Inflammatory bowel disease (IBD) is a persistent and recurrent intestinal disorder that tends to relapse over time. It is typically categorized into two subtypes: ulcerative colitis (UC) and Crohn’s disease (CD)^[1]. The prevalence of IBD continues to increase globally, imposing a significant burden on public health systems; according to the Global Burden of Disease Study, a total of 3,830,000 individuals suffered from IBD in 2021^[2]. Patients with IBD can present with multiple life-threatening complications, including intestinal bleeding, obstruction, perforation, and cancer and extraintestinal manifestations^[3]. Although the etiology of IBD remains to be fully elucidated, its pathophysiology is associated with genetic, immune, and environmental factors^[4]. According to previous epidemiological studies, decreased air quality may contribute to IBD progression, and some harmful air contaminants may be associated with gastrointestinal side effects^[5-7]. The underlying mechanisms behind these pathologies are complex and may include direct deleterious effects on epithelial cells, immune reaction modifications, and gut microbiota regulation^[8].

As an essential component of the air, particulate matter 2.5 (PM_2.5) is a complex mixture of many constituents, including organic carbon, black carbon (BC), sulfate, nitrate, ammonium, and metal elements^[9], and is associated with various disorders, including respiratory, endocrine, cardiovascular, and cerebrovascular diseases^[10]. PM_2.5 absorbance is a measurement of how much light the PM in PM_2.5 can absorb. It is particularly relevant for components such as BC within PM_2.5, which is known for its light-absorbing characteristics and contribution to climate change and health effects^[11]. Therefore, PM_2.5 absorbance is considered as an indicator and proxy of BC, reflecting the carbon fraction concentration in PM_2.5, which could enter the bloodstream and even be disseminated to distant organs, including the intestinal tract^[12]. At present, the potential pathogenicity of PM_2.5 in IBD remains controversial. According to a population-based cohort study, long-term exposure to PM_2.5 was associated with a 20.4% higher risk of developing IBD^[13]. In an ecologic analysis study conducted in the US, the emissions of six air contaminants, including PM_2.5, were significantly correlated with an increased risk of IBD-related hospitalizations^[14]. However, a European nested case-control study across six countries found no evidence showing that participants who suffered from IBD were exposed to higher PM_2.5 levels, with an absence of consistent association^[6]. Given the current inherent limitations of observational studies, including confounding factors, the causal relationship between PM_2.5 and IBD development remains unclear^[15]. Furthermore, although PM_2.5 absorbance is an essential indicator to assess the elemental carbon of PM_2.5, has rarely been considered in the previous studies. Therefore, further high-quality studies are needed to investigate the relationship between IBD and PM_2.5 and PM_2.5 absorbance.

Considering that there are limited studies to build a causal relationship between PM_2.5, PM_2.5 absorbance, and IBD development, we hypothesized a possible causal link between these two air pollutants and IBD. Mendelian randomization (MR) is an instrumental variable (IV) approach that uses single-nucleotide polymorphisms (SNPs) as IVs to deduce the causal relationship between exposures and outcomes. The methodology of the present study is based on Mendel’s second law of genetics, which entails classifying research subjects based on the presence of particular genetic mutations and then contrasting the incidence of outcomes among these groups. SNPs adhere to the principle of random distribution during meiosis, overcoming the disadvantages of traditional observational studies, by minimizing bias due to reverse causality or confounding factors, including environmental exposures and behaviors, given that genetic variations predate the emergence of the disease^[16-18]. Herein, we employed a two-sample MR design, which used outcome- and exposure-related IVs from two population datasets and could enhance the statistical power to investigate the causal relationship between IBD and PM_2.5 and PM_2.5 absorbance. The results of this study may provide essential insights into the causal relationship between PM_2.5, PM_2.5 absorbance, and IBD, ultimately offering a theoretical basis and further research directions for the prevention of this disease.

Figure 1 shows a flowchart of our study design, encompassing the process of identifying IVs, performing MR studies through five distinct approaches, and conducting sensitivity analyses. In order to comprehensively grasp our research framework, it is crucial to elucidate the foundation of MR, which comprises three essential assumptions. The first assumption requires that the genetic variants utilized as IVs should be closely related to exposure factors, and the correlation coefficient is primarily < 5 × 10⁻⁸. The second assumption is that the proposed genetic variants should not be linked to any confounders. The third assumption is that the selected IVs should affect the risk of the outcome only through the risk factors that we are concerned with^[16]. These assumptions ensure that the findings of MR would not be influenced by extraneous confounding factors, including population and environmental characteristics and socioeconomic conditions. Moreover, the possibility of reverse causality is eliminated since genetic variation accounts for the emergence of exposure prior to the outcome, thereby mitigating the shortcomings inherent in conventional epidemiological methodologies. We conducted two-sample MR analysis using publicly available genetic datasets in two genome-wide association studies (GWAS) to identify the causal links between PM_2.5 and UC and CD.

Figure 1.  Study design and workflow of the current study.

The UK Biobank is a large-scale, prospective cohort study encompassing genetic and phenotypic data from approximately 500,000 residents in the UK. This expansive, open-access resource offers a wealth of detailed health and lifestyle information for each participant, including biological data and biomarkers, which has achieved data depth and breadth. Moreover, follow-up information is provided by linking health and medical records for further tracing research^[19,20]. We used the UK Biobank GWAS, which contained 423,796 European participants, to obtain our summary genetic dataset on PM_2.5 (GWAS ID: ukb-b-10817) and PM_2.5 absorbance (GWAS ID: ukb-b-11312). The study, which was based on a cohort study evaluating the effects of air pollution in Europe, utilized the land use regression model to predict PM_2.5 concentrations in participants’ homes^[21]. The mean (± standard deviation) PM_2.5 level in the GWAS was 9.99 (± 1.06) µg/m^3[22]. We searched for genetic variants of UC and CD in the UK Biobank GWAS to reduce potential deviation in population stratification. The CD dataset (GWAS ID: ukb-a-552), including 732 cases of European descent and 336,467 controls, and the UC dataset (GWAS ID: ukb-b-19386), including 1987 cases and 461,023 controls, were defined on the basis of clinical diagnosis. Ethical review and approval are not applicable for our study because all GWAS data on human participants used in this study are publicly available in the IEU OpenGWAS Project (https://gwas.mrcieu.ac.uk/). These datasets can also be downloaded from the GWAS catalog (https://www.ebi.ac.uk/gwas/).

Strict quality control measures were implemented to meticulously select relevant SNPs. First, we grouped the datasets according to the criterion of P < 5 × 10⁻⁸ to filter SNPs that are closely related to exposure factors. Second, independent genetic variants were selected by conducting a clumping process, which involved setting a distance window of 10,000 kilobases and a linkage disequilibrium (LD) coefficient threshold of r² < 0.001. This step was important to avoid any LD between SNPs and to ensure the complete independence of the selected genetic variants^[23]. Third, we retained palindromic SNPs on the basis of the following threshold: minor allele frequency < 0.3^[24]. Notably, if the allele frequency contained in the details of an SNP is close to 0.5, we could hardly pinpoint the minor allele because of sampling variance around the allele frequency. Thus, we excluded such SNPs at the outset of MR analyses to improve the accuracy of our study. Fourth, we estimated the proportion of variance interpreted (R²) and F-statistics for each SNP to predict the power of the selected IVs. Genetic instruments (F-statistics < 10) were regarded as weak instruments, which should be ruled out from MR analysis^[25]. Finally, we used the Mendelian Randomization Pleiotropy RESidual Sum and Outlier (MR‐PRESSO) test to determine potential horizontal pleiotropy and to remove aberrant IVs (outliers) in order to exclude the influence of pleiotropy and estimate the corrected results^[26].

The inverse variance weighting (IVW) test was used to assess the causal relationship between PM_2.5, PM_2.5 absorbance, and IBD. This test is known for its ability to greatly determine causation^[27]. We also complemented our evidence using MR-Egger, weighted mode, weighted median, and simple mode to enhance the precision and stability of our results^[28]. Aside from using the MR‐PRESSO test, we also performed an MR-Egger regression test as a further measure to determine whether any pleiotropy was present among the selected SNPs, thereby increasing the overall credibility and integrity of our findings. If there was no significant difference (P < 0.05) between the MR-Egger intercept and 0, then IVs were considered to not affect IBD risk through other confounding factors^[29]. For the quality of the IVW test and MR-Egger model, Cochran’s Q test was conducted to calculate heterogeneity between screened SNPs. If the P-value from Cochran’s Q test was less than 0.05, then there is heterogeneity among the study results^[30]. We also applied leave-one-out analysis to exclude random deviations from selecting IVs and to identify whether the causal link would be affected by a particular SNP^[31].

All statistical analyses were conducted using R software (version 4.1.3) by utilizing the “TwoSampleMR”^[23] package and “MR-PRESSO”^[26]. Statistical significance was established at P < 0.05.

We used P < 5 × 10⁻⁸ as a threshold for selecting SNPs and conducted the clumping LD process (r² < 0.001). Eight IVs associated with PM_2.5 and five IVs associated with PM_2.5 absorbance were obtained. Notably, we adjusted the criterion to P < 1 × 10⁻⁷ in the PM_2.5 absorbance and UC group to detect more relevant SNPs and enhance the statistical power^[32,33]. Ultimately, seven IVs of CD and five IVs of UC were selected in the MR analyses of PM_2.5 to IBD. In the MR analyses of PM_2.5 absorbance to IBD, four IVs of CD and three IVs of UC were screened. The F-statistic was greater than 10 (ranging from 30 to 69) for all IVs, suggesting the absence of weak IV deviation. No outlier was removed through MR-PRESSO considering the nonsignificant result of the MR-PRESSO global test (P > 0.05). More detailed information about the selected IVs is presented in Supplementary Table S1.

The IVW MR analysis, which was regarded as the primary method to estimate the causal relationship in our study, genetically revealed a significant association between PM_2.5 and UC (odds ratio [OR] = 1.010; 95% confidence interval [CI] = 1.001–1.019, P = 0.020). Such associations were statistically consistent, although nonsignificant in the MR-Egger, weighted mode, weighted median, and simple mode. However, we did not observe a significant association between PM_2.5 exposure and Crohn´s Disease in MR analysis (OR = 1.002; 95% CI = 0.995–1.009, P = 0.516) (Table 1, Supplementary Table S2, Figure 2, and Figure 3). No heterogeneity was observed in the sensitivity analysis (Cochran’s Q: P = 0.732 and P = 0.611) (Supplementary Table S3), and no evidence of horizontal pleiotropy was detected in the MR-Egger regression analyses (Egger intercept: P = 0.312 and P = 0.591) (Supplementary Table S4). The results of leave-one-out analysis, wherein each SNP was deleted individually, were consistent with those of the IVW method. The sensitivity analyses showed that the positive associations found by MR are consistent (Supplementary Table S5 and Supplementary Figure S1).

Figure 3.  Forest plots were used to visualize the causal effect between PM_2.5 and UC (Panel A), PM_2.5 and CD (Panel B), PM_2.5 absorbance and UC (Panel C), PM_2.5 absorbance and CD (Panel D). The y-axis shows the analyses for the effect of each SNP on UC and CD, The red points indicate the IVW and MR–Egger estimates using all SNPs. MR, Mendelian randomization; PM_2.5, particulate matter 2.5; SNP, single nucleotide polymorphism; UC, ulcerative colitis; CD, Crohn’s disease.

We also found a significant genetic correlation between PM_2.5 absorbance and UC (OR = 1.012; 95% CI = 1.004–1.019, P = 0.002) in the IVW test of PM_2.5 absorbance to UC. The ORs of the five models performed in the MR analyses are all consistently positive (Table 1, Supplementary Table S2, Figure 2, and Figure 3). Simultaneously, we found no evidence that the genetic level of PM_2.5 absorbance was significantly associated with CD (OR = 0.998; 95% CI = 0.989–1.008, P = 0.769) (Table 1, Supplementary Table S2, Figure 2, and Figure 3). No heterogeneity was detected in Cochran’s Q tests (P = 0.732 and P = 0.611) (Supplementary Table S4). We also eliminated the probability that the MR-Egger regression test had P-values less than 0.05 with the exclusion of potential pleiotropy (P = 0.396 and P = 0.556) (Supplementary Table S3). Leave-one-out plots predicted that no single IV could independently drive the results of MR. All evidentiary materials ensured the reliability of the results (Supplementary Table S5 and Supplementary Figure S1). We eliminated the possibility that the MR-Egger intercept had P < 0.05 with the exclusion of possible horizontal pleiotropy.

Figure 2.  Scatter plots were used to visualize the causal effect between PM_2.5 and UC (Panel A), PM_2.5 and CD (Panel B), PM_2.5 absorbance and UC (Panel C), PM_2.5 absorbance and CD (Panel D). The x-axis shows the SNP effect and SE on PM_2.5 and its absorbance. The y-axis shows the SNP effect and SE on UC and CD. The regression lines for the inverse-variance weighted (IVW) method, the MR–Egger regression method, the weighted median, the weighted mode, and the simple mode are shown. The slope of each straight line indicates the magnitude of the causal association. PM_2.5, particulate matter 2.5; SNP, single nucleotide polymorphism; SE, standard error; UC, ulcerative colitis; CD, Crohn’s disease.

Figure S1.  Forest plots of the “leave-one-out” sensitivity analyses to demonstrate the impact of individual SNPs on the results. The x-axis shows MR “leave-one-out” sensitivity analyses for PM_2.5 on UC (Panel A), PM_2.5 on CD (Panel B), PM_2.5 absorbance on UC (Panel C), PM_2.5 absorbance on CD (Panel D). The y-axis shows the analyses for the effect of “leave-one-out” of SNPs on UC and CD. The red points indicate the IVW estimates using all SNPs. MR, Mendelian randomization; PM_2.5, particulate matter 2.5; SNP, single nucleotide polymorphism; UC, ulcerative colitis; CD, Crohn’s disease.

The relationship between PM_2.5 and IBD has been debated. For example, some studies indicated that PM_2.5 exposure or inhalation was associated with an increased risk of IBD development and IBD-related hospitalizations^[13,14]. It was also found that a 1-log increase in the density of total criteria pollutant emission was associated with a 40% increase in the rate of UC and CD hospitalizations^[14]. In a Chinese study using the distributed lag nonlinear model, every 10 mg/m³ increase in PM_2.5 was correlated with a higher chance of developing IBD (relative risk = 1.037, 95% CI = 1.005–1.070)^[34]. Similar observations were reported from other regions^[35]. Other studies did not find an association between PM_2.5 and IBD^[6]. However, these conclusions may not be reliable because of the inherent drawbacks of observational study designs.

In order to address these issues, the present study used MR analysis to investigate genuine causal connections. The findings offer conclusive evidence indicating a causative link between PM_2.5, PM_2.5 absorbance, and an elevated risk of UC. The results of the present study provide new evidence for this epidemiological association and the clinical practice of disease prevention.

There are several possible biological mechanisms by which PM_2.5 and PM_2.5 absorbance increase the risk of UC development. One is the direct effects of PM_2.5 on intestinal epithelial cells. We speculated that PM_2.5 would exhibit a similar cytotoxic effect in intestinal epithelial cells since airborne pollutants could induce airway epithelial cell injuries and cause airway cellular apoptosis^[5]. Furthermore, two previous studies using animal models also suggested that PM_2.5 administration might trigger systemic inflammatory response and oxidative stress^[36,37]. In this process, oxygen radicals, inflammatory factors, and cytokines would be released in serum and tissue, further inducing chronic organic injury and increasing the risk of autoimmune conditions (e.g., neuroinflammation, type 1 diabetes, asthma, and systemic inflammatory cardiovascular diseases)^[38-42]. Given that IBD is a chronic condition involving immune-mediated and autoimmune disorders^[43], future studies need to determine whether PM_2.5 may have an immunomodulatory effect on the intestinal mucosa. Another possible mechanism that might mediate the effects of PM_2.5 on the gut is its direct effects on intestinal microecology. PM_2.5 exposure can decrease the abundance and diversity of intestinal microbiota, thus leading to intestinal flora dysfunction^[44-46]. Moreover, it was presumed that PM_2.5 absorbance can lead to oxidative stress, thereby leading to IBD onset. A large cohort study showed that PM_2.5 absorbance was possibly related to increased levels of gamma-glutamyl transferase (GGT)^[47], which is an indicator of liver damage related to oxidative stress, and a hospital-based survey proved that patients with IBD had abnormally increased GGT levels^[48]. Notably, apart from IBD and other respiratory diseases, air pollution can lead to other disorders in our body. With regard to other gastrointestinal diseases, air pollutants can trigger the onset of and exacerbate gastrointestinal inflammation^[45], colorectal cancer^[49], and liver cancer^[8]. Some studies have investigated the potential pathophysiological mechanisms behind the effects of air pollutants. For example, the inhalation of airborne particles can modify gut microbiota, specifically affecting Firmicutes, Acidobacteria, and Proteobacteria, and provoke acute and chronic inflammatory reactions within the intestinal tract^[50]. Furthermore, PM can initiate oxidative-stress-induced cell death in the gastrointestinal epithelium, compromise the integrity of tight junction proteins, and play a role in exacerbating gastrointestinal inflammatory disorders in vitro and in vivo^[51]. PM even has the potential to precipitate systemic metabolic alterations by interfering with glycerophospholipid metabolism and linoleic acid pathways^[50]. From this point of view, air pollution may cause neurotoxicity, including autism spectrum disorder, Alzheimer’s disease, and Parkinson’s disease^[52]. Moreover, positive relationships were found between the incidence of chronic kidney disease and PM_2.5^[53]. According to a combined cross-sectional and retrospective cohort study performed on 8,689 children in China, exposure to outdoor air pollutants, including carbon monoxide, ozone, and PM_2.5, in early life was associated with childhood parasitic infections^[54].

There are also some hypotheses that explain the link between genetic variants and our environmental exposures of interest. The interaction between genetic variations and environmental exposures, known as gene-environment interactions, has been reported to play a critical role in complex human traits and diseases. Specifically, the risk genotype intensifies the adverse impact of environmental risk factors; conversely, exposure to such risk factors can amplify the influence of the risk genotype. Furthermore, environmental exposure and genetic predisposition independently contribute to disease risk and synergistically increase the risk when combined^[55,56]. Epigenetic theory can also explain the relationship between air pollutants and genetic variants. According to multiple experimental studies, PM_2.5 exposure is significantly associated with the methylation of various genes, including those related to aging^[57], and the circadian rhythm^[58]. Thus, extensive research is needed to obtain a clearer understanding of the physiological mechanisms.

Regarding the relationship between PM_2.5 and CD, only a limited number of epidemiological studies have been screened, and their findings are inconsistent. Specifically, a Chinese study indicated that every 10 mg/m³ increase in PM_2.5 was correlated with a higher chance of IBD development and that the effects on CD are more pronounced than those on UC^[34]. However, a European study using multivariable Cox proportional hazards models reported that long-term exposure to airborne pollutants, including PM_2.5, was not correlated with the risk of CD development^[59]. Our study avoided the confounding factors of traditional epidemiological research, addressing the disputes of previous studies. However, more extensive studies are needed to provide more evidence regarding the associations of PM_2.5 and PM_2.5 absorbance with the risk of IBD.

Our MR study has the following key advantages. First, to the best of our knowledge, the present study is the first to analyze the causal relationships between PM_2.5, PM_2.5 absorbance, and IBD using two-sample MR. This resolves the issues of previous epidemiologic studies, compensates for the deficiencies of traditional observational studies, and provides new demonstrations for evaluating the health hazards of environmental pollution. Second, our study profits from large-scale PM_2.5, PM_2.5 absorbance GWAS (n = 423,796 individuals from Europe), and IBD GWAS (CD: n = 337,199 individuals from Europe; UC: n = 463,010 individuals from Europe) datasets, improving the reliability of analyses. Additionally, the potential deviation induced by population stratification has been diminished as the participants are of European descent. Moreover, we used multiple independent SNPs as instruments to eliminate the effect of LD on possible links. Eventually, we performed various methods for MR analyses and conducted comprehensive pleiotropy analyses and heterogeneity tests to assess the MR results. The absence of horizontal heterogeneity and pleiotropy indicated that the impact of nonheritable environmental confounders was minimized and ensured that the stability and credibility of our results are superior to other those of other traditional epidemiological methods.

In the present study, we applied MR analyses to evaluate the potential causal relationships between PM_2.5, PM_2.5 absorbance, and IBD risk. In individuals of European ancestry, we demonstrated that PM_2.5 was associated with a 1.0% higher risk of UC and that PM_2.5 absorbance was linked with a 1.2% higher risk of UC from a genetic perspective. Notably, considering that the impact of genetic variants occurs at birth, MR estimates the long-standing influence of risk factors on outcomes across life, rather than at a single temporal juncture. Thus, unlike traditional observational research, which captures exposures at specific time points, the primary focus of MR is to evaluate the causal hypothesis rather than the precise quantification of the effect size (point estimates and intervals)^[60]. Additionally, no significant association was observed between PM_2.5 exposure, PM_2.5 absorbance, and CD. Our research pioneers the application of MR to establish associations between PM_2.5, PM_2.5 absorbance, and IBD at the genetic level, offering new perspectives for the prevention and management of gastrointestinal diseases. The results may have implications for formulating public health responses and disease prevention strategies (e.g., enforcing more stringent emission regulations, advocating for eco-friendly transportation, enhancing public knowledge of air quality, etc.). Consequently, these measures may help with controlling the incidence of UC.

Despite the advantages, this study also has some limitations. First, the GWAS datasets included in the MR analyses were based on European ancestry. Thus, further studies on individuals from other countries are needed to improve the generalizability of our results. Second, because of the limited information provided by the original analyses and datasets, we could hardly pinpoint the localization of European participants nor obtain demographic information (e.g., age, gender, and socioeconomic status) about them. The composition of PM_2.5 pollutants, particularly concerning the nature of airborne PM, could exhibit considerable diversity across Europe. Meanwhile, demographic factors may influence participants’ susceptibility to diseases. As a result, they might lead to the potential imprecision of our MR results. Third, given the relatively low incidence of IBD, all sample data used in the present study come from European participants, and there might be some inevitable overlap. Considering the limited details available from the database, we were unable to identify and eliminate these overlaps, which might affect the reliability of our results. Finally, it may be very difficult for us to eliminate confounding factors completely and absolutely using summary statistical data for our MR analysis. Hence, we could only draw a preliminary conclusion regarding the causal associations between PM_2.5, PM_2.5 absorbance, and IBD. Further research will focus on conducting more comprehensive studies to collect high-quality evidence concerning the idiographic pathophysiological mechanisms through which air pollution affects IBD risk. This includes expanding the study’s scope to encompass a broader range of air pollutants, identifying possible biomarkers that could shed light on how air contaminants contribute to IBD onset, exploring genetic susceptibilities that may modify the impact of airborne PM, and longitudinal studies to track air pollution over time and their direct correlation with IBD incidence.