Mapping multi-omics characteristics related to short-term PM2.5 trajectory and their impact on type 2 diabetes in middle-aged and elderly adults in Southern China

Full text

Journal of Hazardous Materials 468 (2024) 133784
Available online 15 February 2024
0304-3894/© 2024 Elsevier B.V. All rights reserved.
Research Paper
Mapping multi-omics characteristics related to short-term PM
2.5
trajectory
and their impact on type 2 diabetes in middle-aged and elderly adults in
Southern China
Jia-ting Wang
a
,
1
, Wei Hu
a
,
1
, Zhangzhi Xue
b
,
c
,
1
, Xue Cai
b
,
c
,
1
, Shi-yu Zhang
a
, Fan-qin Li
a
,
Li-shan Lin
a
, Hanzu Chen
a
, Zelei Miao
b
,
c
, Yue Xi
a
, Tiannan Guo
b
,
c
, Ju-Sheng Zheng
b
,
c
,
*
, Yu-
ming Chen
a
,
**
, Hua-liang Lin
a
,
**
a
Guangdong Provincial Key Laboratory of Food, Nutrition and Health, Department of Epidemiology, School of Public Health, Sun Yat-sen University, Guangzhou
510080, China
b
Westlake Center for Intelligent Proteomics, Westlake Laboratory of Life Sciences and Biomedicine, Hangzhou, Zhejiang Province, 310030, China
c
School of Medicine, School of Life Sciences, Westlake University, Hangzhou, Zhejiang Province, 310030, China
HIGHLIGHTS GRAPHICAL ABSTRACT
•Trajectory grouping in short-term PM
2.5
exposures effectively capture variations
in microbiota composition.
•Novel PM
2.5
-related multi-omics bio-
markers well predicted higher T2D
prevalence and incidence.
•The crucial PM
2.5
-related biomarkers
include fecal rhamnose and glycylpro-
line, serum hippuric acid, and protein
TB182.
•The subnetwork correlation provides
critical insights into the molecular
mechanisms underlying PM
2.5
-related
T2D impacts.
ARTICLE INFO
Keywords:
PM2.5
Metagenome
Proteome
Metabolome
Type 2 diabetes
ABSTRACT
The relationship between PM
2.5
and metabolic diseases, including type 2 diabetes (T2D), has become increas-
ingly prominent, but the molecular mechanism needs to be further claried. To help understand the mechanistic
association between PM
2.5
exposure and human health, we investigated short-term PM
2.5
exposure trajectory-
related multi-omics characteristics from stool metagenome and metabolome and serum proteome and metab-
olome in a cohort of 3267 participants (age: 64.4 ±5.8 years) living in Southern China. And then integrate these
features to examine their relationship with T2D. We observed signicant differences in overall structure in each
* Corresponding author at: Westlake Center for Intelligent Proteomics, Westlake Laboratory of Life Sciences and Biomedicine, Hangzhou, Zhejiang Province,
310030, China.
** Corresponding authors.
E-mail addresses: zhengjusheng@westlake.edu.cn (J.-S. Zheng), chenyum@mail.sysu.edu.cn (Y.-m. Chen), linhualiang@mail.sysu.edu.cn (H.-l. Lin).
1
These authors contributed equally to this work.
Contents lists available at ScienceDirect
Journal of Hazardous Materials
journal homepage: www.elsevier.com/locate/jhazmat
https://doi.org/10.1016/j.jhazmat.2024.133784
Received 14 December 2023; Received in revised form 29 January 2024; Accepted 12 February 2024

Journal of Hazardous Materials 468 (2024) 133784
2
omics and 193 individual biomarkers between the high- and low-PM
2.5
groups. PM
2.5
-related features included
the disturbance of microbes (carbohydrate metabolism-associated Bacteroides thetaiotaomicron), gut metabolites
of amino acids and carbohydrates, serum biomarkers related to lipid metabolism and reducing n-3 fatty acids.
The patterns of overall network relationships among the biomarkers differed between T2D and normal partici-
pants. The subnetwork membership centered on the hub nodes (fecal rhamnose and glycylproline, serum hip-
puric acid, and protein TB182) related to high-PM
2.5
, which well predicted higher T2D prevalence and incidence
and a higher level of fasting blood glucose, HbA1C, insulin, and HOMA-IR. Our ndings underline crucial PM
2.5
-
related multi-omics biomarkers linking PM
2.5
exposure and T2D in humans.
1. Introduction
Atmospheric ne particulate matter, particularly particulate matter
with an aerodynamic diameter of ≤2.5 µm (PM
2.5
) is regarded as a key
component of air pollution. PM
2.5
is known to impact a broad range of
health issues, including chronic obstructive pulmonary disease, car-
diometabolic diseases, and cancer [1,2], through mechanisms such as
oxidative stress, systemic inammation, and epigenetic modications
[3,4]. However, the specic multi-omics biomarkers of PM
2.5
exposure,
particularly those relating to host and gut microorganisms, remain
largely unexplored in humans.
Recent evidence suggests that PM
2.5
can disturb gut microbiota in
both experimental animals [5–7] and humans [8,9]. Exposure to PM
2.5
has been associated with decreased microbial alpha diversity [9] and
reduced short-chain fatty acid (SCFA) producing taxa like Bacteroidaceae
[10], which are often inversely related to metabolic syndrome [11].
Furthermore, the abundance of gut genera such as Enorma and afliated
Coriobacteriia, which are involved in various metabolic processes, may
also be affected by PM
2.5
exposure [10,12–14]. However, these studies
have limitations, including small sample size (range: 43−76) or assess-
ments based on long-term average emissions rather than specic daily
exposure measurements. The susceptibility and rapid response of gut
microbiota to environmental changes for several days, including PM
2.5
exposure [12,15], underscore the importance of examining short-term
exposure effects. And short-term study within a single region is
possible to avoid the ecological bias that emerged in multi-center
long-term exposure studies. In addition, PM
2.5
exposure-related fecal
metabolites as a functional readout of the gut microbiome also need to
be further explored [16].
Circulating proteins and metabolites play a crucial role in the path-
ophysiological mechanisms of chronic diseases [17,18]. Exposure to
300
μ
g PM
2.5
/m
3
for 2 h has been found to increase cardiovascular
disease-related proteins such as fractalkine, apolipoprotein B, and ma-
trix metalloproteinase-12 [19]. Moreover, metabolites associated with
short-term PM
2.5
exposure are primarily enriched in pathways like
linoleic acid and sphingolipid metabolism [12,13,20]. Given the limi-
tations of mono-omics studies or those with sparse samples, a compre-
hensive, multi-omics integration from the same samples in a large-scale
study is essential to uncover new insights into the relationship between
PM
2.5
exposure and health.
Type 2 diabetes (T2D) is a systemic metabolic disease with a growing
prevalence of comorbidity rate [21]. Around 20% of the global T2D
burden in 2019 was attributed to PM
2.5
exposure [22]. While the asso-
ciation between long-term PM
2.5
exposure and T2D is established, the
specic omics-biomarkers involved are not fully understood. Moreover,
conducting long-term clinical trials on humans to study the effect of
different PM
2.5
exposure levels on T2D is impractical. Based on the
hypothesis that the impact of short-term and medium to long-term PM
2.5
exposure on biomarkers is largely consistent [23–25], we devote to
explore how short-term PM
2.5
-related biomarkers relate to T2D presence
and incidence.
Previous short-term PM
2.5
exposure studies often overlooked the
trajectory of exposure in the days leading up to specimen collection,
instead using average values over several days. This approach may not
accurately capture the varied impacts of daily exposures at different
times relative to when the specimen is collected. To bridge existing
research gaps, our study embarked on a prospective cohort study
involving 3267 participants. These participants were categorized based
on their PM
2.5
exposure trajectories during the 7 days leading up to the
specimen collection. Our primary objective was to establish a connec-
tion between short-term PM
2.5
exposure trajectories and various multi-
omics characteristics, including stool microbiome and metabolome, as
well as blood proteome and metabolome. Subsequently, we delved into
understanding the associations between PM
2.5
-related biomarkers and
the presence and incidence of T2D among adults in Southern China.
2. Methods
2.1. Study design and participants
This study forms a part of the Guangzhou Nutrition and Health Study
(GNHS), a prospective cohort based in Southern China registered at
www.clinicaltrials.gov (No. NCT03179657) [26]. The cohort included
4048 participants, aged 40−75 years, residing in Guangzhou for at least
ve years. Participants were followed up every three years. For this
multi-omics integrating analysis, we included 3267 participants who
provided stool and/or fasting blood samples between 2014 and 2020
and had detailed residential addresses, as depicted in the Abstract
Graphic. The Ethics Committee of the School of Public Health, Sun
Yat-Sen University, granted approval for the study. Written informed
consent was obtained from all participants.
2.2. PM
2.5
and meteorological parameter estimation
To assign levels of ambient PM
2.5
and meteorological parameters, we
recorded the residential address and sampling time for each participant.
Daily mean PM
2.5
concentrations were sourced from the ChinaHigh-
AirPollutants datasets (CHAP, accessible at https://weijing-rs.github.io
/product.html). Further details on CHAP are available in the original
publications [27,28] and supplementary material. Meteorological data,
including temperature and relative humidity, were obtained from the
fth generation of European Reanalysis dataset, provided by the Euro-
pean Center for Medium-Range Weather Forecasts [29,30]. Participants’
addresses were geocoded into longitude and latitude coordinates using
the Application Programming Interface from Auto Navi Map (https
://www.autonavi.com). Subsequently, a bilinear interpolation method
was utilized to determine individual exposure levels [31], as elaborated
in the supplementary material.
In our preliminary analyses, we explored the explanation of the
differences in the composition of each omics data at different lag periods
(1–30 days) of PM
2.5
exposure (Fig. S1). Considering our initial results,
we elected to conduct our study focusing on a 7-day lag period for short-
term PM
2.5
exposure assessment.
2.3. Collection of metadata and biological sample and T2D diagnosis
Metadata encompassing demographics, anthropometrics, and life-
style factors, which are potentially relevant to gut microbiota and
physiological properties and are unlikely to be an intermediate variable
between PM
2.5
exposure and omics biomarkers, were selected as
J.-t. Wang et al.

Journal of Hazardous Materials 468 (2024) 133784
3
controlled variables [15,32]. These covariables included sex, age, body
mass index (BMI), household income, education level, smoking status,
alcohol consumption, physical activity, and energy intake. Data were
collected by trained staff at follow-up sites. Comprehensive details on
covariate assessment, specimen collection, and T2D-related indices
testing, including fasting blood glucose (FBG), glycated hemoglobin A1c
(HbA1c), insulin, and homeostatic model assessment of insulin resis-
tance (HOMA-IR), are provided in the supplementary material. T2D
cases were dened according to the criteria of the American Diabetes
Association and the World Health Organization, which included one or
more of the following: FBG ≥7.0 mmol/L, HbA1c ≥6.5%, or
self-reported treatment with insulin or other hypoglycemic [33,34].
2.4. Omics measurements
2.4.1. Shotgun metagenomic sequencing and taxonomic and functional
proling
As detailed in our previous paper [35] and supplementary material,
fecal DNA was extracted using the QIAamp DNA Stool Mini Kit (Qiagen,
Hilden, Germany) and quantied with the Qubit quantication system
(Thermo Scientic, Wilmington, DE, US). Shotgun metagenomic
sequencing was conducted on the Illumina HiSeq platform (Illumina
Inc., CA, USA) utilizing a 2 ×300-bp paired-end read protocol, yielding
an average of 44.6 million (minimum, 22.1 million; maximum, 100
million) paired-end raw reads per sample. Taxonomic proling was
performed with MetaPhlAn2 (version 2.6.02) using default settings and
quality-controlled sequencing reads [36]. Functional proling of the
metagenome was carried out using HUMAnN3 (version 2.8.1) [37],
generating microbial pathways based on the MetaCyc database [38,39].
2.4.2. Fecal and blood metabolomics proling
Targeted metabolomics proling of fecal and serum samples was
performed using the Q300 Kit by Metabo-Prole Corp. (Shanghai,
China), as described in our previous study [40] and supplementary
material. This kit allows for the analysis of up to 310 metabolites across
12 biochemical classes. Samples underwent pretreatment, mixing with
internal standards, and analysis via ultra-performance liquid chroma-
tography coupled to tandem mass spectrometry (UPLC-MS/MS). The
system included a C18 analytical column (2.1 ×100 mm, 1.7
μ
M, 40 ◦C).
Quality control using mixed samples was conducted after every 14
samples. Raw data calibration and features annotation with MSI level 1
condence were performed using QuanMET software (v2.0,
Metabo-Prole, Shanghai, China).
2.4.3. Circulating proteomics proling
For proteome analysis, serum samples were analyzed by data-
independent acquisition mass spectrometry (DIA-MS) as previously
described [41,42]. This approach allows for accurate quantication with
high reproducibility in complex proteomes. After digestion, peptides
were analyzed over a 20 min linear liquid chromatography gradient on a
TripleTOF 5600 system (SCIEX, CA, USA) coupled with the Eksigent
NanoLC 400 System (Eksigent, Dublin, CA, USA). A serum spectral li-
brary containing 3474 peptide precursors and 536 unique proteins from
the Swiss-Prot Homo sapiens database [43] was employed for wiff les
analysis using DIA-NN software (version 1.7.12). Further details on
peptide extraction, detection, DIA-NN settings, and quality control are
provided in the supplementary material.
2.5. Statistical analyses
2.5.1. Data preprocessing and transformation
To maintain a focus on widely present variables and to mitigate the
effects of zero ination, we included microbial taxa with a prevalence
greater than 10%. For individuals’ data, metabolites and proteins below
the detection limit were assigned a zero value. Features exhibiting more
than 90% zeros were excluded. Potential batch effects were corrected
using the “Combat” algorithm. Pseudo-counts, one order of magnitude
lower than the minimum detectable level, were assigned to samples
below the detection limit, which minimally altered data distribution.
Transformations included a centered log-ratio for microbial taxonomic
relative abundances and a log transformation for metabolite and protein
concentrations. This transformed data was then scaled to normalize
dimension. All statistical analyses were conducted using R software
(version 4.1.0).
2.5.2. PM
2.5
exposure trajectory grouping
To accurately capture the varying impacts of daily PM
2.5
exposures at
different times prior to specimen collection, our study employed group-
based trajectory modeling (GBTM) using the gbmt package. This method,
initially proposed by Nagin and Land in 1993, is designed to identify
latent groups within the population based on similar time-based
response patterns, thus effectively summarizing complex exposure
data [44,45]. We classied individuals into higher (H-PM
2.5
) and lower
(L-PM
2.5
) exposure groups based on PM
2.5
levels in the 7 days preceding
biological specimen collection. Adhering to standard practices, we set
the polynomial degree of group trajectories at 2 and assigned individuals
to these groups based on their highest estimated probability of
belonging. To validate our model, we ensured the average posterior
probability of assignment (APPA) for individuals in each group exceeded
a threshold of 0.7 [45]. These classications were crucial in all further
analyses of our study.
2.5.3. Differential omics features between PM
2.5
exposure groups
To analyze the β-diversity of each omics category, we used the vegan
package to calculate inter-individual dissimilarity matrices based on
composition. This included the Bray-Curtis distance for metagenome
and metabolome analyses, and the Euclidean distance for the proteome.
Principal coordinates were then derived through classical metric
multidimensional scaling based on these distance matrices. Differential
abundances of omics features between PM
2.5
exposure groups were
identied using the diff-analysis function in the R package Micro-
biotaProcess [46], as detailed in the supplementary material. We applied
a multiple linear regression model to estimate the associations (stan-
dardized mean difference [SMD]) of the identied biomarkers with high
PM
2.5
exposure (H-PM
2.5
), adjusted for potential confounding cova-
riates. Model I accounted for age, sex, BMI, household income, educa-
tion level, smoking status, alcohol consumption, energy intake, physical
activity, and antibiotic use (for gut bacteria). Model II included addi-
tional adjustments for temperature and relative humidity, averaged over
7 days prior to sampling, and treated as natural cubic splines with three
degrees of freedom, to address nonlinear relationships.
Binary logistic regression was employed to establish predictive
models for the screening features of each omics category. This aimed
validate PM
2.5
exposure grouping and assess the predictive ability of the
principal PM
2.5
-related biomarkers in Model II (FDR p <0.1) for T2D
prevalence. The area under the curve of the receiver operating charac-
teristic (ROC-AUC) was used to evaluate the models’ predictive perfor-
mance. Additionally, the c-index of the Cox proportional hazards model
was used to evaluate the prediction of T2D incidence by the principal
PM
2.5
-related biomarkers.
2.5.4. Metabolic pathway enrichment
Pathway enrichment analysis of PM
2.5
-related metabolites was con-
ducted using MetaboAnalyst 5.0, with "Homo sapiens" as the model
organism. A hypergeometric test determined whether these PM
2.5
-
related metabolites were represented in KEGG metabolic pathways more
than expected by chance. Statistical signicance was indicated by p-
values <0.05.
2.5.5. Network correlation analysis
Pairwise partial Spearman’s rank correlation analyses were per-
formed across the sample sets comprising all four omics categories. We
J.-t. Wang et al.

Journal of Hazardous Materials 468 (2024) 133784
4
constructed a network of PM
2.5
-related biomarkers with signicant
correlations (|r| >0.15, FDR p <0.05) using Cytoscape 3.8.2. The top 10
biomarkers, ranked based on a combination of three network node at-
tributes (degree, closeness, and betweenness), were identied as hub
nodes. Subnetworks centered on each of the crucial hub nodes that were
recognized as principal PM
2.5
-related biomarkers in Model II (FDR p <
0.1), were extracted from the overall network. We calculated the cor-
relation between the predicted probability of higher PM
2.5
exposure
(determined by subnetwork membership centered on each crucial hub
node) and the prevalence of T2D, as well as the predicted values of T2D-
related indices (FBG, HbA1c, insulin, and HOMA-IR), as described in the
supplementary material.
3. Results and discussion
3.1. Characteristics of the participants and exposure grouping
Among the 3267 participants (average age: 64.4 ±5.8 years), 67.9%
were female (Table S1). Over an average follow-up period of 5.15 years
(SD: 1.68 years) post-omics determination, there were 608 (18.7%)
prevalent and 263 incident T2D cases. The daily average PM
2.5
exposure
levels of the study population living in Guangzhou (2014–2020) showed
signicant variation, ranging from 5.4 to 244.8
μ
g/m
3
(median: 35.40,
IQR: 23.65). Two distinct PM
2.5
exposure trajectory groups (L-PM
2.5
vs.
H-PM
2.5
, Fig. 1a−1b) were identied in the 7 days preceding biological
specimen collections for both fecal and blood samples. The APPA for
these groupings ranged from 0.94 to 0.97, well above the 0.7 threshold
for model adequacy [45]. The mean (SD) 7-day PM
2.5
concentrations
were signicantly higher in the H-PM
2.5
group compared to the L-PM
2.5
group for both stool (54.8 [17.5] vs. 28.3 [7.2],
μ
g/m
3
) and serum
(52.73 [9.14] vs. 30.03 [7.49],
μ
g/m
3
) samples (p <0.001) (Table S1).
The PM
2.5
level in the L-PM
2.5
was lower than Class I standards of Chi-
nese ambient air quality standards implemented since 2016 and in the
H-PM
2.5
group was lower than Class II [47]. This substantial variability
in exposure allows us to explore the impact of varying short-term
exposure levels on multi-omics biomarkers.
3.2. PM
2.5
trajectories and stool metagenome and metabolome
H-PM
2.5
exposure was associated with signicant differences in gut
microecology β-diversity compared to L-PM
2.5
(Fig. 1c). The 7-day PM
2.5
exposure trajectories accounted for more variance in bacterial β-di-
versity than either 7-day or 1-year cumulative exposure and general
demographic characteristics (Fig. S2). This indicates that short-term
exposure studies provide more insight into microbiota composition
changes when using trajectory grouping instead of cumulative exposure
calculations.
In comparing fecal biomarkers between L-PM
2.5
and H-PM
2.5
groups,
we identied signicant differences in110 out of 426 bacterial taxa and
36 out of 199 fecal metabolites (p<0.05, FDR p<0.1, LDA >2.0,
Fig. 2a, Tables S5−S6). The combined fecal biomarkers of differential
taxa and metabolites accounted for 24.7% of the variance in PM
2.5
Fig. 1. Trajectory groups of estimated 7-day PM
2.5
exposure and the overall variance of omics. PM
2.5
exposure trajectories were estimated for 7 days prior to
sampling based on participant’ residential addresses. Exposure trajectories were modeled separately for stool and serum samples due to subtle differences in sampling
time. A total of 1876 individuals were subject to trajectory grouping for stool samples (a) and 2804 for serum samples (b). The average posterior probability of
assignments for the models ranged from 0.94 to 0.97. (c) Shows the standardized mean difference (SMD) of beta-diversity indices of each omics category (inter-
individual dissimilarity matrix calculated using Bray-Curtis distance for metagenome and metabolome, and Euclidean distance for proteome) between PM
2.5
exposure groups, with the L-PM
2.5
group as a reference. (d) Illustrates the Z-score of the Firmicutes/Bacteroidetes ratio difference between PM
2.5
exposure groups. (e)-
(g) Depict differences in the composition of bacterial phyla, and stool and serum metabolite classications between the H- and L-PM
2.5
groups, respectively. Ab-
breviations: Meta.Gen, Metagenome; Stool.Met, Stool metabolome; Ser.Pro, Serum proteome; Ser.Met, Serum metabolome; Act., Actinobacteria; Pro., Proteobacteria;
Ben., Benzenoids; Car., Carbohydrates; Fat., Fatty acids; Org., Organic acids.
J.-t. Wang et al.

Journal of Hazardous Materials 468 (2024) 133784
5
grouping, with a ROC-AUC of 0.788 (95% condence interval [CI]:
0.760 −0.817) for inverse prediction of PM
2.5
grouping (Fig. 2d−2e).
Eight bacterial taxa from the Bacteroidetes and Actinobacteria phyla and
eight fecal metabolites, including class of amino acids, bile acids, and
carbohydrates, were identied as the principal PM
2.5
-related fecal bio-
markers after adjusting for potential multi-covariates (FDR p<0.1 in
Model II, Fig. 2b−2c). Consistent with previous studies [10,12,13], we
found short-term PM
2.5
exposure was correlated with decreased abun-
dance in the Bacteroidetes phylum and an increase in a branch under the
Actinobacteria phylum, including class Coriobacteriia, order Coriobacter-
iales, family Coriobacteriaceae, genus Enorma, and species Collinsella
massiliensis (Table S5). Additionally, a positive association of 7-day
PM
2.5
exposure with the dominant Firmicutes phylum and a higher Fir-
micutes/Bacteroidetes ratio at the phylum level in H-PM
2.5
(p=4.4 ×10
-5
; Fig. 1d−1e and Table S2) was observed. This might be
supported by a 20-day PM
2.5
exposure experimental study [5]. However,
opposite associations were noted in studies involving medium- or
long-term PM
2.5
exposure with the Firmicutes phylum [8,9]. Exposure to
air pollution is an ongoing process, and animal experiments have
observed some persistent biomarkers in long-term exposure scenarios
[23–25]. The reasons for the inconsistency between this study and
previous medium- or long-term PM
2.5
exposure studies may partially
due to the differences in study regions (a key determinant [48], subjects’
characteristic (like pregnant subjects vs. general older adults), and the
specic roles of taxa within the Firmicutes phylum. Pregnancy is known
to be associated with a profound alteration of the gut microbiota, with a
trend toward a decrease in the abundance of Firmicutes with longer
gestation [49,50]. As dominant intestinal bacteria, Firmicutes include
health-benecial bacteria involved in host nutrition and metabolism
through SCFA synthesis [51], such as Lachnospiraceae and Clostridiaceae,
which were negatively associated with PM
2.5
exposure and mediated
T2D effects [9]. Conversely, some opportunistic pathogens in Firmicutes,
e.g., Enterococcus and Ruminococcus, may be stimulated by PM
2.5
expo-
sure [52]. This was conrmed in our data, where the opportunistic
pathogen species Ruminococcus gnavus was positively associated with
H-PM
2.5
exposure (Table S5). These results underscore the importance of
considering PM
2.5
-related specic organisms with biological functions
within the bacterial phyla generally associated with diseases such as
obesity or metabolic disorders [53].
In this study, the association of several recognized probiotics with
PM
2.5
exposure was notably signicant. Bacteroides thetaiotaomicron
showed an inverse relationship with PM
2.5
, whereas Bidobacterium
longum and Bacteroides dorei were more abundant in the H-PM
2.5
group
(Fig. 2b). The interactions among microbial taxa in complex commu-
nities could be pivotal in responding to external environmental distur-
bances [54]. A co-occurrence pattern of the genus Enorma and
Bidobacterium with incident atrial brillation was observed in a
large-scale population-based study involving 6763 individuals [55].
Fig. 2. Differential stool community characteristics between PM
2.5
trajectory groups. (a) Identies of 110 bacterial taxa with signicant inter-group differences
(p<0.05, FDR p<0.1, LDA >2.0). (b) Presents LDA scores and standardized mean differences (SMD) of microbial biomarkers with FDR p<0.1 between the H- and
L-PM
2.5
groups in Model II. (c) Shows stool metabolites with FDR p<0.1 between the H- and L-PM
2.5
groups in Model II. (d) Indicates the percentage of the variance
of PM
2.5
grouping explained by biomarkers of the fecal microbial metagenome and metabolome, individually and combined. (e) Demonstrates the area under the
curve of the receiver operating characteristic (ROC-AUC) for the predictive models of PM
2.5
grouping validity using screened markers. Covariates in Model I included
age, sex, BMI, household income, education level, smoking status, alcohol consumption, energy intake, physical activity, and antibiotic use (for gut bacteria); Model
II additionally adjusted for temperature and relative humidity as natural cubic splines (with three degrees of freedom).
J.-t. Wang et al.

Journal of Hazardous Materials 468 (2024) 133784
6
Similarly, B. dorei has been shown to promote the proliferation of Bi-
dobacterium in coculture assays [56]. However, the decline in the
B. thetaiotaomicron, a carbohydrate degrading bacterium, along with the
disruption of the entire microbial community, may contribute to meta-
bolic disorders in the gut.
Although there are no reports directly linking PM
2.5
exposure to fecal
metabolites, the PM
2.5
-related fecal metabolic classes identied in this
study (mainly belonging to carbohydrates and amino acids, Fig. 2c)
align closely with those found in blood in prior researches [12,13,20,
57]. All eight principal biomarkers of fecal metabolites in this study
were inversely related to PM
2.5
exposure (SMD [95%CI]: ranging from
−0.200 [−0.360− − 0.040] for guanidoacetic acid to −0.310
[−0.469− − 0.151] for L-proline, Table S6). L-proline is crucial for
maintaining redox and nucleotide homeostasis [58]. Elevated levels of
circulating proline have been observed in patients with T2D, obesity,
and insulin resistance [58]. The reduction of L-proline in stools in the
H-PM
2.5
group implies a potential disruption in PM
2.5
-related excretion
of proline. Ornithine, an essential amino acid for ammonia detoxica-
tion, has been linked to mitigating hepatic pathogenesis induced by
acetaminophen [59]. The carbohydrate N-acetyl-D-glucosamine has
been noted for its indirect role in immune regulation and potential
anti-tumor effects [60]. Rhamnose has been reported to alleviate
symptoms of T2D by modulating the intestinal microbiota [61].
Collectively, the inuence of PM
2.5
on gut microecology seems to
involve disruptions in microbial composition (notably the carbohydrate
metabolism-related B. thetaiotaomicron) and the dysregulation of gut
metabolism, particularly in amino acids and carbohydrates.
3.3. PM
2.5
exposure trajectories and serum proteome and metabolome
Signicant differences in the composition of circulation metabolism
were observed between H-PM
2.5
and L-PM
2.5
groups (Fig. 1c). We
identied 17 differential serum proteins from 386 detected proteins and
30 serum metabolites from 199 targeted metabolites between these
2 PM
2.5
groups (p<0.05, FDR p<0.1, LDA >2.0, Tables S7−S8). After
adjusting for covariates, 12 proteins and 13 metabolites emerged as
principal PM
2.5
-related serum biomarkers (FDR p<0.1 in Model II,
Fig. 3a−3b). Combined, these serum biomarkers explained 19.9% of the
variance in PM
2.5
grouping and yielded a ROC-AUC of 0.765 (95% CI:
0.731–0.798) for their inverse prediction of PM
2.5
grouping
(Figs. 3c–3d).
Serum proteins inversely related to PM
2.5
in this study (SMD:
−0.086− − 0.170, Fig. 3a) are implicated in lipid metabolism (apolipo-
protein C-II [APOC2] and alpha-N-acetylneuraminide alpha-2,8-
Fig. 3. Differential biological features in serum between H- and L-PM
2.5
trajectory groups. (a) Displays LDA scores and SMDs of host serum proteins with signicant
inter-group differences (p<0.05, FDR p<0.1, LDA >2.0) and between the H- and L-PM
2.5
groups in Model II (FDR p<0.1). (b) Present serum metabolites with
signicant differences (p<0.05, FDR p<0.1, LDA >2.0) in inter-group analysis and between the H- and L-PM
2.5
groups in Model II (FDR p<0.1). (c) Indicates the
percentage of variance in PM
2.5
grouping explained by biomarkers of the host serum metabolome, proteome, and their combined effect. (d) Shows the ROC-AUC for
predictive models of PM
2.5
grouping validity based on the screened markers. Covariates adjusted in Model I included age, sex, BMI, household income, education
level, smoking status, alcohol consumption, energy intake, and physical activity. Model II additionally adjusted for temperature and relative humidity as natural
cubic spline (with three degrees of freedom).
J.-t. Wang et al.

Journal of Hazardous Materials 468 (2024) 133784
7
sialyltransferase [SIA8A]), coagulation (bronectin [FINC]), and neuron
navigator 3 (NAV3). Previous small-scale studies (n =14−56) exploring
PM
2.5
-related serum proteins have shown that air pollution exposure can
increase levels of APOM, APOB, brinogen, and sphingomyelin [19,62,
63]. Additionally, PM
2.5
-related disruption of sphingolipid metabolism
was characterized by a reductions in sphingosine and sphingomyelin,
along with increased serum insulin, HOMA-IR, and a decreased insulin
action index [13].
Conversely, serum proteins enriched in H-PM
2.5
(SMD:
0.092−0.150) included carbonic anhydrases 1 (CAH1), hepatocyte nu-
clear factor 4-alpha (HNF4A), mitochondrial ribosomal protein L19
(RM19), phosphate-regulating neutral endopeptidase (PHEX), comple-
ment C1r subcomponent-like protein (C1RL), 182 kDa tankyrase-1-
binding protein (TB182), and adrenocortical dysplasia protein homo-
log (ACD). Elevated levels of these proteins might be associated with an
increased risk of myocardial infarction, cardiac hypertrophy, T2D,
esophageal squamous cell carcinoma, and chronic lymphocytic leuke-
mia [64–66].
The principal PM
2.5
-related serum metabolites were predominantly
fatty acids (SMD: −0.210− − 0.261, Fig. 3b). PM
2.5
exposure was
inversely correlated with serum levels of docosahexaenoic acid (DHA),
n-3 docosapentaenoic acid (DPA), and 10Z-heptadecenoic acid. Similar
ndings of L-PM
2.5
-related DHA and 10Z-heptadecenoic acid were
observed in studies on air pollution [67,68]. DHA and DPA are known to
be benecial for cardiometabolic health and overall survival [69],
suggesting that H-PM
2.5
may adversely affect cardiometabolic health by
reducing circulating n-3 fatty acids.
Furthermore, PM
2.5
exposure showed a positive association with the
phenylalanine metabolite 3-(3-hydroxyphenyl)-3-hydroxypropanoic
acid (HPHPA, SMD: 0.243 [95% CI: 0.080−0.406]) and hippuric acid
(SMD: 0.214 [95% CI: 0.051−0.377]), but an inverse relationship with
pyruvic acid (SMD: −0.164 [95% CI: −0.315− − 0.013]) and N-acetyl-L-
aspartic acid (NAA, SMD: −0.167 [95% CI: −0.329− − 0.006]). Elevated
HPHPA levels have been linked to increased autism risk [70], while
higher serum hippuric acid may be associated with adverse cardiovas-
cular outcomes [71]. Conversely, pyruvic acid, which was also found to
be inversely correlated with ne particulate matter oxidative potential
[72], was reduced in patients with pulmonary arterial hypertension
[73]. PM
2.5
-related disruptions in aspartate metabolism have been
previously proposed [63,74]. The reduction of NAA, a benecial amino
acid metabolite important for neural health [75], was noted as a po-
tential hypoxia biomarker [76]. These ndings suggest the potential
systemic physiological changes associated with short-term PM
2.5
expo-
sure and their implications for health.
3.4. PM
2.5
-related metabolic pathway
Pathway enrichment analysis of the principal PM
2.5
-related metab-
olites identied 3 and 8 signicant metabolic pathways originating from
microbiota metabolism and co-metabolism by both microbiota and host,
respectively (Table S9). These pathways, particularly those involving
amino acid and energy metabolism, align with ndings from previous
studies as summarized in a systematic review on air pollution [63].
These metabolic pathways may provide mechanistic insights into the
PM
2.5
-related functions of gut bacteria (Table S10). For example, our
results indicated that the primary function of B. thetaiotaomicron was
dTDP-L-rhamnose biosynthesis I (Fig. S3, Table S10). The reduced
concentration of fecal rhamnose (Fig. 2c), linked to the pathway of
fructose and mannose metabolism, corresponded with the diminished
presence of B. thetaiotaomicron in the H-PM
2.5
group. This reduction is
likely due to PM
2.5
exposure impacting the gut’s B. thetaiotaomicron,
known for its efcient metabolism of glycans [77]. These ndings sug-
gest that PM
2.5
-induced disturbances in the gut microbiota may illumi-
nate the underlying mechanisms of PM
2.5
-related disorders in
disease-relevant metabolic patterns.
3.5. Connections between PM
2.5
-related multi-omics biomarkers and T2D
The predictive accuracy for T2D prevalence, measured by ROC-AUCs
(95% CI), was 0.627 (0.583−0.670) using combined markers of PM
2.5
-
related fecal microbiota and metabolites, 0.707 (0.662−0.752) using
host serum proteins and metabolites, and 0.761 (0.718−0.803) when
considering all these biomarkers together (Fig. 4a−4b). For T2D inci-
dence prediction, the corresponding c-index (95% CI) values of the Cox
proportional hazards model were 0.656 (0.594−0.719), 0.715
(0.644−0.787), and 0.763 (0.698−0.827) for these biomarkers,
respectively (Fig. 4c). While the association between PM
2.5
exposure and
increased T2D risk is well-documented [22], further research is neces-
sary to elucidate the underlying mechanisms.
In this study, we observed a less robust correlation among PM
2.5
-
related biomarkers within T2D subjects compared to the total partici-
pant group in a network comprising 129 nodes and 509 edges (average
value of absolute partial Spearman’s rank correlation coefcients: 0.168
vs. 0.211, p<0.001, Fig. 4d−4e, Table S11). This was corroborated by
sensitivity analyses involving a random sample of non-T2D subjects
matched in size to T2D group (Fig. S4, Table S11). Among the top ten
hub nodes in the network, four (fecal metabolite rhamnose and glycyl-
proline, serum metabolite hippuric acid, and protein TB182) were
principal PM
2.5
-related biomarkers in Model II (Table S12), identied as
crucial hub nodes.
The subnetwork membership centered on each of these crucial hub
nodes demonstrated effective performance in predicting T2D (ROC-
AUCs: 0.643−0.665, Figs. S6−S7). Notably, the direction of the pre-
dicted probability of H-PM
2.5
exposure was consistent with that of T2D
prevalence and T2D-related indices (FBG, HbA1C, insulin, and HOMA-
IR, Fig. 5 and Fig. S6), although individual associations of these
PM
2.5
-related biomarkers with T2D were not strongly signicant
(Fig. S8).
Fecal rhamnose, one of the crucial hub nodes, correlated with several
PM
2.5
-related bacterial species involved in rhamnose biosynthesis
(Fig. 5a and Table S10). The rhamnose-dopamine receptor D1-PKA axis
has been implicated in thermogenesis and anti-obesity mechanisms
[78]. Polysaccharides rich in rhamnose from Atractylodes chinensis
(DC.) Koidz. has been found to alleviate type 2 diabetic symptoms by
regulating intestinal microbiota and metabolites [61], echoing our
ndings of the combined effects of multiple PM
2.5
-related biomarkers on
T2D.
Serum hippuric acid, another crucial hub node, was signicantly
associated with fecal rhamnose and glycylproline (Fig. 5c). As a novel
PM
2.5
-related biomarker in this study, hippuric acid has been linked to
improved metabolic proles in diabetic mice supplemented with phe-
nolipid named jambone E, which also inhibited gluconeogenesis via
AKT-mediated insulin signaling modulation [79]. Elevated plasma hip-
puric acid levels have been associated with an increased risk of adverse
cardiovascular events in two independent cohorts (GeneBank at the
Cleveland Clinic [n =4000] and the LipidCardio study in Berlin
[n =833]) [71], and variations in hippurate excretion in individuals
with type 2 diabetes have been reported [80,81], though some studies
noted lower urinary hippurate levels in T2D patients [82]. Our ndings
suggested that hippuric acid’s role in the relationship between PM
2.5
exposure and T2D might also depend on its association with fecal bac-
teria and metabolites.
TB182, a novel PM
2.5
-related serum protein and crucial hub node, is
a binding protein of tankyrase 1, implicated in esophageal cancer
tumorigenesis via β-catenin signaling pathway activation [65]. It
correlated signicantly with several PM
2.5
-related bacteria, including
the principal biomarker genus Enorma (Figs. 2b and 5d), which was also
found to be positively associated with PM
2.5
exposure [12,13]. Proteins
involved in immune, carcinogenic, apoptotic, and endocrine signaling
pathways, which are key in diseases like inammatory bowel disease,
colorectal cancer, obesity, and T2D, are often targeted by gut microbes
[83]. Therefore, PM
2.5
exposure may underscore the impact of gut
J.-t. Wang et al.

Journal of Hazardous Materials 468 (2024) 133784
8
Fig. 4. Prediction of T2D prevalence and incidence by PM
2.5
-related biomarkers and their interaction. (a) Shows the ROC-AUC for predictive models of T2D
prevalence using single omics markers. (b) Depicts the ROC-AUC for predictive models of T2D prevalence using a combination of omics markers. (c) Illustrates the C-
index of the Cox proportional hazards model for T2D incidence prediction using PM
2.5
-related biomarkers. (d) Display the network association among the four omics
features (fecal microbial metagenome and metabolome, host serum metabolome and proteome). Connections (edges) represent partial Spearman’s rank correlations
(|r| >0.15), adjusted for age, sex, BMI, smoking status, alcohol consumption, income level, education level, energy intake, physical activity, and antibiotic use (for
gut bacteria correlation) (all FDR p<0.05). Hub nodes with labels are in the top 10 PM
2.5
-related biomarkers, based on combined ranking of three network attributes
(betweenness, closeness, and degree). (e) Portrays the same network association among T2D participants in the study (n =608).
J.-t. Wang et al.

Journal of Hazardous Materials 468 (2024) 133784
9
microbes on host proteins, disrupting the body’s homeostatic regulation.
3.6. Strengths and limitations
This study boasts several strengths. Firstly, we innovatively address
the effects of daily PM
2.5
exposure uctuations by distinguishing 7-day
PM
2.5
exposure trajectory groups, yielding a more nuanced interpreta-
tion of omics variance compared to the cumulative average exposure
over 7 days and 1 year. Secondly, by integrating four-dimensional omics
data from gut microbial and human hosts, including microbial meta-
genomics, stool metabolomes, and host circulating proteomics and
metabolomes, we successfully identied crucial cross-omics PM
2.5
-
related biomarkers and a tightly connected network biomarker. The
ability of these biomarkers to predict T2D prevalence and incidence
offers new mechanistic insights into the link between PM
2.5
and T2D in
humans. Thirdly, the relatively large sample size of this study facilitates
the detection of weaker associations within sparse omics data,
enhancing the reliability of our ndings.
However, the study also has limitations that warrant consideration.
Firstly, the estimation of PM
2.5
exposure based on residential address
may not be as precise as measurements obtained from wearable devices.
Given the large sample size, the use of wearable devices for exposure
estimation was impractical. We attempted to mitigate this limitation by
updating participants’ moving situation from baseline records to mini-
mize the impact of address changes. Secondly, given the diverse com-
ponents of PM
2.5
and their variable toxicity, it’s warranted for future
studies to analyze these components in detail for a more precise un-
derstanding. Thirdly, the study’s participants primarily resided in a
single city in Southern China, potentially limiting the ability to capture
responses to long-term PM
2.5
exposure due to homogeneity in exposure
levels. Our approach to obtaining individualized short-term exposure
levels within long-term sampling at the city level may help avoid the
signicant ecological confounding effects often seen in multi-center
studies. Finally, these short-term exposure-related biomarkers may not
fully represent the long-term effects of PM
2.5
exposure.
4. Conclusions
Our study effectively maps the multi-omics biomarkers in gut
microbiota, fecal metabolites, and host serum, highlighting their rela-
tionship with ambient PM
2.5
exposure trajectories. Notably, we discov-
ered novel biomarkers, such as the disturbance in B. thetaiotaomicron,
and alterations in gut metabolites and serum biomarkers concerning
lipid metabolism and n-3 fatty acids. These ndings shed light on the
molecular mechanisms inuenced by short-term PM
2.5
exposure, offer-
ing critical insights for developing public health interventions. Partic-
ularly, the dysregulation of subnetwork interactions in T2D subjects,
focused on critical biomarkers, provides a foundation for designing
strategies to counteract the effects of PM
2.5
exposure.
Funding
This work was supported by the National Natural Science Foundation
of China [82073546 by Y.-m.C., 81903316 and 82073529 by J.-S.Z.];
the Key Research and Development Program of Guangzhou, China
[202007040003] by Y.-m.C.; and the 5010 Program for Clinical
Research of the Sun Yat-sen University [2007032] by Y.-m.C..
CRediT authorship contribution statement
Wang Jia-ting: Writing – original draft, Investigation, Formal
analysis. Hu Wei: Formal analysis. Xue Zhangzhi: Data curation. Cai
Xue: Data curation. Zhang Shi-yu: Investigation. Li Fan-qin: Project
administration, Investigation. Lin Li-shan: Project administration,
Investigation. Chen Hanzu: Project administration, Investigation. Miao
Zelei: Data curation. Xi Yue: Project administration, Investigation. Guo
Tiannan: Writing – review & editing, Conceptualization. Zheng Ju-
Sheng: Writing – review & editing, Conceptualization. Chen Yu-ming:
Writing – review & editing, Supervision, Resources, Conceptualization.
Lin Hua-liang: Writing – review & editing, Investigation,
Conceptualization.
Declaration of Competing Interest
The authors declare that they have no known competing nancial
interests or personal relationships that could have appeared to inuence
the work reported in this paper.
Fig. 5. Subnetworks centered on each crucial hub node and their relationship with T2D. Displays subnetworks centered on each crucial hub node signicantly
associated with PM
2.5
in Model II (a: fecal rhamnose, b: fecal glycylproline, c: serum hippuric acid, d: serum protein TB182). And the consistency in the direction of
predicted probability for H-PM
2.5
exposure and T2D prevalence, as indicated by PM
2.5
-related biomarkers in subnetwork membership (SNM) centered on each crucial
hub node. Nodes with labels are principle PM
2.5
-related biomarkers in Model II.
J.-t. Wang et al.

Journal of Hazardous Materials 468 (2024) 133784
10
Data Availability
Data will be made available on request.
Appendix A. Supporting information
Supplementary data associated with this article can be found in the
online version at doi:10.1016/j.jhazmat.2024.133784.
References
[1] Global burden of 87 risk factors in 204 countries and territories, 1990–2019: a
systematic analysis for the Global Burden of Disease Study 2019., in: Lancet, 2020,
pp. 1223–1249.
[2] Hystad, P., Larkin, A., Rangarajan, S., AlHabib, K.F., Avezum, ´
A., Calik, K.B.T.,
et al., 2020. Associations of outdoor ne particulate air pollution and
cardiovascular disease in 157436 individuals from 21 high-income, middle-
income, and low-income countries (PURE): a prospective cohort study. Lancet
Planet Health 4, e235–e245. https://doi.org/10.1016/s2542-5196(20)30103-0.
[3] Weichenthal, S.A., Lavigne, E., Evans, G.J., Godri Pollitt, K.J., Burnett, R.T., 2016.
Fine particulate matter and emergency room visits for respiratory illness. Effect
Modication by Oxidative Potential. Am J Respir Crit Care Med 194, 577–586.
https://doi.org/10.1164/rccm.201512-2434OC.
[4] Aghaei-Zarch, S.M., Alipourfard, I., Rasoulzadeh, H., Naja, S., Aghaei-Zarch, F.,
Partov, S., et al., 2023. Non-coding RNAs: an emerging player in particulate matter
2.5-mediated toxicity. Int J Biol Macromol 235, 123790. https://doi.org/10.1016/
j.ijbiomac.2023.123790.
[5] Dai, S., Wang, Z., Yang, Y., Du, P., Li, X., 2022. PM2.5 induced weight loss of mice
through altering the intestinal microenvironment: Mucus barrier, gut microbiota,
and metabolic proling. J Hazard Mater 431, 128653. https://doi.org/10.1016/j.
jhazmat.2022.128653.
[6] Wang, W., Zhou, J., Chen, M., Huang, X., Xie, X., Li, W., et al., 2018. Exposure to
concentrated ambient PM2.5 alters the composition of gut microbiota in a murine
model. Part Fibre Toxicol 15, 1–13. https://doi.org/10.1186/s12989-018-0252-6.
[7] Mutlu, E.A., Comba, I.Y., Cho, T., Engen, P.A., Yazıcı, C., Soberanes, S., et al., 2018.
Inhalational exposure to particulate matter air pollution alters the composition of
the gut microbiome. Environ Pollut 240, 817–830. https://doi.org/10.1016/j.
envpol.2018.04.130.
[8] Gan, Q., Ye, W., Zhao, X., Teng, Y., Mei, S., Long, Y., et al., 2022. Mediating effects
of gut microbiota in the associations of air pollutants exposure with adverse
pregnancy outcomes. Ecotoxicol Environ Saf 234, 113371. https://doi.org/
10.1016/j.ecoenv.2022.113371.
[9] Liu, T., Chen, X., Xu, Y., Wu, W., Tang, W., Chen, Z., et al., 2019. Gut microbiota
partially mediates the effects of ne particulate matter on type 2 diabetes: evidence
from a population-based epidemiological study. Environ Int 130, 1–11. https://doi.
org/10.1016/j.envint.2019.05.076.
[10] Tanya, L.A., Roshonda, B.J., Zhanghua, C., Jeniffer, S.K., Rima, H., Frederick, L.,
et al., 2019. Exposure to trafc-related air pollution and the composition of the gut
microbiota in overweight and obese adolescents. Environ Res 176, 139–148.
https://doi.org/10.1016/j.envres.2017.11.046.Exposure.
[11] He, Y., Wu, W., Wu, S., Zheng, H.M., Li, P., Sheng, H.F., et al., 2018. Linking gut
microbiota, metabolic syndrome and economic status based on a population-level
analysis. Microbiome 6, 172. https://doi.org/10.1186/s40168-018-0557-6.
[12] Li, T., Fang, J., Tang, S., Du, H., Zhao, L., Wang, Y., et al., 2022. PM2.5 exposure
associated with microbiota gut-brain axis: Multi-omics mechanistic implications
from the BAPE study. Innovation 3, 100213. https://doi.org/10.1016/j.
xinn.2022.100213.
[13] Zhao, L., Fang, J., Tang, S., Deng, F., Liu, X., Shen, Y., et al., 2022. PM2.5 and
serum metabolome and insulin resistance, potential mediation by the gut
microbiome: a population-based panel study of older adults in China. Environ
Health Perspect 130, 1–10. https://doi.org/10.1289/EHP9688.
[14] Dong, X., Guthrie, B.G.H., Alexander, M., Noecker, C., Ramirez, L., Glasser, N.R.,
et al., 2022. Genetic manipulation of the human gut bacterium Eggerthella lenta
reveals a widespread family of transcriptional regulators. Nat Commun 13, 7624.
https://doi.org/10.1038/s41467-022-33576-3.
[15] Gacesa, R., Kurilshikov, A., Vich Vila, A., Sinha, T., Klaassen, M.A.Y., Bolte, L.A.,
et al., 2022. Environmental factors shaping the gut microbiome in a Dutch
population. Nature 604, 732–739. https://doi.org/10.1038/s41586-022-04567-7.
[16] Zierer, J., Jackson, M.A., Kastenmüller, G., Mangino, M., Long, T., Telenti, A.,
et al., 2018. The fecal metabolome as a functional readout of the gut microbiome.
Nat Genet 50, 790–795. https://doi.org/10.1038/s41588-018-0135-7.
[17] Gou, W., Yue, L., Tang, X.Y., Wu, Y.Y., Cai, X., Shuai, M., et al., 2022. Circulating
proteome and progression of type 2 diabetes. J Clin Endocrinol Metab 107,
1616–1625. https://doi.org/10.1210/clinem/dgac098.
[18] Talmor-barkan, Y., Bar, N., Shaul, A.A., Shahaf, N., Godneva, A., Bussi, Y., et al.,
2022. Metabolomic and microbiome proling reveals personalized risk factors for
coronary artery disease. Nat Med 28 (2), 295–302. https://doi.org/10.1038/
s41591-022-01686-6.
[19] Mookherjee, N., Ryu, M.H., Hemshekhar, M., Orach, J., Spicer, V., Carlsten, C.,
2022. Dening the effects of trafc-related air pollution on the human plasma
proteome using an aptamer proteomic array: a dose-dependent increase in
atherosclerosis-related proteins. Environ Res 209, 112803. https://doi.org/
10.1016/j.envres.2022.112803.
[20] Wang, J., Lin, L., Huang, J., Zhang, J., Duan, J., Guo, X., et al., 2022. Impact of
PM2.5 exposure on plasma metabolome in healthy adults during air pollution
waves: a randomized, crossover trial. J Hazard Mater 436, 129180. https://doi.
org/10.1016/j.jhazmat.2022.129180.
[21] Chatterjee, S., Khunti, K., Davies, M.J., 2017. Type 2 diabetes. Lancet 389 (10085),
2239–2251. https://doi.org/10.1016/S0140-6736(17)30058-2.
[22] G.D.a.A.P. Collaborators, Estimates, trends, and drivers of the global burden of
type 2 diabetes attributable to PM(2⋅5) air pollution, 1990–2019: an analysis of
data from the Global Burden of Disease Study 2019, in: Lancet Planet Health, 2022,
pp. e586-e600.
[23] Wang, J., Yan, Y., Si, H., Li, J., Zhao, Y., Gao, T., et al., 2023. The effect of real-
ambient PM2.5 exposure on the lung and gut microbiomes and the regulation of
Nrf2. Ecotoxicol Environ Saf 254, 114702. https://doi.org/10.1016/j.
ecoenv.2023.114702.
[24] Liu, W., Zhou, Y., Qin, Y., Li, Y., Yu, L., Li, R., et al., 2020. Sex-specic effects of
PM2.5 maternal exposure on offspring’s serum lipoproteins and gut microbiota. Sci
Total Environ 739, 139982. https://doi.org/10.1016/j.scitotenv.2020.139982.
[25] Fang, Z.F., Wang, Z.N., Chen, Z., Peng, Y., Fu, Y., Yang, Y., et al., 2024. Fine
particulate matter contributes to COPD-like pathophysiology: experimental
evidence from rats exposed to diesel exhaust particles. Respir Res 25, 14. https://
doi.org/10.1186/s12931-023-02623-y.
[26] Ling, C.W., Zhong, H., Zeng, F.F., Chen, G., Fu, Y., Wang, C., et al., 2023. Cohort
prole: Guangzhou nutrition and health study (GNHS): a population-based multi-
omics study. J Epidemiol. https://doi.org/10.2188/jea.JE20230108.
[27] Wei, J., Li, Z., Cribb, M., Huang, W., Xue, W., Sun, L., et al., 2020. Improved 1 km
resolution PM2.5 estimates across China using enhanced space–time extremely
randomized trees. Atmos Chem Phys 20, 3273–3289. https://doi.org/10.5194/
acp-20-3273-2020.
[28] Wei, J., Li, Z., Lyapustin, A., Sun, L., Peng, Y., Xue, W., et al., 2021. Reconstructing
1-km-resolution high-quality PM2.5 data records from 2000 to 2018 in China:
spatiotemporal variations and policy implications. Remote Sens Environ 252,
112136. https://doi.org/10.1016/j.rse.2020.112136.
[29] Urban, A., Di Napoli, C., Cloke, H.L., Kyselý, J., Pappenberger, F., Sera, F., et al.,
2021. Evaluation of the ERA5 reanalysis-based Universal Thermal Climate Index
on mortality data in Europe. Environ Res 198, 111227. https://doi.org/10.1016/j.
envres.2021.111227.
[30] Burkart, K.G., Brauer, M., Aravkin, A.Y., Godwin, W.W., Hay, S.I., He, J., et al.,
2021. Estimating the cause-specic relative risks of non-optimal temperature on
daily mortality: a two-part modelling approach applied to the Global Burden of
Disease Study. Lancet 398, 685–697. https://doi.org/10.1016/S0140-6736(21)
01700-1.
[31] Cai, M., Wei, J., Zhang, S., Liu, W., Wang, L., Qian, Z., et al., 2023. Short-term air
pollution exposure associated with death from kidney diseases: a nationwide time-
stratied case-crossover study in China from 2015 to 2019. BMC Med 21, 32.
https://doi.org/10.1186/s12916-023-02734-9.
[32] Chen, L., Zhernakova, D.V., Kurilshikov, A., Andreu-S´
anchez, S., Wang, D.,
Augustijn, H.E., et al., 2022. Inuence of the microbiome, diet and genetics on
inter-individual variation in the human plasma metabolome. Nat Med 28 (11),
2333–2343. https://doi.org/10.1038/s41591-022-02014-8.
[33] W.H. Organization, Denition and diagnosis of diabetes mellitus and intermediate
hyperglycaemia: Report of a WHO/IDF consultation in, 2006.
[34] A.D. Association, 2020. 2. Classication and Diagnosis of Diabetes: Standards of
Medical Care in Diabetes-2020. Diabetes Care 43, S14–s31. https://doi.org/
10.2337/dc20-S002.
[35] Shuai, M., Zhang, G., Zeng, Ff, Fu, Y., Liang, X., Yuan, L., et al., 2022. Human gut
antibiotic resistome and progression of diabetes, 2104965 Adv Sci, 2104965.
https://doi.org/10.1002/advs.202104965.
[36] Truong, D.T., Franzosa, E.A., Tickle, T.L., Scholz, M., Weingart, G., Pasolli, E.,
et al., 2015. MetaPhlAn2 for enhanced metagenomic taxonomic proling. Nat
Methods 12, 902–903. https://doi.org/10.1038/nmeth.3589.
[37] Abubucker, S., Segata, N., Goll, J., Schubert, A.M., Izard, J., Cantarel, B.L., et al.,
2012. Metabolic reconstruction for metagenomic data and its application to the
human microbiome. PLoS Comput Biol 8, e1002358. https://doi.org/10.1371/
journal.pcbi.1002358.
[38] Caspi, R., Billington, R., Ferrer, L., Foerster, H., Fulcher, C.A., Keseler, I.M., et al.,
2016. The MetaCyc database of metabolic pathways and enzymes and the BioCyc
collection of pathway/genome databases. Nucleic Acids Res 44, D471–480.
https://doi.org/10.1093/nar/gkv1164.
[39] Caspi, R., Billington, R., Fulcher, C.A., Keseler, I.M., Kothari, A.,
Krummenacker, M., et al., 2018. The MetaCyc database of metabolic pathways and
enzymes. Nucleic Acids Res 46, D633–d639. https://doi.org/10.1093/nar/gkx935.
[40] Deng, K., Xu, J.J., Shen, L., Zhao, H., Gou, W., Xu, F., et al., 2023. Comparison of
fecal and blood metabolome reveals inconsistent associations of the gut microbiota
with cardiometabolic diseases. Nat Commun 14, 571. https://doi.org/10.1038/
s41467-023-36256-y.
[41] Xu, F., Yu, E.Y., Cai, X., Yue, L., Jing, L.P., Liang, X., et al., 2023. Genome-wide
genotype-serum proteome mapping provides insights into the cross-ancestry
differences in cardiometabolic disease susceptibility. Nat Commun 14, 896.
https://doi.org/10.1038/s41467-023-36491-3.
[42] Cai, X., Xue, Z., Zeng, F.F., Tang, J., Yue, L., Wang, B., et al., 2023. Population
serum proteomics uncovers a prognostic protein classier for metabolic syndrome.
Cell Rep Med 4 (9), 101172. https://doi.org/10.1016/j.xcrm.2023.101172.
[43] Zhang, Y., Cai, X., Ge, W., Wang, D., Zhu, G., Qian, L., et al., 2022. Potential use of
serum proteomics for monitoring COVID-19 progression to complement RT-PCR
detection. J Proteome Res 21, 90–100. https://doi.org/10.1021/acs.
jproteome.1c00525.
J.-t. Wang et al.

Journal of Hazardous Materials 468 (2024) 133784
11
[44] Nagin, D.S., Land, K.C., 1993. Age, criminal careers, and population heterogeneity:
specication and estimation of nonparametric mixed poisson model. Criminology
31, 327–362. https://doi.org/10.1111/j.1745-9125.1993.tb01133.x.
[45] Nagin, D.S., Odgers, C.L., 2010. Group-based trajectory modeling in clinical
research. Annu Rev Clin Psychol.
[46] Xu, S., Zhan, L., Tang, W., Wang, Q., Dai, Z., Zhou, L., et al., 2023.
MicrobiotaProcess: a comprehensive R package for deep mining microbiome.
Innov (Camb) 4, 100388. https://doi.org/10.1016/j.xinn.2023.100388.
[47] Ambient air quality standards, in, Ministry of Ecology and Environment, China,
2016.
[48] He, Y., Wu, W., Zheng, H.M., Li, P., McDonald, D., Sheng, H.F., et al., 2018.
Regional variation limits applications of healthy gut microbiome reference ranges
and disease models. Nat Med 24, 1532–1535. https://doi.org/10.1038/s41591-
018-0164-x.
[49] Koren, O., Goodrich, J.K., Cullender, T.C., Spor, A., Laitinen, K., B¨
ackhed, H.K.,
et al., 2012. Host remodeling of the gut microbiome and metabolic changes during
pregnancy. Cell 150, 470–480. https://doi.org/10.1016/j.cell.2012.07.008.
[50] Li, G., Yin, P., Chu, S., Gao, W., Cui, S., Guo, S., et al., 2021. Correlation analysis
between GDM and gut microbial composition in late pregnancy. J Diabetes Res
2021, 8892849. https://doi.org/10.1155/2021/8892849.
[51] Fan, Y., Pedersen, O., 2021. Gut microbiota in human metabolic health and
disease. Nat Rev Microbiol 19, 55–71. https://doi.org/10.1038/s41579-020-0433-
9.
[52] Liu, J.X., Yuan, H.Y., Li, Y.N., Wei, Z., Liu, Y., Liang, J., 2022. Ephedra sinica
polysaccharide alleviates airway inammations of mouse asthma-like induced by
PM2.5 and ovalbumin via the regulation of gut microbiota and short chain fatty
acid. J Pharm Pharm 74, 1784–1796. https://doi.org/10.1093/jpp/rgac078.
[53] Takeuchi, T., Kameyama, K., Miyauchi, E., Nakanishi, Y., Kanaya, T., Fujii, T.,
et al., 2023. Fatty acid overproduction by gut commensal microbiota exacerbates
obesity. e369 Cell Metab 35, 361–375. https://doi.org/10.1016/j.
cmet.2022.12.013.
[54] Li, X., Wan, W., Zheng, L., Wang, A., Luo, X., Huang, Q., et al., 2021. Community
assembly mechanisms and co-occurrence patterns of nitrite-oxidizing bacteria
communities in saline soils. Sci Total Environ 772, 145472. https://doi.org/
10.1016/j.scitotenv.2021.145472.
[55] Palmu, J., B¨
orschel, C.S., Ortega-Alonso, A., Mark´
o, L., Inouye, M., Jousilahti, P.,
et al., 2023. Gut microbiome and atrial brillation-results from a large population-
based study. EBioMedicine 91, 104583. https://doi.org/10.1016/j.
ebiom.2023.104583.
[56] Gao, G., Cao, J., Mi, L., Feng, D., Deng, Q., Sun, X., et al., 2021. BdPUL12
depolymerizes β-mannan-like glycans into mannooligosaccharides and mannose,
which serve as carbon sources for Bacteroides dorei and gut probiotics. Int J Biol
Macromol 187, 664–674. https://doi.org/10.1016/j.ijbiomac.2021.07.172.
[57] Li, H., Cai, J., Chen, R., Zhao, Z., Ying, Z., Wang, L., et al., 2017. Particulate matter
exposure and stress hormone levels: a randomized, double-blind, crossover trial of
air purication. Circulation 136, 618–627. https://doi.org/10.1161/
CIRCULATIONAHA.116.026796.
[58] Vettore, L.A., Westbrook, R.L., Tennant, D.A., 2021. Proline metabolism and redox;
maintaining a balance in health and disease. Amino Acids 53, 1779–1788. https://
doi.org/10.1007/s00726-021-03051-2.
[59] Ding, Y., Koda, Y., Shashni, B., Takeda, N., Zhang, X., Tanaka, N., et al., 2023. An
orally deliverable ornithine-based self-assembling polymer nanomedicine
ameliorates hyperammonemia in acetaminophen-induced acute liver injury. Acta
Biomater 15 (168), 515–528. https://doi.org/10.1016/j.actbio.2023.07.005.
[60] Xu, W., Jiang, C., Kong, X., Liang, Y., Rong, M., Liu, W., 2012.
Chitooligosaccharides and N-acetyl-D-glucosamine stimulate peripheral blood
mononuclear cell-mediated antitumor immune responses. Mol Med Rep 6,
385–390. https://doi.org/10.3892/mmr.2012.918.
[61] Zhang, X., Ma, Q., Jia, L., He, H., Zhang, T., Jia, W., et al., 2023. Effects of in vitro
fermentation of Atractylodes chinensis (DC.) Koidz. polysaccharide on fecal
microbiota and metabolites in patients with type 2 diabetes mellitus. Int J Biol
Macromol 253, 126860. https://doi.org/10.1016/j.ijbiomac.2023.126860.
[62] Zhang, Q., Du, X., Li, H., Jiang, Y., Zhu, X., Zhang, Y., et al., 2022. Cardiovascular
effects of trafc-related air pollution: a multi-omics analysis from a randomized,
crossover trial. J Hazard Mater 435, 129031. https://doi.org/10.1016/j.
jhazmat.2022.129031.
[63] Casella, C., Kiles, F., Urquhart, C., Michaud, D.S., Kirwa, K., Corlin, L., 2023.
Methylomic, proteomic, and metabolomic correlates of trafc-related air pollution
in the context of cardiorespiratory health: a systematic review, pathway analysis,
and network analysis. Toxics 11 (12), 1014. https://doi.org/10.3390/
toxics11121014.
[64] Chen, X., Li, Y., Yuan, X., Yuan, W., Li, C., Zeng, Y., et al., 2022. Methazolamide
attenuates the development of diabetic cardiomyopathy by promoting β-catenin
degradation in type 1 diabetic mice. Diabetes 71, 795–811. https://doi.org/
10.2337/db21-0506.
[65] Gao, A., Su, Z., Shang, Z., He, C., Miao, D., Li, X., et al., 2022. TAB182 aggravates
progression of esophageal squamous cell carcinoma by enhancing β-catenin
nuclear translocation through FHL2 dependent manner. Cell Death Dis 13, 900.
https://doi.org/10.1038/s41419-022-05334-2.
[66] Poncet, D., Belleville, A., t’kint de Roodenbeke, C., Roborel de Climens, A., Ben
Simon, E., Merle-Beral, H., et al., 2008. Changes in the expression of telomere
maintenance genes suggest global telomere dysfunction in B-chronic lymphocytic
leukemia. Blood 111, 2388–2391. https://doi.org/10.1182/blood-2007-09-
111245.
[67] Liang, D., Ladva, C.N., Golan, R., Yu, T., Walker, D.I., Sarnat, S.E., et al., 2019.
Perturbations of the arginine metabolome following exposures to trafc-related air
pollution in a panel of commuters with and without asthma. Environ Int 127,
503–513. https://doi.org/10.1016/j.envint.2019.04.003.
[68] Mu, L., Niu, Z., Blair, R.H., Yu, H., Browne, R.W., Bonner, M.R., et al., 2019.
Metabolomics proling before, during, and after the beijing olympics: a panel
study of within-individual differences during periods of high and low air pollution.
Environ Health Perspect 127, 57010. https://doi.org/10.1289/EHP3705.
[69] Jiang, H., Wang, L., Wang, D., Yan, N., Li, C., Wu, M., et al., 2022. Omega-3
polyunsaturated fatty acid biomarkers and risk of type 2 diabetes, cardiovascular
disease, cancer, and mortality. Clin Nutr 41, 1798–1807. https://doi.org/10.1016/
j.clnu.2022.06.034.
[70] Shaw, W., 2010. Increased urinary excretion of a 3-(3-hydroxyphenyl)-3-
hydroxypropionic acid (HPHPA), an abnormal phenylalanine metabolite of
Clostridia spp. in the gastrointestinal tract, in urine samples from patients with
autism and schizophrenia. Nutr Neurosci 13, 135–143. https://doi.org/10.1179/
147683010x12611460763968.
[71] Nemet, I., Li, X.S., Haghikia, A., Li, L., Wilcox, J., Romano, K.A., et al., 2023. Atlas
of gut microbe-derived products from aromatic amino acids and risk of
cardiovascular morbidity and mortality. Eur Heart J 44 (32), 3085–3096. https://
doi.org/10.1093/eurheartj/ehad333.
[72] Tang, Z., Sarnat, J.A., Weber, R.J., Russell, A.G., Zhang, X., Li, Z., et al., 2022. The
oxidative potential of ne particulate matter and biological perturbations in
human plasma and saliva metabolome. Environ Sci Technol 56, 7350–7361.
https://doi.org/10.1021/acs.est.1c04915.
[73] Wawrzyniak, R., Greˇ
sner, P., Lewicka, E., Macioszek, S., Furga, A., Zieba, B., et al.,
2023. Metabolomics meets clinics: a multivariate analysis of plasma and urine
metabolic signatures in pulmonary arterial hypertension. J Proteome Res. https://
doi.org/10.1021/acs.jproteome.3c00255.
[74] Feng, B., Liu, C., Yi, T., Song, X., Wang, Y., Liu, S., et al., 2021. Perturbation of
amino acid metabolism mediates air pollution associated vascular dysfunction in
healthy adults. Environ Res 201, 111512. https://doi.org/10.1016/j.
envres.2021.111512.
[75] Guo, S., Sun, Y., Wu, T., Kwok, L.Y., Sun, Z., Wang, J., et al., 2024. Co-fermented
milk beverage has better stability and contains more health-promoting amino acid
metabolites than single-strain-fermented milk beverage over one-month storage.
Food Chem 430, 136840. https://doi.org/10.1016/j.foodchem.2023.136840.
[76] Sajid, M.I., Nunez, F.J., Amirrad, F., Roosan, M.R., Vojtko, T., McCulloch, S., et al.,
2023. Untargeted metabolomics analysis on kidney tissues from mice reveals
potential hypoxia biomarkers. Sci Rep 13, 17516. https://doi.org/10.1038/
s41598-023-44629-y.
[77] Luis, A.S., Briggs, J., Zhang, X., Farnell, B., Ndeh, D., Labourel, A., et al., 2018.
Dietary pectic glycans are degraded by coordinated enzyme pathways in human
colonic Bacteroides. Nat Microbiol 3, 210–219. https://doi.org/10.1038/s41564-
017-0079-1.
[78] Lv, S., Hu, T., Zhang, R., Zhou, Y., Yu, W., Wang, Z., et al., 2023. Rhamnose
displays an anti-obesity effect through stimulation of adipose dopamine receptors
and thermogenesis. Diabetes 72, 326–335. https://doi.org/10.2337/db22-0552.
[79] Wang, G., Xu, J., Ma, H., Mu, Y., Xu, W., Yan, N., et al., 2023. Phenolipid JE
improves metabolic prole and inhibits gluconeogenesis via modulating AKT-
mediated insulin signaling in STZ-induced diabetic mice. Pharm Res 187, 106569.
https://doi.org/10.1016/j.phrs.2022.106569.
[80] Messana, I., Forni, F., Ferrari, F., Rossi, C., Giardina, B., Zuppi, C., 1998. Proton
nuclear magnetic resonance spectral proles of urine in type II diabetic patients.
Clin Chem 44, 1529–1534.
[81] van Doorn, M., Vogels, J., Tas, A., van Hoogdalem, E.J., Burggraaf, J., Cohen, A.,
et al., 2007. Evaluation of metabolite proles as biomarkers for the
pharmacological effects of thiazolidinediones in Type 2 diabetes mellitus patients
and healthy volunteers. Br J Clin Pharm 63, 562–574. https://doi.org/10.1111/
j.1365-2125.2006.02816.x.
[82] Salek, R.M., Maguire, M.L., Bentley, E., Rubtsov, D.V., Hough, T., Cheeseman, M.,
et al., 2007. A metabolomic comparison of urinary changes in type 2 diabetes in
mouse, rat, and human. Physiol Genom 29, 99–108. https://doi.org/10.1152/
physiolgenomics.00194.2006.
[83] Zhou, H., Beltr´
an, J.F., Brito, I.L., 2022. Host-microbiome protein-protein
interactions capture disease-relevant pathways. Genome Biol 23, 72. https://doi.
org/10.1186/s13059-022-02643-9.
J.-t. Wang et al.