Abstract
Particulate matter (PM) is a ubiquitous component of air pollution that
is epidemiologically linked to human pulmonary diseases. PM chemical
composition varies widely, and the development of high-throughput
experimental techniques enables direct profiling of cellular effects
using compositionally unique PM mixtures. Here, we show that in a human
bronchial epithelial cell model, exposure to three chemically distinct
PM mixtures drive unique cell viability patterns, transcriptional
remodeling, and the emergence of distinct morphological subtypes.
Specifically, PM mixtures modulate cell viability, DNA damage
responses, and induce the remodeling of gene expression associated with
cell morphology, extracellular matrix organization, and cellular
motility. Profiling cellular responses showed that cell morphologies
change in a PM composition-dependent manner. Finally, we observed that
PM mixtures with higher cadmium content induced increased DNA damage
and drove redistribution among morphological subtypes. Our results
demonstrate that quantitative measurement of individual cellular
morphologies provides a robust, high-throughput approach to gauge the
effects of environmental stressors on biological systems and score
cellular susceptibilities to pollution.
Keywords: systems biology, transcriptomics, cell morphology, cell
phenotyping, air pollution
__________________________________________________________________
Significance Statement.
Air pollution is linked to multiple life-threatening pathologies
including lung cancer and atherosclerotic diseases. A core component of
air pollution is particulate matter (PM). Previous studies have not
fully addressed that specific PM compositions elicit unique cellular
responses at single-cell level. Additionally, these studies have
predominantly focused on molecular responses. Combining transcriptional
changes, pathway remodeling, and changes in cellular morphologies, we
show that cell morphology is an indicator of PM-induced cellular damage
and provide a robust framework to profile single-cell responses and
susceptibility to PM mixtures.
Introduction
Ambient air pollution threatens human health through direct links to
chronic illnesses and premature deaths. High pollution levels are
associated with elevated incidences of ischemic heart disease, lung
cancer, aggravated asthma, chronic obstructive pulmonary disease
(COPD), stroke, and adverse birth outcomes ([39]1–6). In 2019, it was
estimated that 6.67 million deaths could be attributed to air pollution
exposure worldwide ([40]7). Particulate matter (PM), which consists of
microscopic solids and liquid droplets, is an important component of
ambient air pollution and is widely studied in the context of global
health ([41]8). These particulates and their precursor chemicals are
emitted from many natural and man-made sources, including volcanic
activity, burning of biomass, vehicle emissions, coal-burning
powerplants, and other industrial activities ([42]9).
Studies have identified strong associations between PM size and
different biological responses ([43]10, [44]11). However, a key
challenge in elucidating the effects of PM exposure is that PM chemical
composition can vary greatly across geographical areas and
environments, as there are various anthropogenic and biogenic
contributors that emit different chemical species ([45]12–14). These
inherent geographical differences of PM can impose challenges toward
understanding the different influences at the cellular and molecular
levels, since the biological effects can vary with chemical
composition. Studies using lung cell models such as A549, BEAS-2B, or
primary airway epithelial cells have demonstrated the impact that
different PM mixtures and pollutants can have on cellular pathway
remodeling ([46]11, [47]15). For example, studies using a bronchial
epithelial cell model, BEAS-2B, exposed to mixtures of either secondary
organic aerosol or aerosolized formaldehyde showed unique molecular
responses and pathway remodeling ([48]16, [49]17). Additionally, other
studies have investigated the induction of oxidative stress due to PM
exposure and have highlighted unique regulatory pathways that
contribute to the proinflammatory response ([50]15, [51]18). However,
advances in high-throughput sequencing and quantitative morphological
phenotyping to investigate the effects of multiple PM mixtures on the
cellular and molecular level could allow for prediction and faster
analysis of the effects of pollution mixtures on human health.
Different exposure methods have also provided insights into the
biological effects of PM, including liquid-submerged exposures
([52]19), air–liquid interface exposures (ALI) ([53]20, [54]21), and
pseudo-air–liquid-interface exposures ([55]19). It is worth noting that
these vary in cost, physiological relevance, and throughput. Studies
have also looked at a variety of environmental pollutants, including
PM[10], PM[2.5], and PM[0.1] (PM with aerodynamic diameters of <10,
2.5, and 0.1 µm, respectively) collected from cities including Beijing,
Milan, Seoul, and others ([56]22–27). Organic and aqueous extractions
of PM have also been investigated along with individual components or
pollution types including secondary organic aerosols, diesel exhaust
particles, volcanic ash, and metals. However, the results of these
studies vary greatly, in part, due to their use of different cell
models, exposure times and protocols, and PM types that are often not
fully characterized. All these factors introduce challenges to drawing
meaningful comparisons of the biological effects of different PM types.
Thus, improving our methods to simultaneously map cellular and
molecular effects of different PM mixtures using new high-throughput
technologies continues to be an important area of research.
Previous studies looking at air-pollution-induced pathway remodeling
via transcriptomics have found changes in regulatory pathways that
control cellular morphology, including significant alterations in
cholesterol synthesis pathways of bronchial epithelial cells that
result in distinct morphological changes ([57]17, [58]28). By
extension, these types of studies indicate that decreases in cell size
could be used as a biomarker of toxicity ([59]17). Overall, cellular
and nuclear morphology is linked to upstream changes in gene expression
and cellular dysfunction ([60]29, [61]30), with significant pathway
remodeling in cell death programs, apoptotic pathways, extracellular
matrix (ECM) interactions, and cytoskeleton structures ([62]30–32). In
the context of aging, changes in cell and nuclear sizes, as well as
irregularities in cell shapes associate strongly with fundamental
defects and senescence ([63]33, [64]34). While long-term pollutant
exposures of lung cells are linked to increased senescence, it is
unclear how short-term exposures modulate cellular responses based on
molecular or morphological phenotypes ([65]35).
Here, we expose the BEAS-2B human bronchial epithelial cell model to
three well-characterized and compositionally unique PM mixtures
available from the National Institute of Standards and Technology
(NIST): urban (SRM1648a), fine (SRM 2786), and diesel exhaust (SRM
2975). These mixtures have differing levels of important components,
such as lead, cadmium, and nitro-polycyclic aromatic hydrocarbons
(nitro-PAHs; Table [66]1 and [67]Dataset S1). Exposures were performed
at multiple concentrations ranging from 31 to 1,000 µg/mL for 24 h to
investigate the effects of multiple PM types on human lung-epithelial
cells. We performed liquid-submerged exposures given the
high-throughput nature of the method relative to ALI. Following
exposures, we measured transcriptional changes to identify specific
PM-composition-dependent remodeling of molecular pathways. In parallel,
we performed morphological analysis of cells at baseline and after PM
exposures to develop a robust single-cell platform to profile cellular
responses and the emergence of functional subtypes of cells. Together
our study provides a multiscale approach to quantify molecular and
morphological responses to several relevant PM mixtures. Additionally,
we show that we can quantify cell morphology to score cellular
susceptibility to PM exposure, offering a new tool for understanding
the cellular effects of environmental stressors. To validate this, we
isolated single-cell clones to show that subclones exhibit differential
morphological responses which associated with differences in
susceptibility.
Table 1.
Select compositional differences between PM types.
Component Urban PM mass fraction (mg/kg) Fine PM mass fraction (mg/kg)
Diesel exhaust PM mass fraction (mg/kg)
Cadmium 73.7 4.34 —
Lead 6,550 286 —
Nitro-PAHs 0.73962 0.99598 45.907
[68]Open in a new tab
Data listed as reported by NIST in mg/kg of total PM mass. No cadmium
or lead concentrations were reported for SRM 2975.
Results
Cellular viability is differentially modulated by unique PM mixtures
We measured the survival of BEAS-2B cells following exposure to three
individual PM mixtures sourced from NIST (urban, fine, and diesel
exhaust) to quantify changes in toxicity to cells. The urban and fine
samples contain PM collected over extended periods from two different
cities, St Louis, Missouri and Prague, Czech Republic, respectively.
The diesel exhaust sample was collected from the exhaust of a
diesel-powered engine. Importantly, these mixtures exhibit major
differences in several components; for instance, the mass fractions of
lead, cadmium, and nitro-PAHs vary by at least an order of magnitude
between at least two of the samples (Table [69]1). A complete
comparison of reported compositional data is given in [70]Dataset S1.
Interestingly, cadmium and lead are both highly toxic metals that can
be found in air pollution from manufacturing of batteries, cigarette
smoke, metal processing, and the production of plastics ([71]9,
[72]36), while nitro-PAHs are primarily emitted from combustion of
diesel fuel and have been shown to have mutagenic and genotoxic
properties ([73]37). Diesel exhaust is a major component of air
pollution in urban areas resulting from heavy traffic, and diesel
engines emit more particles and 10-times higher levels of nitro-PAHs
than gasoline engines ([74]38).
To evaluate the effects of PM exposures on cellular viability, we used
the alamarBlue assay which measures the reductive capacity of cells as
a proxy for viability. We observed that after cells were exposed to PM
for 24 h (Fig. [75]1A), cell populations exhibited PM-type- and
concentration-dependent changes in alamarBlue signal (Fig. [76]1B). For
example, urban PM induced a steady decrease in signal at concentrations
≥250 µg/mL (P < 0.05). However, fine PM induced a significant decrease
in signal only at the highest concentration of 1,000 µg/mL.
Paradoxically, diesel exhaust PM induced an increase in signal across
all concentrations, as shown previously ([77]39). Often interpreted as
an increase in viability, however, the increase in signal could be
associated with metabolic change in the capacity of cells to reduce
resazurin in the alamarBlue. This prompted us to look further into this
increase by evaluating the change in alamarBlue signal at short times.
A 1.5 h exposure to diesel PM showed similar increases, indicating the
cellular metabolism of the exogenous PAHs in this PM sample may be
increasing the alamarBlue signal as a significant change in the number
of cells would not occur at this shorter time point (Fig. [78]S1). It
should be noted that we also tested if the diesel PM was directly
interfering by reducing the resazurin in the alamarBlue assay; for
these experiments, we incubated alamarBlue with cell media with and
without the diesel PM, and we saw no difference in signal (Fig.
[79]S1).
Fig. 1.
[80]Fig. 1.
[81]Open in a new tab
Effects of PM exposure on cell viability. A) Graphical depiction of
submerged PM exposure method. Created with BioRender.com. B) Cell
viability following 24 h exposures to different PM types and
concentrations. Values are percentages of viable cells relative to
unexposed control cells as measured with the alamarBlue assay (n = 7,
error bars represent 1 SD, *P < 0.05 using Student's t-test).
The exposure concentrations of 125 and 500 µg/mL, equivalent to 35.2
and 140.8 µg/cm^2 in terms of deposition over cell growth area, were
chosen for further analysis. These concentrations were chosen based on
previous analyses indicating that 20 µg/cm^2 could be deposited in the
tracheobronchial regions of the lung over a period of 8 h in an urban
environment ([82]40), ∼35.2 µg/cm^2 falls within an expected deposition
amount within areas of the human lung for a 24-h period in an urban
environment, and ∼140.8 µg/cm^2 could be representative of exposure
levels in extremely polluted cities.
Exposure to different PM mixtures induces differential DNA damage responses
and cell death
Interestingly, the alamarBlue assay showed PM-dependent changes in
signal across a wide range of concentrations (up to 500 µg/mL). To
investigate this further, we profiled the DNA damage responses to PM
mixtures. Using confocal microscopy, we measured the accumulation of
the histone phosphorylation γH2AX as a marker of double stranded DNA
breaks, a marker of genotoxicity and cell death ([83]41). We found that
exposures to the different PM mixtures at concentrations of 125 or
500 µg/mL led to differential levels of DNA damage, with exposures
leading to increases in DNA damage relative to the control, unexposed
cells (Fig. [84]2A and B). Additionally, at higher exposure
concentrations, γH2AX staining in some cells becomes pan-nuclear which
is indicative of apoptosis in other cell types (Figs. [85]2A and
[86]S2) ([87]42, [88]43).
Fig. 2.
[89]Fig. 2.
[90]Open in a new tab
PM exposure leads to alteration of apoptotic levels and DNA damage. A)
Representative images of the immunofluorescent staining of γH2AX across
different exposure conditions. B) Violin plots overlaid with box and
whisker plots showing the distribution of average γH2AX intensity
values for cell nuclei in each exposure condition. C) Flow cytometry
analysis of AV–PI apoptosis assay following PM exposure. AV+/PI– and
AV+/PI+ represent early- and late-stage apoptotic cells, respectively,
AV–/PI+ represents dead cells (n = 3, ≥10,000 cells per measurement,
error bars represent the standard error of the mean, *P < 0.05,
comparisons are drawn against the respective control populations using
Student's t-test). The remaining cells in each condition were healthy
(AV–/PI–). D) Correlation between the percentage of dead cells from the
apoptosis assay shown in (A) and the average γH2AX intensity following
PM exposure. Shading represents a 95% CI.
Furthermore, we evaluated whether PM exposures were inducing cell death
via apoptosis or based on nonapoptotic mechanisms. Both apoptotic and
nonapoptotic mechanisms are associated with aberrant levels of DNA
damage. To determine the mode of cell death, we used an annexin V
(AV)–propidium iodide (PI) flow cytometry assay, as previously used to
investigate the mode of cell death in lung cells exposed to PM mixtures
([91]44).
Cells exposed to 125 µg/mL of all PM conditions exhibited small
increases in the population of dead cells (AV–/PI+) and a decrease in
the population of apoptotic cells (AV+/PI– and AV+/PI+; Fig. [92]2C).
AV+/PI– indicated cells were in the early stages of apoptosis, while
AV+/PI+ indicated cells are in later stages of apoptosis or dying due
to loss of membrane integrity. In contrast, exposure to 500 µg/mL
concentration of each PM mixture resulted in an increase in the number
of dead cells (AV–/PI+; Fig. [93]2C). Representative scatter plots of
flow cytometry data from each condition are shown in Fig. [94]S3. We
next compared the trends in the populations across the different PM
mixtures and observed that exposure to different PM compositions led to
different distributions of apoptotic vs. dead cells. For instance,
exposure to the urban PM mixtures resulted in a greater number of dead
cells relative to the fine and diesel mixtures. These results
corroborate the general pattern in viability observed via the
alamarBlue assay, with urban PM inducing the greatest losses of
viability, but better captures loss of viability in the diesel and fine
conditions given the alamarBlue assay also captured increases in
metabolic activity of PM components (Figs. [95]1B and [96]S1).
In the literature, AV–/PI+ events are sometimes reported as cells
undergoing mechanisms of death such as ferroptosis or cells lacking a
cell membrane, resulting in no AV signal ([97]45, [98]46). This would
be consistent with our results given our collection of attached and
detached cells following the 24 h length of exposure. It is therefore
possible that following PM exposures, some cell membranes may have been
fully ruptured, leaving nuclei behind. However, it is also possible
that other forms of cell death are taking place such as ferroptosis,
resulting from the exposure to iron and other metals present in the PM
mixtures.
The levels of DNA damage are also associated with the levels of
observed cell death (Fig. [99]2D). As shown in Fig. [100]2, γH2AX
intensity increases with exposure to increasing PM concentrations for
each of the three PM types. Furthermore, γH2AX intensity positively
correlates (Pearson Coefficient of R = 0.94, P = 0.0014) with the
percentage of dead cells (AV–/PI+) found in the AV–PI data for the same
exposed populations (Fig. [101]2D). Taken together, these data show
increases in cell death and DNA damage levels are observed with
increasing PM concentrations. These levels are also dependent on the PM
composition, as the three mixtures show markedly different trends.
Additionally, the correlation between cell death and γH2AX intensity
points to a framework of DNA damage-associated cell death.
Post-PM exposure transcriptional profiling indicates common and unique gene
expression remodeling
The unique differences observed in viability and the patterns of DNA
damage following PM exposures at 125 and 500 µg/mL prompted us to
investigate whether PM exposures also induced differential molecular
responses. To better understand how the underlying transcriptomic
profiles influence differential viability across PM mixtures, we
assessed changes in gene expression patterns via 3′-TagSeq ([102]47,
[103]48) ([104]Datasets S2–S7). This approach takes advantage of the
poly(A) tail on mRNA for sequencing library preparation, allowing the
accurate quantification of protein-coding transcripts.
We first observed that exposure to urban and fine PM mixtures induced
significant changes in the expression of a greater number of mRNA
transcripts relative to diesel exhaust under the two PM concentrations
tested. Furthermore, the magnitudes of the changes were larger for
cells exposed to urban and fine mixtures than those exposed to diesel
exhaust PM (Fig. [105]3A–F). These observations indicate that the urban
and fine mixtures have a lower threshold for stimulation of cellular
responses. Additionally, the number of genes that were differentially
expressed (DE) by each PM type increased with higher concentrations
(i.e. 125 µg/mL vs. 500 µg/mL exposures; Fig. [106]3A–F). Moreover, the
majority of the genes (at least 67% for each condition) that were up-
and down-regulated in the 125 µg/mL conditions were similarly up- and
down-regulated in the 500 µg/mL condition (Fig. [107]S4), indicating
consistency in the transcriptional responses across different
concentrations of each PM type, with additional pathway activation at
higher concentrations.
Fig. 3.
[108]Fig. 3.
[109]Open in a new tab
Transcriptomic analysis of PM stress reveals unique network remodeling.
A–F) Volcano plots showing significantly DE genes (cutoffs = Log[2]FC >
1, P[adj] < 0.05) relative to control cells after exposure to diesel,
fine, and urban PM at 125 µg/mL (A–C) and 500 µg/mL (D–F) for 24 h. G)
Log[2]Fold changes in expression of select TGF-β-related genes. A gray
background indicates the expression change was not significant (P >
0.05). H and I) Venn diagrams of the significantly DE genes from each
condition. Intersections represent genes that were DE in overlapping
conditions. J–L) Bubble plots showing select enriched GO biological
process terms that are commonly enriched among 2 or more of the
low-level exposure conditions (J), or unique to the low-level fine (K)
or urban (L) exposure conditions.
Additionally, we observed that four mRNAs encoded by the CYP1A1,
CYP1B1, ID1, and ID3 genes were DE postexposure across all conditions
(Fig. [110]3G); two were overexpressed (CYP1A1 and CYP1B1) and two
displayed decreased expression (ID1 and ID3), relative to expression
levels in unexposed control cells. The CYP1A1 and CYP1B1 are members of
the cytochrome P450 (CYP) family that metabolize endogenous compounds
such as fatty acids and steroid hormones ([111]49). Consistent with our
results, these genes are upregulated in human epithelial lung cell
models in response to exogenous PAHs present in PM ([112]15). These
PAHs bind to the cytosolic aryl hydrocarbon receptor, which then
mediates expression of the cytochromes and promotes a proinflammatory
response to induce reactive oxygen species (ROS) production in cells.
ID1 and ID3 are inhibitors of DNA binding proteins that are induced by
transforming growth factor β (TGF-β) and have been implicated in the
regulation of senescence, apoptosis, and cell cycle alterations
([113]50). Moreover, ID1 expression has also been shown to decrease
after exposure to coarse PM (PM with an aerodynamic diameter between
2.5 and 10 µm) ([114]11), but the roles of these genes have been less
defined in the context of air pollution exposures. Importantly, the
increase in expression of the CYP genes and the decrease in expression
of the inhibitor of DNA binding (ID) genes suggest that the response to
organic cyclic compounds as well as alteration of the TGF-β regulatory
pathways is commonly remodeled by these unique PM mixtures. This is
further supported by the differential expression of additional
TGF-related genes (Fig. [115]3G). Interestingly, many TGF-β-related
genes are involved in the regulation of cell morphology and motility.
We next observed that unique mRNAs were significantly DE only when
cells were exposed to certain PM mixtures, but not others (Fig. [116]3H
and I). For example, TNFAIP6, a regulator of the ECM, LCAT, a protein
involved in extracellular metabolism, and CXCL1, a protein involved in
inflammation, are significantly DE under only fine exposure conditions.
However, genes including DDIT4, a protein induced by DNA damage, MT1E,
a protein involved in the cellular response to cadmium, and ACTN4, an
actin-binding protein, are DE under only urban exposure conditions.
This indicates that there could be unique pathway activation that is
dependent upon the PM composition.
Overall, the gene expression patterns observed in cells exposed to fine
and urban PM mixtures exhibit significant pathway remodeling, whereas
cells exposed to diesel exhaust PM exhibit less remodeling. Similarly,
we observed a dose dependence in the extent of pathway remodeling, i.e.
more changes with higher PM concentrations. Finally, we noted that
although expression of a limited set of 4 genes was consistent across
all conditions (i.e. CYP1A1, CYP1B1, ID, and ID3), other genes are DE
in a manner that is dependent on the PM type.
Gene ontology analysis reveals PM-dependent remodeling of apoptosis,
motility, and morphology pathways
To determine the key remodeled pathways postexposure and the extent to
which they were remodeled, we performed Gene Ontology (GO) and pathway
enrichment analysis. We performed this analysis using the
transcriptomics data from cells exposed to urban, fine, and diesel
exhaust PM mixtures at 125 and 500 µg/mL (Figs. [117]3J–L and [118]S5).
The complete list of enriched GO terms for each condition is given in
[119]Datasets S8–S13. Using Enrichr ([120]51), we identified 34
pathways that were significantly enriched (P[adj] < 0.01) in cells
exposed to the 125 µg/mL concentration of both urban and fine PM
mixtures, relative to baseline. We selected 11 nonredundant pathways to
show in Fig. [121]3J. We observed changes in the expression of genes
related to the mitogen-activated protein kinase (MAPK) cascade (e.g.
EDN1, GDF15, TGFB2, ANGPT1, and LIF), epithelial cell proliferation
(e.g. CDKN1C and EPGN), regulation of apoptosis (e.g. FCMR and CITED2),
and cell migration and ECM organization pathways (e.g. SFRP1 and FGG).
It is worth noting that for cells exposed to urban and fine PM at
500 µg/mL, similar pathways were also significantly enriched (P[adj] <
0.01; Fig. [122]S5 and [123]Datasets S11–S13).
Interestingly, we observed few DE genes in cells exposed to the diesel
exhaust PM mixture. Only the “response to organic cyclic compound”
pathway was significantly enriched (P[adj] < 0.01) in cells exposed to
all PM types at the 125 µg/mL concentration. However, this response
appears to be ubiquitous, with the CYP1A1 and CYP1B1 genes increasing
in expression across all conditions postexposure. Similarly, we
identified upregulation of IL1B, which was upregulated in all
conditions except 125 µg/mL diesel exhaust PM.
We also identified key genes exhibiting differential expression across
both the 125 µg/mL urban and fine PM exposure conditions that
contributed to the remodeling of multiple pathways (Figs. [124]3I–J and
[125]S6). For example, IL1A, IL1B, and TGF-β genes were part of several
GO-defined pathways that comprise cytokine signaling cascade and TGF-β
signaling. Other genes involved across many pathways include GAS6,
which is involved in cell growth and migration and cytokine signaling,
and PTK2B, a protein involved in the activation of MAPK signaling and
reorganization of the actin cytoskeleton. These genes are present in
many of the most significantly altered pathways, highlighting their
importance in the biological response to PM exposure.
Finally, we observed that several exclusive GO terms were significantly
enriched (P[adj] < 0.01) in cells postexposure to urban PM at both
concentrations (125 and 500 µg/mL; Figs. [126]3K and [127]S5), which
included unique responses to metal ions. Examples of these pathways
include response to cadmium ion, copper ion, and zinc ion, which
encompass mRNAs encoded by the MT1 family genes (MT1G, MT1E, MT1F, and
MT1M). The patterns of gene expression changes involved in the
regulation of metal ions are consistent with the increase in metal
composition (i.e. cadmium) in the urban PM mixture, relative to the
other mixtures tested (Table [128]1). Similar to the 125 µg/mL urban
exposure, at the 500 µg/mL urban condition, the top significantly
enriched GO term is a response to the metal ion, again indicating the
importance of the increased metal concentrations in the urban PM sample
relative to fine and diesel exhaust.
Taken together, these data indicate that cells differentially regulate
their gene expression patterns in a PM composition-dependent manner.
However, pathways related to cell morphology, and ECM remodeling seem
to be broadly shared across all PM exposure conditions, with pathways
related to apoptosis shared across the urban and fine conditions.
PM compositions drive the emergence of morphological subtypes postexposure
Since unique PM mixtures drive differential responses, particularly in
apoptosis, cytoskeletal structure, and ECM-related pathways, we
wondered whether these responses could be captured by changes in
cellular morphologies across cell populations. Using our BEAS-2B cell
line model, we exposed cells to the same PM mixtures at the same
concentrations and exposure times described above. After exposure,
cells were fixed and stained for F-Actin (488-Phalloidin), DNA (DAPI),
and γH2AX (anti-γH2AX [phospho-S139] antibody; Fig. [129]4A and B). The
Phalloidin and DAPI stains were used to delineate the cell and nuclear
boundaries, and γH2AX to quantify the extent of persistent DNA damage.
For each cell and nuclear boundary, we computed 33 discrete parameters
describing features related to the sizes and shapes of individual cells
(Table [130]S1). Across all conditions, we analyzed ∼13,000 single
cells. To identify whether BEAS-2B cells exhibited morphological
subtypes that changed after PM exposure, we performed dimensional
reduction and clustering analyses on cells analyzed across all
conditions. Using a combination of k-means clustering and uniform
manifold approximation and projection (UMAP), we identified 10 distinct
morphology clusters, each having unique cellular and nuclear
morphological profiles (Fig. [131]4C and D). Furthermore, these 10
morphological clusters can be further grouped into 3 cluster groups
(CGs), with morphology clusters 1, 2, and 5 belonging to CG1,
morphology clusters 3, 4, and 6, belonging to CG2, and morphology
clusters 7–10 belonging to CG3 (Fig. [132]S7).
Fig. 4.
[133]Fig. 4.
[134]Open in a new tab
Morphological analysis reveals distinct cellular subtypes. A) Graphical
depiction of the morphological analysis pipeline. B) Representative
fluorescence microscopy images of cells from each condition. C) UMAP
visualization of the 33 measured morphological parameters for each cell
in every condition. UMAP-1 (x-axis) was negatively correlated with size
and UMAP-2 (y-axis) was positively correlated with cell elongation, or
linearity. k-means clustering was applied to cluster cells of similar
morphologies. D) Representative cellular (outer) and nuclear (inner)
morphologies of cells from each k-means morphology cluster. E–H) Plots
showing the distribution of cells from each respective exposure
condition in red within the UMAP space. I) Heatmap displaying the
enrichment in a number of cells in each morphology cluster for each
exposure condition. The bar graph shows the Shannon entropy for the
distribution of cell morphologies within each exposure group. The
dendrogram identifies clusters with similar morphological features. J)
Heatmap displaying the mean γH2AX in morphology clusters across all
exposure conditions. Mean γH2AX intensity across each exposure
condition (top). γH2AX intensity of all cells within each k-means
cluster (right). Panel A created with BioRender.com.
Next, we asked whether cells exposed to both low (125 µg/mL) and high
(500 µg/mL) concentrations of each PM mixture exhibited differential
abundance of cells across each morphology cluster. Upon comparison, we
observed pronounced shifts in the abundance of cells per morphology
cluster in a PM-dependent manner (Figs. [135]4E–I and [136]S8).
Specifically, when compared with unexposed conditions, cells exposed to
500 µg/mL of urban and fine PM exhibited higher fractions of cells in
clusters 9 and 10, which describe smaller, more rounded morphologies
(Fig. [137]4G–I). However, cells exposed to 500 µg/mL of diesel PM
exhibited higher fractions of cells in clusters 4 and 8, which describe
larger, more elongated cell morphologies (Fig. [138]4F–I). Based on the
observed fractional redistributions among morphology clusters per
condition, we computed the Shannon entropy as a way to estimate
cellular heterogeneity ([139]52). Although cells were redistributed
among morphology clusters per PM conditions, only cells exposed to
urban 500 µg/mL showed a pronounced decrease in heterogeneity relative
to unexposed control cells (Fig. [140]4I).
Taken together, our results indicate that cells exposed to different PM
mixtures drive fractional redistributions among cellular morphology
clusters in a PM-dependent manner. Furthermore, the differential shift
in localization of cells exposed to urban PM (toward small, more
rounded morphologies, especially at the 500 µg/mL exposure
concentration) and diesel PM (toward larger, more elongated
morphologies) point out that these PM mixtures are likely driving
unique responses based on the underlying compositions. Finally, these
results suggest the potential utility of cell morphology cluster
profiles to denote functional subtypes in pre- and postexposed cells.
Morphological clusters are characterized by the extent of persistent DNA
damage
Given that cells exposed to both urban and fine PM exhibited a higher
fraction of cells with smaller, more rounded cell morphologies (Fig.
[141]4G and H) and decreased viability relative to unexposed cells
(Fig. [142]1B), we investigated whether morphology clusters were
associated with persistent DNA damage. Since each cell was costained
for γH2AX, we computed the extent of DNA damage based on the total
nuclear abundance of phosphorylated-H2AX (γH2AX) as previously
performed ([143]53). Comparing cells from all exposure conditions, we
observed a significant increase in the γH2AX content for cells exposed
to 500 µg/mL urban PM relative to unexposed control cells.
Additionally, viewing the cells at higher magnification shows an
increase in the number of γH2AX foci following exposures (Fig.
[144]S2). Furthermore, to test whether cells in different morphology
clusters exhibited different levels of DNA damage, we pooled cells
within each morphology cluster across all conditions and quantified the
levels of γH2AX. Interestingly, we found that cells belonging to
clusters 9 and 10 had the highest levels of γH2AX (i.e. high DNA
damage), with cluster 4 exhibiting the lowest level of damage (Fig.
[145]4J). These results suggest that the identified cell morphology
clusters could be further defined based on the extent of DNA damage and
susceptibility to cell death after PM exposure.
Cadmium drives morphological shifts among functional clusters after PM
exposures
To further test the hypothesis that the chemical compositions of the PM
mixtures drive specific shifts among morphological clusters (i.e.
smaller, rounder, and less viable cells), we systematically
supplemented our PM mixtures with different concentrations of cadmium
chloride (0–25 µM) and lead acetate (0–250 µM) that mimic those used in
other studies ([146]54, [147]55). We selected cadmium (Cd) and lead
(Pb), due to their variable concentrations across different PM mixtures
(Table [148]1), and the pronounced shifts in both the viability and the
morphological shifts when cells were treated with urban PM (urban PM
has the highest concentration of Cd in the tested PM mixtures). First,
we observed a significant decrease (P ≤ 0.05) in viability with
increasing levels of cadmium chloride supplementation across all
conditions tested (Figs. [149]5A and [150]S9). In contrast, cells
exposed to PM mixtures supplemented with lead acetate resulted in
little to no change in viability (Fig. [151]S10).
Fig. 5.
[152]Fig. 5.
[153]Open in a new tab
Cadmium drives morphological shifts in cells. A) Cell viability
following 24 h exposures to different PM types and concentrations
supplemented with cadmium chloride (0–25 µM Cd). Values are percentages
of viable cells relative to unexposed control cells as measured with
the alamarBlue assay (n = 6, error bars represent 1 SD). B–H)
Morphological distributions of cells from each respective exposure
condition with CdCl[2] supplementation displayed across the UMAP space.
Evaluating the morphological effects of cells exposed to increasing
CdCl[2] concentrations across all PM mixtures (Fig. [154]5B–H), we
observed a general tendency toward smaller, rounded morphologies
described by clusters 9 and 10. Cells exposed to 125 µg/mL urban PM and
15 µM Cd resembled the distributions of 500 µg/mL urban either alone or
with 5 or 15 µM cadmium supplementation (Figs. [155]5C, D and
[156]S11B, C). For cells exposed to 125 µg/mL of fine PM, 15 µM cadmium
supplementation led to a great shift relative to the 5 µM. However, in
the 500 µg/mL fine PM conditions, even at 5 µM we observed a shift
toward clusters 9 and 10, with 500 µg/mL fine PM with cadmium
supplementation resembling the 500 µg/mL urban PM conditions (Figs.
[157]5E, F and [158]S11D, E). Finally, cells exposed to 125 µg/mL
diesel PM with 15 µM cadmium exhibited a bi-phasic shift in the
abundance of cells among clusters, with 33.9% of cells shifting toward
clusters 9 and 10. However, cells exposed to 500 µg/mL of diesel and
15 µM cadmium exhibited a similar shift toward clusters 9 and 10 (30.4%
of cells), despite the increased PM concentration (Fig. [159]5G and H).
Collectively, our data indicate that the differential abundance of
cadmium in the different PM mixtures may drive differential toxicity
among PM mixtures. Importantly, these observed correlations between
increased PM toxicity (lower viabilities with cadmium supplementation)
and distinct morphological redistributions among cell populations
suggest the potential for predicting the toxicity and susceptibility of
cells to different PM mixtures using their morphologies.
Single-cell morphology is associated with cellular susceptibility to urban PM
exposures
Since cells exhibited unique morphologies and responses to PM
exposures, we wondered whether cellular morphologies encoded resilience
or reduced susceptibility to PM exposure at the single-cell level. To
test whether the morphologies of cells were associated with the
response to PM exposure, we isolated single-cell clones from the
parental BEAS-2B cell line. Seeding a single cell per well of a 96-well
plate, we generated 12 single-cell clones. Analyzing the morphologies
of each clone, we did not observe any clone localizing specifically to
one morphology cluster (Fig. [160]4C). However, when separating the
individual morphological clusters into the three CGs (CG1, CG2, and
CG3), we observed that some clones differentially occupied one or
multiple of the three CGs (Figs. [161]6A–C, [162]S12A, and [163]S13).
As expected, when we compared the cellular heterogeneities (i.e.
cell-to-cell variations) of the 12 clones relative to the parental, we
observed an overall reduction in the Shannon entropy for each of the
clones, with clone 7 having the lowest heterogeneity (Fig. [164]S14).
Fig. 6.
[165]Fig. 6.
[166]Open in a new tab
Morphology encodes susceptibility to PM exposure. A) The 10 k-means
clusters that are used to define cell morphology can be further grouped
into 3 CGs (CG1–3) using hierarchical clustering. B) Representative
images of each population. C) Clonal populations show enrichment in
different morphological CGs. The distributions of cell morphologies
differ for each clone and the parent cell population from which the
clones were derived. D–G) Upon exposure to urban PM, clones with unique
baseline morphologies show different shifts in morphology. H)
Susceptibility scores for each cell population normalized to the
Parental population. A more positive value indicates a higher
susceptibility to PM stress. I and J) Average γH2AX intensity values in
cell nuclei for populations in the baseline and urban 125 µg/mL
exposure condition. Significance determined by one-way ANOVA test, ***P
< 0.001.
To further test the hypothesis that cellular morphologies governed the
response to PM mixtures, we exposed all clones to the urban PM mixture
at 125 and 500 µg/mL for 24 h. To illustrate unique baseline
morphologies, we selected the parental and three clones that exhibited
differential abundance of cells within the three morphological CGs.
Specifically, at baseline the parental line had a similar abundance of
cells across all three CGs, clone 7 was highly abundant for cells in
CG1, clone 8 was highly abundant in CG2, and clone 1 was highly and
equally abundant in CG1 and CG2 (Fig. [167]6C). Based on these starting
morphologies, we tested the cellular responses to urban PM exposures
(both 125 and 500 µg/mL).
After exposure, single-cell clones showed differences in morphological
distributions. For the parental and isolated clonal populations, cells
exposed to 500 µg/mL of urban PM resulted in a drastic shift toward CG3
and more specifically morphology clusters 9 and 10. However, the major
differences in morphological shifts were observed in the cells exposed
to 125 µg/mL of urban PM (Figs. [168]6D–G and [169]S12B). For the
parental and clone 7 populations, there were very little shifts in the
abundance of cells within the CGs, as shown by the significant overlap
of contours from control (unexposed) and the 125 µg/mL conditions (Fig.
[170]6D and F). Interestingly, for clone 8, at 125 µg/mL urban PM,
there was a significant shift toward CG3 (specifically clusters 9 and
10), while clone 1 showed a shift toward CG2, the intermediate
morphological regime (Fig. [171]6E and G). These results suggest that
clone 8 may be more susceptible to urban PM relative to clone 7, while
clone 1 is moderately susceptible. We proposed the shifts in morphology
(i.e. morphological transitions rather than baseline morphologies)
captured by the UMAP space could be used to quantify cell population
susceptibility to PM exposure using the following equation:
[MATH: SSC=(CG1UL−
CG1C)−(CG1UH−
CG1UL)(CG1UH−
CG1C) :MATH]
(1)
where S[SC] is the susceptibility score, CG[1] represents the fraction
of cells within cluster group 1 at each respective condition (C =
control, UL = urban 125 µg/mL, and UH = urban 500 µg/mL). This equation
quantifies the shift away from the healthy CG1 morphology, giving
higher scores when populations shift away from CG1 at the lower
exposure concentration. The resulting scores (Fig. [172]6H) show a
spectrum of susceptibilities with clone 8 being the most susceptible,
followed by clones 2, 3, and 4 which showed more drastic morphological
shifts at the 125 µg/mL exposure condition. Additionally, clones 10 and
5 had the lowest susceptibility scores.
To further test this notion of susceptibility based on morphological
shifts, we evaluated whether DNA damage responses followed our
susceptibility scoring (Fig. [173]6I and J). Clones 5, 7, and 10 showed
low expression of γH2AX at baseline, and clones 5 and 10 continued to
show the lowest levels following exposure to 125 µg/mL of urban PM,
agreeing with the susceptibility scores. Additionally, clone 1 was
ranked as more susceptible than the parental population and showed
significantly higher expression of γH2AX. Additionally, clone 8
exhibited the highest γH2AX signal at baseline (>2× compared to the
parental, P < 0.001), with a significant increase at 125 µg/mL of urban
PM (P < 0.001).
Taken together, these results point to the notion that quantifiable
shifts in cell morphology profiles can be used as predictors or
biomarkers of PM-induced responses, even at the 125 µg/mL concentration
which did not induce drastic increases in cell death (Fig. [174]2C).
Clones with a high abundance of cells in CG1 and lower numbers of cells
shifting to CG2 or CG3 upon exposure were most resilient to urban PM,
while clones with large shifts toward CG2 and CG3 demonstrated
increasing susceptibility to urban PM. Finally, susceptibility to urban
PM exposures seems to be influenced by the levels of DNA damage.
Discussion
In this study, we demonstrate a multiscale approach to characterize the
unique differences in cellular response to three PM mixtures using
molecular and quantitative morphological analyses. We further
investigated morphological variations across populations of unexposed
and PM-exposed cells to show that cellular morphology is associated
with susceptibility to urban PM exposures and provides mechanistic
insights into variable responses across cell populations. Additionally,
we show that these responses are dependent on the composition of the PM
mixture, for instance, an abundance of cadmium can drive unique
cellular transcriptional responses and morphological changes.
With the emergence of single-cell technologies and deep-learning tools,
there has been a tremendous acceleration in the capacity to quantify
and analyze specific cell states and behaviors across cell populations
([175]56–58). Specifically, analysis of biophysical properties, such as
motility and morphology, offers an efficient method to discretize
functional subtypes of cells ([176]33, [177]59, [178]60). In this work,
we profile the single-cell morphological changes after exposure to
various PM mixtures to quantify cellular responses and identify
cellular properties that are associated with cellular susceptibility to
pollutants. Three major and novel findings of this work include the
following: First, we identified that although there is a common
transcriptomic response to PM in the activation of the P450 family
cytochromes, as shown previously ([179]15, [180]61), the degree of
pathway remodeling is dependent on the PM composition and concentration
of exposure. Second, we show cell morphology is a strong indicator of
response to differential PM exposure. Third, we used single-cell clones
to show that shifts from unique starting morphologies can encode
susceptibility to air pollution exposure. Collectively, our findings
show that cell morphology has the potential to be used as a biomarker
for environmental risk assessment (Fig. [181]7).
Fig. 7.
[182]Fig. 7.
[183]Open in a new tab
Model of exposure susceptibility. Cellular morphology encodes
susceptibility to PM exposure and is dependent upon the interplay
between molecular changes in cells. Created with BioRender.com.
With the development of single-cell technologies, in both
transcriptional and morphological profiling, and the advances in RNA
fluorescence in situ hybridization techniques, further studies could be
performed to more directly link the expression of different transcripts
with morphological features of individual cells. Additionally, the use
of primary cells in a more realistic ECM environment or the use of ex
vivo lung model technologies ([184]62, [185]63), and the testing of
additional freshly collected PM mixtures containing volatile organic
components could further improve the biological context of future work.
Finally, the use of live-cell imaging to monitor cellular changes over
time could lead to a better understanding of cluster stability among
morphological patterns and would further enhance our ability to
determine susceptibility to PM mixtures and other bioactive agents.
Taken together, our data begin to elucidate how different PM mixtures
drive unique changes in morphological and transcriptional signatures,
and how individual cells within a population have differing levels of
susceptibility, encoded in their morphologies. This knowledge could
provide a better understanding of how components of PM such as cadmium
and other metals drive PM toxicity. Furthermore, our platform for
quantifying susceptibility could provide a potential way to investigate
the effects of sensitizers and desensitizers that alter cellular
responses to environmental stress. Finally, our findings could
facilitate the development of morphology-based methods for
characterizing an individual's risk of air pollution exposure.
Materials and methods
Cell culture
BEAS-2B cells (ATCC CRL-9609) were cultured from cryopreserved stocks
in collagen-coated T-75 culture flasks according to ATCC guidelines.
Briefly, cells were seeded at 3,000 cells/cm^2 and cultured in 23 mL of
bronchial epithelial cell growth medium (BEGM) (Lonza, CC-3170),
omitting the addition of the gentamicin–amphotericin aliquot to the
medium, as recommended by ATCC. Cells were grown at 37°C in a
humidified incubator with a 5% CO[2] atmosphere, and complete media
exchanges were performed every 48 h. After approximately 4 days, the
cultures reached ∼70% confluency, and cells were subcultured into
6-well or 96-well plates coated with type 1 collagen (Advanced
BioMatrix, Cat#5005) at 1,500 cells/cm^2, and allowed to attach to the
growth surface for 24 h before exposure to PM.
PM exposure
The BEAS-2B cells were exposed to three PM mixtures collected from
different sources that were purchased from NIST, urban PM (SRM 1648a),
fine atmospheric PM (SRM 2786), and diesel exhaust PM (SRM 2975). Just
before the start of the exposures, the three PM mixtures were weighed
using an analytical balance and suspended in sterile DI H[2]O in
10 mg/mL stock solutions. The suspensions were sterilized by
ultraviolet (UV) irradiation for 30 min as done previously ([186]44).
Serial dilutions were performed with BEGM medium to reach the tested
concentrations between 1,000 and 31 µg/mL. For exposures that contained
supplements of cadmium, cadmium chloride was dissolved in DI H[2]O and
filter sterilized before being added to the PM mixtures at 1,000×
dilutions. To begin the exposures, media from the well plates was
removed and replaced with equal volumes of PM-containing media for
exposed cells or fresh media for unexposed control cells. The cells
were incubated at 37°C in a humidified incubator with a 5% CO[2]
atmosphere for a 24-h exposure period before downstream analysis.
AlamarBlue assay
BEAS-2B cells were seeded in 96-well plates at a density of
10,000 cells/well. Twenty-four hours later, the cells were exposed to
PM mixtures at concentrations ranging from 31 to 1,000 µg/mL as
described above with n = 7 replicates per condition. Following a 24-h
exposure period, the media containing the PM was removed and 100 µL of
fresh BEGM containing 10% alamarBlue (Invitrogen, DAL1025) by volume
was added to each well. Cells were then incubated at 37°C for 2 h in
the dark. Following this incubation, the fluorescence of each well was
measured (Ex. 560/Em. 590) using a BioTek Cytation3 microplate reader.
The fluorescence readouts correspond to cell metabolic activity and
were normalized to the readings from unexposed control cells after
performing background correction by subtracting the fluorescence of
wells containing only the alamarBlue-BEGM mixture.
AV–PI flow cytometry
In this assay, levels of the fluorescein isothiocyanate (FITC)-labeled
AV protein indicate apoptosis as the AV protein binds with high
affinity to the phosphatidylserine that is translocated from the inner
side of the cell membrane to the outer side. Likewise, levels of PI,
which fluoresces upon binding DNA in cells that have ruptured or become
permeable, indicate cell death or cells that are in the latest stages
of apoptosis ([187]46, [188]64). The preparation of cells for flow
cytometry was conducted according to established protocols ([189]64).
Briefly, following the completion of PM exposures using n = 3
replicates, culture media was collected and put on ice to recover
detached cells. Adherent cells were trypsinized and combined with the
collected culture media. The combined cells were washed twice with cold
phosphate-buffered saline (PBS) before proceeding with AV-FITC and PI
staining of 250,000 cells per sample using an eBioscience Annexin V
Apoptosis Detection Kit (ThermoFisher, 88-8005-72). Prepared samples
were analyzed on a Sony Biotechnology MA900 Cell Sorter available
through the Center for Biomedical Research Support at UT Austin. At
least 10,000 cells per replicate were analyzed for AV binding and PI
incorporation.
Cell staining and imaging
Following exposure, cells adhered to cover glass coated with type 1
collagen (Advanced BioMatrix, Cat#5005) were washed with prewarmed PBS
for 5 min then fixed by incubation for 15 min at 37°C in a freshly
prepared, methanol-free 4% formaldehyde solution in PBS. Cells were
rinsed 3× with PBS before being permeabilized by incubation in a 0.1%
Triton-X PBS solution for 4 min. Cells were again rinsed 3× with PBS
and then blocked with 1% bovine serum albumin (BSA) in PBS for 20 min
at room temperature. Cells were incubated with a 1:400 dilution of a
recombinant anti-γH2AX (phospho-S139) antibody (Abcam, ab81299)
overnight at 4°C to visualize the DNA damage biomarker. The next day,
cells were washed 3× with PBS for 5 min and then incubated with a 1:250
dilution of a fluorescently tagged secondary antibody (Goat Anti-Rabbit
IgG H&L [Alexa Fluor 488]; Abcam, ab150077) for 1 h at room
temperature. Cells were then rinsed 3× with PBS and stained with Alexa
Fluor 594 Phalloidin (Invitrogen, A12381) and Invitrogen NucBlue Fixed
Cell ReadyProbes Reagent (DAPI) (Invitrogen, [190]R37606) according to
the manufacturer’s protocols to allow visualization of the F-actin
structure and nuclei, respectively. Microscopy slides were then
assembled using ProLong Gold Antifade Mountant (Invitrogen,
[191]P36930) and were sealed with clear nail polish. Slides were stored
at 4°C until imaging.
Fluorescent images were acquired with a Leica Stellaris 5 Confocal
Microscope at 20× resolution using 3 laser lines (405 Diode: DAPI
Nuclear Stain; 488 Diode: Alexa Fluor 488 secondary antibody targeting
γH2AX; 647 Diode: Phalloidin/Actin Stain). Individual Nuclei/Cell
Boundaries were segmented with CellProfiler ([192]65) in combination
with in-house curation pipelines to ensure well-segmented single cells.
Briefly, an immunofluorescence-focused segmentation algorithm used the
DAPI stain to delineate the nucleus shape and the Phalloidin stain to
delineate the general cell shape. Approximately, 13,000 single cells
spanning all exposure conditions were procured for this work with an
additional 40,000 cells analyzed for single-cell clones.
Data processing and morphological analysis
Thirty-three key morphological parameters were extracted from each cell
using a cell profiler morphological analysis pipeline and accounted for
various degrees of communality as shown by a primary factor analysis
(Table [193]S1 and Fig. [194]S15). Although certain parameters capture
more variance than others, using all parameters with any degree of
variance provides a synergistic approach to better map morphological
heterogeneity (i.e. a parameter with lower magnitude communality may
capture variance not captured by a parameter with a high communality).
To compare morphological parameters of different scales to understand
population variance, all morphological parameters were independently
log normalized. The collection of parameters used in this study
captures the ensemble of outputs of the morphological profiling module
of cell profiler and was hypothesized to capture easily interpretable
morphological heterogeneity in the cell population.
This “normalized” morphological parameter dataset was subsequently used
to construct a 2D-uniform manifold and projection (UMAP) space
([195]66). UMAP is a nonlinear dimensionality reduction algorithm that
seeks to capture the structure of high dimensional data in a
lower-dimensionality space (for this work, the 33-vector space was
simplified down to 2). Each point in the UMAP space represents an
individual cell whose morphological parameters have been transformed
and projected onto the 2D-UMAP space. k-means clustering, an
unsupervised clustering method, was used to identify distinct
morphological groups within the cell dataset from the log-normalized
dataset. An optimal number of clusters, 10, was calculated by a plateau
in the inertia and silhouette values of the k-means algorithm which
partitioned the clusters in an unsupervised manner (Fig. [196]S16). To
quantify morphological heterogeneity, the Shannon entropy for each PM
exposure condition was calculated using the k-means clusters as
follows:
[MATH: S=−∑i=1
10pi⋅l
og(pi) :MATH]
(2)
where S is the Shannon entropy (greater magnitude signifies a more
heterogeneous population) and p[i] is the fraction of the population
that is in morphological cluster i ([197]59). For single-cell cloning
analysis, larger morphological CGs were created to identify overarching
morphological themes of the k-means clusters. Briefly, ward-based
clustering was performed on the average morphological signature across
each k-means cluster, and the analysis identified three morphological
groups that encompassed the k-means clusters.
γH2AX content per cell was analyzed through the mean nuclear intensity
of the fluorescent 488 channel as previously performed ([198]53).
Specifically, the summation of the pixel values (normalized to range
from 0 to 1) of the 488 channel was divided by the pixel area of the
encompassing nuclei. The resulting mean γH2AX expression was then
layered across the UMAP manifold and analyzed per cluster.
Approximately, 13,000 individual cells encompassing all PM exposures
were analyzed for the morphological analysis.
Single-cell cloning and live-cell imaging
Single cells of the BEAS-2B cell line were isolated using a Sony
Biotechnology MA900 Cell Sorter available through the Center for
Biomedical Research Support at UT Austin. Individual cells were sorted
into a 96-well plate and allowed to proliferate. Media exchanges of
BEGM were performed every 48 h. Cell populations were expanded to
collagen-coated 24-well plates, 6-well plates, and finally, T-75 flasks
before freezing cells to create multiple clonal populations. Clonal
populations were then similarly used in experiments as the parental
BEAS-2B population as described above. Approximately, 40,000 single
cells spanning all clones and urban exposure conditions were analyzed.
3′-Tag RNA sequencing
BEAS-2B cells were cultured and exposed to PM as described above.
Following the completion of 24 h exposures to the three PM types at 2
concentrations (125 and 500 µg/mL), RNA extraction was immediately
performed on n ≥ 4 replicates by lysing cells with TRIzol Reagent
(Invitrogen, 15596026). The RNA underwent DNase I treatment and was
purified using a Direct-zol RNA Miniprep Kit (Zymo Research, R2052)
according to manufacturer protocol. The purity of the RNA was confirmed
using a Nanodrop 2000 Spectrophotometer (Thermo Scientific), and RNA
concentration was determined using a Qubit 4 Fluorometer (ThermoFisher)
RNA broad range assay kit (ThermoFisher, [199]Q10210). Before library
preparation, RNA quality was determined using an Agilent Bioanalyzer,
and all samples used for sequencing had an RNA integrity number (RIN)
score >8.80. The RNA was submitted to the University of Texas genomic
sequencing and analysis facility for 3′ RNA-based library preparation
and sequencing based on previously published protocols ([200]47,
[201]48). Libraries were quantified using the Quant-it PicoGreen dsDNA
assay (ThermoFisher) and pooled equally for subsequent size selection
at 350–550 bp on a 2% gel using the Blue Pippin (Sage Science). The
final pools were checked for size and quality with the Bioanalyzer High
Sensitivity DNA Kit (Agilent) and their concentrations were measured
using the KAPA SYBR Fast qPCR kit (Roche). The pooled libraries were
sequenced on a NovaSeq6000 (Illumina) and a sequencing depth of 4.5
million reads per sample was achieved with a single-end, 100-bp read
length.
Differential gene expression analysis
Following sequencing, the raw reads were preprocessed to remove adapter
contamination and trim the unique molecular identifier barcodes, remove
duplicates, and remove poor-quality reads. The Human Reference Genome
was assembled and indexed using
Homo_sapiens.GRCh38.dna.primary_assembly.fa and
Homo_sapiens.GRCh38.104.gtf from Ensembl using the genomeGenerate run
mode in STAR (Spliced Transcripts Alignment to a Reference) version
2.7.0d ([202]67). The filtered reads were then aligned to the generated
genome using STAR. HTSeq ([203]68) was used to count the aligned reads
in each .bam file generated by STAR. The DESeq2 package ([204]69) was
then used to quantify differential gene expression in R ([205]70).
Differential expression was determined for each PM exposure condition
relative to the counts from unexposed control cells. Significantly DE
genes were defined as having a log[2](Fold Change) ≥ 1 and P[adj] <
0.05. GO term analysis was performed using the Enrichr web tool
([206]51) to determine GO Biological Process terms that were
significantly enriched in the sets of significantly DE genes.
Significant GO terms were defined as having P[adj] < 0.01. Chord plots
were constructed using the GOplot package in R.
Statistical analysis
One-way Student's t-tests were used to determine the significance of
the alamarBlue and AV–PI assays. For RNAseq differential expression,
the Wald test and a Bonferroni adjustment were used to determine
adjusted P-values. The linear regression of AV and γH2AX data was
evaluated using Pearson coefficient and P statistics. One-way ANOVA
tests were used to calculate the significance for γH2AX intensities and
all morphological comparisons across cell clusters including Shannon
Entropy.
Supplementary Material
pgad415_Supplementary_Data
[207]Click here for additional data file.^ (61.1MB, zip)
Contributor Information
Sean M Engels, McKetta Department of Chemical Engineering, University
of Texas at Austin, Austin, TX 78712, USA.
Pratik Kamat, Department of Chemical and Biomolecular Engineering,
Johns Hopkins University, Baltimore, MD 21218, USA.
G Stavros Pafilis, McKetta Department of Chemical Engineering,
University of Texas at Austin, Austin, TX 78712, USA.
Yukang Li, Department of Biology, Johns Hopkins University, Baltimore,
MD 21218, USA.
Anshika Agrawal, Department of Chemical and Biomolecular Engineering,
Johns Hopkins University, Baltimore, MD 21218, USA.
Daniel J Haller, Department of Chemical and Biomolecular Engineering,
North Carolina State University, Raleigh, NC 27606, USA.
Jude M Phillip, Department of Chemical and Biomolecular Engineering,
Johns Hopkins University, Baltimore, MD 21218, USA; Institute for
Nanobiotechnology, Johns Hopkins University, Baltimore, MD 21218, USA;
Department of Biomedical Engineering, Johns Hopkins University,
Baltimore, MD 21218, USA; Department of Oncology, Sidney Kimmel
Comprehensive Cancer Center, Baltimore, MD 21231, USA.
Lydia M Contreras, McKetta Department of Chemical Engineering,
University of Texas at Austin, Austin, TX 78712, USA; Institute for
Cellular and Molecular Biology, The University of Texas at Austin,
Austin, TX, 78712, USA.
Supplementary Material
[208]Supplementary material is available at PNAS Nexus online.
Funding
The authors acknowledge the funding support of this study from the
National Institutes of Health R21ES032124 (L.M.C.) and U01AG060903
(J.M.P.), the National Science Foundation EF-2022124 (L.M.C.), The
Johns Hopkins University Older Americans Independence Center of the
National Institute on Aging under award number P30AG021334 (JMP), and a
Hypothesis Fund Award (L.M.C.). This work was supported by the National
Science Foundation Graduate Research Fellowship Program under Grant No.
DGE 2137420 (S.M.E.). Any opinions, findings, and conclusions or
recommendations expressed in this material are those of the authors and
do not necessarily reflect the views of the National Science
Foundation. The authors acknowledge the Texas Advanced Computing Center
(TACC) at UT Austin for providing high-performance computing resources.
TagSeq deduplication was provided by the Bioinformatics Consulting
Group at The University of Texas at Austin Computational Biology and
Bioinformatics Core Facility. RRID:SCR_022688. The authors also thank
the Genomic Sequencing and Analysis Core Facility at UT Austin,
RRID:SCR_021713, for assistance with RNA-sequencing.
Author Contributions
S.M.E., P.K., J.M.P., and L.M.C. designed the study. S.M.E., P.K.,
G.S.P., Y.L., A.A., and D.J.H. performed the experiments. S.M.E. and
P.K. led the analysis of the results. S.M.E., P.K., J.M.P., and L.M.C.
wrote the manuscript. S.M.E., P.K., J.M.P., L.M.C., G.S.P., and D.J.H.
edited the manuscript. L.M.C. and J.M.P. supervised the study.
Preprint
This manuscript was posted on a preprint:
[209]https://doi.org/10.1101/2023.05.17.541204.
Data Availability
The authors declare that all the data supporting the findings of this
study are either available within the paper and its [210]Supplementary
material or available in public repositories. Raw sequencing data are
available at the National Center for Biotechnological Information
(NCBI) short read archive (SRA) under BioProject Accession no.
PRJNA954385. Scripts for morphological analyses can be found at
[211]https://github.com/pkamat22/PM_MORPH.
References