Multi-organ metabolic dysregulation and cecal microbiota alterations following black carbon exposure

Graphical abstract

Highlights

• •Long-term BC exposure alters plasma metabolites in 248,288 UK Biobank participants.
• •Controlled BC exposure in mice induces widespread tissue-specific metabolic shifts.
• •Gut microbiota shows reduced α-diversity with enrichment of Akkermansia and loss of Bacteroides.
• •Histopathology reveals lung inflammation, cardiac edema, and neuronal condensation.
• •Gene expression changes indicate inflammatory activation and impaired intestinal barrier.

Abstract

Background

Black carbon (BC) has been linked to adverse health outcomes, yet underlying mechanisms remain unclear. Integrating metabolomic and metagenomic data across tissues may clarify BC-induced biological pathways.

Methods

We performed human epidemiology and mice experimental approaches. We included 248,288 participants with annual BC exposure estimates and plasma metabolomic profiles. Elastic net regression identified BC-associated metabolites. Male C57BL/6J mice were exposed to filtered air or BC (1 mg/m³, 1 h/day, 5 days/week, 12 weeks). Multi-tissue metabolomics and cecal contents microbiota sequencing were conducted, with histology and gene expression measurements.

Results

In humans, long-term BC exposure significantly altered plasma metabolites, notably increasing saturated fatty acids (β = 0.048), while decreasing docosahexaenoic acid (β = −0.002). Amino acid metabolism was broadly disrupted, involving elevated valine (β = 0.011) and reduced glutamine (β = −0.006). In mice, metabolomic profiling showed organ-specific shifts, including increased glutathione and cortisol in the liver (2.88-fold and 2.06-fold), increased PC(16:0/18:1(9Z)) in the heart (3.22-fold), elevated anandamide and arachidonic acid in the kidney (2.35-fold and 1.48-fold), and decreased multiple fatty acids and lysophospholipids across organs. Cecal microbiota exhibited reduced alpha-diversity (Shannon: 3.67 vs. 4.50, P < 0.05) and taxonomic shifts, including an increased abundance of g_Akkermansia and decrease in g_Bacteroides. Multi-omics integration revealed significant microbiota-metabolome correlations in the cecum and plasma (Mantel r = 0.276, P = 0.012). Histological examination confirmed organ injuries, notably lung inflammation, cardiac edema, and neuronal condensation. Gene expression analysis showed increased Il-6 in the lung (5.35-fold, P = 0.047), increased Mb in the heart (5.18-fold, P = 0.010), and increased Igfbp7 in the kidney (3.03-fold, P = 0.001), while Tjp1 expression in cecum was reduced (0.42-fold, P = 0.004).

Conclusions

Our findings suggest that BC exposure may alter systemic metabolism and gut microbiota, potentially contributing to tissue injury and inflammation. The gut–organ axis could be a target for mitigating BC-related health effects.

Introduction

Black carbon (BC), a key component of fine particulate matter (PM_2.5), is primarily produced through the incomplete combustion of fossil fuels and biomass [1]. Increasing epidemiological evidence has linked BC exposure to a wide range of adverse health outcomes, such as cardiopulmonary diseases, metabolic dysfunction, neurodegenerative conditions [2]. Unlike many other particulate constituents, BC exhibits strong persistence and surface reactivity, enabling it to carry adsorbed organic chemicals and metals, thereby amplifying its toxicity potential [3].

Prior research has shown that BC can penetrate deep into the alveolar regions of the lungs and subsequently translocate into the systemic circulation [4]. Recent studies have reported detectable BC in peripheral blood, liver, kidneys, and even the brain, suggesting widespread biodistribution following inhalation exposure [5], [6], [7]. Emerging evidence also indicates that BC can accumulate in the gastrointestinal tract, either through mucociliary clearance and swallowing of inhaled particles or direct translocation from the bloodstream [8]. These observations collectively imply that BC exposure may exert coordinated effects across multiple organs. However, most existing research has adopted a single-organ or single-endpoint perspective, which limits our understanding of how BC triggers systemic biological responses. The lack of integrative, multi-organ evaluations represents a critical gap in current environmental health research.

Multi-omics technologies provide a means to capture the complexity of biological responses stemming from environmental exposures [9], [10], [11]. Previous studies employing omics approaches have revealed that pollutants can influence metabolic pathways, perturb immune and inflammatory signaling, and alter epigenetic regulatory networks [12], [13]. Yet, these investigations typically focus on individual biomolecular layers (e.g., only metabolites or only gene expression) and are rarely designed to elucidate cross-tissue interactions or dynamic organ-to-organ communication networks. Consequently, the systemic biological mechanisms linking BC exposure to disease susceptibility remain poorly resolved.

Among potential organ systems involved, the gut has emerged as a particularly compelling target. The gut microbiota plays a central role in host metabolic, immune, and neuroendocrine regulation [14]. Perturbations of microbial composition and function have been implicated in the development of cardiometabolic disorders, systemic inflammation, and neurological dysfunction [15]. Accumulating evidence suggests that inhaled air pollutants can alter gut microbiome structure [16], [17]. However, it remains unclear that whether gut microbial alterations represent a downstream consequence of systemic toxicity or a mechanistic driver linking BC exposure to multi-organ effects. Therefore, characterizing gut microbiota changes in conjunction with systemic metabolic disruptions is key to identifying potential causal pathways and cross-organ biological signaling.

To address the gaps in understanding systemic metabolic effects of BC exposure, we first analyzed plasma metabolomic data from over 240,000 UK Biobank participants. The large sample size, standardized exposure assessment, and high-resolution metabolomics in UK Biobank enabled robust identification of metabolic signatures associated with long-term BC exposure. To validate and extend these findings mechanistically, we established a chronic low-dose BC exposure mouse model reflecting occupationally relevant conditions. We performed untargeted metabolomic profiling across multiple tissues and metagenomic sequencing of cecal contents to assess systemic metabolic changes and gut microbial alterations. Integrating population-level evidence with organ-specific and microbiome data, this study aims to clarify the systemic biological impacts of BC exposure and the role of gut microbiota-linked metabolic pathways in shaping cross-organ responses.

Materials and methods

Epidemiological study

Study population

The UK Biobank is a large-scale, prospective cohort study that enrolled approximately 500,000 individuals aged between 37 and 73 years from England, Scotland, and Wales during the period from 2006 to 2010. At baseline, participants completed detailed electronic questionnaires and underwent standardized physical and anthropometric assessments. In addition, biological specimens, including plasma, were collected and stored for long-term research use. For the current analysis, individuals without available nuclear magnetic resonance (NMR)-based metabolomic data (n = 254,123) were excluded. The final analytic sample consisted of 248,288 participants.

BC exposure assessment

In alignment with prior research, elemental carbon was used as a proxy for BC [18], [19], [20]. Long-term concentrations of BC were estimated at a 3 km × 3 km spatial resolution using the EMEP4UK model [21]. This model has previously shown strong agreement with ground-based monitoring data [22]. Participants’ residential address histories, recorded at recruitment and updated thereafter, were geocoded to obtain corresponding latitude and longitude coordinates. Annual exposure levels were estimated using bilinear interpolation based on these geolocated addresses. To derive individual-level long-term exposure, pollutant concentrations were averaged across all recorded addresses, weighted by the duration of residence at each location. For this study, the primary exposure window was defined as the four-year mean pollutant concentration preceding the baseline assessment, which aligns with previous UK Biobank studies investigating associations between long-term air pollution exposure and metabolic profiles [23], [24].

Plasma metabolomic profiling

EDTA plasma samples were collected from participants during baseline visits (2006–2010) and stored for future analyses. Metabolomic profiling was performed by Nightingale Health Ltd. (Finland) using a high-throughput nuclear magnetic resonance (NMR) spectroscopy platform. The platform quantified 251 metabolic biomarkers, including 14 lipoprotein subclasses, various fatty acids and their composition, small molecules (such as amino acids, glycolysis intermediates, and ketone bodies), and 81 derived composite ratios. Among these, 170 biomarkers were reported as absolute concentrations, while 81 were expressed as ratios. A detailed description of the metabolites is provided in Supplementary Materials 2. Further methodological details are available in the UK Biobank’s official documentation and previous publication [25].

Statistical methods

To identify metabolites associated with BC exposure, we used elastic net regression, a regularized machine learning method that combines Lasso and Ridge penalties and is well-suited for high-dimensional data with multicollinearity [26]. Prior to modeling, all metabolic biomarkers (both absolute concentrations and ratios) were log-transformed and standardized to stabilize variance, reduce skewness, and approximate a normal distribution, thereby enhancing model performance and interpretability in subsequent regression and correlation analyses. Standardization also ensured comparability across biomarkers. The optimal regularization parameter (λ) was determined through 10-fold cross-validation, selecting the value corresponding to a mean squared error within one standard error of the minimum [27]. Metabolites with non-zero regression coefficients in the final model were considered relevant to BC exposure. Statistical significance was assessed using a false discovery rate threshold of ≤ 0.05. All analyses were conducted using R software (version 4.1.0).

Experimental study

Animals and BC exposure

Wild-type male C57BL/6J mice (6 weeks old; GemPharmatech, Guangdong, China) were housed under standard conditions and acclimatized for one week before exposure. Printex U carbon black (Printex 90, Germany), a well-established ambient BC model [28], [29], [30], was administered to mice via a whole-body inhalation exposure system (HOPE-MED 8050). Based on national BC exposure standards (4 mg/m³, 8-h TWA) and literature review [31], [32], [33], [34], [35], mice were randomly assigned to filtered air or 1 mg/m³ BC exposure (n = 8/group) for 1 h/day, 5 days/week, over 12 weeks. At study end, mice were anesthetized with tribromoethanol, and cecal contents and tissues were collected, snap-frozen in liquid nitrogen, and stored at −80 °C.

Metagenomics of cecum contents

Genomic DNA from cecal contents was extracted using the FastPure Stool DNA Isolation Kit (MJYH, Shanghai, China). DNA integrity and concentration were assessed using agarose gel electrophoresis and a Qubit Fluorometer to ensure sample quality prior to library construction. Genomic DNA was then fragmented to ∼350 bp using the Covaris M220 system, and paired-end libraries were prepared using the NEXTFLEX Rapid DNA-Seq Kit (Bioo Scientific, USA). Sequencing was performed on the Illumina NovaSeq™ X Plus platform to generate high-throughput paired-end reads. Data quality control was performed using fastp (version 0.20.0; https://github.com/OpenGene/fastp). Adapter sequences at both the 5′ and 3′ ends were trimmed, reads with post-trimming lengths < 50 bp were removed, and reads with average quality scores < Q20 were filtered out to obtain high-quality clean reads. To eliminate host-derived contamination, the clean reads were aligned to the Mus musculus reference genome using BWA (version 0.7.17; https://bio-bwa.sourceforge.net/), and reads mapping to the host genome (along with their paired mates) were removed. The remaining non-host reads were assembled into contigs (≥300 bp) using MEGAHIT, and open reading frames (ORFs ≥ 100 bp) were predicted using Prodigal. A non-redundant gene catalog was constructed using CD-HIT with a 90 % sequence identity threshold. Gene abundances across samples were quantified using SOAPaligner with a 95 % alignment identity cutoff. Taxonomic annotation was performed using DIAMOND against the NCBI NR database (e-value ≤1e-5), and functional annotation was conducted on the Majorbio Cloud platform [36].

Untargeted metabolomics

For solid samples, 30 mg tissue was extracted with acetonitrile:water (4:1, v/v) containing 1 μg/mL internal standards (reserpine, chloramphenicol), homogenized at −10 °C for 30 min, and centrifuged (12,000 rpm, 10 min, 4 °C). Liquid samples (30 μL) were mixed with 120 μL extraction solvent, vortexed, and centrifuged under the same conditions. Supernatants were transferred to vials for LC-MS/MS. A pooled quality control (QC) sample was prepared by combining equal aliquots of all samples and injected every 6–8 runs throughout the analytical sequence to monitor instrument stability and signal drift. Blank samples were analyzed to evaluate background contamination, and internal standards were used to assess extraction efficiency. Metabolomic profiling used a UPLC-Synapt XS system (Waters) with an HSS T3 column (100 × 2.1 mm, 1.8 μm). Mobile phases were 0.1 % formic acid in water (A) and acetonitrile (B). The gradient started at 2 % B, increased linearly to 98 % over 10 min, held for 2 min, then returned to 2 % for re-equilibration. Data were acquired in positive and negative ESI modes (50–1200 Da, ±2.0 kV). LC-MS data were processed using MS-DIAL. Retention time alignment, peak detection, and peak area normalization were performed automatically. Features were retained only if detected in >80 % of QC injections and with a relative standard deviation < 30 % in pooled QC samples. Total ion, base peak, and extracted ion chromatograms overlays of QC injections were used to visually assess chromatographic consistency. Metabolites were annotated using HMDB, and pathway enrichment was performed with MetaboAnalyst.

Hematoxylin-eosin staining

Tissue samples were fixed in 10 % formalin, paraffin-embedded, sectioned, and deparaffinized with xylene. After rehydration through graded ethanol, sections were stained with hematoxylin, differentiated in acid alcohol, blued, then counterstained with eosin. Finally, sections were dehydrated, cleared in xylene, and mounted with neutral resin.

RNA extraction and qRT-PCR

RNA was extracted from tissues using TriQuick Reagent (Solarbio). 1000 ng RNA was reverse-transcribed to cDNA with ReverTra Ace® qPCR RT Master Mix (TOYOBO). qRT-PCR was performed using SYBR Green Premix Pro Taq HS (Accurate Biology) on a QuantStudio 6 Pro (Thermo Fisher) following standard protocols. Actb was the reference gene, and relative expression was calculated by the 2^^(−ΔΔCt) method. Primer sequences are listed in Supplementary Material 1: Table S1.

Statistical analysis

For metagenomics analysis, alpha-diversity, differential abundance, functional contribution, and KEGG pathway enrichment analyses were performed to evaluate cecal microbiota composition and potential functional changes following BC exposure. Details of these bioinformatics analyses, including specific statistical tests, visualization approaches, and software used, are provided in Supplementary Material 1: Supplementary Methods. For metabolomics data, metabolites were deemed statistically significant with a variable importance in projection score  ≥1 and a fold change threshold of >1.2 or <0.8. To investigate the relationships between cecal microbiota and multi-organ metabolic profiles, we first evaluated the global concordance between microbial community structure and metabolomic profiles across organs using Mantel tests. The overall spatial alignment between the two datasets was further assessed with Procrustes analysis. We then performed Spearman correlation analyses to identify specific genus–metabolite associations in each organ. Finally, distance-based redundancy analysis (db-RDA) was applied to identify metabolite features that explained variation in microbiota composition. Detailed analytical procedures are provided in Supplementary Material 1: Supplementary Methods. qPCR results were expressed as mean ± SEM. Between-group differences were analyzed using unpaired Student's t-tests, with statistical significance defined as P ≤ 0.05.

Results

Altered plasma metabolic profile associated with long-term BC exposure in human

The mean of annual average estimates of BC was 1.08 ± 0.36 μg/m³. Table 1 summarizes the baseline characteristics of 248,288 UK Biobank participants stratified by mean BC exposure. The higher BC exposure group was characterized by a younger age, a higher proportion of non-White participants, and greater socioeconomic deprivation compared to the lower exposure group (56.11 vs. 57.06 years; 98.04 % vs. 91.12 %; and −2.05 vs. −0.57, respectively). Long-term BC exposure significantly altered plasma metabolite profiles across major biochemical categories (Supplementary Material 1: Table S2). Fatty acid metabolism showed the strongest alterations: saturated fatty acids (β = 0.048) and d linoleic acid (β = 0.028) were positively associated with BC exposure, accompanied by higher PUFA/MUFA (β = 0.013) and omega-6/omega-3 ratios (β = 0.012), whereas DHA% was slightly reduced (β = −0.002). Amino acid metabolism exhibited broad disruption, most notably affecting glutamine and branched-chain amino acids, including increased valine (β = 0.011) and histidine (β = 0.001), while glutamine (β = −0.006) and tyrosine (β = −0.004) were lower. Glycolysis-related metabolites also demonstrated consistent changes, with higher lactate (β = 0.018) and pyruvate (β = 0.015), and lower citrate (β = −0.005). In addition, sensitivity analyses using 1-year, 2-year, and 3-year exposure windows for BC yielded metabolic signatures that were consistent with our primary findings (Supplementary Material 1: Tables S3−S5).

Table 1.

Baseline characteristics of participants stratified by mean BC exposure (1.08 µg/m³).

	Total (n = 248,288)	Low BC exposure (n = 132,913)	High BC exposure (n = 115,375)
Age at recruitment	56.62 ± 8.08	57.06 ± 7.92	56.11 ± 8.22
Sex
Female	131,531 (52.98 %)	70,545 (53.08 %)	60,986 (52.86 %)
Male	116,757 (47.02 %)	62,368 (46.92 %)	54,389 (47.14 %)
Ethnicity
White	235,439 (94.82 %)	130,309 (98.04 %)	105,130 (91.12 %)
Asian	5260 (2.12 %)	996 (0.75 %)	4264 (3.7 %)
Black	3087 (1.24 %)	312 (0.23 %)	2775 (2.41 %)
Mixed	1330 (0.54 %)	440 (0.33 %)	890 (0.77 %)
Other	2048 (0.82 %)	452 (0.34 %)	1596 (1.38 %)
Missing	1124 (0.45 %)	404 (0.3 %)	720 (0.62 %)
Education
Any school degree	93,630 (37.71 %)	51,285 (38.59 %)	42,345 (36.7 %)
College education	78,003 (31.42 %)	40,759 (30.67 %)	37,244 (32.28 %)
Vocational qualification	16,810 (6.77 %)	9373 (7.05 %)	7437 (6.45 %)
Other	56,920 (22.92 %)	30,278 (22.78 %)	26,642 (23.09 %)
Missing	2925 (1.18 %)	1218 (0.92 %)	1707 (1.48 %)
Income
Less than 18,000 £	49,649 (20 %)	25,194 (18.96 %)	24,455 (21.2 %)
18,000–30,999 £	54,987 (22.15 %)	30,614 (23.03 %)	24,373 (21.13 %)
31,000–51,999 £	54,911 (22.12 %)	30,515 (22.96 %)	24,396 (21.14 %)
52,000–100,000 £	41,810 (16.84 %)	22,923 (17.25 %)	18,887 (16.37 %)
Greater than 100,000 £	10,429 (4.2 %)	5207 (3.92 %)	5222 (4.53 %)
Missing	36,502 (14.7 %)	18,460 (13.89 %)	18,042 (15.64 %)
Deprive index	−1.37 ± 3.07	−2.05 ± 2.72	−0.57 ± 3.24
Smoking status
Never	135,077 (54.4 %)	74,315 (55.91 %)	60,762 (52.66 %)
Previous	85,950 (34.62 %)	45,955 (34.58 %)	39,995 (34.67 %)
Current	26,038 (10.49 %)	12,154 (9.14 %)	13,884 (12.03 %)
Missing	1223 (0.49 %)	489 (0.37 %)	734 (0.64 %)
Intake alcohol
Never	19,985 (8.05 %)	9109 (6.85 %)	10,876 (9.43 %)
Occasional	56,868 (22.9 %)	29,263 (22.02 %)	27,605 (23.93 %)
Moderate	122,481 (49.33 %)	68,259 (51.36 %)	54,222 (47 %)
Heavy	48,475 (19.52 %)	26,154 (19.68 %)	22,321 (19.35 %)
Missing	479 (0.19 %)	128 (0.1 %)	351 (0.3 %)
Physical activities
Low	38,424 (15.48 %)	20,105 (15.13 %)	18,319 (15.88 %)
Moderate	81,120 (32.67 %)	43,400 (32.65 %)	37,720 (32.69 %)
High	80,538 (32.44 %)	44,292 (33.32 %)	36,246 (31.42 %)
Missing	48,206 (19.42 %)	25,116 (18.9 %)	23,090 (20.01 %)

BC concentration measurement and mice weight

Monitoring of BC concentrations in the exposure chamber confirmed stable aerosol generation throughout the experimental period, averaging 1.02 ± 0.04 mg/m³ (Supplementary Material 1: Fig. S1A). Body weight was monitored regularly and was comparable between the BC-exposed and control groups (29.42 g vs. 29.83 g at 17 weeks, P = 0.667), indicating that BC exposure did not affect overall weight gain in mice (Supplementary Material 1: Fig. S1B).

Alterations in cecal microbiota composition after BC exposure in mice

We first evaluated whether BC exposure alters the microbial community in the cecal. After removing low-quality sequences, an average of 98.59 % of the raw sequencing reads were retained for subsequent analysis (Supplementary Material 1: Table S6). The microbial alpha-diversity in cecal content, assessed via Shannon Diversity Index analysis, showed significantly lower value in the BC group compared to controls (3.67 vs. 4.50, P _FDR = 0.024, Fig. 1A). Genus-level operational taxonomic unit (OTU) analysis showed that most OTUs were shared between groups; however, 64 OTUs (2.37 %) were unique to the BC group, while 177 OTUs (6.54 %) were exclusive to the control group (Fig. 1B). Detailed results of differential taxa are provided in Supplementary Materials 3. Distinct differences in genus-level taxa were observed between the two groups. For example, g_Mycoavidus, g_Halorubrum, and g_Oligella were among those found exclusively in control samples, whereas g_Pisciglobus, g_Sinirhodobacter, and g_Exilispira were representative of those detected only in the BC-exposed group. At the genus level, g_Akkermansia was the prominent genus in the cecal followed by g_Eubacterium, g_Escherichia, and g_Bacteroides. Notably, the relative abundance of g_Akkermansia was elevated in the BC group (Fig. 1, Fig. 1). To quantify the impact of individual taxa on group differences, Linear Discriminant Analysis (LDA) was performed. g_Bacteroides emerged as the most influential genus in the control group, whereas g_Escherichia had the greatest contribution in the BC group (Fig. 1E). KEGG pathway analysis (level 3 clustering) demonstrated group-specific functional profiles (Fig. 1F). Metabolic pathways were predominantly enriched, including: biosynthesis of secondary metabolites, microbial metabolism in diverse environments, and biosynthesis of cofactors. g_Akkermansia was the major contributor to these functions (Fig. 1G). Specific KEGG pathways were differentially enriched between groups; for example, the phosphotransferase system was more abundant in controls, while teichoic acid biosynthesis was enriched in the BC group (Fig. 1H).

Fig. 1.

Alterations in Cecal Microbiota Composition Following BC Exposure. (A) Shannon α-diversity index showing differences in genus-level diversity between the Control and BC exposure groups. (B) Venn diagram displaying the unique and shared genera between the two groups. The overlapping region represents common genera, while the non-overlapping areas indicate group-specific genera. Numbers denote the corresponding genus counts. (C) Genus-level community composition. The bar plot shows the top 20 most abundant genera, with low-abundance genera grouped as “Others.” The x-axis represents samples, and the y-axis indicates the relative abundance of each genus. Different colors correspond to different genera, and bar height reflects their proportional abundance. (D) Differences in average relative abundance between the Control and BC groups. The x-axis shows the log₂ fold change in genus abundance, and the y-axis lists the corresponding genera. Genera enriched in the BC group are shown in BC color, while those enriched in the Control group are shown in Control color. (E) Linear discriminant analysis (LDA) showing LDA scores of differentially abundant genera identified by LEfSe. Higher LDA scores indicate a greater contribution of that genus to between-group differentiation. (F) Genus-level Circos plot illustrating the distribution of dominant genera between the two groups. The left semicircle represents genus composition within each group, while the right semicircle indicates the distribution of dominant genera across groups. Colored ribbons link groups to genera, and ribbon width reflects the relative abundance of each genus. (G) Functional contribution analysis based on genus-level clusters of orthologous groups annotation. The bar plot illustrates the contributions of dominant genera to various functional categories. (H) KEGG pathway enrichment analysis. The x-axis represents the reporter score, and the y-axis lists the enriched KEGG pathways. Bar colors denote the groups in which the pathways are enriched. n = 8 per group. *P < 0.05, **P < 0.01.

Metabolomic profiles of major organs after BC exposure in mice

We then performed untargeted metabolomic profiling. Quality control using overlaid total ion, base peak, and extracted ion chromatograms of pooled QC injections showed consistent chromatographic performance throughout the analytical sequence (Supplementary Material 1: Figs. S2–S3). Untargeted metabolomic profiling revealed alterations in metabolite abundance following BC exposure (Supplementary Material 1: Table S7). Taurocholic acid was markedly elevated in plasma (2.44-fold) and content (1.66-fold), while glutathione (2.88-fold) and cortisol (2.06-fold) surged in liver. Heart tissue exhibited profound up-regulation of PC (16:0/18:1(9Z)) (3.22-fold) and taurine (1.67-fold). Kidney showed increased anandamide (2.35-fold) and arachidonic acid (1.48-fold), whereas spleen demonstrated significant rises in 1-palmitoylphosphatidylcholine (2.03-fold) and estrone (1.75-fold). Conversely, widespread down-regulation affected fatty acids: palmitoleic (spleen, 0.28-fold), linoleic (spleen, 0.28-fold; liver, 0.64-fold; content, 0.79-fold), and succinic acid (content, 0.40-fold; brain, 0.61-fold). LysoPC species were consistently reduced in plasma (LysoPC(0:0/18:0), 0.36-fold) and liver. Spleen displayed striking suppression of multiple lipids, including myristic acid (0.23-fold) and γ-linolenic acid (0.41-fold), suggesting tissue-specific lipid metabolism dysregulation.

Correlation between altered multi-organ metabolic dysregulation and cecal microbiota

We next examined the correlation between microbial composition and metabolomic alterations across multiple organs (Supplementary Material 1: Table S8, Figs. S4–S12). We first assessed the overall association between the gut microbiota and host metabolome across different organs. Mantel tests revealed significant correlations between microbial composition and metabolomic profiles in the cecum (r = 0.276, P = 0.012) and plasma (r = 0.229, P = 0.044), but not in other organs. This spatial congruence in the cecum was further supported by Procrustes analysis, which showed a significant correlation between the cecal microbiota and metabolome (M² = 0.753, R = 0.497, P = 0.049).

To identify specific contributors to these global associations, we performed Spearman correlation analysis. In the cecum, several significant genus-metabolite pairs were identified. For instance, the abundance of g__Dorea was positively correlated with gamma-Linolenic acid (ρ = 0.641, P = 0.007), whereas g__Akkermansia was negatively correlated with the same metabolite (ρ = −0.621, P = 0.010) and with LysoPC(17:0/0:0) (ρ = −0.600, P = 0.014). Additionally, g__Desulfovibrio was negatively correlated with succinic acid (ρ = −0.597, P = 0.015) and a bile acid (ρ = −0.503, P = 0.047). In the brain, g__Bacteroides showed a strong positive correlation with adenosine (ρ = 0.753, P = 7.6 × 10⁻⁴) and a negative correlation with palmitic acid (ρ = −0.621, P = 0.010).

To determine the key metabolites driving variation in microbial community structure, we performed db-RDA. Organ-specific analyses identified several such drivers: gamma-Linolenic acid was a significant explanatory variable in the cecum (r² = 0.478, P = 0.015). In the brain, palmitic acid, cortisone, succinic acid, and PC(16:0/18:1(9Z)) were significantly associated with microbial variation (r² = 0.436–0.497, all P < 0.05). Significant associations were also found in the heart (hyodeoxycholic acid, r² = 0.541, P = 0.005) and lung (D-alpha-aminobutyric acid, r² = 0.550, P = 0.005).

Organ injury characterization by histology and molecular markers following BC exposure in mice

To investigate the systemic toxic effects of BC exposure, HE staining was performed on major organs including the heart, liver, spleen, lung, kidney, and brain (Fig. 2). Compared to control mice, pulmonary tissue from BC-treated mice exhibited alveolar wall thickening, infiltration of inflammatory cells, and partial alveolar collapse, whereas normal alveolar architecture was observed in controls. Cardiac tissue from BC-exposed mice showed mild interstitial edema, whereas the control group displayed a well-organized and compact myocardial structure. In the brain, histological analysis revealed nuclear condensation in the BC-treated group. Upon histopathological examination, no apparent damage or morphological changes were observed in the liver, spleen, or kidney of BC-exposed mice.

Fig. 2.

H&E Staining Reveals Histopathological Changes in Major Organs Following BC Exposure. Representative H&E staining images of the heart, liver, kidney, cecum, brain, lung, and spleen from control and BC-exposed mice, scale bars = 50 µm.

We further analyzed the mRNA expression levels of gene markers indicative of organ injury (Fig. 3). Although most markers showed a trend toward altered expression following BC exposure, the changes did not reach statistical significance. Notably, Mb expression in the heart and Igfbp7 expression in the kidneys were significantly increased in the BC group (5.18-fold, P = 0.010; and 3.03-fold, P = 0.001, respectively). In the lungs, Il-6 expression was also significantly elevated (5.35-fold, P = 0.047). Additionally, Tjp1, a gene associated with intestinal barrier integrity, was significantly downregulated in the cecum following BC exposure (0.42-fold, P = 0.004).

Fig. 3.

Relative mRNA expression of organ injury markers following BC exposure. Relative mRNA levels of injury-related genes in the heart, liver, kidney, cecum, brain, lung, and spleen were determined by qRT-PCR. Data are presented as mean ± SEM (n = 8 per group). Statistical significance was determined by an unpaired t-test compared to the control group.

Discussion

This study presents the complex interactions between inhaled black carbon (BC) exposure, gut microbiota dysbiosis, and systemic metabolic disruption across multiple organs. Our findings reveal that BC inhalation triggers interconnected changes in microbial communities and metabolite pathways, culminating in organ-specific injury. By integrating human metabolomic data from the UK Biobank with comprehensive murine multi-organ omics analyses, we provide cross-species evidence that BC exposure disrupts metabolic homeostasis in diverse tissues.

Across tissues we observed a metabolic signature of oxidative stress: major antioxidants and stress‐related metabolites were elevated, while energy intermediates and vitamin cofactors were depleted. For example, tissue levels of taurine and glutathione (key thiol antioxidants) were strongly increased in BC‐exposed mice. This parallels prior findings that particulate pollution perturbs glutathione and taurine metabolism [37]. Similarly, cortisol (the principal glucocorticoid in rodents) was markedly higher after BC exposure, indicating activation of the HPA stress axis. Indeed, epidemiologic and animal studies link fine PM exposure with elevated systemic cortisol (corticosterone) levels [38]. These coordinated changes suggest that BC inhalation triggers a global stress response, mobilizing redox defenses and stress hormones. Notably, we also observed increases in specific lipids associated with inflammation: arachidonic acid and multiple phosphatidylcholine (PC) species were upregulated. Arachidonic acid is released by PLA2 activity and serves as a precursor to pro‐inflammatory eicosanoids [39], while PCs are major membrane components that can modulate immunity [40]. In fact, oxidative stress is known to activate PLA2, hydrolyzing membrane phospholipids to yield free fatty acids and lysophospholipids [37]. Our data showed an accumulation of arachidonic acid but, interestingly, a depletion of many other free fatty acids and lysoPCs. This may reflect rapid further metabolism or sequestration of these lipids under chronic BC stress. In any case, the lipidomic pattern, with elevations in PCs and arachidonate in contrast to reductions in generic fatty acids and lysolipids, is consistent with air pollutant-induced membrane turnover and inflammatory signaling [41].

Simultaneously, we observed perturbations of energy and vitamin metabolism. Succinate, a TCA cycle intermediate, was reduced in multiple organs, suggesting mitochondrial dysfunction or enhanced succinate consumption by inflammatory cells [42]. Riboflavin also declined systemically. Riboflavin is a cofactor for many redox enzymes, and its depletion could reflect increased utilization under oxidative stress or impaired synthesis by gut microbes [43]. Indeed, gut bacteria are a source of B vitamins, so gut dysbiosis could contribute [43]. The rise in taurocholic acid, a taurine-conjugated bile acid, is notable. This may indicate altered bile acid metabolism by the liver (cholestatic stress) or reduced microbial deconjugation in the gut [44]. Recent omics work showed that chronic PM_2.5 exposure causes hepatic bile acid accumulation and dysregulation of bile acid pathways, which could underlie the taurocholate surge [45]. In sum, the metabolomic alterations across organs, including elevated taurine, glutathione, cortisol and arachidonate-containing lipids, together with reduced fatty acids, succinate and riboflavin, suggest a coordinated oxidative and inflammatory metabolic response to pollutant exposure [41].

Histological analysis confirmed significant lung injury: alveolar walls were thickened and infiltrated with immune cells. These changes match classic pollutant‐induced pneumonitis. Jo et al. exposed mice to chronic PM and similarly found inflammatory cell infiltration and thickened alveolar walls in lung tissue [46]. We also saw elevated Il-6 mRNA in lung, reflecting local cytokine production. Particulate matter is known to provoke pulmonary IL-6 release – Budinger et al. showed that PM_2.5 instillation into mouse lungs dramatically raises bronchoalveolar IL-6 [47]. IL-6 in the lung can have systemic sequelae and itself indicates substantial lung immune activation [48]. Thus, our lung data confirm that BC triggers a classic inflammatory injury in the airway, with structural damage and pro‐inflammatory cytokine upregulation.

Evidence of secondary injury, likely mediated by systemic inflammation and stress hormones, emerged in other organs. In the heart, we observed myocardial edema on H&E slides and upregulation of Mb mRNA. Myoglobin is a sensitive marker of cardiac muscle stress [49], implying BC-induced cardiomyocyte strain. While direct citations for myoglobin in PM exposure are scarce, cardiovascular epidemiology firmly links particulate exposure to heart damage: chronic PM increases the risk of ischemic heart disease and infarction [50]. The edema we see likely reflects vascular leakage or inflammation in the myocardium. Together with elevated cortisol and systemic inflammation, these findings imply that BC exposure exerts cardiotoxic effects.

In the kidney, expression of Igfbp7 was significantly higher in BC‐exposed mice. IGFBP7 is a recognized early biomarker of renal tubular stress and injury [51]. Its elevation is observed in acute kidney injury models and pollutant toxicity: for example, mice chronically exposed to cadmium (a heavy metal pollutant) show strong upregulation of renal IGFBP7 [52]. By analogy, the IGFBP7 increase here suggests that BC components (or secondary inflammatory factors) induce early tubular injury. Our data indicate subclinical renal damage, which aligns with known nephrotoxic effects of air pollution, mediated by oxidative stress and inflammation [53].

Pathological changes were also evident in brain tissue: nuclear condensation of cortical cells, indicative of neuronal damage or apoptosis. Particulates can affect the brain by crossing the blood–brain barrier or via systemic inflammation [54]. Studies have demonstrated that inhaled nanoparticles and secondary cytokines (TNF-α, IL-6) activate microglia and induce oxidative in the brain [55]. Our histological finding is consistent with these mechanisms: BC likely triggered neuroinflammation, leading to neuronal injury. Though we did not measure brain cytokines, the systemic inflammatory milieu (elevated IL-6, altered metabolites) and direct particle effects likely contributed to neural harm.

The gut was profoundly affected via two key pathways: impaired barrier integrity and altered microbial composition. First, tight-junction gene Tjp1 was downregulated in the cecum, suggesting impaired epithelial barrier [56]. This “leaky gut” phenotype is reported in other PM exposure models. Xie et al. found that chronic PM_2.5 inhalation destroyed the colonic mucus layer and reduced expression of Tjp1 and occluding [57]. Similarly, we saw lowered Tjp1, implicating loss of gut barrier function. Such barrier disruption would allow translocation of bacteria or LPS, fueling systemic inflammation [56]. Second, the cecal microbiome shifted markedly. Alpha-diversity was significantly decreased in BC‐exposed mice, consistent with other pollution studies. For example, gut microbiota diversity and composition changes were observed in both humans and mice exposed to high levels of PM [58], [59]. In parallel, genus‐level shifts were notable: Akkermansia was enriched and Bacteroides was depleted in BC mice. The increased abundance of Akkermansia muciniphila is notable, as this mucin-degrading bacterium commonly expands when the intestinal mucus layer or nutrient environment is altered. In Xie et al.’s PM study, Akkermansia abundance rose dramatically [57]. This matches our finding and suggests that BC may alter mucus or nutrient availability in favor of Akkermansia. In contrast, Bacteroides, which are major saccharolytic commensals involved in carbohydrate fermentation, were substantially reduced. Our observation agrees with other rodent studies: mice exposed to chronic PM had lower Bacteroides levels [60]. Since Bacteroides species produce short‐chain fatty acids (like acetate and propionate), their loss could impair gut metabolic output and anti‐inflammatory signaling [61]. Altogether, the microbiome changes form a feedback loop: pollutant-induced inflammation and mucus changes favor certain bacteria (Akkermansia, Lactococcus) which in turn modulate host metabolism and immunity.

The observed correlations between cecal microbiota and multi-organ metabolites suggest a potential role of gut microbial shifts in shaping systemic metabolic profiles. Spearman correlation analysis revealed links between microbial taxa and organ metabolites. db-RDA indicated that variations in microbial community structures were significantly associated with specific metabolites. Mantel and Procrustes analyses further supported moderate concordance between microbial and metabolomic profiles in the cecum and plasma. Integrating these findings, a multi-organ cascade emerges. Inhaled BC first injures the lung, driving a strong local immune response (IL-6, oxidative stress) and releasing systemic mediators. Simultaneously, some particles are swallowed or reach the gut via the circulation [8], perturbing the microbiome and breaking down barrier function. The disrupted gut then contributes endotoxins and altered metabolites back into the circulation. Meanwhile, circulating stress hormones and cytokines (including cortisol, IL-6) affect distal organs. In total, BC exposure sets off a systemic perturbation: oxidative damage and inflammation at the portal (lung), dysbiosis in the gut, and metabolic dysregulation in multiple organs.

Our observations align with known toxicological mechanisms of air pollution. BC carries reactive compounds (organic and inorganic) that generate free radicals. The upregulation of antioxidant metabolites (glutathione, taurine) in our study is a classic response to such ROS exposure [3], [62]. Elevated cortisol and inflammatory lipids further signal an activated stress/inflammatory state. The microbiome findings underscore the gut–lung axis: polluted air, once inhaled, alters gut ecology and vice versa, as has been increasingly recognized in recent literature [32], [63], [64]. Organ‐specific vulnerabilities are also evident: the lung is the direct target, the heart and brain are sensitive to inflammation and hypoxia, the kidney filters circulating toxins, and the gut both encounters swallowed particles and responds to systemic inflammation. The observed histologic and molecular changes in each organ reflect these pathways.

This study has several limitations. First, the murine exposure model, while allowing controlled BC dosing and mechanistic insights, may not fully recapitulate the complexity of human environmental exposures, which often involve mixed pollutants and variable concentrations. Second, our BC exposure dose and schedule, though informed by regulatory standards and literature, represent a simplified scenario that may not reflect chronic low‐level or intermittent exposures experienced by human populations. Third, the multi‐organ metabolomic and transcriptomic analyses were performed at a single terminal time point, limiting our ability to capture dynamic temporal changes or early‐stage responses. Fourth, while our gut microbiota analyses revealed clear compositional shifts, causal links between specific microbial changes and host metabolic or inflammatory alterations remain to be confirmed, for example via germ‐free or microbiota‐transfer experiments. Finally, although we integrated cross‐species evidence from human metabolomics and murine models, differences in physiology, metabolism, and microbiota between species may limit direct translational inference. Addressing these limitations in future longitudinal, multi‐pollutant, and mechanistic studies will help clarify the pathways by which BC drives systemic effects.

Conclusion

Our work redefines air pollution toxicity as a gut-orchestrated systemic crisis, in which inhaled pollutants disrupt microbial balance, barrier integrity, and metabolic networks to affect distant organs. By combining human epidemiological insight with murine mechanistic evidence, we identify consistent metabolic signatures and microbial shifts linked to BC exposure. These findings emphasize the need for microbiota-targeted or metabolite-based interventions to mitigate the far-reaching consequences of air pollution on human health.

Compliance with ethics requirements

The epidemiological study using UK Biobank data was approved by the National Health Service North West Centre for Research Ethics Committee (REC reference: R21/NW/0157), and all participants provided written informed consent. All experimental animal procedures complied with the ARRIVE guidelines and were approved by the Animal Care & Welfare Committee of the Guangzhou Center for Disease Control and Prevention (Guangzhou Health Supervision Institute) (Grant no. 2023061).

CRediT authorship contribution statement

Shengtao Wei: Data curation, Formal analysis, Methodology, Writing – original draft, Writing – review & editing. Wenxue Li: Investigation, Methodology, Writing – review & editing. Shanshan Ran: Writing – review & editing. Jingyi Zhang: Writing – review & editing. Zilong Zhang: Writing – review & editing. Zijun Yang: Writing – review & editing. Fei Tian: Writing – review & editing. Lan Chen: Writing – review & editing. Peng Hu: Writing – review & editing. Jun Yuan: Project administration, Resources, Supervision, Writing – review & editing. Hualiang Lin: Conceptualization, Data curation, Project administration, Supervision, Funding acquisition, Writing – review & editing.

Funding

This study was supported by the National Natural Science Foundation of China (No.82373534), and the GuangDong Basic and Applied Basic Research Foundation (No.2024A1515010465).

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

The following are the Supplementary data to this article: