartorius (Göttingen, Germany).

Cell Culture

Primary human bronchial epithelial cells (Lonza, Basel, Switzerland) were grown in bronchial epithelial cell growth medium with growth factor supplements (PromoCell, Heidelberg, Germany). Cells were immortalized after transfection of genes encoding cyclin-dependent kinase-4 and human telomerase reverse transcriptase (60). Mouse primary tracheal epithelial cells were grown in complete epithelial cell medium with accompanying frozen supplement kit (Cell Biologics, Chicago, IL, USA). All cells were grown at 37°C in a humidified atmosphere of 5% CO2 in air. Cells were grown on 6-well plates for eDNA collection, 96-well plates for oxidative stress and DNA release measurements, and 2-well chamber slides (Laboratory-Tek, VWR International, Chicago, IL, USA) for caspase-3 assays.

Oxidative Stress Measurements

The generation of Reactive Oxygen Species (ROS) was measured utilizing a ROS Assay kit (Cell Biolabs, San Diego, CA, USA) and cells grown to confluence on a 96-well plate. Cells were preincubated for 30 minutes with NAC, GSH, or HBSS alone. Fluorescent intensity was continuously measured at 480nm/530nm every minute for 10 minutes after Alternaria or PEI extract in HBSS was added to cells by a Flexstation 3 Microplate Reader (Molecular Devices, San Jose, CA, USA). HBSS alone was used as a control.

DNA Release Measurements

The kinetics of eDNA release was measured utilizing the SpectraMax Quant dsDNA Assay kit (Molecular Devices) and a confluent 96-well plate of cells. Fluorescent intensity was continuously measured by the Flexstation 3 Microplate Reader at 350nm/460nm every 5 minutes for 2 hours. HBSS alone was used as a control.

Caspase-3 Activity

Cells were incubated with 5 μmol Caspase-3 dye reagent for 30 minutes and then exposed to PEI (5ug/mL) in HBSS with or without calcium. HBSS alone was used as a control. A Nikon fluorescence microscope with 40x fluorescence objective was used for live-cell imaging at excitation/emission wavelengths of 500nm/530nm. Phase contrast images were also taken at identical locations to delineate individual cells.

eDNA Sample Collection and Purification

Apical HBSS was collected from immortalized human bronchial epithelial cells (hBECs) and primary mouse airway epithelial cells (mAECs) exposed to Alternaria (30ug/mL) or PEI (5μg/mL). Concentrations were selected based upon dose-response data generated in previous work (3, 6) and capillary gel electrophoresis to ensure cell viability throughout the collection process. Bronchoalveolar lavage (BAL) fluid was also collected from female wild-type (WT) BALB/c mice (n = 3 / treatment) from The Jackson Laboratory (Bar Harbor, Maine, USA) after intranasal administration of Alternaria or PEI for 4 hours. Following intranasal administration, mice were humanely euthanized via pentobarbital with trachea cannulization to collect BAL fluid using 0.5mL PBS. All samples were stored at −80°C until downstream processing. Animals were maintained and all experiments were performed with the approval of the Institutional Animal Care and Use Committee of Mayo Clinic.

Extracellular DNA (eDNA) was collected from apical HBSS and BAL fluid and purified using the Genomic DNA Clean & Concentrator (Zymo Research, Irvine, California, USA). Genomic DNA (gDNA) was collected from untreated mAECs and hBECs to serve as controls by using the Quick-DNA HMW MagBead Kit (Zymo Research) to lyse confluent cells. DNA was eluted into elution buffer, quantified using a Qubit dsDNA HS Assay kit (Invitrogen, Carlsbad, CA, USA), and stored at −80°C until pooling for sequencing. DNA was then pooled to create 3, ~1mg aliquots for each combination of sample type and treatment, with 1 aliquot of eDNA submitted for capillary gel electrophoresis using a Qiaxcel instrument (QIAGEN, Hilden, Germany) prior to sequencing.

Sequencing of eDNA Fragments

Aliquots of eDNA were prepared for sequencing using Oxford Nanopore’s Native Barcoding Kit (Oxford Nanopore Technologies, OX4 4DQ, UK) and associated manufacturer’s protocol (NBA_9168_v114_revP_20Nov2024), effectively avoiding any further fragmentation of eDNA or amplification that could create artificial biases in the characterization of fragment distribution in the genome. Aliquots were combined during the library preparation process to create 3 barcoded, multiplexed preps sequenced for 72 hours each on an Oxford Nanopore minION R10.4.1 flow cell utilizing minKNOW software (v6.2.14).

Characterization of eDNA

Sequencing reads were pooled and basecalled utilizing dorado (v1.0.0, Oxford Nanopore Technologies) super-accurate model with modified bases (v5.0, Oxford Nanopore Technologies) and aligned to the latest human and mouse reference genomes (GRCh38 and GRCm39). Default settings were used for all software and scripts in downstream analyses except where specified otherwise. Differences in sequencing depth between conditions were accounted for by standardizing sample size down to the lowest condition’s number of reads for analyses of methylation (~0.5 million reads), as well as repetitive element and motif analyses (~1 million reads). Transcription start site (TSS) coordinates were downloaded for the latest human and mouse assemblies from the refTSS database (61). Promoter regions were defined as the regions 500bp downstream of the TSS and 2000bp upstream of the TSS. The bedtools (62) intersect function was used to compare the coordinates of mapped reads to the coordinates of promoters and TSS sites. Any amount of overlap was reported for read coordinates corresponding with TSS or promoter coordinates and expressed as a percentage of total reads. R studio (v2024.04) was utilized (63) to assign genic annotations to reads through use of the R package annotatr (64) using the build_annotations and annotate_regions functions in combination with the dependencies rtracklayer (65), dplyr (66), and GenomicRanges (67) to summarize and tabulate annotations. De novo motif analysis of fragments of eDNA was performed using HOMER (v5.1)(10), with randomized background distributions accounting for GC content, CpG density, and genomic feature distributions. Enriched motifs were defined by as those with a threshold of fdr-adjusted q<0.05 compared to background distributions, a lack of redundancy with other motifs, and unrelated to simple repeats. Abundances of identified motifs in reads was estimated by HOMER’s scanMotifGenomeWide.pl, utilizing the output motifs from the findMotifs.pl script.

To resolve potential mapping artifacts and blackout areas in reference assemblies in repetitive regions, reads were re-aligned to the human or mouse telomere-to-telomere (T2T) assemblies (8, 9). Mapped read coordinates were compared with the coordinates of repetitive DNA regions via bedtools intersect, with overlaps counted at progressive/sequential threshold values for each individual read from 0–100% (i.e. at least 10% of the read must overlap with repetitive region coordinates). Bedtools shuffle was utilized to create a size-matched, randomly shuffled control distribution for each exposure type to mimic a randomly distributed fragmentation pattern. Repeatmasker repetitive regions annotations (7) for the mm39 reference genome and the grch38 human reference genome assembly were downloaded from UCSC genome table browser (68). Repeatmasker was then used to determine the average proportions and types of repetitive elements inherent to fragments of eDNA, excluding RNA repetitive elements and simple repetitive elements.

Epigenetic Characterization of eDNA

Methylation at CpG sites was calculated on an individual read (fragment) basis using Oxford Nanopore Technology’s modkit extract function (v0.5.1) with the traditional presetting applied, default confidence filtering criteria based upon subsampling of reads, and a required minimum coverage of 3 applied. The hg38 and mm39 genomes were used as the references. Control genomic DNA from untreated cells was size-matched down to 2500bp to match average fragmentation patterns of eDNA, with control fragments therefore containing comparable amounts of CpG sites per read compared to eDNA fragments. Bedtools shuffle was then used prior to methylation tabulation to create matching-sized control fragments randomly distributed throughout the genome to compare to eDNA fragments. A list of promoter coordinates (as defined above) was created based upon the known genes of the hg38 and mm39 reference genomes and used as an input parameter in modkit extract to calculate methylation at CpG sites for fragments that were mapped specifically to promoter regions. Methylation per read is expressed as the average proportion of CpG sites per read that are methylated. Entropy at CpG sites per individual read was calculated in bits from the output tables of modkit extract using the formula as previously described (69), and is expressed as the average entropy rate per read. Max entropy per read was calculated by substituting max entropy values and multiplying by the amount of CpG sites per fragment. Methylation status and entropy at CpG sites surrounding the TSS was calculated by creating 50, 100bp windows or bins up- and down-stream of the TSS of known genes from mm39 and grch38. These coordinates were then provided as an input parameter to modkit extract or modkit entropy, with CpG site methylation and entropy averaged across these windows.

Statistical Analysis

Statistical analyses were performed by GraphPad Prism v10 software (GraphPad Software, La Jolla, California, USA). Specific statistical tests are indicated in the relevant figures.

Results

Genomic DNA is fragmented and released from airway epithelial cells following aeroallergen and nanoparticle-induced oxidative stress:

Acute exposure of hBECs and mAECs to Alternaria stimulated production of reactive oxygen species (ROS; Fig. 1A) and calcium-dependent extracellular export of genomic DNA in a manner that was previously shown to preserve cell viability (Fig. 1C) (5, 6). PEI also stimulated ROS production (Fig. 1B) and genomic DNA release (Fig. 1D), and removal of extracellular Ca²⁺ partially blocked DNA release. Previously, Alternaria (6) and PEI (3) were shown to induce sustained increases in intracellular [Ca²⁺] which caused noncanonical activation of caspase 3 in Alternaria treated cells, triggering DNA fragmentation (6). Here, we show a similar effect produced by PEI (Fig. 1E–H). Extracellular DNA (eDNA) fragments collected from apical media of cultured human airway epithelial cells and bronchoalveolar lavage (BAL) fluid from BALB/c mice displayed a range of sizes between 1000–5000 bp following Alternaria exposure. PEI exposure produced similarly sized fragments with an additional band at ~35K bp (Fig. 1I).

Fig. 1. Alternaria and PEI produced oxidative stress, fragmentation and extracellular release of genomic dsDNA.

Alternaria (A) and PEI (B) increased ROS generation in HBECs. Pretreatment with ROS scavengers N-acetylcysteine (NAC) or glutathione (GSH) blocked ROS production. DNA release into the apical media (eDNA) from mAECs exposed to Alternaria (C) or hBECs treated with PEI (D). Removal of extracellular Ca²⁺ inhibited the response. E/F. Fluorescently labeled caspase-3 substrate loaded into hBECs was cleaved and translocated into the nucleus after PEI treatment (G, H). Panels F and H show the fluorescent signal before and after PEI treatment while E and G show phase contrast overlays. (I) Capillary gel images displaying eDNA fragment sizes from hBECs and mAECs exposed to Alternaria or PEI, with calculated median peak sizes indicated.

eDNA fragments contain large amounts of repetitive and centromeric sequences:

To characterize released DNA, fragments of eDNA from hBECs and mAECs exposed to Alternaria or PEI were mapped to GRCh38 (https://ftp.ncbi.nlm.nih.gov/genomes/all/GCF/000/001/405/GCF_000001405.26_GRCh38/) and GRCm39 (https://ftp.ncbi.nlm.nih.gov/genomes/genbank/vertebrate_mammalian/Mus_musculus/reference/GCA_000001635.9_GRCm39/), respectively. A large number of fragments were mapped to repetitive genomic regions either partially or in totality (Fig. 2A–B). To further investigate the nature of the repetitive content, eDNA fragments were analyzed with Repeatmasker (7), excluding simple repeats, and compared to size-matched random controls. The average fragment of eDNA was comprised of ~48–54% repetitive DNA (Fig. 2E–J), with 5–37% satellite DNA associated with centromeres. The average randomly shuffled control was composed of 30–38% repetitive DNA, with a notably smaller amount of satellite sequence compared to fragments of eDNA (Fig. 2C & 2D). Because centromeric regions are particularly difficult to map, fragments were re-aligned to the human and mouse telomere-to-telomere (T2T) references genomes (8, 9). Up to 50% of fragments associated with the smaller human chromosomes from hBEC eDNA were traced to centromeric regions (Fig. 2K and Fig. 2L), though lower percentages were observed for eDNA from mAECs (Supplemental Fig. S1). Approximately 50% of eDNA specifically derived from chromosomes #21 and #22 were traced to centromeric regions.

Fig. 2. eDNA fragments released by Alternaria or PEI exposure are enriched for repetitive content.

Intersections of mapped fragments with Repeatmasker repeat annotations, expressed as the proportion of the fragment that consists of repetitive DNA, are plotted for eDNA released by Alternaria (A) and PEI (B) and for randomized control fragments. Average proportions of each category of repeat per fragment are displayed for the hBEC random shuffle (C), the mAEC random shuffle (D), hBECs exposed to Alternaria (E) or PEI (F), mAECs exposed to Alternaria (G) or PEI (H), and mBAL fluid from mice exposed to Alternaria (I) or PEI (J). Histograms in (K) and (L) represent the percentage of fragments for each individual chromosome that mapped to the centromere region for Alternaria and PEI eDNA, respectively. “X” denotes the X chromosome.

eDNA is composed of transcriptionally active regions of the genome:

In addition to determining the repetitive content of eDNA fragments, potential intersections with annotations for known promoters, transcription start sites (TSSs), and transcription factor binding site motifs were also investigated. In hBECs, ~5% of the fragments were mapped to TSS regions and ~15% were mapped to promoter regions (Fig. 3A) following Alternaria and PEI exposure. Alternaria and PEI-induced fragments from mAECs and mBAL fluid contained ~10% TSS and ~20% promoter sequences (Fig. 3B & C). De novo motif analysis using HOMER (10) identified two enriched motifs within fragments of eDNA, with 60–80% of eDNA from hBECs containing at least 1 of each motif (Fig. 3D & 3F). In mAECs and mBAL fluid, 20–60% of fragments contained at least one of these same motifs (Fig. 3E & G). When these motifs were compared to the database of known motifs compiled by Dfam (11) and HOMER, the motif in panel D was most similar to transcription factor binding sites of the NFAT family (NFATC2 or NFATC4, Score: 0.72, where 1 is a perfect match, Offset: 2) and the motif in panel E was most similar to the transcription factor binding site for Stat5b (Score: 0.79, Offset: 0).

Fig. 3. eDNA fragments induced by Alternaria or PEI contain features of transcriptionally active regions of the genome.

Histograms show the percentage of total fragments containing at least one TSS or promoter sequence for hBEC eDNA (A), mAEC or mBAL eDNA evoked by Alternaria (B), and mAEC and mBAL eDNA evoked by PEI (C). (D-G) Motifs were identified in fragments of eDNA. De-novo motif searching results were filtered using a threshold of q < 0.05 and curated to remove simple repeats. Forward and reverse complement nucleotide logos are displayed for each identified motif, with the height of each letter indicating the probability of the nucleotide at each position. Histograms for each treatment group indicate the percentage of total fragments that contained at least one of the relevant motifs. Each identified motif was investigated in eDNA from hBECs (D & F) and from mAECs (E & G).

Alternaria and PEI induce release of DNA fragments from chromosomal fragile sites:

Mapping fragments to non-repetitive genomic regions revealed the genes most represented in fragments of eDNA. In hBECs, most of the genic fragments from both Alternaria and PEI eDNA stemmed from chromosome 20 and a few other chromosomes (Table 1). Though the specific genes targeted for export varied between human and mouse airways, the top genes exported after Alternaria or PEI exposure were remarkably comparable regardless of exposure or sample type (Supplemental Tables S1 and S2). When just the subset of fragments that contained promoters were considered, a similar bias towards chromosome 20 in hBEC eDNA was observed after Alternaria or PEI exposure (Supplemental Table S3). Similarly, the most frequently exported promoters in mAECs and mBAL fluid showed a bias towards chromosome 7 (Supplemental Tables S4 and S5) indicating a species difference.

Table 1.

Top exported genes in hBECs following exposure to Alternaria or PEI.

Gene Name	Chromosome	Position in List of Top Exported Genes (Alternaria)	# of Reads Mapped in Human Genome (Alternaria)	Position in List of Top Exported Genes (PEI)	# of Reads Mapped in Human Genome (PEI)
SEL1L2	20	1	5118	1	730
MACROD2	20	2	3989	2	570
DLG2	11	3	3319	6	453
LOC613266	20	4	3281	4	471
LOC124904873	2	5	3205	5	457
PTPRD	9	6	3153	3	482
TENM2	5	7	2971	15	396
LRP1B	2	8	2917	13	399
MACROD2-AS1	20	9	2915	10	418
CSMD3	8	10	2868	8	430
PCDDH15	10	11	2851	9	426
EYS	6	12	2848	16	382
FLRT3	20	13	2819	11	405
MACROD2-IT1	20	14	2808	12	399
CTNNA3	10	15	2737	22	359
DAB1	1	16	2588	17	380
PDE4D	5	17	2543	33	335
RBFOX1	16	18	2503	32	335
CCSER1	4	19	2501	30	337
GRID2	4	20	2488	18	376

Methylation patterns of eDNA are consistently altered:

Individual eDNA fragments were hypomethylated across CpG sites compared to control genomic DNA (gDNA) in hBECs (Fig. 4A), mAECs and mBAL fluid (Fig. 4B). When the subset of fragments that mapped to promoters were considered, eDNA was substantially more hypomethylated compared to corresponding promoter sequences from control gDNA (Fig. 4C–D). The pattern of epigenetic modification was evaluated by determining methylation entropy at CpG sites across each read. The mean entropy of individual eDNA fragments was consistently and considerably lower relative to corresponding control gDNA (Fig. 4E–G). Interestingly, fragments of eDNA released in response to Alternaria, showed that CpG sites surrounding (± 5000 bp) the TSS were hypermethylated compared to corresponding control gDNA (Fig. 4H & 4I). Methylation entropy in the same region was decreased for Alternaria-induced eDNA fragments (Fig. 4J & 4K), indicating a consistent pattern of hypermethylation around the TSS. However, hypermethylation at CpG sites surrounding TSS released in response to PEI exposure was not observed.

Fig. 4. eDNA fragments have altered methylation levels and methylation patterns.

Methylation status at CpG sites was assessed on an individual fragment basis globally (A & B) and for the subset of fragments mapping to promoter regions (C & D), with bars representing the average proportion of methylated CpG sites per read. Average methylation entropy at CpG sites for hBEC eDNA (E), mAECs (F) and mBAL fluid (G) was also graphed on an individual fragment basis. Maximum entropy/fragment for each group is denoted by “MAX”. (*) denotes statistically significant comparisons with the corresponding sequence in control gDNA. Average methylation at CpG sites surrounding (± 5000 kb) the TSS for hBECs (H) and mAECs (I). Average methylation entropy at CpG sites surrounding (± 5000 kb) the TSS for hBECs (J) and mAECs (K). Each point represents the average methylation or methylation entropy of CpG sites located within each 100 bp window surrounding the TSS. Position “0” represents the TSS.

Discussion

The predominant feature shared by all Alternaria eDNA, and most of the eDNA produced by PEI, is the presence of repetitive content. While about 20–30% of eDNA was completely repetitive, the majority of fragments consisted of hybrid fragments of repetitive DNA and non-repetitive genomic sequence. Furthermore, satellite DNA was remarkably prevalent in eDNA, with the average fragment being composed of ~5–37% satellite repeats. Satellite repeats normally comprise ~6% of the entire human genome, with centromeric satellites comprising ~3% (9, 12, 13). When fragments were separated by chromosome and mapped to centromeric regions, smaller chromosomes, such as #21 and #22, in the human genome contained larger proportions of centromeric fragments, with remarkable overlap between Alternaria and PEI eDNA distributions. Notably, the smaller gene-rich chromosomes were previously found to be the most frequently associated with human eDNA when reads were normalized to chromosome size (6). Though the proportions of centromeric fragments per chromosome were more variable in mAECs and mBAL, this observation is likely due to differing centromeric positions and the structural organization of mouse and human chromosomes, though the amount of total satellite repeats in both are comparable (8). Prior observations in microglia reported transient, DNA-damage induced DNA release into the cytoplasm with a similar bias towards repetitive DNA (14), though these repetitive sequences were traced to intergenic elements and not centromeric regions. Repetitive parts of the genome are particularly vulnerable to DNA damage and double-stranded breaks due to inherent challenges of accurately replicating repetitive content, as well as the formation of secondary structures that can impair DNA replication and repair machinery (15).

The characterization of eDNA traced to nonrepetitive regions (in any proportion) revealed the presence of promoter and TSS regions within eDNA fragments. Comparisons of eDNA associated with exposures to Alternaria or PEI in airway epithelial cells revealed remarkable consistency in proportions of eDNA fragments associated with promoter or TSS regions. Additionally, these patterns were maintained regardless of whether an in vitro or an in vivo method was used for exposures or when comparisons were made between species. These results echo prior observations utilizing amplicon-based sequencing methods, where promoter enrichment was examined through comparison of eDNA fragments to size-matched random control distributions (6). Based on prior observations of transcriptional activity during Alternaria exposure (6) and known associations between increased gene expression and chromatin accessibility, we hypothesized that areas of active transcription would be more readily accessible to endonucleases activated during exposure. However, these areas of the genome only represented 2.18% of HBE reads in the current study. These results indicate that observed patterns in eDNA localization within the genome cannot solely be attributed to changes in gene expression or chromatin accessibility related to the exposures themselves, since a random fragmentation process dependent only upon chromatin accessibility would preferentially affect these regions.

De novo motif analysis revealed two transcription factor binding motifs inherent to eDNA, with high prevalence of both motifs in hBEC eDNA suggesting that a significant fraction of fragments contain both motifs. Though these motifs differed from those previously observed in DNA released into the cytoplasm (14), this is likely due to differences in mechanisms underlying the release of DNA. The motifs observed in eDNA were most similar to known patterns for transcription factor binding sites that are typically induced during inflammation (16–19), or are induced by Alternaria or PEI exposure (3, 5, 6). The open and more accessible chromatin configuration associated with promoter and TSS regions may make them more accessible to endonuclease fragmentation or other types of DNA damage (6, 20–22). While these motifs are typically found in promoter regions, they can also be found throughout the genome in areas without transcriptional activity (23, 24). Further investigation into the mechanism underlying eDNA export is required to determine whether these motifs play a functional role in the export process.

Further investigation into the nonrepetitive content of eDNA fragments revealed that several of the most highly represented human genes in eDNA are positioned within chromosomal fragile sites, which are considered hotspots of genomic instability due to their susceptibility to replication stress and breaks during DNA replication (25–28). For example, the genes MACROD2, MACROD2-AS1, MACROD2-IT1, FLRT3, LOC613266 and LOC124904873 each map to the same band of human chromosome 20 (Chr20p12). Some of these genes (MACROD2-AS1, MACROD2-IT1, LOC613266 and LOC124904873) are antisense or intronic long non-coding RNAs (lncRNAs) that flank or are embedded in genes in the MACROD2 cluster. This locus is an established deletion hotspot in certain cancers and developmental disorders and is strongly represented in human eDNA (29–31). Other highly exported DNA fragments also mapped to genes associated with fragile sites including DLG2 (Chr11q14), LRP1B (Chr2q22), PTPRD (Chr9p24), CSMD3 (Chr8q23), RBFOX1 (Chr16p13) and GRID2 (Chr4q22)(25, 32–35). The genes LRP1B, CSMD3, PTPRD, DLG2 and RBFOX1 in particular are associated with fragile sites because they are some of the largest genes in the genome (27, 36). Though eDNA was traced to some fragile sites in the genome, these fragile sites were not the ones most frequently associated with Caspase-activated DNase activity or apoptosis (37).

Another feature of the most prevalent human genes represented in eDNA fragments relates to a neurodevelopmental gene bias, where DLG2, GRID2, RBFOX1, CTNNA3, TENM2, FLRT3 and DAB1 encode for proteins involved in neuronal adhesion, synaptic scaffolding and axon guidance (38–44). The top twenty promotor and TSS regions present in fragments of human eDNA also displayed a frequent occurrence of the MACROD2 cluster, consistent with this locus being a hotspot for oxidative stress-induced mobilization and extracellular release of DNA fragments. Additionally, several predicted lncRNAs associated with the MACROD2 fragile region (LOC124904873, LOC613266, LOC124902727, LOC101927908, LOC105377709, LOC124901127, LOC105375972 and LOC105378311 and the microRNA MIR7151) were also identified as having promoter and/or TSS sites present within eDNA, indicating that most of the eDNA promoters originated from Chr20p12. Other large genes where eDNA promoters and TSS sites were identified included TENM2 (Chr5q34), LRP1B (Chr2q22) and PCDH15 (Chr10q21), with two of these genes (TENM2 and PCDH15) involved in neurodevelopment. Therefore, the loss of genomic sequence related to these genes may have detrimental functional effects without intervention and repair.

The most highly exported regions from mBAL fluid and mAECs shows that nine (Chl1, Fgfr2, Sfi1, Tenm2, Nkx1–2, Hmx3, Gpr26, Tacc2, and Lsamp) of the top twenty exported genes overlap between the two conditions, and are comprised of several large genes (Cntnap2, Csmd1, Ptprd, Dmbt1, Tenm2, Lsamp) located in fragile regions of the mouse genome (25, 45–48). As observed with human eDNA, these genes are also late replicating, susceptible to deletions and possess a low density of replication origins. Furthermore, several of the mapped genes from mBAL and mAECs exhibit a neuro-adhesion or synaptic bias as was described earlier for a subset of human eDNA gene fragments. A key difference in the list of genes from BAL fluid was the presence of genes functioning as metabolic and signaling regulators including OAT, HTRA1 and LHPP (45), which were not in the top twenty exported genes from mAECs. Furthermore, the genes from mAECs contained several GM and RIK loci genes, which represent predicted or uncharacterized genes. In contrast, only one predicted gene (2310057M21RK) was identified in the list of most abundant exported genes from BAL fluid. One probable reason for differences between mAECs and BAL fluid may be due to the differentiation state of airway epithelial cells in vivo compared to actively dividing mAECs in culture. However, despite these differences, it remains notable that both BAL fluid and mAECs share many of the same top exported genes and that the pattern of large genes located in fragile regions of the genome with a bias towards involvement in neurodevelopment corresponds with the pattern observed in human eDNA.

Because epithelial cells exposed to Alternaria or PEI remain viable following extracellular release of their DNA fragments, alterations or perturbations of the epigenetic state of their genomes may have lasting impacts on transcriptional regulation and the stability of chromatin structure. Fragments of eDNA exhibited less methylation at CpG sites than fragments of control gDNA, though the decrease in methylation was greater for groups treated with Alternaria compared to PEI. Fragments mapped to promoter regions were even more hypomethylated in all eDNA fragments, contrasting sharply with reports of hypermethylation of promoter regions after direct oxidative damage with hydrogen peroxide (49). These alterations in methylation were also remarkably consistent when the distribution of overall methylation patterns was assessed via methylation entropy, with all eDNA displaying a decreased mean entropy compared to both their theoretical maximum per individual fragment and to fragments of control gDNA. The assessment of CpG sites surrounding transcription start sites revealed a consistent Alternaria-specific hypermethylation of CpG sites surrounding transcription start sites throughout the genome, with decreased randomness approaching the TSS from either direction compared to control gDNA CpG methylation patterns. In an exogenous promoter model, repaired double-stranded breaks in promoters with sustained activity of proteins associated with the repair process were epigenetically silenced through hypermethylation of CpG islands (50). A potential explanation for the differences in methylation surrounding the TSS between Alternaria and PEI eDNA may therefore be linked to the fragmentation pattern associated with PEI eDNA, where CpG islands within larger fragments may be less accessible to DNA methyltransferases. Further study of epigenetic remodeling associated with Alternaria or PEI exposures should be accompanied by study of transcriptional activity to shed light on how epigenetic changes are potentially associated with the repair process.

Regardless of the relative proportions of repetitive content within eDNA fragments, the bulk of eDNA was mapped to regions of the genome that are particularly vulnerable to damage. Repetitive regions, transcription start sites, and chromosomal fragile sites are all locations that are susceptible to DNA breakage, though the specific characteristics that render them vulnerable vary. An important consequence of repetitive DNA accumulation in the nucleus is the potential for secondary structures to form such as hairpins and G-quadruplexes (15, 51), which are known to impede repair of double-strand breaks (52, 53). The presence of repetitive DNA may also reduce availability of DNA repair proteins like Ku70/Ku80, XRCC4, or DNAPKcs, which can contribute to aberrant end-joining or repeat expansions (54, 55), suggesting that extracellular release of DNA fragments from the nucleus may serve to promote DNA repair after oxidative-stress induced damage, ensuring near term cell survival. However, loss of repetitive regions including SINEs, LINEs, and LTRs may disrupt chromatin structure and contribute to genome instability (15, 56, 57). Furthermore, loss of centromeric DNA can lead to chromosome structural alterations, replication fork collapse and aneuploidy (58, 59). Thus, extracellular release of repetitive and fragile site DNA fragments from the nucleus may initially be associated with cell survival, but the loss of repetitive regions may cause genome instability in the long term, which could lead to chromosomal anomalies with pathogenic consequences.

Supplementary Material

Supplemental figures and tables have been deposited in Figshare and are available at https://doi.org/10.6084/m9.figshare.32588991.

Data and materials availability:

The raw sequence data generated in this study has been deposited in the Sequence Read Archive (SRA), with the study ID PRJNA1345762.