Introduction

Chronic Obstructive Pulmonary Disease (COPD) is a progressive and debilitating respiratory condition affecting millions globally1,2. Characterized by persistent airflow limitation and respiratory symptoms, COPD is one of the leading causes of morbidity and mortality worldwide3. The World Health Organization projects that by 2030, COPD will become the third leading cause of death, highlighting the urgent need for enhanced understanding, prevention, and treatment strategies4. This escalating health crisis calls for innovative approaches to unravel the complex pathophysiology of COPD and develop more effective therapeutic interventions.

The pathogenesis of COPD is intricate, involving an interplay of genetic predispositions, environmental exposures, and lifestyle factors5. While smoking remains the predominant risk factor, increasing evidence points to the significant roles of air pollution, occupational exposures, and genetic susceptibility in the development and progression of the disease6,7,8. The heterogeneous nature of COPD, characterized by varying phenotypes and endotypes, further complicates disease comprehension and presents challenges in developing targeted therapies. Despite advances in symptomatic management, current COPD treatments remain largely palliative, focusing primarily on bronchodilation and inflammation reduction9. There is an unmet need for therapies that can modify disease progression, slow its advancement, and improve long-term clinical outcomes10. Addressing this gap underscores the importance of identifying novel therapeutic targets and developing interventions that focus on the underlying molecular mechanisms of COPD.

In recent years, there has been growing interest in the application of metabolomics to better understand complex diseases such as COPD11. Metabolomics, the comprehensive study of small-molecule metabolites in biological systems, provides a valuable insight into the biochemical processes underlying disease states12. By analyzing the metabolic profiles associated with COPD, researchers aim to identify novel biomarkers for early diagnosis, discover new therapeutic targets, and gain a deeper understanding of the molecular mechanisms driving disease progression13. The metabolome, as the final downstream product of gene expression and environmental interactions, offers a real-time snapshot of an individual’s physiological state14.

The advent of genome-wide association studies (GWAS) and Mendelian randomization (MR) techniques has provided researchers with powerful tools to explore the causal relationships between metabolites and disease outcomes. These methods allow the use of genetic variants as instrumental variables to infer causal associations, facilitating the disentanglement of the complex factors contributing to COPD pathogenesis15. Furthermore, the integration of metabolomic data with genetic information offers the potential to identify metabolic pathways and rate-limiting enzymes as targets for therapeutic intervention16. This systems biology approach provides a more holistic understanding of the disease process compared to traditional reductionist methods. Investigating the interplay between lifestyle factors, metabolic profiles, and COPD risk is another critical area of research. While the detrimental effects of smoking on lung health are well-established, the impact of other modifiable lifestyle factors on COPD-related metabolites remains poorly understood17. Elucidating these relationships could enable the development of targeted lifestyle interventions as part of a comprehensive strategy for COPD prevention and management. The potential for lifestyle modifications to influence disease-related metabolites presents a promising avenue for non-pharmacological interventions in COPD care.

In this context, our study aims to employ advanced analytical techniques to explore the intricate relationships between metabolites, genetic factors, and COPD risk. By combining two-sample Mendelian randomization, co-localization analysis, and protein-protein interaction network analysis, we seek to identify metabolites that are causally associated with COPD risk, elucidate the genetic architecture underlying these associations, uncover key metabolic pathways and rate-limiting enzymes involved in COPD pathogenesis, explore interactions between these pathways and existing drug targets, and assess the impact of lifestyle factors on COPD-related metabolites. This comprehensive approach promises to enhance our understanding of COPD biology and holds significant clinical implications. By identifying novel biomarkers and therapeutic targets, our findings could inform the development of personalized prevention strategies and targeted treatments for COPD. The integration of metabolomic data with genetic and lifestyle information may allow for the stratification of COPD patients into distinct subgroups, facilitating more tailored treatment approaches and improving clinical outcomes18,19. Moreover, by clarifying the impact of lifestyle factors on disease-related metabolites, this research may provide a scientific foundation for incorporating lifestyle interventions into COPD management. Understanding how modifiable behaviors influence the metabolic pathways involved in COPD could empower both patients and healthcare providers to make informed decisions about lifestyle modifications that complement traditional medical treatments20.

In the era of precision medicine, the integration of metabolomic, genetic, and lifestyle data offers a robust approach to unraveling the complexities of COPD. By bridging the gap between basic science and clinical application, this study aims to contribute to the development of more effective, personalized approaches to COPD prevention, diagnosis, and treatment. Identifying novel drug targets through the exploration of metabolic pathways and their interactions with existing therapies could accelerate drug discovery efforts and lead to more targeted interventions.

Method

Study design

This study first explored the association between metabolites and COPD through a two-sample Mendelian randomization analysis. Co-localization analysis was then employed to evaluate the relationship between metabolites and susceptibility to COPD. Following this, we identified pathways in which the positively associated metabolites were enriched, with a particular focus on finding the rate-limiting enzymes involved in these pathways. To further investigate potential therapeutic targets, we used protein-protein interaction (PPI) network analysis to explore interactions between these rate-limiting enzymes and existing drug targets.

Finally, the Mendelian randomization method was systematically applied to examine the relationship between healthy lifestyle factors and the six FDR-corrected metabolites. The objective was to provide supportive validation for our colocalization findings, helping to distinguish true causal metabolites from those confounded by lifestyle behaviors (Fig. 1).

Fig. 1
Fig. 1
Full size image

Study Workflow. (1) Screening: MR analysis was performed to screen 1,400 metabolites (GWAS Catalog) against COPD risk (FinnGen). (2) Causal Confirmation: Carnitine C14 and 3-hydroxyoleoylcarnitine were identified using stringent colocalization (SuSiE, PP4 > 0.8). (3) Mechanism Exploration: Rate-limiting enzymes (ACACA, ACACB, ELOVL6) were identified, and a PPI network revealed interactions between ACACA/ACACB and the Salbutamol target, ADRB2. (4) Confounder Check: Pleiotropy was excluded by assessing 17 lifestyle factors. (5–6) Validation: Results were validated using multi-omics data from human COPD tissues (GEO) and in vitro experiments demonstrating that Salbutamol restores fatty acid homeostasis via the ADRB2-ACACA axis.

Data sources

Data for 1,400 metabolites were obtained from the GWAS catalog, with stringent inclusion criteria applied, including a significance threshold of 5e-8, R² < 0.001, and a linkage disequilibrium (LD) window of 10,000 kb21. The study analyzed 1,091 metabolites and 309 metabolite ratios derived from plasma samples, which were categorized into eight superpathways: lipid, amino acid, xenobiotics, nucleotide, cofactor and vitamins, carbohydrate, peptide, and energy pathways22.

These metabolites were assessed in the Canadian Longitudinal Study on Aging (CLSA) using GWAS. The analysis identified 690 metabolites significantly associated with 248 loci and 143 metabolite ratios linked to 69 loci. Of the 850 known metabolites, 81 newly discovered metabolites were evaluated, resulting in 85 unique associations. This dataset offers valuable insights into the genetic architecture underlying metabolite levels and identifies potential therapeutic targets for metabolic diseases.

Data on COPD were sourced from FinnGen R11, while lifestyle-related data were obtained from the IEU openGWAS database23,24.

MR analysis

Two-sample Mendelian randomization (MR) was performed using the “TwoSampleMR” R package to evaluate the causal relationships between 1,400 metabolites and COPD. Instrument validity was first ensured by clumping all SNPs using a stringent r^2 < 0.001 and a 10,000 kb window to achieve independence. The F-statistic for each instrument was calculated to assess instrument strength, excluding any with F < 10.

Our analytical strategy was contingent on the number of available instrumental variables (SNPs) for each metabolite. (A) For metabolites associated with only one valid SNP, the Wald ratio method was applied18. (B) In cases where multiple independent SNPs were available, the random-effects inverse-variance weighted (IVW) method was planned as the primary analysis25. For these multi-SNP exposures, a comprehensive suite of sensitivity analyses (including Cochran’s Q test, MR-Egger regression, the Weighted Median and Mode methods, and MR-PRESSO) was also specified to assess heterogeneity and horizontal pleiotropy. As reported in the Results (Fig. 3), all six metabolites that ultimately passed FDR correction were instrumented by a single SNP. Consequently, the Wald ratio was the only applicable method for these primary findings. Finally, the Steiger directional test was performed to confirm the direction of effect (i.e., that metabolites influence COPD risk, not vice-versa). To account for multiple testing across all 1,400 metabolites, the Benjamini-Hochberg method was used to adjust p-values and control the false discovery rate (FDR)26.

Colocalization analysis

We utilized the coloc R package coupled with the SuSiE framework to identify shared causal variants, a method designed to robustly account for multiple association signals within genomic regions27. Analyses were performed within a ± 250 kb window around lead SNPs, consistent with locus-scale strategies to capture regional linkage disequilibrium (LD)28,29. We applied robust prior probabilities (p1 = 10− 4, p2 = 10− 4, p12 = 10− 5) following established sensitivity guidelines30. GWAS summary statistics were harmonized to the hg38 build using the 1000 Genomes European reference panel. To distinguish true shared signals from distinct variants in LD (high PP3), we strictly defined colocalization as a posterior probability of a shared variant (PP4) > 0.8, abandoning the composite PP3 + PP4 metric30,31,32.

Metabolite pathway analysis and identification of rate-limiting enzymes

The SUPER-PATHWAY and SUB-PATHWAY classifications for all metabolites were identified from the original metabolite dataset. Based on these classifications, we systematically searched for the rate-limiting enzymes within each pathway, particularly those regulating key steps in the identified SUB-PATHWAYS33. These rate-limiting enzymes are crucial as they control the metabolic flux through pathways and serve as potential targets for therapeutic interventions.

Drug targets and PPI network analysis

We initially used DrugBank to compile a list of drug targets for commonly used COPD treatments. We then evaluated the potential relevance of proteins identified as therapeutic targets for COPD, focusing on those with robust supporting evidence. Any protein targeted by approved or investigational drugs was classified as a potential drug target, and detailed information on these drugs was recorded34.

Subsequently, a protein-protein interaction (PPI) network was constructed using the STRING database (version 12.0)35. The network was built using a minimum required interaction score of high confidence (0.700), incorporating all seven evidence channels (Textmining, Experiments, Databases, Co-expression, Neighborhood, Gene Fusion, and Co-occurrence) to ensure the reliability of all reported edges.

The impact of lifestyle on metabolites

We conducted Mendelian randomization analysis to investigate the influence of lifestyle factors on COPD-related metabolites36. The primary objective was to validate our colocalization findings. We hypothesized that metabolites identified as confounded (high PP3) would associate with their known behavioral sources (e.g., caffeine, smoking), while metabolites confirmed as causal (high PP4) would associate with relevant metabolic or smoking-related factors. A total of 17 lifestyle factors (detailed in Supplementary Material 1) were analyzed to evaluate their causal relationships with COPD-related metabolites.

The methodology employed in this analysis was consistent with the approach used for Mendelian randomization across the entire metabolome, ensuring robustness and reliability. All statistical analyses were conducted using R software version 4.4.2, providing a comprehensive framework for assessing the influence of lifestyle factors on COPD-related metabolites37.

Validation using external transcriptomic datasets

To robustly validate the expression patterns of the identified key genes (ACACA, ACACB) and drug target (ADRB2) in human COPD samples, we retrieved three independent gene expression datasets (GSE10006, GSE19407, and GSE20257) from the Gene Expression Omnibus (GEO) database. These datasets included lung tissue samples from COPD patients and healthy controls. To eliminate non-biological variability between different studies, we merged the datasets and performed batch correction using the “sva” R package (Combat method). Principal Component Analysis (PCA) was utilized to visualize the data distribution before and after batch correction. Subsequently, differential expression analysis was conducted to compare the mRNA levels of ACACA, ACACB, and ADRB2 between the COPD and control groups, using Student’s t-test with a significance threshold of P < 0.05.

In vitro experimental validation

Cell culture and treatment: Human bronchial epithelial cells (BEAS-2B and 16HBE) were purchased from the American Type Culture Collection (ATCC). Cells were cultured in DMEM supplemented with 10% fetal bovine serum (FBS) and 1% penicillin/streptomycin at 37 °C in a 5% CO₂ atmosphere. To construct an in vitro COPD model, cells were exposed to Cigarette Smoke Extract (CSE). The experimental groups were divided into: Control, CSE stimulation (2.5%), and CSE + Salbutamol (10 µM) treatment.

Quantitative Real-Time PCR (qRT-PCR): Total RNA was extracted using the TRIzol reagent (Invitrogen). Reverse transcription was performed using a PrimeScript RT Reagent Kit (Takara). qPCR was conducted using SYBR Green I (Takara) on a LightCycler 480 system. The specific primer sequences used in this study are listed in Supplementary Table S1. Relative mRNA expression was calculated using the 2−△△Ct method, normalized to GAPDH.

Western Blotting: Cells were lysed in RIPA buffer containing protease and phosphatase inhibitors. Proteins were separated by SDS-PAGE and transferred onto PVDF membranes. The membranes were blocked and incubated with primary antibodies against ACACA (proteintech,#21923-1-AP), phospho-ACACA (Abcam, ab68191), and GAPDH (Abcam, ab8245) overnight at 4 °C, followed by incubation with HRP-conjugated secondary antibodies. Protein bands were visualized using an ECL system, and band intensity was quantified using ImageJ software.

Statistical Analysis: Data are presented as mean ± standard deviation (SD). For comparisons among multiple groups (Control vs. CSE vs. CSE + Sal), a one-way analysis of variance (ANOVA) followed by Tukey’s post-hoc test was performed. A P-value < 0.05 was considered statistically significant.

Results

Mendelian randomization analysis of metabolites and COPD

In the initial analysis, we identified significant correlations between 59 metabolites and COPD (P < 0.05, Supplementary Material 2). After adjusting for the false discovery rate (FDR), six metabolites remained significantly associated with COPD (FDR < 0.05, Figs. 2 and 3).

Figure 2 illustrates the distribution of these metabolites, highlighting the six initial candidates that passed the FDR threshold. Figure 3 presents the odds ratios (OR) and confidence intervals (CI) for the six key metabolites that retained statistical significance after FDR correction, including: Carnitine C14 (OR = 1.638, 95% CI: 1.360–1.974, P < 0.001); Paraxanthine (OR = 0.704, 95% CI: 0.584–0.850, P < 0.001); 3-methylxanthine (OR = 1.460, 95% CI: 1.192–1.787, P < 0.001); Indolin-2-one (OR = 0.792, 95% CI: 0.700–0.897, P < 0.001); 6-hydroxyindole sulfate (OR = 0.701, 95% CI: 0.579–0.848, P < 0.001); and 3-hydroxyoleoylcarnitine (OR = 1.388, 95% CI: 1.166–1.652, P < 0.001).

These six metabolites were subjected to further validation using stringent colocalization analysis.

Fig. 2
Fig. 2
Full size image

Volcano plot depicting the causal relationships between 1,400 metabolites and chronic obstructive pulmonary disease (COPD). The horizontal dashed line marks the false discovery rate (FDR < 0.05) significance threshold. Metabolites significantly associated with an increased risk of COPD (FDR < 0.05) are highlighted in blue, while those associated with a decreased risk of COPD are shown in red. The size of the point reflects the variance explained (PVE). The six labeled metabolites passed the initial FDR threshold, but only Carnitine C14 and 3-hydroxyoleoylcarnitine were subsequently confirmed to have a shared causal variant with COPD.

Fig. 3
Fig. 3
Full size image

Forest plot illustrating the causal relationships between FDR-corrected metabolites and COPD. Odds ratios (OR) and 95% confidence intervals (CI) for the six key metabolites with significant associations are presented, highlighting key metabolites such as Carnitine C14, Paraxanthine, and 3-methylxanthine.

Co-localization results

We performed conditional colocalization analysis on the six FDR-corrected metabolites using a ± 250 window and the SuSiE method. Under the stringent criterion of PPH4 > 0.8, only two metabolites demonstrated robust evidence of colocalization with COPD: Carnitine C14 (PPH4 = 0.925) and 3-hydroxyoleoylcarnitine (PPH4 = 0.855) (Fig. 4, Supplementary Material 3).

The remaining four metabolites, including Paraxanthine and 3-methylxanthine, failed to meet the PP4 threshold. For these loci, the results were overwhelmingly dominated by high PP3 (≈ 93%), suggesting that their genetic associations with COPD and metabolite levels are likely driven by distinct causal variants in linkage disequilibrium. This confirms our focus on the two lipid-related metabolites.

To rule out reverse causality, we conducted the Steiger test, which confirmed the correct causal direction (i.e., metabolites influencing COPD) for all six initial associations. This directionality is meaningful only for the two metabolites subsequently confirmed by colocalization (Carnitine C14 and 3-hydroxyoleoylcarnitine) (Supplementary Material 4).

Fig. 4
Fig. 4
Full size image

Colocalization of genetic loci for COPD and plasma metabolites using LocusComparer. Scatter plots display the concordance of association signals between metabolite levels (x-axis) and COPD risk (y-axis) within locus windows of ± 250 kb. Each point represents a SNP, positioned according to its –log10(P) for both traits. Lead SNPs are highlighted by purple diamonds and labeled with their rsIDs, whereas point colors indicate linkage disequilibrium (LD; r²) with the lead SNP. Panels (A–F) correspond to: (A) Carnitine C14, (B) Paraxanthine, (C) 3-methylxanthine, (D) Indolin-2-one, (E) 6-hydroxyindole sulfate, and (F) 3-hydroxyoleoylcarnitine. Although all loci exhibit overlapping association patterns, strong colocalization evidence (PP4 > 0.8) supports shared causal variants for COPD with Carnitine C14 (A) and 3-hydroxyoleoylcarnitine (F), whereas other metabolites likely reflect distinct causal signals in LD.

Search for metabolic pathways and rate-limiting enzymes

Based on the co-localization results, we selected six metabolites and identified their corresponding pathways. For the SUPER-PATHWAY classification, both Carnitine (C14) and 3-hydroxyoleoylcarnitine were assigned to the lipid pathway, specifically within the fatty acid metabolism subpathway. The other four metabolites (Paraxanthine, 3-methylxanthine, Indolin-2-one, 6-hydroxyindole sulfate) were classified under the external ingestion pathway. Paraxanthine and 3-methylxanthine, for instance, are key metabolites of caffeine and nicotine, respectively, reflecting behavioral factors that may confound simple causal interpretation.

In terms of SUB-PATHWAYS, Carnitine (C14) belongs to the Fatty Acid Metabolism (Acid Carnitine, Long Chain Saturated) subpathway, regulated by the rate-limiting enzymes ELOVL6, ACACA, and ACACB. Similarly, 3-hydroxyoleoylcarnitine is part of the Fatty Acid Metabolism (Acid Carnitine, Hydroxy) subpathway, also regulated by the enzymes ACACA and ACACB (Supplementary Material 5).

As depicted in Fig. 5, the protein-protein interaction (PPI) network highlights the interactions between these rate-limiting enzymes and other key proteins, revealing potential therapeutic targets in COPD-related fatty acid metabolism. The network analysis also highlights an interaction between ACACB and ADRB2, the primary target of Salbutamol, suggesting that therapeutic strategies aimed at modulating ADRB2 may hold promise for COPD treatment.

Fig. 5
Fig. 5
Full size image

PPI network analysis illustrating the interactions between existing drug targets and rate-limiting enzymes in the fatty acid metabolism pathway associated with metabolites linked to COPD. The key enzymes ACACB, ELOVL6, and ACACA regulate the levels of Carnitine (C14) and 3-hydroxyoleoylcarnitine. The network indicates a significant interaction between ACACB and Salbutamol, which targets ADRB2, suggesting potential therapeutic interventions involving ADRB2 in COPD treatment.

Drug targets and PPI network construction

To further identify potential therapeutic targets for chronic obstructive pulmonary disease (COPD), we selected five drugs currently used to treat COPD and retrieved their corresponding targets from the DrugBank database (Supplementary Material 6). Upon analysis, we found that ACACB and ACACA, key rate-limiting enzymes in fatty acid metabolism, display high-confidence interactions within a network containing major targets for COPD medication (Fig. 5). The visualization highlights a significant interaction between ADRB2 and other proteins, establishing its close network proximity to the core metabolic enzyme ACACB. The network analysis reveals a functional connectivity between ADRB2 and the metabolic enzymes. Rather than a direct physical interaction, the network implies a signaling cascade likely mediated by Protein Kinase A (PKA/PRKACA). Specifically, activation of ADRB2 is known to stimulate PKA, which can phosphorylate and inhibit ACACA/ACACB. This suggests that targeting ADRB2 could modulate the fatty acid metabolism pathway via this phosphorylation axis. However, the other drug targets included in the analysis did not show direct interactions with the identified rate-limiting enzymes. The network diagram in Fig. 5 illustrates these interactions, highlighting the potential significance of the ADRB2-ACACB axis for further investigation, particularly in its capacity to link existing bronchodilatory therapy with metabolic regulation.

The impact of lifestyle on metabolite regulation

The lifestyle factor analysis provided further validation for our colocalization-based categorization of the six metabolites (Fig. 6). For the two metabolites confirmed as causal, we identified relevant metabolic and lifestyle links: cereal intake influenced Carnitine C14 levels, while both dried fruit intake and ‘never smoking’ status were associated with 3-hydroxyoleoylcarnitine levels. Crucially, for the four metabolites excluded due to confounding, our analysis linked them directly to their suspected confounding sources. Associations were found for 3-methylxanthine (a caffeine/theophylline metabolite), Paraxanthine (a caffeine/nicotine metabolite), 6-hydroxyindole sulfate, and Indolin-2-one (e.g., with water intake or cheese intake). These findings support their classification as ‘external ingestion’ metabolites whose initial MR signals likely reflected confounding by dietary or smoking behaviors.

Fig. 6
Fig. 6
Full size image

Heatmap illustrating the impact of 17 lifestyle factors on COPD-related metabolites identified through Mendelian randomization. Significant associations are marked with an asterisk (*), indicating the influence of specific lifestyle factors on metabolite levels. Notable findings include associations between cheese intake and 3-methylxanthine levels, cereal intake and Carnitine C14 levels, and dried fruit intake and 3-hydroxyoleoylcarnitine levels. Other significant associations include fruit smoothie intake with 6-hydroxyindole sulfate levels, oily fish intake with Indolin-2-one levels, water intake with Paraxanthine levels, and never smoking with 3-hydroxyoleoylcarnitine levels. These findings provide supportive evidence for the colocalization analysis, helping to distinguish the two causal metabolites (Carnitine C14, 3-hydroxyoleoylcarnitine) from the four metabolites whose associations likely reflect lifestyle confounding (e.g., Paraxanthine, 3-methylxanthine).

Validation of metabolic gene signatures in multi-center cohorts

To verify the clinical relevance of our bioinformatics findings, we integrated three GEO datasets (GSE10006, GSE19407, GSE20257). As shown in Fig. 7A-B, the “ComBat” algorithm successfully removed batch effects, creating a unified dataset suitable for comparative analysis. In the merged cohort, differential expression analysis revealed distinct alterations in the fatty acid metabolism axis (Fig. 7C). Specifically, the mRNA levels of the rate-limiting enzymes ACACA and ACACB were significantly upregulated in COPD tissues compared to controls (Fig. 7D, P < 0.05), consistent with the predicted accumulation of fatty acid metabolites. Conversely, the expression of ADRB2 (the target of Salbutamol) was significantly downregulated in the COPD group (P < 0.01). These transcriptomic signatures in human tissues strongly corroborate our MR and colocalization findings, suggesting a transcriptional reprogramming of lipid metabolism in COPD.

Fig. 7
Fig. 7
Full size image

Validation of key gene expression in external COPD cohorts. (A) Principal Component Analysis (PCA) plot of three GEO datasets (GSE10006, GSE19407, GSE20257) before batch correction, showing distinct separation by study. (B) PCA plot after batch correction using the ComBat algorithm, demonstrating successful integration of the datasets. (C) Volcano plot of the merged dataset illustrating differentially expressed genes between COPD and normal lung tissues. (D) Boxplots comparing the expression levels of ACACA, ACACB and ADRB2 in normal vs. COPD tissues. Statistical significance was determined by t-test (* P < 0.05, ** P < 0.01, *** P < 0.001).

Salbutamol restores fatty acid metabolic homeostasis in vitro

We further investigated the mechanistic link between the bronchodilator Salbutamol and the identified metabolic pathway using an in vitro COPD model. In BEAS-2B cells, CSE stimulation significantly upregulated the mRNA expression of ACACA and ACACB while suppressing ADRB2 (Fig. 8A), replicating the expression pattern observed in human tissues. Mechanistically, ACACA activity is inhibited by phosphorylation. Our Western blot analysis demonstrated that CSE exposure markedly reduced the phosphorylation of ACACA (p-ACACA) relative to total ACACA (t-ACACA), indicating aberrant activation of fatty acid synthesis (Fig. 8B-C). Crucially, treatment with Salbutamol reversed this effect, significantly restoring p-ACACA levels and increasing the p-ACACA/t-ACACA ratio (P < 0.001). These findings in both BEAS-2B and 16HBE cell lines provide experimental evidence that Salbutamol—traditionally acting via ADRB2—can modulate lipid metabolism by promoting the inhibitory phosphorylation of ACACA, thereby potentially counteracting the lipid dysregulation in COPD.

Fig. 8
Fig. 8
Full size image

Salbutamol modulates the ACACA signaling axis in CSE-induced airway epithelial cells. (A) Relative mRNA expression of ACACA, ACACB, ADRB2 in BEAS-2B cells under Control, CSE (Cigarette Smoke Extract), and CSE + Salbutamol (Sal) conditions, determined by qRT-PCR. (B) Representative Western blot images showing the protein levels of phosphorylated ACACA (p-ACACA), total ACACA (t-ACACA), and GAPDH in BEAS-2B and 16HBE cells across the treatment groups. (C) Quantitative analysis of the p-ACACA/t-ACACA ratio from Western blot replicates. Data are presented as mean ± SD. (* P < 0.05, ** P < 0.01, *** P < 0.001; ns, not significant).

Discussion

This comprehensive study provides novel insights into the complex interplay between metabolites, genetic factors, and lifestyle in the context of COPD. By employing a multi-faceted approach that integrates Mendelian randomization (MR), co-localization analysis, and protein-protein interaction (PPI) network analysis, we have identified several key findings that enhance our understanding of COPD pathogenesis and suggest potential avenues for therapeutic intervention.

Our initial MR analysis identified 59 metabolites associated with COPD, of which six metabolites remained statistically significant after false discovery rate (FDR) correction. Crucially, the subsequent stringent conditional colocalization analysis (PP4 > 0.8) refined this list to just two key metabolites: Carnitine C14 and 3-hydroxyoleoylcarnitine. This methodological filter increases the certainty of our causal claims. The exclusion of the other four metabolites is particularly informative, as their high PP3 values suggest that their initial MR association was likely confounded by distinct genetic signals in the region, supporting the necessity of applying stringent PP4 criteria38,39. The association with Carnitine C14, in particular, may reflect impaired long-chain fatty acid oxidation in COPD patients, which could contribute to energy imbalance and cellular stress in lung tissues40.

We initially observed negative associations for four metabolites: Paraxanthine, Indolin-2-one, and 6-hydroxyindole sulfate. Paraxanthine, a known metabolite of caffeine, has demonstrated anti-inflammatory properties in previous studies41. Its inverse relationship with COPD risk in our analysis suggests a potential protective effect, warranting further investigation. Similarly, the negative associations with Indolin-2-one and 6-hydroxyindole sulfate, both involved in tryptophan metabolism, hint at possible alterations in this pathway in COPD patients42. However, the rigorous conditional colocalization analysis failed to support a shared causal variant (PPH4 < 0.06) for any of these four metabolites, indicating that their MR associations were likely driven by distinct genetic signals (high PPH3 > 0.92). This finding is critical, especially since these are classified as external ingestion metabolites. Paraxanthine, in particular, is a major metabolite of nicotine, and 3-methylxanthine (another excluded metabolite) is derived from caffeine, both strongly reflecting smoking and dietary behaviors that are potent confounders for COPD. Thus, these initial MR signals were likely artifacts of pleiotropic confounding by lifestyle factors, and we exclude them from our final causal conclusions concerning COPD pathogenesis.

The co-localization analysis provided robust evidence for a shared genetic architecture between the identified metabolites and COPD across multiple genomic windows. This genetic overlap strengthens the case for a causal relationship between these metabolites and COPD susceptibility. Additionally, the Steiger directional test confirmed the absence of reverse causality, further validating the reliability of our findings18. Our pathway analysis revealed that both Carnitine C14 and 3-hydroxyoleoylcarnitine are part of the lipid super-pathway, specifically within fatty acid metabolism. The identification of ELOVL6, ACACA, and ACACB as rate-limiting enzymes in these pathways highlights potential therapeutic targets. ACACA and ACACB catalyze the committed and rate-limiting step of de novo fatty acid synthesis, while ELOVL6 governs the specific elongation of long-chain fatty acids, thereby acting as key flux controllers of this metabolic fate. These enzymes play critical roles in fatty acid elongation and synthesis, suggesting that modulating these processes could impact COPD risk or disease progression43,44. Our findings suggest a ‘metabolic traffic jam’ mechanism in COPD. Biologically, ACACA catalyzes the formation of malonyl-CoA, a substrate for fatty acid synthesis. The observed upregulation of ACACA suggests a state of enhanced lipogenesis, which likely contributes to the lipid accumulation and metabolic dysregulation in COPD tissues. The elevation of plasma Carnitine C14 observed in our MR analysis likely reflects the accumulation and leakage of intermediate acylcarnitines resulting from this impaired metabolic flux. While this pathway focuses on synthesis and elongation, we acknowledge that core fatty-acid β-oxidation and transport enzymes, such as ACADVL, also play indispensable roles in acylcarnitine dynamics. Our focus remains on the enzymes supported by the genetic colocalization signal, but future functional studies should explore the interplay between ACACA/ACACB and these transport regulators. Focusing on rate-limiting enzymes is particularly valuable from a drug development perspective. By targeting these key regulatory points within metabolic pathways, significant alterations in metabolite levels—and consequently, disease risk or progression—may be achieved. This approach aligns with the growing interest in metabolic reprogramming as a therapeutic strategy across various diseases, including respiratory disorders45.

The identification of rate-limiting enzymes offers tangible targets for intervention. Our PPI network analysis initially constructed a high-confidence interaction bridge between ACACA/ACACB and ADRB2, the primary clinical target of the bronchodilator Salbutamol. While Salbutamol is classically prescribed for smooth muscle relaxation, our in vitro validation (Fig. 8) reveals a novel metabolic dimension to its efficacy. We observed that CSE exposure significantly reduced the inhibitory phosphorylation of ACACA in bronchial epithelial cells, effectively ‘releasing the brake’ on fatty acid synthesis and mimicking the metabolic dysregulation observed in COPD46. Strikingly, Salbutamol treatment successfully restored ACACA phosphorylation levels and rescued the expression of ADRB2. This provides the first experimental evidence that β2-adrenergic stimulation can modulate lipid metabolism via downstream signaling—likely through the PKA-AMPK axis to promote ACACA phosphorylation—thereby inhibiting the overactive fatty acid synthesis characteristic of COPD pathology47,48. This finding suggests that the clinical benefits of Salbutamol may stem partially from restoring metabolic homeostasis, beyond its established bronchodilatory effects. Furthermore, identifying this interaction provides a rationale for investigating additional drugs that modulate fatty acid metabolism as potential therapeutics for COPD. Although the other drug targets analyzed did not exhibit direct interactions with the identified rate-limiting enzymes, this does not diminish their significance. Instead, it underscores the complexity of COPD pathophysiology and the necessity of multi-targeted therapeutic strategies49. The network analysis offers a foundation for future studies to explore indirect interactions and downstream effects of existing COPD medications on metabolic pathways.

Our analysis of lifestyle factors (Sect. “The impact of lifestyle on metabolite regulation”, Fig. 6) provides a final layer of validation for our findings. The associations found for the four excluded metabolites—such as Paraxanthine (a nicotine/caffeine metabolite) and 3-methylxanthine (a caffeine metabolite)—reinforce their nature as ‘external ingestion’ metabolites50. Their association with lifestyle factors supports the conclusion that their initial MR signals were artifacts of pleiotropic confounding by dietary and smoking behaviors, justifying their exclusion. In stark contrast, the associations identified for the two causal metabolites point towards genuine metabolic and lifestyle links51. The link between ‘never smoking’ and 3-hydroxyoleoylcarnitine levels reinforces the established importance of smoking in COPD pathogenesis and its interplay with fatty acid pathways. Similarly, the link between cereal intake and Carnitine C14 suggests a potential dietary-metabolic axis influencing this pathway. This analysis underscores the critical importance of using stringent colocalization to separate true, targetable causal pathways from associations merely confounded by lifestyle52.

The findings of this study have several important clinical implications. First, the identification of specific metabolites associated with COPD risk highlights potential biomarkers for early disease detection and prognosis. These metabolites could be integrated into screening panels to identify individuals at higher risk of developing COPD or to monitor disease progression in diagnosed patients53. Second, the elucidation of key metabolic pathways and rate-limiting enzymes involved in COPD pathogenesis offers new targets for drug development. The focus on fatty acid metabolism, in particular, suggests that therapies targeting lipid homeostasis could be promising approaches for COPD treatment54. Future research should investigate the efficacy of compounds that modulate the activity of ELOVL6, ACACA, and ACACB in preclinical models of COPD. Third, the observed high-confidence interactions between metabolic pathways and existing drug targets (specifically the ADRB2-ACACB axis) provide a rationale for exploring combination therapies and drug repurposing strategies. Targeting both bronchodilation and metabolic pathways simultaneously could enhance therapeutic efficacy in COPD treatment55. Finally, the insights gained into the relationship between lifestyle factors and COPD-related metabolites offer opportunities for evidence-based lifestyle interventions. These findings could inform dietary guidelines and lifestyle recommendations for COPD prevention and management, potentially improving patient outcomes through non-pharmacological approaches56.

To further verify the robustness of these findings across different cohorts, we validated the expression of key pathway genes in three independent external datasets from the GEO database (Fig. 7). The transcriptomic analysis of human lung tissues confirmed that ACACA and ACACB are significantly upregulated in COPD patients, while ADRB2 is downregulated, mirroring the metabolic signature predicted by our MR analysis and observed in our cell models. This cross-validation between genetic causality (MR), transcriptomic signatures (GEO), and functional assays (Western Blot) strongly supports the biological validity of the identified fatty acid metabolism axis. However, we acknowledge that our genetic instruments were primarily derived from European-ancestry populations. While the biological consistency across our multi-omics validation is reassuring, future studies incorporating multi-ancestry GWAS are essential to determine whether these metabolic causal links and therapeutic targets are generalizable to non-European populations.

However, these findings must be interpreted within the context of their limitations, which in turn define the priorities for future research. First, a major limitation is that both metabolite GWAS (CLSA) and COPD outcome data (FinnGen R11) were exclusively European-ancestry, precluding generalization to non-European populations. Genetic variant frequencies, linkage disequilibrium patterns, and environmental modifiers (e.g., smoking topography, biomass exposure, caffeine intake) differ substantially across ancestries, potentially altering the identified causal relationships.

Second, a key methodological limitation arises from our initial MR screening. All six metabolites that passed FDR correction were instrumented by only a single SNP. This restricted the MR analysis to the Wald Ratio method and precluded the use of standard, multi-instrument sensitivity analyses (e.g., MR-Egger regression) to assess horizontal pleiotropy. The Wald Ratio results, in isolation, are therefore susceptible to bias from potential pleiotropic effects of the single instrument. To address this specific vulnerability, we did not rely on the MR results alone; instead, we implemented stringent colocalization analysis (SuSiE, PP4 > 0.8) as the decisive validation step. This was essential for distinguishing true causal relationships from confounding, and successfully filtered out four metabolites (e.g., Paraxanthine, 3-methylxanthine) that showed high confounding signals (PP3 > 0.92). Third, while we robustly validated the expression levels of key enzymes at both mRNA and protein levels, we did not directly measure fatty acid oxidation rates or real-time metabolic flux (e.g., using Seahorse assays or isotope tracing). Therefore, the functional consequences of the observed enzyme alterations are inferred based on established biochemical pathways.

Future research should focus on validating these findings in larger, diverse populations and exploring the functional consequences of altered metabolite levels in COPD57. Longitudinal studies examining how metabolite profiles change over the course of COPD progression could provide valuable insights into disease mechanisms and identify stage-specific therapeutic targets58. Additionally, interventional studies targeting the identified metabolic pathways and evaluating their impact on COPD outcomes are warranted.

Conclusions

In conclusion, this study represents a significant step forward in our understanding of the metabolic underpinnings of COPD. By integrating metabolomic, genetic, and lifestyle data, we have uncovered novel insights into COPD pathogenesis and identified potential therapeutic targets and lifestyle interventions. These findings lay the groundwork for developing more personalized and effective approaches to COPD prevention and treatment, ultimately aiming to improve outcomes for the millions of individuals affected by this devastating disease worldwide.