The CLEC7A Y238∗ variant and susceptibility to pneumonia: a prospective population-based cohort study and mechanistic investigation

Summary

Background

DECTIN-1, a C-type lectin receptor encoded by CLEC7A, mediates antifungal immunity, but its antibacterial role remains unclear. We investigated, using human genetic and animal studies, whether the common loss-of-function CLEC7A Y238∗ variant (rs16910526) influences susceptibility to bacterial respiratory infection, and sought to identify the bacterial ligand mediating DECTIN-1 recognition.

Methods

We analysed Y238∗ and pneumonia risk in the Malmö Diet and Cancer Study, a prospective Swedish cohort, using Cox models with age as timescale (28,126 participants, 5493 events; 26,601 and 5190 fully adjusted) over a mean 17·3 years of follow-up. We challenged wild-type, Clec7a^+/−, and Clec7a^−/− mice with Streptococcus pneumoniae and non-typeable Haemophilus influenzae (NTHi). Affinity proteomics, ELISA, flow cytometry, and NF-κB reporter assays identified and validated the ligand.

Findings

Y238∗ carriers had elevated pneumonia risk in adjusted Cox models (HR 1·12, 95% CI 1·04–1·20; p = 0·002; concordant competing-risk sHR 1·12, 1·04–1·20). Clec7a-deficient mice showed gene-dose-dependent impairment of S. pneumoniae clearance, impaired NTHi clearance, and blunted cytokine responses. Affinity pull-down identified elongation factor Tu (EF-Tu) as a DECTIN-1 ligand that bound the carbohydrate-recognition domain and activated NF-κB signalling when clustered; orthologues from four bacterial species also bound.

Interpretation

These findings support a role for DECTIN-1 in pulmonary antibacterial defence and suggest that inherited deficiency increases pneumonia susceptibility in humans. EF-Tu is a candidate ligand that provides mechanistic support for the association.

Funding

Swedish Research Council, Knut and Alice Wallenberg Foundation, Swedish Heart Lung Foundation, Swedish Government Funds for Clinical Research (ALF), Skåne County Council's Research and Development Foundation.

Research in context.

Evidence before this study

We searched PubMed from database inception to Apr 1, 2026, without language restrictions, using the terms (“CLEC7A” OR “Dectin-1”) AND (“Y238∗” OR “rs16910526” OR “bacteria” OR “pneumonia” OR “bacterial infection”). The search returned 157 records. Multiple studies documented interactions between DECTIN-1 and mycobacteria. Two reports linked the CLEC7A Y238∗ variant to clinical outcomes in patients at high risk of fungal infection: invasive fungal disease in patients with acute myeloid leukaemia receiving induction chemotherapy, and chronic lung allograft dysfunction and death in lung transplant recipients, where Y238∗ carriage was associated with increased risk of fungal pathogens. No study had examined the association between CLEC7A variation and bacterial pneumonia risk in the general population. Separately, although bacterial elongation factor Tu (EF-Tu) has documented surface localisation and moonlighting functions in host–pathogen interactions, it had not been identified as a ligand for DECTIN-1.

Added value of this study

We provide population-level genetic, experimental, and biochemical evidence linking DECTIN-1 to pulmonary antibacterial defence. In a prospective cohort of more than 28,000 individuals, the CLEC7A Y238∗ variant was associated with a 12–13% increase in pneumonia risk, an effect observed for respiratory infection and robust across competing-risk and stratified analyses. Case-only analyses indicated that this reflected susceptibility rather than disease severity. In complementary murine pneumonia models, DECTIN-1 deficiency caused gene-dose-dependent impairment of pulmonary bacterial clearance after challenge with Streptococcus pneumoniae, and impaired clearance of non-typeable Haemophilus influenzae. We also identified EF-Tu as a conserved bacterial ligand for DECTIN-1 that binds the carbohydrate-recognition domain and activates NF-κB signalling.

Implications of all the available evidence

These findings support a broader role for DECTIN-1 in host defence against bacterial as well as fungal respiratory infection. The identification of EF-Tu as a conserved bacterial ligand extends the known ligand repertoire of DECTIN-1 beyond β-glucans and offers a molecular mechanism for the contribution of DECTIN-1 to pulmonary antibacterial immunity. A common loss-of-function variant carried by approximately one in seven Europeans is associated with susceptibility to pneumonia independent of established clinical risk factors, identifying a previously unrecognised host genetic determinant of bacterial respiratory infection in the general population.

Introduction

Lower respiratory tract infections remain a leading cause of infectious disease mortality worldwide, with pneumonia accounting for over two million deaths annually.¹ The principal bacterial pathogens driving community-acquired pneumonia, Streptococcus pneumoniae and non-typeable Haemophilus influenzae (NTHi), cause a substantial proportion of severe cases requiring hospitalisation.² Susceptibility to these infections depends on pathogen virulence, prior exposure, vaccination status, and host immune response. Genetic variation in innate immune genes contributes to individual risk,^3,4 yet the contribution of specific pattern-recognition receptor (PRR) variants to pneumonia susceptibility in the general population remains poorly defined.

DECTIN-1, a C-type lectin receptor encoded by CLEC7A, is expressed on myeloid cells and pulmonary epithelium.⁵ Best characterised for its antifungal role, DECTIN-1 recognises β-glucans in fungal cell walls and drives Syk–CARD9-dependent cytokine production.⁶ A common nonsense variant, Y238∗ (rs16910526), introduces a premature stop codon that renders the receptor non-functional. Its allele frequency is 7·5% globally, ranging from 9·5% in south Asians and 8·0% in non-Finnish Europeans to 2·0% in African/African-American and 0·02% in east Asian populations (gnomAD v4·1).⁷ Y238∗ was first characterised in a family with recurrent mucocutaneous fungal infections in which homozygotes showed clinical immunodeficiency.⁸ Monocytes from heterozygous carriers secrete less cytokine, and neutrophils produce less reactive oxygen species.^8,9 Surface receptor levels are not reduced: the premature stop removes the epitope recognised by the standard flow-cytometry antibody, and western blotting with a stalk-region antibody shows normal DECTIN-1 protein.¹⁰ Y238∗ and other loss-of-function variants in the DECTIN-1 pathway have also been associated with severe disseminated coccidioidomycosis, with impaired β-glucan-induced DUOX1-dependent oxidative responses in carriers.¹¹

Beyond its antifungal role, DECTIN-1 has been implicated in the recognition of bacterial pathogens. Early work showed that the receptor cooperates with TLR2 to drive macrophage cytokine responses to several mycobacterial species,¹² and that DECTIN-1 directly binds live Mycobacterium tuberculosis and contributes to dendritic-cell IL-12 production.¹³ More recently, mycobacterial α-glucans were identified as a direct DECTIN-1 ligand, extending receptor engagement beyond β-glucan.¹⁴ DECTIN-1 also mediates proinflammatory responses to non-typeable H. influenzae.^5,15 In non-fungal contexts, the CLEC7A Y238∗ variant has been examined in inflammatory bowel disease and rheumatoid arthritis without finding an association.^16,17

Whether Y238∗ carriage influences bacterial respiratory infection in the general population has not been tested, and no bacterial protein ligand for DECTIN-1 has been identified.

Here, we show that the CLEC7A Y238∗ variant is associated with increased pneumonia risk in a prospective cohort of more than 28,000 individuals and validate these findings in murine pneumonia models that demonstrate gene-dose-dependent impairment of host defence against S. pneumoniae, and impaired host defence against NTHi. We further identify bacterial elongation factor Tu (EF-Tu), a highly conserved surface-exposed protein, as a DECTIN-1 ligand that binds the carbohydrate-recognition domain and activates downstream NF-κB signalling.

Methods

Study design

This study combined human population genetics, murine pneumonia models, and biochemical characterisation to determine how reduced DECTIN-1 function affects pneumonia susceptibility and to identify bacterial EF-Tu as a DECTIN-1 ligand. We first analysed the CLEC7A Y238∗ variant in a prospective population-based cohort, then validated clinical findings in knockout mouse models of S. pneumoniae and NTHi pneumonia, and finally identified and characterised EF-Tu as a DECTIN-1 ligand using affinity proteomics, binding assays, and cell-based reporter systems. In vivo, age-matched male and female C57BL/6-background mice (Clec7a^+/+, Clec7a^+/−, Clec7a^−/−) were inoculated intranasally with S. pneumoniae (strains EF3030 or D39) or NTHi, with PBS-inoculated animals as controls. Prespecified endpoints were 24 h (lung neutrophil infiltration by Ly6G immunohistochemistry with automated QuPath quantification) and 48 h (bacterial burden in lung, bronchoalveolar lavage fluid (BALF), and nasopharynx; BALF cytokines), with predefined humane endpoints. One animal was found dead before the experimental endpoint and was excluded from all analyses. Two further animals were excluded from the bronchoalveolar lavage flow-cytometry analysis only, because counting-bead recovery fell outside the accepted window. No other animals were excluded. Exact sample sizes and replicates are reported in figure legends.

Human cohort and genotyping

We analysed the Y238∗ variant (rs16910526) in the Malmö Diet and Cancer Study (MDCS), a prospective population-based cohort of 29 296 genotyped participants examined between 1991 and 1996, with baseline recruitment beginning on Mar 4, 1991.18, 19, 20 Participants provided questionnaire data, underwent baseline measurements, and gave blood samples. Eligible participants were residents of Malmö, men born between 1923 and 1945 and women born between 1923 and 1950, who attended the baseline examination and gave written informed consent. The present analysis is a secondary genetic analysis of that cohort: participants with genotype data and successful register linkage were eligible, and individuals lacking genotype or covariate data were excluded only from the analyses requiring them. Base Cox models used 28,126 complete cases, and fully adjusted models used 26,601 participants with complete genotype and covariate data (Figure S1, Appendix p 2). Participants were followed through linkage to the Swedish National Patient Register and Cause of Death Register until emigration, death, or Dec 31, 2023. Pneumonia and bacterial pneumonia diagnoses were identified using ICD-9 and ICD-10 codes (Table S1, Appendix p 21). Participant sex was derived from the Swedish personal identity number recorded at inclusion rather than self-reported, and gender identity was not collected. This variable therefore cannot distinguish sex from gender, nor identify participants whose gender identity differs from their registered sex. Race and ethnicity were not recorded by the MDCS and are therefore not reported. DNA was extracted from stored blood samples collected at baseline, and genome-wide genotyping of the cohort was performed using the Illumina GWAS chip (GSA v1 array), as previously described.²¹

Genotyping was performed by the Regeneron Genetics Center according to the manufacturer's standard protocols. Autosomal variants were retained if they had a minor allele frequency above 1%, missingness below 1%, and no departure from Hardy–Weinberg equilibrium at p < 1 × 10⁻¹⁵; individuals with more than 10% missing genotype data, and duplicated or blacklisted samples, were excluded. Genotypes were phased with Eagle v2·3 and imputed against the HRC r1·1 (2016) European reference panel in three batches. Y238∗ (rs16910526) was analysed from this imputed dataset, in which no calls were missing, and its genotype distribution did not deviate from Hardy–Weinberg equilibrium (χ² = 0·16, 1 df, p = 0·69). Of the 29 296 genotyped participants, 28,175 reached age 65 during follow-up and were eligible for age-as-timescale analyses. Genotype distribution in this cohort was AA (wild type) n = 24,214 (85·9%), AC (heterozygous) n = 3816 (13·5%), CC (homozygous) n = 145 (0·5%). For bacterial pneumonia, the refined fully adjusted cohort comprised 23,790 participants after exclusion of individuals with unspecified pneumonia diagnoses from the control group.

Bacteria and culture conditions

Non-typeable H. influenzae (NTHi) 3655 wild type and mutants were routinely cultured in brain heart infusion (BHI) liquid broth supplemented with NAD and haemin (both at 2 μg/mL) or on chocolate agar plates at 37 °C in a humid atmosphere containing 5% CO₂. S. pneumoniae EF3030 (serotype 19F)²² and D39 (serotype 2)²³ were grown statically in Todd Hewitt broth supplemented with 0·5% yeast extract (THY; Sigma Aldrich, St. Louis, MO) at 37 °C until OD₆₀₀ of 0·3–0·6. For in vivo challenge experiments, bacteria were concentrated by centrifugation and saved as frozen stocks at −80 °C in 15% glycerol.

Eukaryotic cell lines

The mucoepidermoid bronchial carcinoma cell line NCI–H292 (CRL-1848, RRID:CVCL_0455) and the type II alveolar epithelial cell line A549 (CCL-185, RRID:CVCL_0023) were from ATCC (Manassas, VA). Cells were grown at 37 °C, 5% CO₂ and cultured in RPMI-1640 with l-glutamine supplemented with 10% (v/v) heat-inactivated foetal bovine serum (FBS), 100 U/mL penicillin, and 100 μg/mL streptomycin.

A549 cells overexpressing DECTIN-1 were transfected as previously described.⁵ Briefly, cells were transfected with the plasmids pUNO1 (control vector; InvivoGen, San Diego, CA) or pUNO1-hDectin1a (bearing the human DECTIN-1 gene CLEC7A isoform a; InvivoGen). Transfectants were cultured in Dulbecco's modified Eagle's medium (DMEM; high glucose with Glutamax) supplemented with 10% (v/v) FBS and 30 μg/mL Blasticidin S (InvivoGen).

Secreted embryonic alkaline phosphatase (SEAP) reporter HEK 293 cells expressing the human DECTIN-1a gene (HEK-Blue hDectin-1a, RRID:CVCL_VV19) and the parental NF-κB-inducible SEAP cell line (HEK-Blue Null1-v, RRID:CVCL_A7YL), both derived from HEK 293 cells, were purchased from InvivoGen. Both cell lines were cultured in DMEM high glucose supplemented with 10% (v/v) FBS, 100 U/mL penicillin, 100 μg/mL streptomycin, 2 mM l-glutamine, 1 μg/mL puromycin, and either 1 × HEK-Blue CLR Selection (hDectin-1a) or 100 μg/mL Normocin (Null1-v) (InvivoGen). In all experiments, the level of DECTIN-1 surface expression was confirmed using flow cytometry. NCI–H292 and parental A549 cells were authenticated by ATCC and together with the A549-derived transfectants STR-authenticated from our lab stocks by Eurofins. The A549-derived transfectants were not STR-authenticated. The HEK-Blue reporter lines underwent vendor functional quality control rather than STR authentication. All cell lines were tested for mycoplasma at approximately two-month intervals throughout the study using MycoStrip (InvivoGen), a loop-mediated isothermal amplification assay targeting the 16S rRNA gene of Mollicutes, on antibiotic-free culture supernatant; all tests were negative.

Mouse infection models

For initial experiments with S. pneumoniae EF3030, six- to ten-week-old C57BL/6J wild-type (stock #000664, RRID:IMSR_JAX:000664) and Clec7a^−/− (B6129S6-Clec7a^tm1Gd, stock #012337, RRID:IMSR_JAX:012337) mice were purchased from Jackson Laboratories (Bar Harbor, ME). For subsequent experiments comparing all three genotypes, mice were intercrossed to generate Clec7a^+/+, Clec7a^+/−, and Clec7a^−/− littermates, ensuring a matched genetic background. Littermates of all three genotypes were infected together within each experiment, so allocation to infection group and timepoint followed litter and cage rather than a randomisation sequence. All animals were housed and bred at Lund University in filter-top cages with standard laboratory chow and water provided ad libitum, a 12 h light/dark cycle, under specific-pathogen-free conditions, and acclimatised for at least 7 days before experiments. Age-matched male and female mice were included in all experiments. DECTIN-1 genotype was validated by PCR.

For bacterial infection experiments, mice were anaesthetised by inhalation of isoflurane (Forene; Abbott, Wiesbaden, Germany) and inoculated intranasally with one of the following: 10⁷ CFU S. pneumoniae EF3030 in 40 μL PBS, 2·2 × 10⁷ CFU S. pneumoniae D39 in 40 μL PBS (from frozen stock), or 1·2 × 10⁸ CFU NTHi in 50 μL PBS (cultured for 3 h, resuspended in PBS). Mice were monitored twice daily for clinical signs of infection (piloerection, lethargy, hunched posture). Body temperature was assessed using infrared thermometry of the ventral abdomen. Humane endpoints included temperature below 30 °C, severe lethargy, or inability to reach food and water; mice meeting these criteria were euthanised immediately. Experimental endpoints were 24 h (immunohistochemistry) or 48 h (bacterial burden, cytokines) postinfection. Uninfected control animals received 40 or 50 μL sterile PBS. At 48 h postinoculation, animals were euthanised by CO₂ asphyxiation. For bronchoalveolar lavage (BAL), animals were placed in supine position and the trachea was incised. Bronchoalveolar lavage fluid (BALF) was collected by two washes of 200 μL ice-cold PBS into the respiratory tract. Cells in BALF were separated by centrifugation and immediately stained and analysed; the remaining fluid was plated for CFU counting or stored at −80 °C for cytokine analysis. Tissues were harvested as previously described.²⁴ Briefly, lung lobes were excised by opening the thoracic cavity. Maxillary bones and the skull between the eyes were cut and tissue in the nasal conchae was collected with forceps. Lungs and nasopharyngeal tissues were mechanically homogenised in 1 mL sterile PBS. Bacterial burden in tissue homogenates and BALF was determined by plating serial dilutions on blood agar (S. pneumoniae) or chocolate agar (NTHi) plates and counting colonies after overnight incubation.

For immunohistochemical analysis of neutrophil infiltration, wild-type, Clec7a^+/−, and Clec7a^−/− mice (n = 5 per genotype) were euthanised at 24 h postinfection with S. pneumoniae D39 (2·2 × 10⁷ CFU). Infection was confirmed by nasopharyngeal culture during tissue harvest. Immediately after euthanasia, lungs were instilled with 4% paraformaldehyde via the trachea, then fixed in 4% paraformaldehyde for 48 h. Fixed lungs were paraffin-embedded and processed for immunohistochemistry. Uninfected control mice received PBS and were processed identically.

For EF-Tu challenge experiments, mice received endotoxin-free recombinant EF-Tu or ovalbumin control intranasally, and BALF cytokines were measured at 24 h by cytometric bead array.

Immunohistochemistry

Paraffin-embedded lung sections were cut at 4 μm thickness using a Microm microtome (Cellab). Heat-induced epitope retrieval was performed using Envision Flex Target Retrieval Solution, high pH (K8010, Dako/Agilent) in a PT-Link module (Dako/Agilent). Immunohistochemical staining was performed using the EnVision Flex kit (K8010, Dako/Agilent) in an Autostainer Link 48 system (Dako/Agilent) according to the manufacturer's instructions. The primary antibody used was rat anti-mouse Ly6G (clone RB6-8C5, catalogue LS-C112469, LifeSpan BioSciences, Seattle, WA; RRID:AB_470492, dilution 1:250). The secondary antibody was rabbit anti-rat IgG (catalogue BA-4001, Vector Laboratories, Burlingame, CA; RRID:AB_10015300, 1:200). Sections were counterstained with haematoxylin. Stained sections were imaged at 20 × magnification using an Aperio ScanScope CS scanner (Leica Biosystems, Wetzlar, Germany). Analysis of brightfield (haematoxylin–3,3′-diaminobenzidine [DAB]) images was performed using QuPath version 0·6·0. Lung tissue sections were manually annotated to exclude airways and large blood vessels, and Ly6G-positive cells were detected using the “Positive cell detection” tool. Haematoxylin optical density (OD) at threshold 0·1 was used to detect all cells within the annotated region, and mean DAB OD at threshold 0·15 was used for Ly6G detection (Table S2, Appendix p 22). The percentage of Ly6G-positive cells was calculated as the ratio of DAB-positive cells to total detected cells. Fields shown in the figures were selected by a fixed rule: among windows lying within lung parenchyma, the one whose Ly6G⁺ proportion is closest to that section's own measured value, which places every field within 0·06 percentage points of its section.

Neutrophil quantification by flow cytometry

In a separate cohort, mice were challenged intranasally with 40 μL S. pneumoniae D39 as described above and harvested 24 h later. Circulating and vascular-marginated leucocytes were labelled in vivo by intravenous injection of 1 μg anti-CD45-PE (clone 30-F11; eBioscience 12-0451-82, RRID:AB_465668) in 200 μL buffer via the tail vein, 3 min before euthanasia. Cervical dislocation was not used, to avoid pulmonary haemorrhage.

Bronchoalveolar lavage was performed through a tracheal incision by instilling 200 μL cold FACS buffer (PBS with 1% BSA and 1 mM EDTA), aspirating, re-instilling the same volume and aspirating again. The thorax was then opened and the pulmonary circulation perfused through the right ventricle with 7–10 mL cold PBS containing 0·5 mM EDTA until the lungs blanched. Lavage cells were pelleted, treated with ACK lysis buffer for 1 min and resuspended in FACS buffer. Lungs were minced and digested in 5 mL RPMI-1640 containing 150 U/mL collagenase and 25 U/mL DNase IV for 60 min at 37 °C with rotation, passed through a 100 μm strainer, and treated with ACK lysis buffer where the pellet remained red.

Cells were stained with Zombie Aqua viability dye (BioLegend), blocked with anti-CD16/32, and stained with anti-CD45-FITC (clone 30-F11; BD Pharmingen 553079, RRID:AB_394609), anti-CD11b-PE-Cy7 (clone M1/70; eBioscience 25-0112-82, RRID:AB_469588), anti-Ly6G-APC (clone 1A8; BD Pharmingen 560599, RRID:AB_1727560) and anti-F4/80-PE-Cy5 (clone BM8; eBioscience 15-4801-81). Single-stain compensation controls were acquired for every fluorochrome, and a control in which the intravenous label was omitted confirmed that the vascular gate reflected the injected antibody rather than background staining. Samples were acquired on a BD FACSymphony A1 cytometer running FACSDiva 9·0·2.

The reported Ly6G⁺ populations were gated as singlets, then live cells, then CD45⁺ leucocytes, then Ly6G against the intravenous CD45 label. The full gating hierarchy, with event counts and percentages of each parent gate, is shown in Figure S2 (Appendix p 3); the extravascular population is the Ly6G⁺ fraction not carrying the intravenous CD45 label. CD11b and F4/80 were acquired but were not used to define the reported population. Perfusion does not clear the marginated pulmonary pool, which is why the intravenous label is required: a median 65·5% of lung Ly6G⁺ cells carried it in uninfected mice against 21·3% after infection (p = 0·011). In lavage fluid no intravascular gate is applied, because lavage fluid contains no vasculature, and the CD45 gate is not applied either, because alveolar-macrophage autofluorescence overlaps the CD45⁺ population; the lavage denominator is therefore live single cells rather than leucocytes.

Absolute numbers were determined against a counting-bead standard: 50 μL AccuCheck Counting Beads (Life Technologies PCB100, lot 2360140, nominal 1054 microspheres/μL) was added to each sample immediately before acquisition. Every fold change, ratio and rank statistic reported here is independent of the bead concentration, which enters only as a constant multiplier. Tubes were accepted when bead recovery fell within a window of 800 to 1300 beads. One animal was found dead before the endpoint and was excluded from all analyses, and two further animals were excluded from the lavage analysis alone, with bead recoveries of 199 and 382.

Cytokine quantification

Cytokine concentrations in mouse BALF samples were analysed using the Mouse Inflammation Cytometric Bead Assay (CBA) kit (Becton Dickinson). Concentrations were calculated using FlowJo software and R by extrapolating from a ten-point standard for each cytokine (TNF-α, IFN-γ, MCP-1, IL-10, IL-6, and IL-12p70). In infection experiments, mice inoculated with PBS were negative controls, whereas in experiments with administration of recombinant EF-Tu into the mouse lung, endotoxin-free ovalbumin (InvivoGen) was the negative control protein.

Recombinant protein expression and purification

Polyhistidine-tagged recombinant proteins produced in this study included different truncated fragments of human DECTIN-1a (UniProtKB accession number Q9BXN2): full-length extracellular domain (residue T66–M247), C-type lectin domain (CTLD; Y127–M247), and stalk domain (T66–I126); and orthologues of EF-Tu from non-typeable NTHi 3655 (CAH0450365), S. pneumoniae 19F (ATCC 49619; BBG39415), E. coli K12 (P0CE47), and L. monocytogenes CCUG 15526 (KFL19772). All three DECTIN-1a fragments were expressed in the same vector and carry an identical C-terminal FLAG and 6 × His tag in the same position, so tag identity and position are constant across the domain-mapping comparison.

The expression and purification of His-tagged recombinant proteins were performed as previously described.²⁵ Briefly, the open reading frames of EF-Tu orthologues were amplified using the Expand High Fidelity PCR System (Roche, Basel, Switzerland) from bacterial genomic DNA using gene-specific primers carrying restriction enzyme cutting sites (Table S3, Appendix p 23). Purified PCR products were digested with specific restriction enzymes and cloned into expression vector pET26b(+) (Novagen, Darmstadt, Germany). The truncated fragments of DECTIN-1a were inserted into pET26b as pKR-Dectin constructs during this study. The resulting constructs were transformed into ClearColi BL21(DE3) (Research Corporation Technologies, AZ) for protein expression with reduced endotoxin contamination. Recombinant protein expression in E. coli was induced with 1 mM isopropyl β-d-1-thiogalactopyranoside at OD₆₀₀ = 0·5 for 3 h at 37 °C. Expression cultures were harvested, sonicated, and centrifuged to retrieve cell lysates. His-tagged proteins were purified from the cell lysates using immobilised metal affinity chromatography (HisTrap FF Crude; Cytiva, Uppsala, Sweden) according to the manufacturer's instructions. Purified proteins were dialysed against PBS, and protein concentrations were estimated at absorbance 280 nm with NanoDrop (Thermo Scientific). Protein purity was analysed by SDS-PAGE and the integrity of His-tagged protein was verified with horseradish peroxidase (HRP)-conjugated anti-His polyclonal antibody (ab1187, RRID:AB_298652; Abcam, Cambridge, UK) by western blotting (Figure S3, Appendix p 5).

EF-Tu sequence conservation analysis

EF-Tu protein sequences from 18 respiratory bacterial pathogens were retrieved from UniProt and verified against the entry for each named strain; the species, strains and accession numbers are listed in Table S4 (Appendix p 24). Sequences were aligned with AlignSeqs in the DECIPHER R package (v3·6·0) under R 4·5·3, using progressive alignment with iterative refinement. Pairwise amino-acid identity is the number of matching residues divided by aligned positions excluding gaps, and hierarchical clustering used Ward's method (ward.D2) on one minus fractional identity. Per-position conservation was scored by Shannon entropy, normalised from 0 to 1, and mapped onto the E. coli EF-Tu structure (AlphaFold model AF-P0CE47-F1), rendered with 3Dmol.js.

Antibody preparation

Rabbit anti-EF-Tu polyclonal antibodies were generated as previously described.²⁵ Briefly, rabbits were immunised subcutaneously with 200 μg of recombinant EF-Tu emulsified in 0·5 mL saline and 0·5 mL incomplete Freund's adjuvant. Booster immunisations were administered three times at four-week intervals, using alum as an adjuvant. Blood was collected two weeks after the final immunisation. The resulting anti-EF-Tu polyclonal antibodies were affinity-purified using EF-Tu conjugated to CNBr-activated Sepharose (GE Healthcare Biosciences, Chicago, IL). The rabbit anti-Protein D polyclonal antibody used as a control in the adhesion and internalisation assays was raised and affinity-purified by the same methods, using recombinant Protein D as the immunogen and as the affinity ligand.

NiMag bead coating

MagneHis nickel beads (Promega, Madison, WI) were separated from storage buffer using a DynaMag-2 magnetic rack (Thermo Fisher, Uppsala, Sweden) and washed three times with 500 μL PBS. Purified protein was added at the desired protein-to-bead ratio. For the uncoated control, PBS was added instead of protein. Beads were incubated overnight at 4 °C on a rolling table. After incubation, the supernatant was removed, beads were washed as above, resuspended in PBS, and stored at 4 °C. Coating efficiency was assessed by SDS-PAGE (Figure S4, Appendix p 6).

EF-Tu identification by affinity purification and mass spectrometry

Recombinant full-length human DECTIN-1a coupled to MagneHis nickel beads was used as bait in affinity pull-down assays with NTHi 3655 whole-cell lysates. NTHi 3655 was cultured in BHI medium supplemented with NAD and haemin (2 μg/mL each) to mid-log phase (OD₆₀₀ ≈ 0·5), harvested, washed with PBS, and resuspended in lysis buffer (PBS, 10 mM imidazole) to an OD₆₀₀ of 5. Cells were lysed by sonication (Bioruptor; Diagenode, Denville, NJ) and the lysate clarified by centrifugation twice (16,000×g, 4 °C). MagneHis Ni-NTA magnetic beads (V8560; Promega) were coated with recombinant human DECTIN-1a overnight at 4 °C in coating buffer (PBS, 0·5 M NaCl, 10 mM imidazole) and incubated with 442 μg of bacterial lysate protein overnight at 4 °C; parallel controls comprised empty beads incubated with lysate and DECTIN-1-coated beads incubated without lysate. Beads were washed three times with wash buffer (PBS, 10 mM imidazole), bound proteins eluted with elution buffer (PBS, 500 mM imidazole), and eluates pooled for mass spectrometry. Three biological replicates were performed for each condition.

Samples were digested with trypsin using a modified filter-aided sample preparation protocol.²⁶ Eluates were denatured in 8 M urea, 100 mM Tris–HCl (pH 8·0), reduced with 15 mM DTT, loaded onto 10 kDa microfiltration filters (Microcon-10, Merck Millipore, Rahway, NJ), alkylated with 15 mM iodoacetamide in 100 mM Tris–HCl (pH 8·0), washed extensively with 100 mM ammonium bicarbonate, and digested with trypsin (Merck) at 37 °C for 16 h, with tryptic peptides recovered by micropurification. Peptides were analysed on an EASY-nLC 1200 system (Thermo Scientific) coupled to an Orbitrap Eclipse Tribrid mass spectrometer (Thermo Scientific), using a 2 cm trap column (100 μm i.d.) and a 15 cm analytical column (75 μm i.d.) packed in-house with ReproSil-Pur C18-AQ 3 μm resin (Dr. Maisch, Ammerbuch, Germany), eluted at 250 nL/min over a 35 min gradient from 6% to 44% phase B (0·1% formic acid, 80% acetonitrile) followed by a 30 s ramp to 100% B and 5 min at 100% B.

Proteins were identified and quantified in Proteome Discoverer 2·5 (Thermo Scientific) against UniProt proteomes of E. coli BL21(DE3) (v2024-01-24), H. influenzae 3655 (v2024-01-24), and Homo sapiens (v2023-06-28), with precursor mass tolerance 10 ppm, fragment mass tolerance 0·02 Da, static carbamidomethyl (+57·021 Da, C), dynamic oxidation (+15·995 Da, M), and trypsin allowing one missed cleavage. Peptides and proteins were filtered to a 1% false discovery rate by a target-decoy strategy with Benjamini-Hochberg correction; label-free quantification used unique peptides only, and proteins quantified by fewer than three unique peptides or specific to E. coli were excluded. Enrichment is the fold-change between DECTIN-1-coated beads and empty bead controls, with relative abundance expressed in parts per million, and a protein was considered significantly enriched at an adjusted p value below 0·05 with detection in all three biological replicates. All raw and processed proteomics files are deposited in PRIDE (PXD070889).

Enzyme-linked immunosorbent assays (ELISA)

For evaluating the binding of recombinant EF-Tu (NTHi and orthologues), Nunc PolySorp plates (VWR, Radnor, PA) were coated with 25 nM Fc-hDectin-1a (InvivoGen) in 0·1 M Tris (pH 9·0) overnight at 4 °C, washed with PBST, and blocked with 2·5% BSA/PBS for 1 h at room temperature. After washing, 10 nM EF-Tu orthologues in PBS supplemented with 1% BSA were added, and the plates were incubated for 1 h at room temperature. Following another wash step, plates were incubated with a 1:5000 dilution of anti-His-HRP-conjugated antibody in 2·5% BSA +0·05% Tween-20 for 1 h at room temperature. ELISA substrate (tetramethylbenzidine in potassium citrate) was added, and reactions were stopped after 15 min with H₂SO₄. Optical density at 450 nm was measured using a Tecan Sunrise plate reader (Männedorf, Switzerland). Protein E, an unrelated His-tagged control protein, was applied at matched concentrations on the same plate as a negative control. Optical densities are reported as measured; no blank or control subtraction was applied. Wells receiving no EF-Tu measured plate and detection background. Concentration-dependent binding was characterised by fitting a four-parameter log-logistic model to the EF-Tu series using the drc package (v3·0-1) in R, with the half-maximal effective concentration (EC₅₀) and its 95% CI obtained by the delta method.

Flow cytometry analyses

For binding experiments, NCI–H292 or A549 epithelial cells were detached with Accutase, washed with PBS supplemented with 1% BSA, and then incubated with laminarin (Sigma–Aldrich; 50 μg/mL) or PBS with 1% BSA for 30 min at room temperature, followed by incubation with recombinant EF-Tu. NCI–H292 cells received a titration of 1 and 2 μg/mL with a matched no-EF-Tu control plotted as the 0 μg/mL condition; A549 transfectants received a single dose alongside empty-vector controls. Rabbit anti-EF-Tu polyclonal antibody was added, followed by a FITC-conjugated swine anti-rabbit polyclonal antibody (Dako Agilent, Santa Clara, CA). Each experiment included a control stained with both antibodies but with EF-Tu omitted, which defines the non-specific antibody signal for this indirect format, and a control incubated without either antibody. Data were acquired on a FACSVerse flow cytometer and analysed with FlowJo software (v10·6·1; Becton Dickinson). The FITC-positive gate was defined once per experiment and applied unchanged to all tubes acquired that day. Binding is reported as the percentage of gated cells falling in the FITC-positive gate.

Analysis of DECTIN-1 signalling

For stimulation experiments, cells were grown to 80% confluency, detached with PBS, and seeded in a 96-well plate at 4 × 10⁵ cells/mL in complete DMEM without antibiotics, and incubated with recombinant endotoxin-free protein in solution or coated on beads. SEAP levels in the supernatant of HEK-Blue hDectin-1a cells were quantified using QUANTI-Blue Solution (InvivoGen) according to the manufacturer's instructions. Briefly, QUANTI-Blue reagents were thawed, combined with sterile water, and incubated at room temperature for 10 min prior to use. Next, 180 μL of QUANTI-Blue Solution was added to each well of a 96-well plate, followed by 20 μL of culture supernatant. After incubation at 37 °C for 3 h, SEAP activity was measured by assessing optical density at 620–655 nm using a Tecan Spark plate reader (Tecan Group, Männedorf, Switzerland). An increase in absorbance indicated SEAP activity and DECTIN-1a receptor activation. The positive controls TNF-α (10 ng/mL; InvivoGen) and zymosan (50 μg/mL; InvivoGen; DECTIN-1 agonist) were included in all experiments.

Bacterial adhesion and internalisation assays

A549 cells transfected with pUNO1-hDectin1a or pUNO1 control vector were seeded in 24-well plates and grown to confluence for two days after subculture. NTHi 3655 was cultured for 3 h in brain heart infusion broth supplemented with NAD and haemin, washed, and resuspended in PBS containing 1% BSA to OD₆₀₀ 1·0. Bacteria were pre-incubated with affinity-purified rabbit anti-EF-Tu polyclonal antibody, rabbit anti-Protein D antibody, or PBS for 45 min at room temperature, then washed and resuspended in culture medium. Antibodies were used at a saturating amount established by prior titration. Where indicated, cells were pre-treated with 50 μg/mL laminarin for 30 min. Bacteria were added at a multiplicity of infection of 100 in 0·5 mL RPMI and incubated at 37 °C in 5% CO₂ for 1 h (adhesion) or 2 h (internalisation). For adhesion, monolayers were washed five times with PBS, detached with trypsin, and resuspended in PBS containing 20% FCS. For internalisation, monolayers were washed twice with PBS, treated with 100 μg/mL gentamicin for 2 h to kill extracellular bacteria, washed twice more, detached with trypsin, and lysed in 0·3% saponin for 5 min on ice with vortexing each minute. Lysates were plated in serial dilution on chocolate agar and colonies counted after overnight incubation. Counts are expressed as a percentage of the control-vector untreated well of the same experiment. Ratios were log₂-transformed and compared by paired two-sided t test with the independent experiment as the unit of replication. n = 5 independent experiments for internalisation, n = 4 for the laminarin condition, and n = 3 for adhesion.

CLEC7A tissue and cell-type expression

CLEC7A tissue expression was taken from GTEx v8 median transcript expression across 25 human tissues, reported as transcripts per million. Cell-type expression in human lung was taken from single-cell RNA sequencing of Tabula Sapiens as provided by the Human Protein Atlas v25·0, reported as normalised counts per million.

Ethics

The Malmö Diet and Cancer Study was approved by the research ethics committee of the Faculty of Medicine, Lund University, on Feb 14, 1990 (LU 51–90), conditional on the consent of the Swedish Data Inspection Agency, which was granted. The present analysis was approved by the Swedish Ethical Review Authority on Feb 9, 2022 (Dnr 2022-00191-01). Written informed consent was obtained from all participants at inclusion. The human study was conducted in accordance with the principles of the Declaration of Helsinki. An opt-out option was published on the Lund University website (LUPop) to inform participants about the inclusion of additional outcomes not specified in the original study consent.

All animal experiments were conducted in accordance with Swedish legislation on animal protection and were approved by the Regional Ethical Committee for Animal Experimentation at Lund University, Sweden (permits 5.8.18-08146/2017, approved May 24, 2017, and 5.8.18-19202/2023, approved Dec 20, 2023). Animal experiments are reported in accordance with the ARRIVE guidelines.

Statistics

Statistical analyses were performed using R version 4·5·1 (R Foundation for Statistical Computing, Vienna, Austria). Human genetics analyses used the survival package (v3·8-3) and the cmprsk package (v2·2-12) for time-to-event modelling, with standard R packages for data handling and visualisation. Mouse experiment analyses used rstatix (v0·7·2) and car (v3·1-3) for non-parametric and parametric testing. For mouse experiments, n represents individual mice; for in vitro experiments, n represents independent biological replicates.

Human genetics analyses used Cox proportional hazards regression with age as the timescale. Participants entered the risk set at age 65 years, or at their age at baseline examination if they were recruited later, and were followed to the first event of the endpoint under analysis, death, emigration, or Dec 31, 2023, whichever came first. Base models adjusted for sex, BMI, and genetic principal components (PC1–PC3, to account for population stratification). Fully adjusted models additionally included COPD/asthma and immunocompromising conditions, with stratification by smoking status and diabetes mellitus when proportional hazards assumptions were violated. Proportional hazards assumptions were tested using Schoenfeld residuals (Figure S5, Appendix p 7). The analysis set was not filtered for relatedness or genetic ancestry; population structure was addressed by adjustment for principal components PC1–PC3. Complete model outputs are provided in Tables S5 and S6 (Appendix pp 25–26). Genotype was modelled as an additive per-C-allele term in the primary analysis; as a secondary analysis it was modelled as a three-level factor with AA as the reference, tested against the covariate-only model with a two-degree-of-freedom likelihood ratio test (Table S7, Appendix p 27). Because the endpoint is the first diagnosis recorded from age 65 years onward, participants with an earlier diagnosis of the endpoint cannot contribute an incident event, and excluding them alters the denominator only. Baseline characteristics were compared across genotypes using one-way ANOVA for continuous variables and chi-square or Fisher's exact tests for categorical variables, with Fisher's exact applied when expected cell counts were less than 5. The human analyses report a single prespecified primary association, between Y238∗ and pneumonia; the remaining endpoints, subgroups and sensitivity analyses are secondary and are reported without adjustment for multiple comparisons. Unadjusted event-free survival curves were estimated by the Kaplan–Meier method with age as the timescale, accommodating delayed entry. All-cause mortality was analysed as an outcome using the same specification, with death as the event. Competing risks of death were modelled using Fine–Gray subdistribution hazards regression with a participant-clustered robust variance, as one participant contributes multiple rows to the risk-set expansion; cumulative incidence functions were estimated using the Aalen-Johansen estimator, which accommodates delayed entry. Covariate-adjusted survival curves were derived using marginal standardisation. Mouse data were analysed using Mann–Whitney U tests or Kruskal–Wallis tests with Dunn's post-hoc correction when data were non-normally distributed, and with unpaired t tests or one-way ANOVA when Shapiro–Wilk testing supported normality. Where the design required it, other procedures were used and are named in the corresponding figure legend: a linear model on log₁₀ values with experiment as a fixed blocking factor for immunohistochemical neutrophil counts; Spearman rank correlation for the relationship between neutrophil counts and bacterial burden; Wilcoxon rank-sum tests for reporter-cell validation; paired t tests on log₂ ratios, two-sided except for the prespecified one-sided comparison of bead-immobilised EF-Tu against control, for within-experiment internalisation and NF-κB contrasts; and two-way ANOVA with Tukey's HSD for cell-binding flow cytometry. Corrections for multiple comparisons are stated in the corresponding legend. Statistical significance was defined as p < 0·05. Missing data were handled by complete-case analysis; participants with missing genotype or covariate data required for a given model were excluded from that model, and no imputation was used in the primary analyses. No participant was missing genotype data. Missing covariate data excluded 1574 of 28,175 participants (5·6%) from the fully adjusted models, predominantly for missing smoking status (1533 participants); body mass index alone accounted for a further 40, and one participant was missing all three genetic principal components. The low exclusion rate and balanced baseline characteristics across genotypes suggest minimal selection bias. As a sensitivity analysis, all four primary Cox models were refitted using the missing-data indicator method, with missing body mass index set to the cohort median and carried as an indicator term, and missing smoking status entered as an additional stratum; results are reported in Table S7 (Appendix p 27). The human analysis used all eligible MDCS participants rather than a sample of predetermined size, and murine group sizes were based on prior experience with pneumonia models. Randomisation was not applicable to the cohort, where the exposure was an inherited genotype. Investigators were not blinded to genotype. Human outcomes were identified from register-linked ICD codes. In the murine experiments, bacterial burden, clinical monitoring, cytokine measurement, and the selection and annotation of histological sections were done unblinded; subsequent detection of Ly6G-positive cells used automated QuPath analysis with the thresholds given in the Appendix, which reduces observer judgement during counting. Genotyped participants were eligible if they reached age 65 years and had subsequent follow-up, all models used complete cases, and outcome-specific exclusions are detailed in Figure S1 (Appendix p 2). Animal inclusion and exclusion criteria, and the three animals excluded, are stated under Study design.

This observational genetic-association study is reported in accordance with the STROBE and STREGA reporting guidelines.²⁷

Role of funders

The funders had no role in study design, data collection, data analysis, interpretation, writing of the report, or the decision to submit for publication.

Results

Y238∗ carriers have increased pneumonia risk

To assess the clinical relevance of CLEC7A variation in respiratory infection, we analysed the Y238∗ variant in the MDCS cohort. Genotype distribution in the age-eligible cohort (n = 28,175) followed Hardy–Weinberg expectations: 85·9% wild type (AA), 13·5% heterozygous (AC), and 0·5% homozygous (CC), yielding a minor allele frequency of 7·3%, consistent with the 8·0% reported for non-Finnish Europeans in gnomAD v4·1.⁷ Baseline characteristics were balanced across genotype groups (Table 1). Mean age at cohort entry was 58·2 (SD 7·6) years, 60·6% of participants were female, and mean follow-up was 17·3 (SD 7·0) years.

Table 1.

Baseline characteristics of the base analytical cohort by CLEC7A Y238∗ genotype.

Characteristic	All (n = 28,126)	AA (wild-type) (n = 24,174)	AC (n = 3808)	CC (n = 144)	p value
Participant characteristics
Age at entry, years, mean (SD)	58·2 (7·6)	58·2 (7·6)	58·2 (7·6)	57·4 (7·7)	0·397
Female sex, n (%)	17,038 (60·6)	14,653 (60·6)	2295 (60·3)	90 (62·5)	0·823
BMI, kg/m2, mean (SD)	25·8 (4·0)	25·8 (4·0)	25·9 (4·1)	25·7 (3·8)	0·736
Current smoker, n (%)	7243 (27·2)	6210 (27·2)	994 (27·6)	39 (28·9)	0·755
Former smoker, n (%)	9156 (34·4)	7837 (34·3)	1268 (35·3)	51 (37·8)	0·361
Never smoker, n (%)	10,202 (38·4)	8823 (38·6)	1334 (37·1)	45 (33·3)	0·115
Comorbidities
COPD or asthma, n (%)	1766 (6·3)	1504 (6·2)	250 (6·6)	12 (8·3)	0·428
Diabetes mellitus, n (%)	1283 (4·6)	1102 (4·6)	179 (4·7)	2 (1·4)	0·174
Immunocompromising conditions, n (%)	1809 (6·4)	1553 (6·4)	246 (6·5)	10 (6·9)	0·965
Heart disease, n (%)	874 (3·1)	747 (3·1)	124 (3·3)	3 (2·1)	0·738
Any malignancy, n (%)	2984 (10·6)	2587 (10·7)	382 (10·0)	15 (10·4)	0·458
Follow-up (from age 65)
Follow-up from age 65, years, mean (SD)	17·3 (7·0)	17·4 (7·0)	17·3 (6·9)	16·3 (7·2)	0·200
Age at end of observation, years, mean (SD)	82·3 (7·0)	82·4 (7·0)	82·3 (6·9)	81·3 (7·2)	0·200
Death during follow-up, n (%)	15,692 (55·8)	13,469 (55·7)	2146 (56·4)	77 (53·5)	0·651
Outcomes
Pneumonia (all-cause), n (%)	5493 (19·5)	4649 (19·2)	814 (21·4)	30 (20·8)	0·00750
Bacterial pneumonia, n (%)a	2537 (10·1)	2148 (9·9)	376 (11·1)	13 (10·4)	0·0996

Data are n (%) unless otherwise stated. The table describes the base analytical cohort (n = 28,126), the population of the base Cox models. Smoking percentages are taken over the 26,601 participants with a recorded smoking status, not over the full cohort.

BMI = body mass index. COPD = chronic obstructive pulmonary disease.

^aBacterial pneumonia is analysed in a refined cohort of 25,145 participants, which excludes controls with unspecified-pneumonia ICD codes.

During follow-up, 5493 pneumonia events occurred in the base analytical cohort of 28,126 participants, and 5190 events in the 26,601-participant fully adjusted cohort (Table S8, Appendix p 29). Using Cox proportional hazards models with age as the timescale, we found that Y238∗ carriers had significantly elevated pneumonia risk in a base model adjusted for sex, BMI, and genetic principal components (HR 1·13, 95% CI 1·06–1·21, p < 0·001; Fig. 1A). This association persisted after further adjustment: in the fully adjusted model, which stratified by smoking status and diabetes and additionally adjusted for COPD, asthma, and immunocompromising conditions, the hazard ratio was 1·12 (95% CI 1·04–1·20; p = 0·002; complete model outputs in Table S5, Appendix p 25). When we restricted the analysis to bacterial pneumonia specifically, the effect remained consistent (HR 1·13, 95% CI 1·01–1·25; p = 0·028; Fig. 1A). Kaplan–Meier curves confirmed that risk allele carriers accumulated pneumonia at a higher rate than non-carriers throughout the entire follow-up period (Fig. 1B). Proportional hazards assumptions were met for the CLEC7A genotype in all models (Figure S5, Appendix p 7).

Fig. 1.

Loss-of-function variant in CLEC7A (Y238∗) is associated with increased pneumonia risk. (A) Forest plot showing hazard ratios (HR) with 95% CI for the association between rs16910526 (CLEC7A Y238∗) and pneumonia outcomes under an additive genetic model (per C allele). Cox proportional hazards regression used age as the timescale, with participants entering the risk set at the later of age 65 years or their baseline examination. Base models adjusted for sex, BMI, and genetic principal components (PC1–3). Adjusted models were stratified by smoking status and diabetes mellitus, with covariate adjustment for COPD/asthma and immunocompromising conditions. Sample sizes and event counts are shown. Genotype-specific hazard ratios with AA as the reference category, and sensitivity analyses excluding early events, excluding prior pneumonia, and with extended covariate adjustment, are given in Table S7 (Appendix p 27). (B) Kaplan–Meier event-free probability curves for all-cause pneumonia (left; n = 28,175; 5509 events) and bacterial pneumonia (right; n = 25,186; 2548 events) by genotype. Age as timescale, with entry at the later of age 65 years or their baseline examination. (C) Aalen-Johansen cumulative incidence of pneumonia and of death without pneumonia, by genotype, accounting for death as a competing event and for delayed entry. Genotypes in the Kaplan–Meier and competing-risk cohort (n = 28,175): AA (wild type, n = 24,214), AC (heterozygous, n = 3816), CC (homozygous, n = 145). Numbers at risk are shown below every panel in (B) and (C). Data are from the Malmö Diet and Cancer Study, a population-based prospective cohort of 28,175 European participants followed through Swedish national registers. CI, confidence interval.

Given the long follow-up and older cohort, we performed Fine–Gray competing risk regression to account for the competing risk of death. These models supported the primary findings for all-cause pneumonia (base model sHR 1·12, 95% CI 1·04–1·20, p = 0·002; fully adjusted sHR 1·11, 95% CI 1·03–1·19, p = 0·005). For bacterial pneumonia the base model gave sHR 1·12 (95% CI 1·01–1·23; p = 0·036), and the fully adjusted model attenuated to sHR 1·11 (95% CI 1·00–1·23; p = 0·050; Figure S6 and Table S6, Appendix pp 8, 26). All-cause mortality did not differ across genotype groups (fully adjusted Cox model with death as the outcome: HR 1·03 per C allele, 95% CI 0·98–1·07, p = 0·25; global test across the three genotypes p = 0·48; Fig. 1C). Stratified Cox models showed consistent risk elevation across sex, age groups, and comorbidity strata, with no significant interactions (Fig. 2A). The hazard ratio remained elevated among participants without chronic respiratory disease and among never-smokers. Covariate-adjusted cumulative incidence curves confirmed a gene-dose-dependent increase in the probability of pneumonia (Fig. 2B) and bacterial pneumonia (Fig. 2C) over the 25-year follow-up period. The association was specific to respiratory infections; Y238∗ was not associated with urinary tract infections or sepsis (Figure S7, Appendix p 9). The risk increase was of similar magnitude for hospitalised disease (inpatient pneumonia HR 1·12, 95% CI 1·02–1·22, p = 0·012) and somewhat larger for prolonged pneumonia-related hospitalisation (HR 1·16, 95% CI 1·02–1·31, p = 0·019; Figure S8, Appendix p 10). Case-only analyses restricted to participants who developed pneumonia showed no association between Y238∗ and inpatient admission, prolonged hospitalisation, 30-day mortality, recurrence, or bacterial aetiology, suggesting that the increased risk reflects susceptibility to infection rather than disease severity.

Fig. 2.

Stratified analyses confirm CLEC7A Y238∗-pneumonia association across clinical subgroups. (A) Forest plot showing hazard ratios (HR) with 95% CI for the association between rs16910526 and pneumonia across prespecified subgroups under an additive genetic model (per C allele). Models adjusted for sex, BMI, genetic principal components (PC1–3), COPD/asthma, and immunocompromising conditions, stratified by smoking status and diabetes mellitus (except where these variables define the subgroup). P_interaction values are from Wald tests of genotype-by-subgroup interaction terms. No significant interactions were detected (p-range 0·12–0·94). Sample sizes and event counts are shown. (B) Covariate-adjusted cumulative incidence of all-cause pneumonia by genotype (n = 26,601; 5190 events; Cox regression p = 0·002). (C) Covariate-adjusted cumulative incidence of bacterial pneumonia by genotype (n = 23,790; 2401 events; Cox regression p = 0·028), using a cohort excluding participants with unspecified pneumonia diagnoses. Adjusted curves (B, C) represent marginal standardisation (G-computation) predictions from the stratified Cox models. Genotypes: AA (wild type, blue), AC (heterozygous, teal), CC (homozygous, red). CI, confidence interval.

To assess whether the MDCS findings replicated in independent populations, we examined summary statistics from published genome-wide association studies and public biobank resources. In the Sakaue et al. cross-population meta-analysis (n = 480,299), rs16910526 showed a concordant association with all-cause pneumonia (OR 1·05, 95% CI 1·01–1·09, p = 0·026).²⁸ In UK Biobank SAIGE summary statistics for bacterial pneumonia (n = 405,248; 6710 cases; GWAS Catalog accession GCST90436210), the direction of effect was concordant with MDCS, although the confidence interval crossed the null (OR 1·03, 95% CI 0·97–1·10; Figure S9, Appendix p 11).

Clec7a deficiency impairs pulmonary clearance of S. pneumoniae

To determine whether DECTIN-1 deficiency directly impairs antibacterial pulmonary defence, we challenged mice with S. pneumoniae in two complementary pneumonia models (Fig. 3A). We first used the low-virulence serotype 19F strain EF3030, which establishes stable respiratory tract infection without rapid systemic dissemination. Wild-type and Clec7a^−/− mice were inoculated intranasally and assessed at 48 h postinfection. Clec7a^−/− mice developed more severe clinical symptoms, including marked lethargy, with several reaching the ethical endpoint for termination (Fig. 3B), and had significantly higher lung bacterial burdens than wild-type controls (p = 0·006; Fig. 3C). Nasopharyngeal colonisation was comparable between genotypes (p = 0·62; Fig. 3D), indicating that the clearance defect was specific to the lower airways rather than reflecting a generalised impairment of mucosal defence.

Fig. 3.

Gene-dose-dependent Streptococcus pneumoniae clearance defect in Clec7a-deficient mice. (A) Schematic of the intranasal infection protocol using S. pneumoniae strains EF3030 (serotype 19F) or D39 (serotype 2). (B) Lung culture outcomes in wild-type and Clec7a^−/− mice 48 h after EF3030 infection (10⁷ CFU), shown as 100% stacked bars (no growth, mild positive, severe positive); n = 31 (wild type), n = 33 (Clec7a^−/−). (C) Lung bacterial load (CFU/mL) after EF3030 infection; p = 0·006, Mann–Whitney U test. Each symbol represents one mouse; horizontal bars indicate medians. ND, not detected (below limit of detection). (D) Nasopharyngeal bacterial load after EF3030 infection, in the same animals as (B) and (C); n = 31 (wild type), n = 33 (Clec7a^−/−); p = 0·62, not significant. (E) Lung bacterial load in wild-type, Clec7a^+/−, and Clec7a^−/− mice 48 h after D39 infection (2·2 × 10⁷ CFU), showing gene-dose-dependent impairment: p = 0·002 (wild type vs Clec7a^−/−), p = 0·04 (Clec7a^+/− vs Clec7a^−/−); n = 10 (wild type), n = 10 (Clec7a^+/−), n = 7 (Clec7a^−/−). (F) BALF bacterial load after D39 infection: p = 0·02 (wild type vs Clec7a^−/−), p = 0·03 (Clec7a^+/− vs Clec7a^−/−). (G) Neutrophil infiltration in the lung 24 h after D39 infection, assessed by anti-Ly6G immunohistochemistry (Ly6G⁺ % of detected cells); n = 5 per genotype. Genotypes did not differ significantly in neutrophil infiltration (p = 0·25). Columns, geometric means; error bars, back-transformed SEM of the log₁₀ values. The Clec7a^−/− to wild-type ratio was 3·75 (95% CI 0·56–25·10). Logarithmic y-axis; columns start at the axis minimum. (H) Representative lung sections 24 h after infection: Ly6G by immunohistochemistry (DAB, brown) with haematoxylin counterstain (blue), genotype by column and treatment by row. Images are representative of all 23 sections analysed, comprising n = 5 infected animals per genotype quantified in (G) and PBS controls. Scale bars: 100 μm. (I) Ly6G⁺ cells in the lung and bronchoalveolar lavage fluid 24 h after D39 infection, by flow cytometry against a counting-bead standard. In the lung the count is extravascular, meaning Ly6G⁺ cells outside an intravenously administered anti-CD45 label; in lavage fluid no intravascular gate is applied. Squares, uninfected; circles, infected; grey lines, pooled medians across the groups spanned by the arrow. Infection raised Ly6G⁺ cells 11·1-fold in the lung (p < 0·0001) and 193-fold in bronchoalveolar lavage fluid (p < 0·0001); genotypes did not differ (p = 0·74 for lung, and p = 0·30 for BALF). Lung n = 11 uninfected and 14 infected; BALF n = 10 uninfected and 13 infected. Kruskal–Wallis test with Dunn's post-hoc test (E, F); linear model on log₁₀ values with experiment as a fixed blocking factor (G); Kruskal–Wallis test across genotypes and Mann–Whitney U tests for the infection effect (I); Mann–Whitney U test (C, D). Significance threshold p < 0·05. ns, not significant. BALF, bronchoalveolar lavage fluid. CI, confidence interval. DAB, 3,3′-diaminobenzidine.

To assess whether partial DECTIN-1 deficiency, modelling the heterozygous state prevalent in human carriers, also affects infection outcomes, we challenged wild-type, Clec7a^+/−, and Clec7a^−/− littermates with the highly virulent capsular serotype 2 strain D39. Lung bacterial loads followed a gene-dose-dependent pattern: half of wild-type mice cleared the infection completely, Clec7a^+/− mice displayed intermediate bacterial burdens with 40% clearance, and none of the Clec7a^−/− mice cleared the bacteria (Clec7a^−/− vs wild type p = 0·002; Clec7a^−/− vs Clec7a^+/− p = 0·04; Fig. 3E). This gradient was mirrored in BALF, where Clec7a-deficient mice again carried significantly higher bacterial loads than wild-type and heterozygous animals (Fig. 3F).

To characterise the innate cellular response underlying impaired clearance, we quantified neutrophil infiltration by Ly6G immunohistochemistry and, in a separate cohort, by flow cytometry. Neutrophil infiltration at 24 h postinfection did not differ between genotypes (n = 5 per genotype; p = 0·25; Fig. 3G and H; Figure S10, Appendix p 12). Flow cytometry yielded concordant results by an independent route. Intravenous anti-CD45 given prior to harvest separated circulating from tissue neutrophils: 65·5% of lung Ly6G⁺ cells lay within the vasculature in uninfected mice against 21·3% after infection (p = 0·011). Cells counted after infection were thus predominantly extravascular. Extravascular Ly6G⁺ cells rose 11·1-fold in the lung (p < 0·0001) and 193-fold in BALF (p < 0·0001), yet did not differ between genotypes in either compartment (p = 0·74 lung, p = 0·30 BALF; Fig. 3I; Figure S2, Appendix p 3).

Recruitment tracked bacterial burden rather than genotype, rising with BALF colony counts in the airway (ρ = 0·66, p = 0·015) and in the same direction in lung (ρ = 0·53, p = 0·053; Figure S11, Appendix p 13). Clec7a-deficient mice therefore carried higher bacterial loads despite a neutrophil response that we could not distinguish from that of wild type. Group sizes remain small: a genotype difference smaller than roughly ten-fold in the lung, or twenty-fold in BALF, may therefore have gone undetected.

Clec7a deficiency impairs NTHi clearance and cytokine responses

To assess whether DECTIN-1 contributes to defence against a second major respiratory pathogen, we challenged wild-type, Clec7a^+/−, and Clec7a^−/− mice with NTHi strain 3655 (Fig. 4A). At 48 h postinfection, Clec7a^+/− mice had higher lung bacterial loads than wild-type controls (p = 0·04), and Clec7a^−/− mice showed a similar trend that did not reach statistical significance (p = 0·051). BALF bacterial loads were similarly elevated in Clec7a^−/− mice (p = 0·02; Fig. 4B).

Fig. 4.

Clec7a-deficient mice show impaired clearance and attenuated cytokine responses during Haemophilus influenzae pneumonia.(A) Lung bacterial burden (CFU/mL) in wild-type (Clec7a^+/+), Clec7a^+/−, and Clec7a^−/− mice 48 h after intranasal infection with 1·2 × 10⁸ CFU non-typeable Haemophilus influenzae (NTHi): p = 0·04 (wild type vs Clec7a^+/−), p = 0·051 (wild type vs Clec7a^−/−; Mann–Whitney U test). Each symbol represents one mouse (n = 9 wild type, 9 Clec7a^+/−, and 10 Clec7a^−/−), each genotype including two uninfected animals given PBS alone; horizontal bars indicate medians. ND, not detected. (B) BALF bacterial burden from the same mice (n = 9, 9, and 10): p = 0·02 (wild type vs Clec7a^−/−, Mann–Whitney U test). (C) Cytokine concentrations (ng/mL) in BALF quantified by cytometric bead array; n = 7 (wild type), 6 (Clec7a^+/−), and 7 (Clec7a^−/−) infected mice. DECTIN-1 deficiency significantly reduced IFN-γ, MCP-1, and IL-6 in Clec7a^−/− vs wild-type mice (Mann–Whitney U test with Benjamini-Hochberg correction; all p = 0·034). MCP-1 in Clec7a^+/− mice was not significantly reduced relative to wild type (p = 0·054). TNF-α, IL-10, and IL-12p70 showed no significant genotype-associated differences. Colour scale and cell values indicate mean concentration (ng/mL). Asterisks denote significant differences vs wild type. BALF, bronchoalveolar lavage fluid.

The cytokine response to NTHi infection differed markedly by genotype. Wild-type mice mounted elevated levels of IFN-γ, MCP-1, and IL-6 compared to Clec7a-deficient animals (all p = 0·034; Fig. 4C). In contrast, both heterozygous and homozygous Clec7a-deficient mice showed minimal cytokine production. The similarly attenuated cytokine responses observed in both genotypes suggest that partial loss of DECTIN-1 function may be sufficient to impair the pulmonary innate immune response to NTHi.

Bacterial EF-Tu binds DECTIN-1 and activates NF-κB signalling

To identify bacterial ligands of DECTIN-1, we performed unbiased affinity pull-down assays using recombinant human DECTIN-1 as bait to capture potential interacting proteins from NTHi whole-cell lysates. Mass spectrometry analysis revealed multiple enriched bacterial proteins (Figure S12, Appendix p 14), 228 of which met the enrichment criteria; the 25 most abundant are listed with their enrichment values in Table S9 (Appendix p 30). EF-Tu was consistently enriched across all three replicates (3·0-fold over background; Benjamini-Hochberg adjusted p = 0·011) and was the most abundant protein captured, accounting for 15·3% of the total bacterial signal and 3·9 times more than the next protein. EF-Tu has documented surface localisation in multiple bacterial species and known moonlighting functions in host–pathogen interactions.^29,30

Human DECTIN-1 is a C-type lectin receptor with an extracellular C-type lectin-like domain (CTLD) linked to the membrane via a stalk (Fig. 5A). ELISA confirmed dose-dependent, saturable binding of recombinant EF-Tu to Fc-tagged human DECTIN-1, while a control protein showed no binding (Fig. 5B). EF-Tu also bound both human and mouse DECTIN-1 in a single-dose ELISA assay (Figure S13, Appendix p 15), consistent with conservation of the interaction across species. Domain-mapping experiments using truncated DECTIN-1 variants localised the interaction to the CTLD: EF-Tu bound both the full-length ectodomain and the isolated CTLD, with negligible binding to the stalk region (Fig. 5C).

Fig. 5.

Bacterial elongation factor Tu binds human DECTIN-1 and activates NF-κB in a CLEC7A gene-dose-dependent manner. (A) Domain-coloured structure of human DECTIN-1: C-type lectin-like domain (CTLD), stalk, transmembrane domain (TMD), and intracellular domain. AlphaFold Protein Structure Database model AF-Q9BXN2-F1 (UniProt Q9BXN2), rendered in PyMOL. (B) Concentration-dependent binding of recombinant bacterial EF-Tu to human DECTIN-1 measured by ELISA (OD₄₅₀ nm), fitted with a four-parameter log-logistic model (EC₅₀ = 0·10 μg/mL, dashed line); Protein E is a negative control. (C) EF-Tu binding to recombinant DECTIN-1 variants (full-length ectodomain, CTLD, or stalk) at increasing concentrations (5, 10, 20 nM), demonstrating CTLD-specific binding. (D) Cross-species binding of recombinant EF-Tu orthologues (from H. influenzae, E. coli, S. pneumoniae, and L. monocytogenes) to human DECTIN-1 measured by ELISA (OD₄₅₀ nm); control protein shows minimal signal. (E) Flow-cytometric analysis of EF-Tu binding to NCI–H292 lung epithelial cells, detected with rabbit anti-EF-Tu and a FITC-conjugated secondary antibody, across EF-Tu concentrations with laminarin competition (reduced binding; p = 0·034, two-way ANOVA with Tukey's post-hoc test). (F) EF-Tu binding to A549 cells transfected with human DECTIN-1 or control vector, detected with rabbit anti-EF-Tu and a FITC-conjugated secondary antibody; DECTIN-1 expression increases EF-Tu binding (p < 0·001) and laminarin partially blocks binding (p = 0·005); two-way ANOVA with Tukey's post-hoc test. (G) Internalisation of live non-typeable H. influenzae 3655 into A549 cells transfected with human DECTIN-1 or control vector (pUNO1), assessed by a gentamicin protection assay; each symbol represents one independent experiment. DECTIN-1 vs control vector, within each condition: p = 0·001 for untreated, p = 0·36 with anti-EF-Tu polyclonal antibodies (pAb), p = 0·043 with the anti-Protein D control antibody (α-PD), and p = 0·71 with laminarin; ns, not significant. Within DECTIN-1-expressing cells, comparing each treatment with untreated, anti-EF-Tu reduced internalisation to 37% of untreated (p = 0·031) and laminarin gave a comparable reduction (p = 0·049); the anti-EF-Tu reduction was specific to DECTIN-1-expressing cells (p = 0·011). Paired two-sided t tests on log₂ ratios throughout; n = 5, except n = 4 for laminarin. (H) NF-κB reporter assay (SEAP) in HEK-Blue hDectin-1a vs HEK-Blue Null1-v cells, comparing empty beads, soluble EF-Tu, EF-Tu-coupled beads, soluble Protein F, and Protein F-coated beads. EF-Tu-coupled beads selectively activate NF-κB in DECTIN-1-expressing cells (p = 0·043, paired one-sided t-test), whereas empty beads or Protein F-coated beads do not. (I) Heatmap of cytokine responses in BALF from wild-type, Clec7a^+/−, and Clec7a^−/− mice 24 h after inhalation of recombinant EF-Tu (n = 7 per group), quantified by cytometric bead array. Colour scale indicates fold change relative to ovalbumin control. Asterisks on wild-type values indicate cytokines significantly elevated in wild-type but not Clec7a^−/− mice (pairwise Mann–Whitney U tests, unadjusted; IL-6, p = 0·007; IL-12p70, p = 0·017). Data are mean ± SEM. Sample sizes are n = 3 independent experiments per point (B, C); n = 4 replicates per orthologue (D); n = 6 independent experiments at 0 and 2 μg/mL and n = 5 at 1 μg/mL (E); n = 4 wells from two independent experiments per condition (F); and in (H) n = 3 for empty beads and for EF-Tu-coupled beads, n = 2 for soluble EF-Tu, n = 4 for soluble Protein F, and n = 10 for Protein F-coated beads pooled across four coating doses in three independent experiments. BALF, bronchoalveolar lavage fluid. SEM, standard error of the mean.

EF-Tu is highly conserved across bacterial species owing to its essential role in translation elongation.³¹ Sequence alignment of 18 respiratory bacterial pathogens revealed 66–98% amino acid identity, with the nucleotide-binding motifs near-invariant (Figure S14 and Table S4, Appendix pp 16, 24). Recombinant EF-Tu orthologues from S. pneumoniae, E. coli, and L. monocytogenes all demonstrated comparable binding to human DECTIN-1 (Fig. 5D). Epithelial cell lines were used as tractable systems in which receptor expression could be controlled. DECTIN-1 is expressed on bronchial and alveolar epithelium in human lung and mediates NTHi-induced cytokine responses in primary bronchial epithelial cells.⁵ Flow cytometry confirmed EF-Tu binding to NCI–H292 lung epithelial cells, and this binding was inhibited by the DECTIN-1 antagonist laminarin (p = 0·034; Fig. 5E). In A549 cells overexpressing DECTIN-1, with surface expression confirmed against control-transfected cells (Figure S15, Appendix p 18), EF-Tu binding was significantly increased (p < 0·001) and again reduced by laminarin (p = 0·005; Fig. 5F). Although laminarin significantly reduced EF-Tu binding, inhibition was incomplete, indicating that DECTIN-1 contributes substantially to EF-Tu recognition while not excluding interactions with additional cell-surface receptors.

Surface EF-Tu on intact bacteria contributes to DECTIN-1-dependent uptake. Three times more live NTHi 3655 were internalised by A549 cells expressing DECTIN-1 than by control-vector cells (p = 0·001, five independent experiments; Fig. 5G). Pre-incubating the bacteria with affinity-purified rabbit anti-EF-Tu antibody reduced internalisation into DECTIN-1-expressing cells to 37% of untreated (p = 0·031) and removed their excess uptake, leaving no detectable difference from control-vector cells (p = 0·36). This reduction was specific to DECTIN-1-expressing cells (p = 0·011). In parallel, using an anti-Protein D control antibody, bacterial uptake remained higher in DECTIN-1-expressing cells than in control-vector cells (p = 0·043). Pre-treating the cells with laminarin, which blocks the receptor rather than the ligand, gave a comparable result (p = 0·049). All contrasts in this paragraph are paired two-sided t tests on log₂ ratios (Fig. 5G). Anti-EF-Tu pAb did not reduce total cell-associated bacteria in either cell line (Figure S16, Appendix p 19), which argues against a loss of bacterial attachment.

DECTIN-1 activation requires clustering of two or more receptors via its hemITAM motif, typically induced by larger particulate ligands.³² Consistent with this, soluble EF-Tu did not activate NF-κB in HEK-Blue hDectin-1a reporter cells. However, bead-immobilised EF-Tu, with coating verified for both proteins (Figure S4, Appendix p 6), induced significant NF-κB activation in DECTIN-1-expressing cells (p = 0·043) but not in control cells lacking the receptor (Fig. 5H). To confirm in vivo relevance, wild-type, Clec7a^+/−, and Clec7a^−/− mice were challenged intranasally with endotoxin-free recombinant EF-Tu, with ovalbumin as a molecular-weight-matched control. BALF cytokines showed a gene-dose-dependent pattern: IL-6 and IL-12p70 were reduced in Clec7a^−/− mice compared to wild type (p = 0·007 and p = 0·017, respectively, by unadjusted pairwise Mann–Whitney U test), with intermediate responses in heterozygotes (Fig. 5I).

Discussion

Our combined genetic, experimental, and biochemical data show that DECTIN-1 contributes to pulmonary antibacterial defence and identify bacterial EF-Tu as a functional ligand for this receptor, while not excluding participation of additional receptors. The CLEC7A Y238∗ loss-of-function variant was associated with a 12–13% increase in pneumonia risk in a prospective cohort of over 28,000 individuals, murine models showed impaired pulmonary bacterial clearance, gene-dose-dependent in the pneumococcal model, and affinity proteomics identified EF-Tu as a conserved DECTIN-1 ligand that activates NF-κB signalling.

Previous clinical studies associated the CLEC7A Y238∗ variant with susceptibility to fungal infection, including mucocutaneous candidiasis, phaeohyphomycosis, severe coccidioidomycosis and invasive aspergillosis, while its contribution to bacterial respiratory infection has remained largely unexplored.^8,10,11,33 Our data extend the relevance of CLEC7A to pneumonia in an unselected general population. The association remained consistent across adjusted Cox models, competing-risk regression, and stratified analyses. Its pattern of association with pneumonia but not with all-cause mortality, and with respiratory rather than non-respiratory infections, further supports a link between DECTIN-1 deficiency and pneumonia susceptibility.

The identification of EF-Tu as a DECTIN-1 ligand provides a plausible molecular mechanism linking receptor deficiency to impaired bacterial sensing. EF-Tu is a universally conserved GTPase constituting 5–10% of total bacterial protein, with documented surface localisation and moonlighting functions in host–pathogen interactions across diverse species.29, 30, 31 The requirement for ligand immobilisation to trigger NF-κB signalling aligns with the known hemITAM clustering mechanism of DECTIN-1 activation,³² and masking surface EF-Tu on intact bacteria reduced DECTIN-1-dependent uptake (Fig. 5G). DECTIN-1 therefore senses EF-Tu presented on the bacterial surface rather than as a soluble protein. Cross-species binding of EF-Tu orthologues from both Gram-positive and Gram-negative pathogens suggests that EF-Tu is a broadly conserved pathogen-associated molecular pattern, extending the ligand repertoire of DECTIN-1 beyond β-glucans. Laminarin did not abolish EF-Tu binding in either cell line, and the receptor-dependent component measured against control-transfected cells exceeded the laminarin-displaceable component. EF-Tu may therefore engage sites on DECTIN-1 that laminarin does not fully occupy, or additional host receptors alongside DECTIN-1. Neither nasopharyngeal colonisation in Clec7a-deficient mice (Fig. 3D) nor sepsis risk in Y238∗ carriers (Figure S7, Appendix p 9) differed from control, despite EF-Tu binding conserved across species, consistent with a role confined to the lower airway rather than to mucosal defence generally. In GTEx v8, lung has the second-highest median CLEC7A expression of 25 tissues (52·1 TPM, after whole blood at 60·3 TPM), and at single-cell resolution this expression is concentrated in lung myeloid cells (macrophages, monocytes, and neutrophils), with lower expression in airway epithelium (Figure S17, Appendix p 20). These findings suggest that the pulmonary vulnerability associated with the CLEC7A Y238∗ variant reflects impaired bacterial recognition principally by lung myeloid cells, with a possible additional contribution from airway epithelium, during the initial encounter with inhaled bacteria.

The gene-dose-dependent phenotype observed in the pneumococcal model mirrors the partial loss-of-function effects seen in human carriers. The Y238∗ variant produces a non-functional receptor, and monocytes from heterozygous carriers show diminished cytokine secretion despite normal receptor protein levels.^8,10 In our murine models, even heterozygous animals exhibited impaired bacterial clearance and blunted cytokine responses, compared to wild-type littermates. Neutrophil recruitment to the infected lung did not differ between genotypes, by immunohistochemistry (p = 0·25) or by flow cytometry, while bacterial clearance was impaired. That dissociation would be consistent with a defect in effector function rather than mobilisation, although these data cannot establish it. DECTIN-1 engagement on neutrophils directly triggers reactive oxygen species production,³⁴ and neutrophils from Y238∗ carriers generate less reactive oxygen species in response to mycobacteria.⁹ Moreover, β-glucan-mediated DECTIN-1 signalling reprogrammes neutrophil function through trained immunity,³⁵ raising the possibility that loss of DECTIN-1 impairs the functional maturation of recruited neutrophils, resulting in a non-productive inflammatory response despite adequate cellular mobilisation.

Our study has several strengths. The MDCS is a prospective population-based cohort with register linkage rather than a case–control sample, and its scale, over 28,000 participants with a mean follow-up of 17 years, provided sufficient power to assess a common genetic variant in relation to a frequent clinical outcome. We tested a single prespecified association in a variant whose loss-of-function mechanism is already established, rather than screening for one. The estimate held across Cox and Fine–Gray models, stratified subgroups, and sensitivity analyses, and the clinical risk factors we could measure did not explain it (Table S7, Appendix p 27). Case-only analyses separated susceptibility from severity, and two independent datasets reproduced the direction of effect (Figure S9, Appendix p 11). Combining human genetics, murine infection models, and biochemical ligand characterisation within a single study strengthens biological plausibility.

Our study has several limitations. First, the epidemiological analysis was conducted in a European cohort, limiting generalisability to populations where Y238∗ allele frequency differs, from 8·0% in non-Finnish Europeans to 2·0% in African/African-American and 0·02% in east Asian populations (gnomAD v4·1).⁷ Second, pneumonia ascertainment relied on ICD codes from healthcare registers; milder infections not leading to clinical contact may be missed, which would bias effect estimates toward the null. Third, the Y238∗ allele is rare in the homozygous state, so the 135 CC homozygotes in the fully adjusted cohort contributed 29 pneumonia events and the genotype-specific estimate for CC is imprecise (Table S7, Appendix p 27); the additive per-allele model, which uses the full sample, is therefore our primary analysis. Vaccination status, frailty, socioeconomic position, kidney function, aspiration risk, and dispensed medications are not included in this study's register linkage or ethical approval and could not be evaluated as confounders. Fourth, EF-Tu is essential for viability and cannot be selectively deleted, so we masked surface EF-Tu on intact bacteria with polyclonal antibodies rather than testing an EF-Tu-deficient mutant. Antibody masking is steric and cannot exclude contributions from other bacterial surface ligands or other host receptors, the blocking experiments used an epithelial transfectant rather than myeloid cells, and the equivalent experiment was not performed in S. pneumoniae. Because Clec7a deletion in our mouse model is whole-body, our data cannot apportion the in vivo phenotype between epithelial and myeloid compartments. Whether EF-Tu is the principal ligand driving the in vivo phenotype therefore remains open. Fifth, investigators were not blinded to mouse genotype: bacterial burden, clinical monitoring, cytokine measurement, and the selection and annotation of histological sections were all done unblinded, and only the subsequent detection of stained cells was automated. The neutrophil comparisons were also small: at the observed between-animal spread, a genotype difference smaller than roughly ten-fold in lung or twenty-fold in BALF would not have been detected, so we report no detectable difference rather than equivalence. Sixth, C57BL/6 mice express a naturally occurring truncated Clec7a splice variant with reduced responsiveness to β-glucan ligands relative to strains carrying the full-length receptor,^36,37 so the murine phenotypes we report are measured against an already attenuated baseline. This makes the observed clearance defect a conservative estimate rather than an inflated one, but it limits extrapolation to full-length human DECTIN-1. We also did not test cells from Y238∗ carriers, so the link between the human genetic association and the murine phenotype rests on convergence between independent lines of evidence rather than on a direct functional test in carrier cells. Finally, residual confounding from population stratification may persist despite adjustment for genetic principal components. Replication in non-European cohorts, structural characterisation of the EF-Tu–DECTIN-1 binding interface, and investigation of whether this pathway can be therapeutically targeted warrant investigation.

Although our findings support a protective role for DECTIN-1 in pulmonary antibacterial defence, the consequences of DECTIN-1 signalling are context dependent: studies in tumour models report both beneficial and detrimental immunoregulatory effects depending on tissue microenvironment and cellular composition, so therapeutic modulation of this pathway will require disease-specific consideration.^38,39

The common CLEC7A Y238∗ variant is associated with increased susceptibility to bacterial pneumonia in humans, and impaired pulmonary bacterial clearance in Clec7a-deficient mice, gene-dose-dependent in the pneumococcal model, supports a causal interpretation of DECTIN-1 in respiratory antibacterial defence. The identification of EF-Tu as a conserved bacterial ligand provides a plausible molecular basis for this recognition, extending the role of DECTIN-1 from fungal immunity to antibacterial defence in the general population.

Contributors

KR, OT, AMB, PFZ, HS, KL and LYY conceived and designed the study. OT, VÖ, LYY, APH, Y-CS, CS, and BS performed laboratory investigations and experiments. OT, AM, AG, and LYY did the formal analysis. OT, AM, AG, and CS curated the data. KR, OM, KS, APH, AMB, and KL provided resources. OT and VÖ prepared the figures. OT, KR, VÖ, and LYY wrote the first version of the manuscript. KR, KL, AMB, KS, and OM supervised the work. KR acquired funding. All authors contributed to revising the manuscript and approved the final version for submission. All authors had full access to all the data in the study and had final responsibility for the decision to submit for publication. OT and KR directly accessed and verified the underlying data of the study.

Data sharing statement

Individual participant data from the Malmö Diet and Cancer Study analysed in this study are not publicly available due to the sensitive nature of personal health data and Swedish data protection legislation. Procedures for sharing data, analytical methods, and study materials for the purpose of reproducing the results can be arranged by contacting the Malmö Diet and Cancer Study Steering Committee; details are available at https://www.malmo-kohorter.lu.se/malmo-kost-cancer-mkc. The mass-spectrometry proteomics data generated in this study have been deposited in the PRIDE repository under accession PXD070889 (https://www.ebi.ac.uk/pride/archive/projects/PXD070889) and are publicly available. Recombinant protein constructs, primary mouse experiment data, and analysis code are available from the corresponding author on reasonable request. Requests for data sharing should be directed to the corresponding author at kristian.riesbeck@med.lu.se.

Declaration of interests

OT, Y-CS, and KR are named inventors on a patent application related to EF-Tu-based vaccine delivery. All other authors declare no competing interests.