Chronic obstructive pulmonary disease (COPD) is most often caused by long-term cigarette smoke (CS) inhalation and includes chronic inflammation, airway remodelling and emphysema leading to progressive lung function impairment which persists after smoking cessation.1 Over 210 million people globally have COPD, causing >3.3 million deaths annually although even these are substantially under-reported.2 Pharmacological treatments have little efficacy in reversing disease, suppressing its progression or preventing mortality and have significant adverse effects.3 Mechanisms of COPD pathogenesis remain poorly understood and no singular driver of disease has been discovered, suggesting that chronic effects from a range of factors drive disease pathogenesis over many years.4 Development of effective therapies will require targeting this array of factors.

The microbiome is under intense investigation for its immunoregulatory capacity and association with disease.5 Most COPD studies have focused on respiratory microbiota, demonstrating increased bioburden, reduced diversity and enrichment of Firmicutes and Proteobacteria.5

The gut hosts the largest and most diverse microbiome of the human body that, depending on its composition, can drive or suppress inflammation, including in the lung.6 Alterations in the gut microbiome from antibiotics or diet influence asthma development and respiratory infections.6 Chronic CS-exposure induces gastrointestinal histopathology,7 and patients with COPD have increased risk of inflammatory bowel diseases6 and altered gut microbiome composition.8

Transfer of whole microbial communities from healthy individuals through faecal microbial transfer (FMT) is an effective therapy in patients with Clostridioides difficile colitis, but its implementation in other diseases is not as well supported.9 FMT may therefore have cost-effective benefits in COPD, but further examination is required. A study using a short-term smoke and poly I:C (TLR3 agonist) model showed that FMT prevented emphysema and identified associations with Bacteroidaceae and Lachnospiraceae families by 16S rRNA gene sequencing,10 but important controls and analyses linking taxa to pathogenesis were lacking and the model was not representative of human CS-induced COPD. Similarly, Parabacteroides goldsteinii lipopolysaccharide (LPS) had protective effects in a murine model of COPD, but broader associations between microbiota and disease were not identified.11 Crucially, neither study identified bacteria associated with COPD in human studies8 or demonstrated human translation.

We hypothesised that gastrointestinal microbiota contribute to COPD pathogenesis, and provide a detailed characterisation of the gastrointestinal microbiome in experimental CS-induced COPD using multiomics. FMT protected against experimental COPD through changes in microbiota that correlated with key disease features. Proteomics and metabolomics implicated restoration of bacterial complex carbohydrate metabolism in the protective effects, supported by interventional studies in experimental COPD and human patients with COPD.

MethodsMice, CS-exposure, microbiome transfer and diet studies

Female C57BL/6 mice (3–5 weeks old) from the University of Newcastle Animal Service Unit (Newcastle, Australia) underwent 3 weeks of baseline microbiome normalisation by transferring soiled bedding and co-housing. Microbiome normalisation was not performed for validation experiments (8 weeks CS).

Mice were exposed to normal room air or CS from 12 3R4F cigarettes (University of Kentucky, Lexington, Kentucky, USA) twice per day, 5 days per week, for 8 or 12 weeks as previously described.12–20 FMT was administered twice per week by transfer of soiled bedding or oral gavage of faecal supernatants from age-matched air-exposed mice to CS-exposed mice and vice versa (online supplemental figure S1). Experimental controls were used as donors for the FMT to ensure the use of age-matched donors which had undergone microbiome normalisation, minimising variability and confounding factors (online supplemental figure S1).

For antibiotic depletion experiments, mice received antibiotics (ampicillin, metronidazole, neomycin, gentamicin (1 g/L) and vancomycin (0.5 g/L)) in drinking water for 7 days.21 Antibiotics were removed and mice received FMT from a separate cohort of air-exposed or CS-exposed donors four times in 10 days. FMT recipient mice were not exposed to CS directly. For diet studies, mice were fed a conventional semi-pure diet (AIN93G) or a resistant starch diet (SF11-025; Specialty Feeds, WA, Australia) ad libitum commencing 2-week prior to CS-exposure and maintained until the end of experiment. Faeces were collected weekly, with at least three fresh faecal pellets collected and stored at −80°C until processing. Airway inflammation, lung and colon histopathology, lung function and gene expression analyses were assessed as previously described.7 12–20 22 Detailed methods are provided in the online supplemental file 1.

All experiments were approved by University of Newcastle Animal Ethics Committee.

Analysis of gut microbiome

Faecal microbiome composition, including metagenomics and 16S rRNA gene sequencing, were analysed as previously described8 with detailed methods provided in the online supplemental file 1. Briefly, α-diversity was calculated using QIIME V.1.8.0,23 principal component analysis performed using the R package vegan V.2.5-124 within metagenomeSeq V.1.22.025 and differential abundance determined using DESeq2 with Benjamini-Hochberg adjustment. Sparse Partial Least-Squares Discriminant Analysis (sPLS-DA) was conducted using the R package mixOmics V.6.3.2.26 Host phenotypes were tested for association with microbiome composition using the envfit function within the vegan R package. Spearman’s rho was calculated using the ‘corr.test’ function within the R package psych V.1.8.12.27 Non-random relationships between taxa were assessed using the CoNET Cytoscape application.28

Cytometry by time-of-flight analysis

For time-of-flight mass cytometry, single cell suspensions of bone marrow, blood and spleen cells were stained with cell cycle marker iododeoxyuridine (IdU), viability marker cisplatin and surface and intracellular antibodies (online supplemental table S19). Expression levels of 38 markers were measured using a Helios instrument (Fluidigm), transformed with a logicle transformation29 and used for Uniform Manifold Approximation and Projection for Dimension Reduction (UMAP) using the naïve R implementation (umap V.0.2.7). Cells clusters were calculated using Rphenograph (Github JinmiaoChenLab V.0.99.1).30 Detailed methods are provided in the online supplemental file 1.

Cell culture

Raw 264.7 monocytes (1×106) were seeded into 12-well plates. Cells were incubated in media alone, or 1:1000 dilutions of sterile-filtered faecal homogenate (100 mg/mL) from CS-exposed or air-exposed mice for 24 hours. During the final 4 hours, LPS (1 µg/mL), monophosphoryl lipid A (4 µg/mL), lipoteichoic acid (1 µg/mL) were added to cell media. Tumour necrosis factor (TNF)-α protein in culture media was assessed using DuoSet ELISA kits (R&D Systems).

Flow cytometry

Colon21 and lung tissue13 18 were digested to produce single cell suspensions and stained for flow cytometry as previously described, before analysis using an LSRFortessa cytometer (BD Biosciences) and FlowJo software (TreeStar). Details are provided in the online supplemental file 1.

Proteomics of mouse faeces

Faecal proteins were analysed by nanoflow liquid chromatography (Ultimate 3000 RSLCnano, Thermo Scientific) with peptides introduced via an Easy-Spray Nano source coupled to a Q-Exactive Plus Quadrupole Orbitrap mass spectrometer (Thermo Scientific) and subjected to data dependent tandem mass spectrometry (MS/MS). Raw files were processed in Proteome Discoverer V.2.1 using the Sequest HT algorithm,31 searched against the mouse gut microbiota GigaDB database32 and integrated with metagenomics results using the psych package in R33 and Diablo mixOmics package. Detailed methods are provided in the online supplemental file 1.

Metabolomics

Metabolomics of caecum contents was performed by Metabolon (Durham, North Carolina, USA) using the Global HD4 MS platform as previously described.8 Detailed methods are provided in the online supplemental file 1.

Human interventional study

Sixteen patients with COPD were recruited from June 2019 to March 2020 at the Respiratory Investigation Unit at The Prince Charles Hospital (TPCH) with written and informed consent. Nine participants were randomised into the intervention arm (inulin 10 g daily for 4 weeks) and seven to the placebo group (maltodextrin 10 g daily for 4 weeks). Age, gender, body mass index (BMI), smoking history (pack years) and spirometry (forced expiratory volume in 1 s/forced vital capacity (FEV1) and FEV1% predicted) were similar between the two groups (table 1).

Table 1

Patient characteristics

Statistical analysis

Except where specified, data was analysed in GraphPad Prism V.9.0 (San Diego, California, USA). Outliers were identified by Grubbs test, and excluded from analysis only if investigators had reported technical errors during sample collection/processing before analysis. Data was analysed by one-way analysis of variance (ANOVA) with Holm-Šídák’s post hoc test or Student’s t-test, or non-parametric equivalents, for data from mice. Data from the patient with COPD cohort was analysed by Mann-Whitney test (continuous variables) or χ2 test (categorical variables).

ResultsFMT alleviated dysbiosis and disease features in experimental COPD

To define changes in microbiome composition and the effects of FMT, C57BL/6 mice underwent microbiome normalisation to control for variability in starting microbiome composition before exposure to mainstream CS through the nose only for 12 weeks (12 wk CS),12–15 18–20 with a subset of mice modelling smoking cessation with 8 weeks of CS-exposure followed by 4 weeks of rest (8 wk CS+4 wk rest). Mice were treated via passive FMT through transfer of soiled bedding from air-exposed mice to CS-exposed mice and vice versa, with controls maintained in their own bedding.

Shotgun metagenomics of faecal samples collected prior to interventions (week 0) and at completion (week 12) recovered 74 metagenome-assembled genomes (MAG)>80% complete (dereplicated at 95% identity) representing 12 families (online supplemental tables 1–4). Microbiome composition was assessed with public genomes within Genome Taxonomy Database release 06-RS202.34 Metagenomic sequencing did not yield adequate sequencing coverage for further investigation of non-bacterial components, with viral signatures detected in <15% of samples and fungal signatures not identified in any samples.

All experimental groups experienced a shift in microbiome composition associated with the maturation of the mice between week 0 and 12 (from 6 to 18 weeks old), with increased α-diversity, enrichment of species belonging to the genus Prevotella and family Muribaculaceae and depletion of Akkermansia and Duncaniella (a member of Muribaculaceae) species (online supplemental figure S2A-C, online supplemental table S5).

While different experimental groups had similar α-diversity at week 12, microbiome composition differed significantly (online supplemental figure S2D,E p=0.01, PERMANOVA of Bray-Curtis distances). CS-exposure was associated with increased Phocaeicola vulgatus, Muribaculaceae species Duncaniella sp001689575 and Amulumruptor sp001689515, Desulfovibrionaceae species Mailhella sp003512875 and Akkermansia muciniphila and decreased Lachnospiraceae species Muricomes sp001517425 (figure 1A,B, online supplemental tables S6, S7). CS-exposed FMT mice maintained increased abundance of most CS-associated species compared with air-exposed mice, except for the Lachnospiraceae member UBA3282 sp009774575 (MAG FTS36, online supplemental table 6). This species, along with CS-associated Mailhella sp003512875 (MAG FTS70), was identified in a multivariate analysis-derived signature distinguishing 12-week CS-exposed mice from both air-exposed and 12-week CS-exposed FMT mice (figure 1C,D, online supplemental table S8). Both species were increased with CS-exposure and decreased with FMT, consistent with a role in disease (figure 1E). Other species contributing to the FMT-associated signal were decreased with CS-exposure and increased with FMT, including obligate anaerobes Duncaniella dubosii, P. goldsteinii, Muribaculum intestinale and Mucispirillum schaedleri (figure 1D, online supplemental table S8). Similar evidence of transfer of anaerobic species to air FMT mice (online supplemental table S6) indicates that bedding swaps were an effective method of FMT, including of anaerobes.

Figure 1

Faecal microbiota transfer (FMT) alleviated dysbiosis and disease features in severe experimental COPD and smoking cessation. Mice were exposed to CS or normal air for 12 weeks, or 8 weeks CS followed by 4 weeks with normal air (8 wk CS+4 wk rest). Mice also received FMT through transfer of soiled bedding or were maintained in their own bedding (Control) for 12 weeks. (A−D) Faecal samples were collected at the end of experiment (week 12) and analysed using shotgun metagenomics. (A) Multivariate analysis (sPLS-DA) of CS-exposed and air-exposed mice at week 12 demonstrating distinction between the groups along component 1 based on relative abundance at the genome level, centred log-ratio transformed and (B) species contributing to that distinction. (C) Multivariate analysis (sPLS-DA) of 12-week CS-exposed control and FMT treated mice, and air-exposed mice, demonstrating distinction between CS-exposed control and FMT mice along component 2 and (D) species contributing to that distinction. (E) Relative abundance of Lachnospiraceae FTS36 and Mailhella sp003512875 FTS70 were increased with CS-exposure, which were alleviated by FMT. (F–N) Hallmark features of COPD assessed at the completion of the experiment (week 12). FMT alleviated CS-induced increases in (F–H) total leucocytes, neutrophils and macrophages in bronchoalveolar lavage fluid (BALF) and (I) parenchymal inflammation in 12-week CS and 8-week CS+4 week rest mice. FMT also alleviated CS-induced increases in (J–K) emphysema-like alveolar enlargement, and (I–N) lung function parameters of compliance, volume and total lung capacity in 12-week CS mice. N=8 per group (A–N). Data presented as mean±SEM. *=p<0.05; **=p<0.01; ***=p<0.001; ****=p<0.0001 using one-way analysis of variance with Holm-Sidak’s post hoc analysis (F–N). COPD, chronic obstructive pulmonary disease; CS, cigarette smoke; sPLS-DA, sparse Partial Least-Squares Discriminant Analysis.

CS-exposure induced lung inflammation characterised by increased total leucocytes, neutrophils and macrophages in bronchoalveolar lavage fluid (BALF) and immune cells in parenchyma (figure 1G,H), as well as emphysematous alveolar destruction (figure 1I,J). BALF macrophages remained elevated after smoking cessation at levels comparable to mice exposed to CS for 12 weeks, while other measures of inflammation and alveolar destruction were present but had reduced severity. Impaired lung function, with increased compliance, volume and total lung capacity, was observed after 12 weeks CS-exposure (figure 1K,M).

FMT-treated mice had significantly lower BALF and parenchymal inflammation in mice exposed to CS for 12 weeks, or with smoking cessation (figure 1E–H). Unlike smoking cessation, FMT reduced BALF macrophages and the combination of smoking cessation and FMT had additive effects by further reducing total leucocytes and parenchymal inflammation. Most importantly, FMT alleviated both emphysema and impaired lung function after CS for 12 weeks (figure 1I–M).

Thus, FMT alleviated hallmark features of COPD and improved the resolution of chronic inflammation with smoking cessation.

FMT alters systemic manifestations of COPD

Given our evidence of CS-induced systemic comorbidities,7 12 we assessed gut manifestations of COPD and systemic leucocyte populations in mice exposed to CS for 12 weeks, with and without FMT. CS-induced increases in colonic submucosal fragmentation and vascularisation, and colon tissue messenger RNA expression of the microbe-sensing pattern recognition receptors Tlr3 and Tlr4 was prevented by FMT, with similar trends for Tlr2 (figure 2A–E). No significant differences were observed in expression of Tlr9 (figure 2F).

Figure 2

Faecal microbial transfer (FMT) alleviated cigarette smoke (CS)-induced colon histopathology and microbial sensor expression changes. (A–F) Mice were exposed to CS or normal air for 12 weeks, and received FMT through transfer of soiled bedding or were maintained in their own bedding (Control). FMT alleviated CS-induced increases in (A) submucosal fragmentation, (B) vascularisation and expression of (C) Tlr2 (non-significant), (D) Tlr3 and (E) Tlr4 in colon tissue. (F) Expression of Tlr9 was not altered by either CS or FMT. N=7–8 per group. Data presented as mean±SEM. *=p<0.05 using one-way analysis of variance with Holm-Sidak’s post hoc analysis. mRNA, messenger RNA.

We employed deep immune profiling using cytometry by time-of-flight with 38 immune cell markers in the bone marrow, blood and spleen. Dimensionality reduction and clustering of bone marrow cells identified 20 clusters with coherent protein expression, but few were altered (figure 3A–C). CS-exposure reduced the abundance of B cells, likely driven by a downregulation of proliferating (IdU+) B cells, but these effects were not alleviated by FMT (figure 3D). In blood, 25 clusters were identified (figure 3E–G). B cells were reduced by CS-exposure and partially restored by FMT (figure 3H). CS-exposure also increased blood Ly6Clo monocytes, and FMT restored their abundance to levels in air-exposed mice (figure 3H). In the spleen, 23 clusters were identified (figure 3I–K) with CD8+ conventional dendritic cells (cDCs) and B cells reduced while CD107+ progenitors and Ly6Clo monocytes were increased by CS-exposure (figure 3L). FMT increased the abundance of CD8+ cDCs, CD107+ progenitors and Ly6Clo monocytes, partially restoring the CS-induced depletion of CD8a+ cDCs but enhancing the impact of CS-induced increases in CD107+ progenitors and Ly6Clo monocytes.

Figure 3

Cigarette smoke (CS)-exposure altered systemic leucocyte populations, and faecal microbial transfer (FMT) alleviated depletion of some blood and splenic populations. (A-I) Mice were exposed to CS or normal air for 12 weeks and received FMT through transfer of soiled bedding or were maintained in their own bedding (Control). Cell abundances were quantified using cytometry by time-of-flight. (A,B) Dimensionality reduction and clustering of bone marrow cells identified 20 clusters of cells (C) based on the scaled mean marker expression. (D) Uniform Manifold Approximation and Projection (UMAP) and a confusion matrix of the normalised frequencies of clusters demonstrated CS-induced depletion of B cells and iododeoxyuridine+ B cells which was not alleviated by FMT. (E,F) Dimensionality reduction and clustering of blood cells identified 20 clusters of cells (G) based on the scaled mean marker expression. (H) UMAP and a confusion matrix of the normalised frequencies of clusters demonstrated a CS-induced depletion of B cells and increase of Ly6Clo monocytes, which were alleviated by FMT. (I,J) Dimensionality reduction and clustering of splenic cells identified 23 clusters of cells (K) based on the scaled mean marker expression. (L) UMAP and a confusion matrix of the normalised frequencies of clusters demonstrated CS-induced depletion of CD8+ conventional dendritic cells (cDCs) and B cells, and a CS-induced increase of CD107+ progenitors and Ly6Clo monocytes. FMT increased the abundance of CD8+ cDCs, CD107+ progenitors and Ly6Clo monocytes compared with CS control mice, partially reducing CD8+ cDCs but enhancing the CS-induced increases in CD107+ progenitors and Ly6Clo monocytes. N=6 per group.

Thus, FMT alleviated CS-induced colonic pathology and TLR expression, as well as CS-induced depletion of blood B cells and splenic CD8+ cDCs and increases in blood Ly6Clo monocytes.

Correlations of lung and gut pathology with microbiota

We assessed the role of the gut microbiome in pathogenesis by fitting phenotypic data to a β-diversity ordination analysis, identifying significant associations with lymphocytes and emphysema consistent with CS-induced separation in microbiome composition (figure 4A). When smoking cessation groups were excluded to assess associations in more severe disease, total leucocytes and neutrophils were also associated with microbiome composition (figure 4B). Correlations between individual species and phenotypes identified positive correlations between Mailhella sp003512875 (FTS70) and inflammation, emphysema, colon vascularisation and Tlr4 expression (figure 4C, online supplemental table S9). Muribaculaceae species Amulumruptor sp001689515 and A. muciniphila positively correlated with lymphocytes and emphysema while A. muciniphila_A correlated with lymphocytes and colon Tlr3 expression. M. intestinale and Muribaculaceae species UBA7173 sp001689685 were negatively correlated with inflammation, emphysema and colon Tlr4 expression indicating a protective role.

Figure 4

Cigarette smoke (CS)-associated bacterial species positively correlate with phenotypic measurements. (A–C) Mice were exposed to CS (12 wk CS) or normal air (Air) for 12 weeks, or 8 weeks CS followed by 4 weeks with normal air (8 wk CS+4 wk rest). Mice also received faecal microbiota transfer (FMT) of soiled bedding or were maintained in their own bedding (Control). (A, B) Non-metric multidimensional scaling (NMDS) ordination plot of Bray-Curtis distances with phenotypic variables significantly associated with CS and/or FMT fitted to axes. Lung compliance was excluded due to colinearity with volume. (A) Lymphocytes and alveolar diameter were significantly associated with microbiome composition (Benjamini-Hochberg, p<0.05). (B) A similar analysis was performed excluding smoking cessation mice (8 wk+4 wk rest), which also identified total leucocytes in bronchoalveolar lavage (BALF) and neutrophils. (C) Spearman’s correlation between phenotypic data and relative abundance of genomes separating CS-exposed mice (red bar) from air-exposed mice or CS-exposed mice receiving FMT (blue bar). Squares marked with an internal asterisk have p<0.05, demonstrating key associations with bacteria. N=8 per group.

CS-induced changes in microbiota impair responsiveness to TLR4 agonists

Given that CS-associated taxa correlated with colonic TLR expression, we assessed the effects of CS-induced dysbiosis on immune responses in vitro. RAW264.7 mouse monocytes were cultured for 24 hours in media alone, or in sterile-filtered homogenates of faeces from mice exposed to CS or air for 12 weeks. During the final 4 hours of incubation, TLR4 agonists LPS and monophosphoryl lipid A (MPLA), or the TLR2 agonist lipoteichoic acid (LTA) were added to directly stimulate TLRs. Faeces from either CS-exposed or air-exposed mice induced TNF-α production compared with media alone, but this was significantly lower with faeces from CS-exposed mice (online supplemental figure S3). LPS and MPLA induced TNF-α production at comparable levels in cells with media or faeces from air-exposed mice, but faeces from CS-exposed mice suppressed LPS-induced and MPLA-induced TNF-α. There were no significant differences in LTA-induced TNF-α production, suggesting that the impairment may be specific to TLR4.

CS-associated microbiota reduce colon innate immune cells but promote lung inflammation

Chronic CS exposure induces gastrointestinal inflammation,7 and the suppression of TLR4 and other innate responses by CS-associated microbiota may provide a competitive advantage in this environment. To further explore the impact of CS-associated microbiota on innate immunity in vivo, mice received antibiotics in drinking water for 7 days to deplete microbial communities. After antibiotics were removed, microbial communities were allowed to recover naturally or were recolonised through oral gavage of faecal homogenates from air or CS-exposed mice. Donors for FMT were mice from separate experiments, and experimental mice in this model received no direct CS exposure. Innate immune cell profiles in the colon and lung were assessed by flow cytometry after 10 days.

There were no changes in innate immune populations when antibiotic-depleted microbiota were recovered naturally or were recolonised with microbiota from air-exposed mice (figure 5A–G; online supplemental figure S4). However, recolonisation with faeces from CS-exposed mice reduced the number of colonic monocytes, immature macrophages (Ly6C+major histocompatibility complex (MHC)-II+) and mature macrophages (Ly6C-MHC-II+), including reduced slow-turnover (T-cell immunoglobulin and mucin domain containing (TIM)4-CD4+) and long-lived, self-renewing tissue resident macrophages (TIM4+CD4+; figure 5A-E). Similarly, colonic neutrophils and eosinophils were also depleted in these mice (figure 5F,G). To identify the impact of gut microbiota on the gut-lung axis in the absence of direct CS exposure, lung granulocyte populations were also assessed. Recolonisation with faeces from CS-exposed mice promoted lung inflammation, increasing interstitial macrophages and eosinophils (figure 5H,I; online supplemental figure S5).

Figure 5

Innate immune cells are reduced in the colon but increased in the lungs after colonisation with microbiota from CS-exposed mice. Microbiota were depleted by antibiotics (ABX) for 7 days before mice were recolonised with faeces from separate air-exposed (Air FMT) or CS-exposed donors (CS FMT) for 10 days. A subset of mice received vehicle (Veh, phosphate-buffered saline+0.05% L-cysteine) to model natural recolonisation. Innate immune cells in the colon and lung were enumerated by flow cytometry. Colon (A) monocytes, (B) immature macrophages and (C) mature macrophages were reduced by CS FMT, including reduction (D) in slow turnover and (E) long-lived, tissue resident macrophages. (F) Colonic neutrophils and (G) eosinophils were also reduced by CS FMT. CS FMT increased (H) lung interstitial macrophages and (I) eosinophils. N=9–11 per group. Data presented as mean±SEM. *=p<0.05; **=p<0.01; ***=p<0.001; using one-way analysis of variance with Holm-Sidak’s post hoc analysis. CS, cigarette smoke; FMT, faecal microbial transfer; MHC, major histocompatibility complex.

Thus, microbiota from CS-exposed mice directly exerts differing effects across the gut-lung axis, with suppressed innate immune responses in the colon but induction of macrophage-driven and eosinophil-driven inflammation in the lungs, even in the absence of CS exposure.

Mutual exclusion contributes to the protective effects of FMT

FMT reduced the abundance of Lachnospiraceae member UBA3282 sp009774575 and Mailhella sp003512875 in CS-exposed mice (figure 1C,D, online supplemental table S8). While FMT-mediated modification of local and systemic immune responses may suppress pathogenic taxa, competition from bacteria introduced through FMT may also contribute to protective effects. To assess the role of competitive inhibition, we identified non-random patterns of co-occurrence in our metagenomics data set using the CoNET Cytoscape application.28

Of the 424 relationships identified, there were 308 co-presence and 116 mutual exclusion relationships (figure 6; online supplemental table S10). Most CS-associated taxa, including P. vulgatus, Duncaniella sp001689575, Amulumruptor sp001689515, Mailhella sp003512875 and A. muciniphila displayed co-present relationships with other CS-associated taxa but no evidence of mutual exclusion with air-associated or FMT-associated taxa. Additionally, Lachnospiraceae member UBA3282 sp009774575 and air-associated Muricomes sp001517425 were not identified in any co-present or mutually exclusive relationships. However, 15 non-random patterns indicating mutual exclusion (13% of all mutual exclusions observed) were identified with Mailhella sp003512875, including with Air/FMT-associated taxa Duncaniella sp001689575, D. dubosii and UBA7173 sp001689685. Thus, while not explaining all protective effects, FMT introduced taxa with which mutual exclusion relationships to Mailhella sp003512875 were observed.

Figure 6

Mailhella sp003512875 (blue ellipse; SB7116_S70_bin_2) has patterns of mutual exclusion with air/FMT-associated taxa. Mice were exposed to CS or normal air for 12 weeks, or 8 weeks CS followed by 4 weeks with normal air (8 wk CS+4 wk rest). Mice also received FMT through transfer of soiled bedding or were maintained in their own bedding (Control) for 12 weeks. Faecal samples were collected at the end of experiment (week 12) analysed using shotgun metagenomics, and non-random patterns of co-occurrence identified using the CoNET Cytoscape application. Patterns of co-occurrence (308, green lines) were more abundant than patterns of mutual exclusion (116, red lines). CS-associated taxa Phocaeicola vulgatus, Duncaniella sp001689575, Amulumruptor sp001689515, Mailhella sp003512875 and Akkermansia muciniphila displayed co-present relationships with other CS-associated taxa but no evidence of mutual exclusion with air-associated or FMT-associated taxa. Lachnospiraceae member UBA3282 sp009774575 was not identified in any patterns of co-present or mutually exclusive relationships. However, Mailhella sp003512875 had 15 non-random patterns indicating mutual exclusion, including with Air/FMT-associated taxa Duncaniella sp001689575, Duncaniella dubosii and UBA7173 sp001689685. N=48 (eight samples per group from six experimental groups).

CS-associated taxa are not introduced during smoking

Bacteria have been identified in CS by both culture-dependent and independent methodologies.35 36 To determine whether CS-associated taxa were introduced to mice directly during CS-exposure, smoke from a 3R4F research cigarette was bubbled through sterile phosphate-buffered saline (PBS). This process was repeated independently with four different cigarettes and independent negative controls. Aliquots were subjected to DNA extraction and analysis by 16S rRNA gene sequencing. After quality filtering and adapter trimming, reads from both CS extract (CSE) and PBS samples were exceptionally low (115.8±20.22 reads vs 329.8±503.6 reads). A total of 55 amplicon sequencing variants (ASVs) were identified, none of which were present in >1 CSE sample while absent from PBS. Due to the low biomass, ASVs were collapsed to genera/family level and relative abundances were broadly similar between CSE and PBS samples, although there was significant inter-sample variability (online supplemental figure S6A,B). Notably, there was no evidence that CS-associated taxa (eg, Duncaniella sp001689575 and Amulumruptor sp001689515, Mailhella sp003512875, A. muciniphila and UBA3282 sp009774575) were enriched in CSE samples. CSE were also cultured on Yeast Casitone Fatty Acids agar under anaerobic conditions,37 but no colonies were observed after 48 hours.

Experimental COPD and FMT were consistently associated with bacterial taxa

To validate our findings and explore temporal changes in the microbiome up until disease onset, mice were exposed to CS for 8 weeks with passive FMT through transfer of soiled bedding. Changes in gut microbiota were profiled weekly using 16S rRNA gene sequencing of fresh faecal samples collected directly from each mouse for the duration of the experiment (online supplemental tables 11-12). Microbiota composition in different groups were distinguishable from week 1 and continued to transition over time (online supplemental figure S7A-C). Multivariate analysis using sPLS-DA revealed CS-associated separation at week 8 (online supplemental figure S7D) driven primarily by five ASVs, including two Muribaculaceae family members which increased in CS-exposed mice (without FMT) from week 3 but only increased in CS-exposed mice receiving FMT after 6 weeks and decreased in subsequent weeks (online supplemental figure S7E,F).

We also explored microbiome changes in gastrointestinal tissues collected at week 8 and observed similar CS-associated separation in colons, driven by the previously identified Muribaculaceae ASVs plus Desulfovibrionaceae and Lachnospiraceae species (online supplemental figure S8A,D). CS-associated separation was less clear in the caecum (online supplemental figure S8B,E) and was not evident in the ileum (online supplemental