Introduction

Chronic obstructive pulmonary disease (COPD) is a leading cause of chronic morbidity and premature death.1 Airflow limitation is only part of the picture. COPD also carries systemic inflammation, oxidative stress, and disturbed innate and adaptive immunity.2 Periodontitis is a chronic inflammatory disease of the tooth-supporting tissues, and it shares several of these features, including persistent host-microbial interaction and tissue-destructive immune activation.3,4 Epidemiological surveys repeatedly report that the two conditions occur together.5,6 What such surveys cannot settle is direction. Smoking, age, socioeconomic position, oral-health behaviour and differences in treatment act on both diseases at once, so observational estimates remain open to confounding.

Work on the oral-lung axis has mostly run in one direction, treating the mouth as a source of respiratory injury through aspiration of oral pathogens or spillover of inflammatory mediators.7,8 The reverse route is biologically plausible but far less studied. COPD may alter haematopoiesis, circulating leukocyte states and inflammatory set points, and these changes could in turn weaken periodontal tissue homeostasis and blunt the host response to chronic microbial challenge.9 Under that reading COPD would not simply coexist with periodontitis; COPD-associated systemic immune priming might raise susceptibility to periodontal inflammation and to defective resolution.10,11 We treat this as a hypothesis to be tested rather than an established sequence.

Two complementary tools can be brought to bear on that hypothesis. Mendelian randomization (MR) uses inherited genetic variants as instrumental variables, which reduces confounding and reverse causation and makes it suited to questions of direction.12 Single-cell transcriptomics works at a different level. It identifies the cell populations that carry a cross-disease signal and helps separate systemic immune input states from local tissue consequences.13–15 Each approach has a well-recognised blind spot: MR is silent about cells, while descriptive single-cell atlases are silent about direction of effect. Studies of the COPD-periodontitis relationship have so far applied the two separately. The novelty of the present work lies in coupling them within a single analytical chain - genetic evidence on direction, pathway and effector-gene prioritization, single-cell localization of the implicated lineage, and a targeted in vitro test of the prioritized genes.16–20 We use this chain to ask whether COPD liability increases periodontitis risk once smoking is accounted for, and to identify the immune program most consistently implicated in that relationship.

MethodsStudy Design

The analysis proceeded in stages (Figure 1). Two-sample MR first tested whether genetic liability to COPD is associated with the risk of chronic periodontitis. Variants underlying the causal estimate were then mapped to genes for pathway enrichment analysis. SMR with HEIDI testing followed,21 using blood cis-eQTLs to prioritize candidate effector genes. Single-cell transcriptomic datasets from COPD and from periodontitis were then analysed to identify the immune lineage carrying the cross-disease signal. Within this framework the COPD peripheral-blood dataset was read as the systemic input state and the periodontitis gingival dataset as the local tissue outcome state. The two datasets come from different cohorts and different tissues, so this pairing is an analytical construct and not a directly observed trajectory. Finally, macrophages were exposed to Porphyromonas gingivalis LPS to test whether the prioritized genes are suppressed during a periodontal inflammatory challenge. Instrument selection, two-sample harmonization, sensitivity analyses and the reporting of causal estimates followed established MR practice, with attention to instrument relevance, exchangeability, exclusion restriction and cautious causal interpretation.

Figure 1 Schematic workflow of the integrative multi-omic study design. Stage 1, two-sample Mendelian randomization (MR) testing the effect of chronic obstructive pulmonary disease (COPD) liability on chronic periodontitis (CP), with multivariable MR used to adjust for smoking-related traits. Stage 2, pathway enrichment analysis of the variants underlying the causal estimate and of disease-specific variants prioritized by summary-data-based MR for COPD and CP separately. Stage 3, eQTL-based prioritization of candidate protective genes within the implicated pathways. Stage 4, reanalysis of public single-cell RNA sequencing datasets and in vitro RT-qPCR validation in RAW264.7 macrophages. Solid arrows indicate the analytical sequence and dashed arrows the assumptions of the MR framework (association of instruments with the exposure, and absence of an independent path from instruments to the outcome or through confounders). The lower panel is a schematic summary of the proposed, hypothesis-generating pathway and does not depict experimentally demonstrated causal steps.

Genetic Data Sources

Summary-level COPD association statistics were obtained from the IEU OpenGWAS/GWAS Catalog dataset ebi-a-GCST90018807; the harmonized SNP file contained 468,475 participants. Chronic periodontitis summary statistics were obtained from the FinnGen endpoint finn-b-K11_PERIODON_CHRON. For multivariable MR we used smoking initiation (IEU OpenGWAS ieu-b-4877), a GWAS of ever versus never regular smoking derived from the GSCAN smoking-behaviour resource. Smoking initiation was chosen as the primary smoking-adjustment phenotype because it indexes genetic liability to taking up smoking. It therefore speaks directly to the concern that COPD instruments may tag smoking-related liability rather than COPD-specific biology. Past tobacco smoking was added as a second, sensitivity phenotype. cis-eQTL instruments for SMR were obtained from the eQTLGen blood resource. All genetic resources were derived from predominantly European-ancestry populations.

Three single-cell datasets were reanalyzed. GSE262668 provided gingival tissue from patients with periodontitis (PD1, PD2, PD4) and from periodontally healthy controls (HC1, HC3, HC4). GSE174609 provided peripheral blood mononuclear cells (PBMCs) from patients with periodontitis and from controls. GSE249584 provided peripheral blood single-cell data from five patients with COPD (COPD1-COPD5) and five controls (HC1-HC5). All three datasets are deposited in the Gene Expression Omnibus as droplet-based single-cell RNA sequencing (scRNA-seq) libraries. GSE249584 was therefore used to define COPD-associated immune cell states, and chromatin accessibility was not examined at any point in this study.

Mendelian Randomization and Enrichment Analyses

Genetic instruments for COPD were selected from the exposure GWAS and harmonized with the periodontitis GWAS. The primary causal estimate was obtained with the inverse-variance weighted (IVW) method. Because COPD instruments might capture smoking-related genetic liability, we performed MVMR with COPD liability and smoking initiation (IEU OpenGWAS ID: ieu-b-4877) entered simultaneously as exposures and chronic periodontitis as the outcome. Smoking initiation served as the primary smoking covariate because it represents genetic liability to active smoking uptake and sits upstream of smoking-related COPD. A second MVMR model substituted past tobacco smoking as a sensitivity analysis. MVMR estimates were obtained using IVW, MR-Egger, LASSO and median-based approaches. Cochran’s Q statistics assessed heterogeneity. Conditional F-statistics were computed for each exposure in every MVMR model to evaluate instrument strength (Figure 2). Following standard practice we treated a conditional F-statistic below 10 as indicating weak conditional instruments; estimates from such models are reported as smoking-adjusted associations and not as smoking-independent causal effects.

Figure 2 Mendelian randomization analysis and genetic architecture of the COPD-periodontitis link. (A) Forest plot of the single-nucleotide polymorphisms (SNPs) used as instrumental variables; the two red estimates at the foot of the panel are the pooled MR-Egger and inverse-variance weighted (IVW) effects. (B) Funnel plot used to assess heterogeneity across instruments. (C) Scatter plot of SNP effects on the exposure (COPD) against SNP effects on the outcome (chronic periodontitis); the coloured lines are the fitted slopes for each MR method. (D) Summary of the MR estimates (IVW fixed and multiplicative random effects, MR-Egger, weighted median, simple mode, weighted mode) with odds ratios and 95% confidence intervals. (E and F) Genetic architecture of chronic periodontitis: Manhattan plot of gene-level associations (E) and dot plot of the corresponding Gene Ontology enrichment (F). In (E), the red ellipses and the red gene labels mark the two prioritized candidate protective genes, ALOX15B and HSP90AA1. (G and H) Genetic architecture of COPD: Manhattan plot (G) and Gene Ontology dot plot (H). In (G), the blue ellipses and the blue gene labels mark the three loci showing the strongest COPD associations (FAM13A, HLA-DQB1 and HYKK). (I) Gene-concept network of the pathways enriched among the variants linking COPD to periodontitis. (J) Gene Ontology enrichment of the same gene set across the three ontologies (BP, biological process; CC, cellular component; MF, molecular function), showing convergence on MHC class II antigen presentation and arachidonic acid metabolism. The circled red and blue regions in (E) and (G) are visual highlights of the genes named in this legend; they carry no statistical meaning and no additional threshold is implied.

Effector-Gene Prioritization

To prioritize candidate effector genes within implicated pathways, we performed eQTL-based SMR using genetically predicted blood gene expression as the exposure and chronic periodontitis as the outcome. HEIDI testing was applied to reduce the likelihood that observed associations were explained by linkage rather than a shared causal variant.

Single-Cell Transcriptomics

Single-cell data were processed in R with Seurat. For every dataset, cells were filtered on the number of detected genes, total UMI counts and the percentage of mitochondrial transcripts; the retained matrices were log-normalized, scaled on highly variable genes, reduced by principal component analysis and then clustered on a shared-nearest-neighbour graph with UMAP used for visualization. Clusters were annotated with canonical lineage markers, and cluster-defining genes were obtained with FindAllMarkers (Figures S1–S3 for the periodontitis datasets and Figure S4–S6 for the COPD dataset). Gingival tissue (GSE262668) and periodontitis PBMCs (GSE174609) were used to define the cell type-specific expression of HSP90AA1 and ALOX15B. Intercellular communication was inferred with CellChat and compared between disease and control samples. Pathway activity scores for arachidonic acid metabolism and for glycolysis/gluconeogenesis were computed per cell and compared across cell types (Figure 3). The COPD blood dataset (GSE249584) was processed through the same pipeline and used to characterize the systemic myeloid state. No chromatin-accessibility (scATAC-seq) data were generated or analysed at any stage of this study.

Figure 3 Altered intercellular communication and metabolic state of the macrophage lineage in COPD blood and in periodontitis gingiva. (A, C and E) were generated from the COPD peripheral-blood dataset (GSE249584); (B, D and F) from the periodontitis gingival dataset (GSE262668). (A and B) Circle plots of the differential number (left) and the differential strength (right) of inferred cell-cell interactions between disease and control; red edges indicate interactions increased in disease, blue edges those decreased in disease, and edge width is proportional to the size of the difference. (C and D) Number of inferred interactions shown separately for healthy controls (HC) and for disease (COPD or periodontitis, PD). (E and F) Heatmaps of overall signalling strength by pathway and cell type in HC and in disease; colour denotes relative strength, and the marginal bar plots give the summed strength per cell type (top) and per pathway (right). (G-J) Single-cell pathway activity in gingival tissue: arachidonic acid metabolism (G, I) and glycolysis/gluconeogenesis (H, J) in healthy control (G and H) and periodontitis (I and J) samples, each shown as a UMAP feature plot (upper) and a violin plot by cell type (lower). (K–N) The same two pathways in the periodontitis PBMC dataset (GSE174609): arachidonic acid metabolism (K, M) and glycolysis/gluconeogenesis (L, N) in controls (K and L) and in periodontitis (M and N). Colour scales are independent for each panel and are not comparable between panels.

In vitro Transcriptional Validation

RAW264.7 murine macrophages were obtained from the American Type Culture Collection (ATCC, Manassas, VA, USA; catalogue number TIB-71) and were maintained in Dulbecco’s modified Eagle medium supplemented with 10% fetal bovine serum at 37 °C under 5% CO2. Cells were stimulated with Porphyromonas gingivalis LPS at 1 μg/mL for 24 h to model a periodontal microbial inflammatory challenge. Relative mRNA expression of the prioritized genes was quantified by RT-qPCR in four independent biological replicates per condition. The experiment was designed to test whether candidate protective genes are vulnerable to local pathogen-associated stimulation in macrophages; it was not designed to reproduce the full COPD to periodontitis trajectory.

Statistical Analysis

MR analyses relied primarily on the IVW method to estimate causal effects, with MR-Egger, weighted median and weighted mode used as sensitivity analyses for horizontal pleiotropy and heterogeneity. Instrument strength was summarized with F-statistics for univariable MR and with conditional F-statistics for MVMR. For the in vitro experiments, data are presented as mean ± standard deviation (SD), and differences between control and LPS-stimulated groups were assessed with Student’s t-test. All tests were two-sided and P < 0.05 was considered statistically significant.

ResultsCOPD Genetic Liability Is Associated with an Increased Risk of Chronic Periodontitis

Two-sample MR showed that genetic liability to COPD was associated with a higher risk of chronic periodontitis (Figure 2A–D). The primary IVW estimate (OR = 1.282, 95% CI 1.097–1.499, P = 0.002) was consistent across sensitivity analyses (weighted median OR = 1.362, P = 0.003; weighted mode OR = 1.406, P = 0.004). MR-Egger gave no evidence of directional horizontal pleiotropy (P = 0.369) (Tables S1–S3).

Smoking is a major risk factor for both diseases, so we used MVMR to ask whether the association survived adjustment for smoking-related traits. The primary model adjusted for smoking initiation (IEU OpenGWAS ieu-b-4877), chosen to represent genetic liability to active smoking uptake. In this model COPD remained associated with chronic periodontitis by MVMR-IVW (OR = 1.183, 95% CI 1.054–1.328, P = 0.0043), MVMR-Egger (OR = 1.255, 95% CI 1.090–1.444, P = 0.0016), MVMR-LASSO (OR = 1.183, 95% CI 1.054–1.328, P = 0.0043) and MVMR-median (OR = 1.255, 95% CI 1.059–1.488, P = 0.0088). Smoking initiation itself was not independently associated with chronic periodontitis in the same model. Cochran’s Q test gave no evidence of heterogeneity (Q = 195.0775, P = 0.9758). The conditional F-statistic was 5.02 for COPD and 17.09 for smoking initiation. Conditional instrument strength was therefore adequate for smoking initiation but weak for COPD, which is an important caveat: with a conditional F-statistic below 10 the COPD estimate in this model may be biased towards the confounded observational association. We accordingly read this result as smoking-initiation-adjusted evidence and not as proof of a smoking-independent COPD effect (Table S4).

A supplementary MVMR model including COPD liability and past tobacco smoking yielded consistent results. COPD remained associated with chronic periodontitis using MVMR-IVW (OR = 1.235, 95% CI 1.101–1.387, P = 3.34×10−4), MVMR-Egger (OR = 1.377, 95% CI 1.056–1.796, P = 0.018), MVMR-LASSO (OR = 1.235, 95% CI 1.101–1.387, P = 3.34×10−4), and MVMR-median (OR = 1.340, 95% CI 1.133–1.584, P = 6.26×10−4). Past tobacco smoking was not independently associated with chronic periodontitis in the same model (MVMR-IVW OR = 1.015, 95% CI 0.921–1.119, P = 0.767). Cochran’s Q test did not indicate heterogeneity (Q = 41.00, P = 0.785). The conditional F-statistics were 24.17 for COPD and 5.32 for past tobacco smoking, indicating robust conditional instrument strength for COPD but limited instrument strength for the past-smoking covariate (Table S5).

Cross-Disease Pathway Analysis Highlighted Antigen Presentation and Arachidonic-Acid Metabolism

To define the biology behind this association we first characterized the genetic architecture of each disease separately (Figure 2E–H) and then focused on the variants linking COPD to periodontitis (Figure 2I and J). The cross-disease enrichment profile was not confined to broad inflammatory categories. The most coherent signals converged instead on major histocompatibility complex class II antigen presentation and on arachidonic acid metabolism. HLA-DQA1 supported the antigen-presentation component, whereas CYP-family annotations supported lipid-mediator biology. One reading of this pattern is that COPD-associated periodontitis susceptibility involves specific immune-regulatory and lipid-resolution pathways rather than non-specific systemic inflammation alone. Enrichment analysis is descriptive, so we present that reading as a hypothesis rather than a demonstrated mechanism (Tables S6–S9).

SMR Prioritized HSP90AA1 and ALOX15B as Candidate Protective Genes

SMR identified HSP90AA1 and ALOX15B as the leading candidate genes. Higher genetically predicted expression of HSP90AA1 was associated with lower periodontitis risk (β = −0.305, SE 0.118, P = 0.0099; HEIDI P = 0.557), as was higher genetically predicted expression of ALOX15B (β = −0.433, SE 0.163, P = 0.0078; HEIDI P = 0.354). HSP90AA1 is linked to antigen-processing machinery and ALOX15B to lipid oxidation and pro-resolving mediator synthesis, which places both genes at the interface between inflammatory activation and inflammatory resolution. Because SMR tests association between genetically predicted expression and disease risk, these genes are prioritized as candidates and not established as functional effectors (Table 1; Table S10).

Table 1 Summary-Data-Based Mendelian Randomization (SMR) Analysis Integrating Blood Cis-eQTLs with Chronic Periodontitis, and Validation in the Periodontitis Single-Cell RNA Sequencing (scRNA-Seq) Macrophage Dataset

Single-Cell Profiling Implicates the Macrophage Lineage and an Altered Metabolic State

The monocyte-macrophage lineage was the compartment most consistently linked to the genetic findings. In the COPD blood dataset, enrichment within this lineage highlighted myeloid differentiation, ATP metabolic process, cytokine regulation and antigen presentation (Figure S4–S6). Comparative CellChat analysis of the two single-cell datasets is summarized in Figure 3A–F; (A, C and E) describe the COPD blood dataset and (B, D and F) the periodontitis gingival dataset. In COPD, relative gains were seen in MHC-I, MHC-II, MIF, PECAM1, GALECTIN, SELPLG, PARs, IL16, ITGB2 and ICAM signalling. Ligand-receptor analysis showed stronger TNF-TNFRSF1A/B, MIF-CD74/CXCR4 or CD44, ITGB2-ICAM1/2, IFNG-IFNGR and HLA class II-CD4 interactions, a pattern compatible with heightened leukocyte adhesion and antigen-presentation circuits.

Gingival single-cell analysis identified macrophages as a central inflammatory hub in periodontitis as well. Relative to healthy tissue, disease-associated programs were enriched for positive regulation of cytokine production, myeloid differentiation, leukocyte activation, ATP metabolic process and response to lipopolysaccharide (Figures S1–S3). Communication analysis showed increased HLA-A/B/C/E-to-CD8 T-cell interactions, CLEC2C/D-KLRB1 signalling, ITGB2-CD226 signalling, PTPRC-CD22 signalling and CCL5-CCR1 axes, while stromal-like compartments showed reduced collagen-SDC1/CD44/integrin communication. Taken together these changes are consistent with a loss of tissue-supporting communication alongside stronger immune-activating cross-talk.

The metabolic analysis pointed in the same direction. Disease macrophages showed higher glycolytic activity and a relative loss of oxidative phosphorylation-linked features in both the gingival and the PBMC datasets (Figure 3G–N). In the periodontitis macrophage differential-expression output, HSP90AA1 was significantly downregulated, with an average log2 fold change of −1.806 and an adjusted P value of 2.09×10−12 (Table S11). ALOX15B did not emerge as a strong differential transcript in that table, and feature plots showed detectable ALOX15B transcripts in only a small minority of cells; HSP90AA1, by contrast, was broadly expressed and enriched in the macrophage/monocyte compartment of both tissues (Figure 4A–D). The evidence for ALOX15B therefore rests mainly on the genetic analyses rather than on its single-cell expression profile.

Figure 4 Candidate protective genes HSP90AA1 and ALOX15B in the macrophage lineage and their response to a periodontal inflammatory stimulus. (A) UMAP embedding of the gingival single-cell dataset (GSE262668), annotated by cell type. (B) UMAP embedding of the periodontitis PBMC dataset (GSE174609), annotated by cell type. (C) Feature plots of ALOX15B (upper) and HSP90AA1 (lower) in the gingival dataset. (D) The corresponding feature plots in the PBMC dataset. In (C) and (D), red denotes cells with detectable transcripts and grey denotes cells without detectable transcripts; HSP90AA1 was broadly expressed and enriched in the macrophage/monocyte compartment, whereas ALOX15B was detected in only a small minority of cells in either dataset. (E) Forest plot of the eQTL-based MR estimates: genetically predicted expression of ALOX15B (odds ratio [OR] = 0.773, 95% confidence interval [CI] 0.646–0.924) and of HSP90AA1 (OR = 0.817, 95% CI 0.717–0.932) was associated with a lower risk of periodontitis (OR < 1). Red squares are point estimates, horizontal bars are 95% confidence intervals, and the dashed vertical line marks OR = 1. (F) RT-qPCR of ALOX15B and HSP90AA1 in murine RAW264.7 macrophages after stimulation with Porphyromonas gingivalis lipopolysaccharide (LPS, red) relative to unstimulated controls (CON, black). Bars show the mean, error bars the standard deviation, and each open circle one independent biological replicate (n = 4 per group). Asterisks denote the significance of the difference between CON and LPS: *P < 0.05; **P < 0.01.

Inflammatory Stimulation Suppressed Both Candidate Genes in Macrophages

eQTL-based MR agreed with the SMR results: genetically predicted expression of ALOX15B (OR = 0.773, 95% CI 0.646–0.924) and of HSP90AA1 (OR = 0.817, 95% CI 0.717–0.932) was associated with a lower risk of periodontitis (Figure 4E). We then asked whether the same genes respond to a periodontal inflammatory stimulus. RAW264.7 macrophages were exposed to P. gingivalis LPS, and transcript levels of both genes fell relative to unstimulated controls (Figure 4F). The assay therefore shows that a periodontal pathogen-associated stimulus can suppress both candidates in macrophages. It does not, by itself, show that either gene mediates the COPD-periodontitis relationship. Read together, the MR, SMR, single-cell and in vitro results are compatible with a model in which COPD liability predisposes to periodontitis by weakening a macrophage program needed for coordinated antigen presentation and inflammatory resolution. Testing that model directly will require experiments that combine a COPD-like systemic exposure with a periodontal challenge.

Discussion

This study provides integrative evidence for a smoking-adjusted association between COPD liability and periodontitis risk, and it points to macrophage reprogramming as the most coherent cellular candidate linking the systemic and periodontal immune compartments. The univariable MR estimate indicated a positive COPD-periodontitis association. That association persisted in MVMR after adjustment for smoking initiation, a phenotype chosen to capture genetic liability to active smoking uptake, and it was directionally consistent after adjustment for past tobacco smoking. These analyses speak directly to the concern that COPD-associated variants merely tag smoking-related liability. They do not settle it. The conditional F-statistic for COPD in the smoking-initiation model fell below the conventional threshold of 10, so weak-instrument bias towards the observational association cannot be excluded, and residual confounding by smoking intensity, duration or nicotine dependence remains possible.22 We therefore describe the finding as a smoking-adjusted association rather than as evidence of a fully smoking-independent causal effect.

How does this sit alongside earlier work? Observational studies have repeatedly reported that periodontitis is more common, and more severe, in people with COPD, and a small randomised trial reported that periodontal treatment reduced exacerbation frequency.3,5,6 Most of that literature has been read in the oral-to-pulmonary direction, with aspiration and systemic spillover of oral pathogens as the proposed mechanism.7,8,20 Our estimates are compatible with those reports but reverse the emphasis, and they align with mechanistic work showing that chronic inflammatory disease can train myelopoiesis and leave a durable imprint on circulating myeloid cells.9 Our design cannot establish that the relationship runs in one direction only. A bidirectional model, in which each disease raises liability to the other, and a shared-susceptibility model, in which common genetic or environmental factors drive both, both remain plausible and would require reciprocal Mendelian randomization in adequately powered periodontitis datasets to test.

The systemic-to-local transition is central to how this work is interpreted, and it deserves careful wording. In COPD peripheral blood, monocyte-macrophage populations showed enhanced antigen-presentation, adhesion, interferon-related and cytokine-associated signalling - features of a primed circulating myeloid compartment rather than a homeostatic one.22–29 In gingival tissue affected by periodontitis, macrophages displayed a more advanced inflammatory phenotype with glycolytic bias, amplified immune communication and reduced stromal-supportive interactions. A parsimonious reading is that COPD-associated systemic myeloid priming raises the likelihood that recruited monocyte-derived macrophages enter the periodontal lesion in a state favouring inflammatory persistence. We want to be explicit that this is an inference drawn across two independent cohorts and two different tissues, not a trajectory observed in matched individuals. Alternative explanations fit the same data. Shared genetic susceptibility acting independently in lung and gingiva, convergent responses to a common environmental exposure, or reprogramming of tissue-resident gingival macrophages could each produce the pattern we describe. Trajectory inference, paired sampling of blood and gingiva from the same patients, or fate-mapping in animal models would be needed to distinguish these possibilities.

HSP90AA1 and ALOX15B fit this framework, though their status should not be overstated. Higher genetically predicted expression of both genes was associated with lower periodontitis risk (Figure 4E), and HSP90AA1 localized to the monocyte-macrophage lineage (Figure 4A–D). HSP90AA1 is plausibly connected to proteostasis and to antigen-processing competence under inflammatory stress, while ALOX15B belongs to lipid-mediator pathways involved in inflammatory termination and resolution.30 Suppression of both transcripts after P. gingivalis LPS stimulation shows that a periodontal microbial challenge can weaken candidate protective nodes in macrophages (Figure 4F). That experiment models only the local hit. It includes no COPD-like systemic exposure such as cigarette-smoke extract, COPD patient serum or inflammatory conditioning of monocytes before LPS challenge, and it was performed in a murine cell line rather than in human macrophages. The perturbation assay therefore supports the periodontal relevance of the two genes without establishing the full causal sequence from COPD to periodontitis. On the present evidence HSP90AA1 and ALOX15B are best described as candidate protective mediators rather than validated mechanistic effectors, and loss-of-function or overexpression studies in macrophages would be the natural next step.31–33

Read at the level of biology rather than statistics, the data favour a more specific account of the COPD-periodontitis relationship than diffuse inflammation. Altered antigen handling, increased leukocyte adhesion and communication, metabolic stress and impaired resolution recur across the genetic and single-cell layers. In periodontal tissue this macrophage-centred process runs alongside reduced matrix-supportive signalling from stromal compartments, which suggests that immune dysregulation and tissue-homeostatic failure advance together. If it holds up, this framework may help explain why some patients with COPD prove more susceptible to destructive periodontal inflammation even after shared behavioural risk factors are taken into account. The translational reading is correspondingly modest: patients with COPD may warrant closer periodontal surveillance, and macrophage resolution programs represent a target worth exploring, but neither claim follows from these data without prospective clinical confirmation.

Several limitations should be acknowledged. First, although MVMR adjusted for smoking initiation and for past tobacco smoking, residual confounding by other smoking-related phenotypes - lifetime smoking exposure, cigarettes per day, pack-years, nicotine dependence and passive smoking - cannot be excluded, and the conditional F-statistic for COPD in the smoking-initiation model was below 10, which calls for cautious causal interpretation. Second, the outcome GWAS was the FinnGen chronic periodontitis endpoint, which is less clinically granular than full-mouth examination-based phenotypes and is therefore vulnerable to outcome misclassification. Third, the genetic resources were drawn mainly from European-ancestry datasets, so generalization to other populations is not warranted. Fourth, the single-cell analyses used separate cohorts and tissues, sample numbers were small, and the proposed systemic-to-local macrophage trajectory is consequently inferential rather than directly observed. Fifth, the RAW264.7 LPS assay models a periodontal microbial challenge but not the COPD exposure state; a dual-hit design using cigarette-smoke extract, COPD serum or hypoxia-oxidative stress preconditioning, ideally in human macrophages, would be required to recapitulate the proposed transition more directly.

Conclusion

This study supports a smoking-adjusted genetic association in which COPD liability is associated with increased susceptibility to periodontitis. Two observations are directly supported by the data. Univariable MR indicated a positive COPD-periodontitis association, and that association persisted in MVMR after adjustment for smoking initiation and for past tobacco smoking, albeit with weak conditional instrument strength in one model. Integrative genetic analyses implicated major histocompatibility complex class II antigen presentation and arachidonic-acid metabolism, and single-cell analyses localized the convergent signal predominantly to the monocyte-macrophage compartment. Beyond that, the interpretation is hypothesis-generating. The proposed systemic-to-local macrophage transition, and the roles assigned to HSP90AA1 and ALOX15B as candidate protective mediators, are inferred from data collected in separate cohorts and tissues and are not established here. We therefore present these findings as a foundation for the mechanistic and translational studies that would be needed to test the proposed pathway.

Declaration of Generative AI and AI-Assisted Technologies in the Writing Process

During the preparation of this work the authors used Claude (Opus 5) and ChatGPT (GPT-5.5) solely to improve the readability and language of the manuscript. After using these tools, the authors reviewed and edited the content as needed and take full responsibility for the content of the publication. No generative artificial intelligence tool was used to generate, analyse or interpret the study data, to draw any scientific conclusion, or to produce any figure or table.

Abbreviation

COPD, chronic obstructive pulmonary disease; MR, Mendelian randomization; GWAS, genome-wide association study; SMR, summary-based Mendelian randomization; SNP, single nucleotide polymorphism; IVW, inverse-variance weighted; LPS, lipopolysaccharide; PBMC, peripheral blood mononuclear cell; MVMR, multivariable Mendelian randomization; eQTL, expression quantitative trait locus; HEIDI, heterogeneity in dependent instruments; OR, odds ratio; CI, confidence interval; scRNA-seq, single-cell RNA sequencing; scATAC-seq, single-cell assay for transposase-accessible chromatin sequencing; GEO, Gene Expression Omnibus; HC, healthy control; PD, periodontitis; MHC, major histocompatibility complex; UMAP, uniform manifold approximation and projection; RT-qPCR, reverse transcription quantitative polymerase chain reaction; ATCC, American Type Culture Collection; SD, standard deviation.

Data Sharing Statement

All datasets analysed in this study are publicly available. COPD summary statistics (ebi-a-GCST90018807), chronic periodontitis summary statistics (finn-b-K11_PERIODON_CHRON) and smoking initiation summary statistics (ieu-b-4877) were obtained from IEU OpenGWAS (https://gwas.mrcieu.ac.uk/). cis-eQTL data were obtained from eQTLGen (https://www.eqtlgen.org/). Single-cell RNA sequencing data are available from the Gene Expression Omnibus under accession numbers GSE262668, GSE174609 and GSE249584. Additional data supporting the findings of this study are available from the corresponding author upon reasonable request.

Ethics Statement

This study analysed only publicly available, de-identified data. No new human participants were recruited, no new biospecimens or personal information were collected, no intervention was performed, and no identifiable individual-level information was accessed. The study protocol, data-source description and supporting materials were reviewed by the Human Research Ethics Committee of the Second Affiliated Hospital of Zhejiang University School of Medicine (Hangzhou, China), which determined on 11 August 2026 that the work is exempt from institutional ethics review and from the requirement for informed consent, and that no separate ethics approval number is issued for exempt research of this kind. The determination was made under Article 32 of the Measures for Ethical Review of Life Science and Medical Research Involving Humans (Guo Wei Ke Jiao Fa [2023] No. 4, issued 18 February 2023, China), items 1 and 2 of which permit exemption for research using lawfully obtained public data and/or anonymized information. A copy of the written determination has been provided to the Editorial Office. Data use was limited to the scope permitted by each database or provider and remained consistent with the applicable original informed consent and ethics approval; ethical approval and informed consent were obtained by the investigators of each original study that generated the genome-wide association and Gene Expression Omnibus datasets reanalysed here. RAW264.7 is a commercially available established cell line purchased from the American Type Culture Collection (ATCC, Manassas, VA, USA; catalogue number TIB-71), and its use required no additional ethical or institutional review board approval.

Acknowledgments

We thank the investigators, participants and consortia who generated and shared the public GWAS and Gene Expression Omnibus datasets analysed in this work, including FinnGen, the GSCAN consortium, eQTLGen, and the contributors to GSE262668, GSE174609 and GSE249584.

Author Contributions

Zelong Hu, Rui Mu and Shijia Huang contributed equally to this work. All authors made a significant contribution to the work reported, whether that is in the conception, study design, execution, acquisition of data, analysis and interpretation, or in all these areas; took part in drafting, revising or critically reviewing the article; gave final approval of the version to be published; have agreed on the journal to which the article has been submitted; and agree to be accountable for all aspects of the work.

Funding

National Natural Science Foundation of China (General Program): Grant/Award Number 82071094.

Disclosure

The authors report no conflicts of interest in this work.

References

1. Li W, Bai J, Ge Y, Fan Y, Huang Q, Deng Z. Association between compound dietary antioxidant index and all-cause and cancer mortality in patients with chronic obstructive pulmonary disease: results from NHANES. Front Med.2025;12;1544841. 1999-2018. doi:10.3389/fmed.2025.1544841

2. Moutsopoulos NM, Kling HM, Angelov N, et al. Porphyromonas gingivalis promotes Th17 inducing pathways in chronic periodontitis. J Autoimmun. 2012;39(4):294–14. doi:10.1016/j.jaut.2012.03.003

3. Zhou X, Han J, Liu Z, Song Y, Wang Z, Sun Z. Effects of periodontal treatment on lung function and exacerbation frequency in patients with chronic obstructive pulmonary disease and chronic periodontitis: a 2-year pilot randomized controlled trial. J Clin Periodontol. 2014;41(6):564–572. doi:10.1111/jcpe.12247

4. Usher AK, Stockley RA. The link between chronic periodontitis and COPD: a common role for the neutrophil? BMC Med. 2013;11(13):241. doi:10.1186/1741-7015-11-241

5. Takeuchi K, Matsumoto K, Furuta M, et al. Periodontitis is associated with chronic obstructive pulmonary disease. J Dent Res. 2019;98(5):534–540. doi:10.1177/0022034519833630

6. Saito M, Shimazaki Y, Yoshii S, Takeyama H. Periodontitis and the incidence of chronic obstructive pulmonary disease: a longitudinal study of an adult Japanese cohort. J Clin Periodontol. 2023;50(6):717–726. doi:10.1111/jcpe.13801

7. Hurst JR, Perera WR, Wilkinson TM, Donaldson GC, Wedzicha JA. Systemic and upper and lower airway inflammation at exacerbation of chronic obstructive pulmonary disease. Am J Respir Crit Care Med. 2006;173(1):71–78. doi:10.1164/rccm.200505-704OC

8. Andreassen H, Vestbo J. Chronic obstructive pulmonary disease as a systemic disease: an epidemiological perspective. Eur Respir J Suppl. 2003;46:2s–4s. doi:10.1183/09031936.03.00000203

9. Li X, Wang H, Yu X, et al. Maladaptive innate immune training of myelopoiesis links inflammatory comorbidities. Cell May. 2022;185(10):1709–1727.e1718. doi:10.1016/j.cell.2022.03.043

10. Zhang Y, Shen Z, Yang J, et al. Single-cell spatial atlas of smoking-induced changes in human gingival tissues. Int J Oral Sci. 2025;17(1):60. doi:10.1038/s41368-025-00385-5

11. Labaki WW, Rosenberg SR. Chronic obstructive pulmonary disease. Ann Intern Med. 2020;173(3):Itc17–itc32. doi:10.7326/aitc202008040

12. Sanderson E, Glymour MM, Holmes MV, et al. Mendelian randomization. Nat Rev Meth Primers. 2022;2. doi:10.1038/s43586-021-00092-5

13. Robinson JH, Delvig AA. Diversity in MHC class II antigen presentation. Immunology. 2002;105(3):252–262. doi:10.1046/j.0019-2805.2001.01358.x

14. van den Hoorn T, Paul P, Jongsma ML, Neefjes J. Routes to manipulate MHC class II antigen presentation. Curr Opin Immunol. 2011;23(1):88–95. doi:10.1016/j.coi.2010.11.002

15. Vander Lugt B, Khan AA, Hackney JA, et al. Transcriptional programming of dendritic cells for enhanced MHC class II antigen presentation. Nat Immunol. 2014;15(2):161–167. doi:10.1038/ni.2795

16. Anwar-mohamed A, Zordoky BN, Aboutabl ME, El-Kadi AO. Alteration of cardiac cytochrome P450-mediated arachidonic acid metabolism in response to lipopolysaccharide-induced acute systemic inflammation. Pharmacol Res. 2010;61(5):410–418. doi:10.1016/j.phrs.2009.12.015

17. Li G, Gu L, Zhao F, et al. WFDC12-overexpressing contributes to the development of atopic dermatitis via accelerating ALOX12/15 metabolism and PAF accumulation. Cell Death Dis. 2023;14(3):185. doi:10.1038/s41419-023-05686-3

18. Ménégaut L, Laubriet A, Crespy V, et al. Inflammation and oxidative stress markers in type 2 diabetes patients with Advanced Carotid atherosclerosis. Cardiovasc Diabetol. 2023;22(1):248. doi:10.1186/s12933-023-01979-1

19. Yang X, Chen H, Shen W, et al. FGF21 modulates immunometabolic homeostasis via the ALOX15/15-HETE axis in early liver graft injury. Nat Commun. 2024;15(1):8578. doi:10.1038/s41467-024-52379-2

20. Gomes-Filho IS, da Cruz SS, Trindade SC, et al. Important evidence of the oral-lung axis, especially during the coronavirus pandemic. Oral Dis. 2022;28 Suppl 2:2636–2638. doi:10.1111/odi.14346

21. Zhu Z, Zhang F, Hu H, et al. Integration of summary data from GWAS and eQTL studies predicts complex trait gene targets. Nat Genet. 2016;48(5):481–487. doi:10.1038/ng.3538

22. Fan Z, Chen R, Xie X, et al. AIM2-driven inflammation in periodontitis: mechanisms and systemic implications. J Inflamm Res. 2025;18:6983–6997. doi:10.2147/jir.S505907

23. Mengistu DT, Curtis JL, Freeman CM. A model of dysregulated crosstalk between dendritic, natural killer, and regulatory T cells in chronic obstructive pulmonary disease. Trends Immunol. 2024;45(6):428–441. doi:10.1016/j.it.2024.04.010

24. Booth S, Hsieh A, Mostaco-Guidolin L, et al. A single-cell atlas of small airway disease in chronic obstructive pulmonary disease: a cross-sectional study. Am J Respir Crit Care Med. 2023;208(4):472–486. doi:10.1164/rccm.202303-0534OC

25. Kammerl IE, Hardy S, Flexeder C, et al. Activation of immune cell proteasomes in peripheral blood of smokers and COPD patients: implications for therapy. Eur Respir J. 2022;59(3):2101798. doi:10.1183/13993003.01798-2021

26. Ricci F, Wollmer P, Engström G, Fedorowski A, Hamrefors V. Dysregulation of the “inflammatory reflex” with abnormal neurohumoral activation may contribute to proinflammatory activity driving the progression of COPD. Eur Respir J. 2018;51(6):1800806. doi:10.1183/13993003.00806-2018

27. Miklós Z, Horváth I. The role of oxidative stress and antioxidants in cardiovascular comorbidities in COPD. Antioxidants. 2023;12(6). doi:10.3390/antiox12061196

28. Shih YM, Chang YJ, Cooke MS, et al. Alkylating and oxidative stresses in smoking and non-smoking patients with COPD: implications for lung carcinogenesis. Free Radic Biol Med. 2021;20(164):99–106. doi:10.1016/j.freeradbiomed.2020.12.442

29. Nataf S, Guillen M, Pays L. Common neurodegeneration-associated proteins are physiologically expressed by human b lymphocytes and are interconnected via the inflammation/autophagy-related proteins TRAF6 and SQSTM1. Front Immunol. 2019;10(2704). doi:10.3389/fimmu.2019.02704

30. Zhao T, Bernstein KE, Fang J, Shen XZ. Angiotensin-converting enzyme affects the presentation of MHC class II antigens. Lab Invest. 2017;97(7):764–771. doi:10.1038/labinvest.2017.32

31. Saare M, Tserel L, Haljasmägi L, et al. Monocytes present age-related changes in phospholipid concentration and decreased energy metabolism. Aging Cell. 2020;19(4):e13127. doi:10.1111/acel.13127

32. Suraneni MV, Schneider-Broussard R, Moore JR, et al. Transgenic expression of 15-lipoxygenase 2 (15-LOX2) in mouse prostate leads to hyperplasia and cell senescence. Oncogene. 2010;29(30):4261–4275. doi:10.1038/onc.2010.197

33. Kusmartsev S. Enhanced 15-lipoxygenase activity and elevated eicosanoid production in kidney tumor microenvironment contribute to the inflammation and immune suppression. Oncoimmunology. 2012;1(2):249–251. doi:10.4161/onci.1.2.18502