Disease-associated gut microbiome and metabolome changes in patients with chronic obstructive pulmonary disease

Chronic obstructive pulmonary disease (COPD) is the third commonest cause of death globally, and manifests as a progressive inflammatory lung disease with no curative treatment. The lung microbiome contributes to COPD progression, but the function of the gut microbiome remains unclear. Here we examine the faecal microbiome and metabolome of COPD patients and healthy controls, finding 146 bacterial species differing between the two groups. Several species, including Streptococcus sp000187445, Streptococcus vestibularis and multiple members of the family Lachnospiraceae, also correlate with reduced lung function. Untargeted metabolomics identifies a COPD signature comprising 46% lipid, 20% xenobiotic and 20% amino acid related metabolites. Furthermore, we describe a disease-associated network connecting Streptococcus parasanguinis_B with COPD-associated metabolites, including N-acetylglutamate and its analogue N-carbamoylglutamate. While correlative, our results suggest that the faecal microbiome and metabolome of COPD patients are distinct from those of healthy individuals, and may thus aid in the search for biomarkers for COPD.

metagenomic genome-based mapping counts (raw data underlying figures 2-7) are provided as a Source Data File. The reference human genome used in this study (Homo_sapiens.GRCh38) is available at https://www.ncbi.nlm.nih.gov/assembly/2334371. Reference bacterial genomes are available from https:// www.ncbi.nlm.nih.gov/assembly/. Additional databases used in this study are available as follows: SILVA v132 (https://www.arb-silva.de/no_cache/download/ archive/release_132/Exports/), GTDB 03-RS86 (https://data.ace.uq.edu.au/public/gtdb/data/releases/release86/86.0/) and 04-R89 (https://data.ace.uq.edu.au/ public/gtdb/data/releases/release89/89.0/), dbCAN v6 (http://bcb.unl.edu/dbCAN2/download/Databases/), Pfam r31 (ftp://ftp.ebi.ac.uk/pub/databases/Pfam/ releases/Pfam31.0/), TIGRFAM v15 (ftp://ftp.jcvi.org/pub/data/TIGRFAMs/) and UniProt UniRef100 (ftp://ftp.uniprot.org/pub/databases/uniprot/previous_releases/ release-2017_06/). It is unfortunately not possible for us to perform a statistical calculation to estimate the appropriate number of participants required for our proposed study as we have no human fecal microbiota data in hand (necessary for performing a power test). Published studies looking at fecal microbiota vary widely in sample size and do not indicate how the participant numbers were devised. Published studies investigating differences in fecal microbiota between healthy individuals and those with inflammatory bowel disease (IBD) range widely from only 5 healthy controls and 11 IBD patients (Gophna 2006 J Clin Microbiol 44:4136) to 65 healthy controls and 96 IBD patients (Takaishi 2008 Int J Med Microbiol 298:463). A study investigating the fecal microbiota of a family of 8 persons (e.g. shared environment and diet) with that of a cohort of 155 unrelated persons reported that while each member of the family had a microbiota that was "personalised" and distinct from the community, the constituent species of the microbiota were consistent for each person but the abundances of the individual species populations were variable day to day (Schloss 2014 Microbiome 2:25).Furthermore, recently published study investigating microbial imbalance in IBD utilized only 30 individual fecal samples. The study reported IBD patients showed a less diverse gut microbiome compared to healthy individuals (Alam 2020 Gut pathogens 12,1) Based on the literature, we feel that 50 controls and 50 COPD patients is a good middle ground between the extremes in sample size. No statistical test carried out. As no study of the gastrointestinal microbiome in COPD had been performed, and no preliminary data was available, sample size was selected based upon a review of the literature in the study.
No data was excluded.
Each sample from the study cohort was sequenced using both 16S rRNA gene amplicon and metagenomic sequencing of the same DNA extraction producing overlapping microbiome profiles. Results were then validated using metagenomic sequencing of an independent validation cohort, with 16% of bacterial species identified in the study cohort as significantly altered in association with COPD also identified in the validation cohort.
As this was an observational study, there was no assignment of patients. Analyses were adjusted for covariates (age, sex, BMI) using a linear model.
Extractions were performed by a technician unfamiliar with the project design and were assigned a de-identified sample code prior to sequencing. Blinding was carried out for extractions and followed up with de identified sample code prior to 16S rRNA sequencing and metagenomics. For process of analysis blinding was done for unsupervised analysis for 16S rRNA sequencing and metagenomics data. For other downstream analysis blinding was not done as knowledge of the sample group was essential to the analysis. For metabolomics samples were assigned a unique identifier associated with source identifier only into the Metabolon LIMS system for extractions. The identifier was tracked for analysis process. For analysis unsupervised method blinding was done and for downstream subsequent analysis blinding was not done as knowledge of the sample group was essential to the analysis. Note that full information on the approval of the study protocol must also be provided in the manuscript.

Clinical data Policy information about clinical studies
All manuscripts should comply with the ICMJEguidelines for publication of clinical research and a completedCONSORT checklist must be included with all submissions.

Clinical trial registration
Twenty-eight COPD patients and 29 healthy controls were recruited from John Hunter Hospital, Belmont District Hospital, Newcastle Community Health Centre, Westlakes Community Health centre and Hunter Medical Research Institute (Newcastle, Australia). Participants were attending pulmonary rehabilitation programmes at these sites, with a defined diagnosis of COPD. COPD participants all were >40 years old and had a previous history of smoking. Healthy controls were adults >40 years old with no history of cardiac or respiratory disease, and with normal lung function measured by spirometry (FEV1/FVC ratio >0.7 and FEV1 >80% predicted). There were a greater percentage of females in the HC group (66%) relative the COPD group (54%), and the mean ages were 60.4 years for the HC group (median 62) and 67 for the COPD group (median 68).
For the independent validation cohort, sixteen COPD patients and twenty two healthy individuals were recruited from the thoracic outpatient clinic at The Prince Charles Hospital (Brisbane Australia) and from the general population, respectively. Patients had a defined diagnosis of COPD. For validation cohort, COPD patients were former smokers of !10 years, who are recruited during stability (>4 weeks since an exacerbation). Healthy controls were adults >40 years old with no history of cardiac or respiratory disease. There were a greater percentage of females in the HC group (68.18%) relative the COPD group (43.75%), and the mean ages were 60.72 years for the HC group and 71.44 for the COPD group. Average BMI was 23.15 for the HC group and 29.11 for the COPD group.
For study cohort, twenty-eight COPD patients and 29 healthy controls were recruited from John Hunter Hospital, Belmont District Hospital, Newcastle Community Health Centre, Westlakes Community Centre and Hunter Medical Research Institute (Newcastle, Australia). The participants were approached by the clinical research office. They were invited to take part in the study and provided with an information and consent form. If they consented to participate, they undertook the procedures as outlined in the methods section. With consent, details in regards their disease and relevant clinical history was confirmed from medical records. For healthy controls, the study was advertised to local community groups and volunteers on the Hunter Medical Research Institute registry. If individuals provided consent, they were contacted by the lead researcher who assessed eligibility.
For the validation cohort, sixteen COPD patients and twenty two healthy individuals were recruited from The Prince Charles Hospital outpatient clinics. The eligibility of the subject was assessed by the lead researcher, and confirmed by a respiratory physician, who then referred the patients for the study. The lead researcher then obtained written informed consent from the subject before data or sample collection. For the healthy participants, the study was advertised to local community groups. These participants contacted the lead researcher to participate in the study.
This two-step recruitment process limits the individual bias in the subject recruitment process and impact on the results.
Approval was obtained from the Human Ethics Research Committees of the Hunter New England Local Health District (14/08/20/3.02) and the University of Newcastle (H-2015-0006).
For the validation cohort, ethics approval was obtained from The Prince Charles Hospital Human Research Ethics Committee (HREC/18/QPCH/234) and the University of Queensland (2108001673/HREC/18/QPCH/234).For the validation cohort, written and informed consent was obtained before any data or sample collection.