Distinctive features of the oropharyngeal microbiome in Inuit of Nunavik and correlations of mild to moderate bronchial obstruction with dysbiosis

Inuit of Nunavik are coping with living conditions that can influence respiratory health. Our objective was to investigate associations between respiratory health in Inuit communities and their airway microbiome. Oropharyngeal samples were collected during the Qanuilirpitaa? 2017 Inuit Health Survey and subjected to metagenomic analyses. Participants were assigned to a bronchial obstruction group or a control group based on their clinical history and their pulmonary function, as monitored by spirometry. The Inuit microbiota composition was found to be distinct from other studied populations. Within the Inuit microbiota, differences in diversity measures tend to distinguish the two groups. Bacterial taxa found to be more abundant in the control group included candidate probiotic strains, while those enriched in the bronchial obstruction group included opportunistic pathogens. Crossing taxa affiliation method and machine learning consolidated our finding of distinct core microbiomes between the two groups. More microbial metabolic pathways were enriched in the control participants and these were often involved in vitamin and anti-inflammatory metabolism, while a link could be established between the enriched pathways in the disease group and inflammation. Overall, our results suggest a link between microbial abundance, interactions and metabolic activities and respiratory health in the Inuit population.

Out of the 1110 Inuit participants in the Qanuilirpitaa? 2017 survey for respiratory health, 125 participants had a forced expiratory volume in one second (FEV 1 ) and forced vital capacity (FVC) ratio below 0.7 that was used to define bronchial obstruction 24 .More than 800 participants had a FEV 1 /FVC ratio above 0.7.We selected 93 participants that met the spirometric criteria for bronchial obstruction.They were matched for age, sex, and geographical location with 105 participants with normal spirometry which was defined as a FEV 1 /FVC ratio > 0.7 and a FVC > 80% predicted (Table S1).The majority (84%) of participants in both groups were smokers.The 198 OP swabs were processed to extract the metagenomes, which were sequenced on a NovaSeq 6000 platform.An average of 104 millions paired-ends reads was obtained per sample.This allowed the detection of 879 taxonomic groups among samples, twenty-six of which were shared by all samples.The most abundant phyla in both the control and mild-to-moderate bronchial obstruction groups were the Firmicutes (70%), the Actinobacteria (18%), the Bacteroidetes (7%) and the Proteobacteria (4%).The three main classes were the Bacilli and Negativicutes, both belonging to the Firmicutes phylum, and the Actinobacteria part of the Actinobacteria phylum (Fig. 1).A total of 509 bacterial species were detected among the 148 bacterial genera detected in the samples (Fig. S1 and Table S2).The genera Streptococcus and Veillonella (both from the Firmicutes phylum), Actinomyces and Rothia (both from the Actinobacteria phylum) and Prevotella (from the Bacteroidetes phylum) were the five most abundant in both groups of participants (Fig. 2 and Fig. S1).

The Inuit oropharyngeal microbiome is distinct from other oropharyngeal microbiomes
The most abundant phyla, classes and genera observed in the respiratory microbiome of Inuit are consistent with other reports which studied the respiratory microbiome in various populations 17,23,25,26 .We further compared the OP microbiome of the Inuit population with other studies that used a similar approach (OP microbiome, Illumina metagenomic sequencing) and with sequencing data readily available.The studies retained involved US and European (Germany) participants studied in the context of (1) schizophrenia 27 , (2) a Neisseria meningitidis outbreak 28 and (3) critically ill COVID-19 patients 29 , as well as their controls.We also included OP samples archived in the Human Microbiome Project (portal.hmpdacc.org),for a total of 327 OP microbiomes external to the current study.Men and women were included in similar proportion and they aged from 18 and over.The Inuit participants aged from 16 and over.More information on these external samples is given in Table S3.The β-diversity of the Inuit OP microbiomes was significantly distinct (p-value = 0.001) and these formed a separate cluster from the US/German samples when visualized by multi-dimensional scaling (Fig. 3A).To exclude bias because of the various diseases, we conducted an analysis with only the control groups of the various studies and found similar cluster separations (Fig. S2) further indicating that the Inuit OP microbiome is indeed distinct.We next investigated the Inuit and US/German OP samples by linear discriminant analysis effect size (LEfSe) at high stringency (LDA = 3.8, p-value < 0.001) to determine the features that most likely explain the distinctive nature of the Inuit OP microbiome.We identified signature taxa for both the Inuit and the US/German groups (Fig. 3B).The Inuit group had a higher abundance of two phyla: Firmicutes (including Veillonella) and Actinobacteria.Two genera, Streptococcus and Lactobacillus from the Bacilli class (Firmicutes phylum) were associated to the Inuit group.Another distinctive feature of this group was the enrichment of two genera belonging to the Actinobacteria phylum, Actinomyces and Bifidobacterium.In contrast, the OP microbiome from the US/German group was enriched in Proteobacteria and notably Gammaproteobacteria and Betaproteobacteria.The genus Neisseria belonging to the Betaproteobacteria class was also highlighted as more abundant in this group.Another class, Bacteroidia, with the two genera Prevotella and Porphyromonas was a considerable feature of US/German group compared to the Inuit group (Fig. 3B).
Similar results were obtained when considering the US and German populations separately.These two populations were distinct from each other and from the Inuit population and clustered separately when visualized by multi-dimensional scaling (Fig. S3A).A SIMPER post-hoc test further revealed that the genera Actinomyces, Bifidobacterium, Lactobacillus, Gemella and Streptococcus, all part of the most discriminant taxa in the LEfSe described above, were among the 9 shared genera that best explained the dissimilarity between the Inuit and US metagenomes, and between the Inuit and German metagenomes.Importantly, none of these shared genera could explain the dissimilarity between the US and German samples.

Comparison between the control and bronchial obstruction groups
We next compared the OP microbiome from the control and bronchial obstruction groups in the Inuit participants to determine whether the microbial community differs according to the presence or absence of bronchial obstruction.We estimated α-diversity to assess the species richness within the OP community of participants from the bronchial obstruction and control groups, and β-diversity to evaluate the similarity of the OP communities between the two groups.No substantial difference in microbial α-diversity was detected between the two groups, whether this was measured by the Chao1, Shannon (Fig. 4A), Simpson or ACE indexes (Fig. S4).Similarly, none of these four α-diversity indexes showed substantial differences between the microbiota derived from participants of the three geographical regions of Nunavik (Table S1 and Fig. S4).However, differences in α-diversity were observed between men and women according to the ACE and Chao1 indexes (Fig. S4C,D), irrespective of their respiratory health status.The α-diversity was also found to decrease as participants were getting older according to the four indices (Fig. S4A-D).According to the abundance-based Bray-Curtis dissimilarity as the distance method, a difference in β-diversity was observed between the control and bronchial www.nature.com/scientificreports/obstruction groups (p-value = 0.009), suggesting differences in microbial composition (Fig. 4B).This β-diversity was not, however, observed when using the Jaccard coefficient (Fig. S5A).Based on this coefficient, there was a significant difference in β-diversity between men and women (p-value = 0.001) (Fig. S5C), as well as between age groups (p-value = 0.028) (Fig. S5D).When looking for taxonomic features responsible for the β-diversity difference between the control and bronchial obstruction groups (Fig. 4B) we found that the Negativicutes, which were among the most abundant bacterial classes in the Inuit group (Fig. 1), were significantly enriched in the control group (p-value < 0.01) (Fig. 5A).This is most likely due to the increased abundance in the controls of the genus Veillonella, and more specifically of the species V. atypica (p-value < 0.01) and V. dispar (p-value < 0.05), as found by LefSe analysis (LDA ≥ 3) (Fig. 5A).The genera Lactobacillus, Megasphaera and Bifidobacterium, along with the species Megasphaera micronuciformis, Actinomyces graevenitzii, Bifidobacterium longum, Lactobacillus gastricus and Lactobacillus kalixensis were also enriched (LDA ≥ 3, p-value < 0.05 or p-value < 0.01) in the control group (Fig. 5A).Overall, those species belong in majority to the Firmicutes phylum.In the bronchial obstruction group, an increased abundance was detected for the species Granulicatella elegans (as well as its associated Granulicatella genus and Carnobacteriaceae family) and Lactobacillus iners (as well as its associated Lactobacillus genus and Lactobacillaceae family), both belonging to the Firmicutes phylum (Fig. 5A).The Streptococcus genus along with 6 species belonging to the Proteobacteria phylum (including two Haemophilus species), were also more abundant in the bronchial obstruction group (Fig. 2 and Fig. 5A).
We next measured the correlation between the species highlighted by our LEfSe analysis.We found that species enriched in the control group are frequently associated together.For example, A. graevenitzii, B. longum, M. micronuciformis, and V. atypica are positively correlated with at least three of the nine species found to also be enriched in the control group (Fig. 5B).Interestingly the presence of V. atypica seems to be negatively correlated  www.nature.com/scientificreports/with at least three species (G.elegans, Neisseria mucosa, and Neisseria sicca) that are enriched in the bronchial obstruction group (Fig. 5B).Species found in the bronchial obstruction group are also more likely to be correlated, with N. mucosa associating with four out of eight bacteria that are enriched in that group (Fig. 5B).N. elongata, found to be enriched in the control group, is a patent exception as it was not positively correlated with any of the nine other species enriched in the control group, but it was in contrast positively correlated with two species found enriched in the bronchial obstruction group (Fig. 5B).A co-occurrence network generated by Sparse Correlations for Compositional data analysis further reinforced our observation on the correlation between Neisseria species, G. elegans and K. denitrificans in the bronchial obstruction group and between Veillonella species, A. graevenitzii and M. micronuciformis in the control group (Fig. S6).

Convergence of biological findings through alternative bioinformatic analyses
While MetaPhlAn3 is widely used for microbiome analysis, comparison with other algorithms is warranted in light of some of reported differences in microbiome analysis that are software-dependent 30,31 .We reinvestigated the whole metagenome data set by using the computational tool mOTUs2.0.This tool highlighted 502 species and the most abundant phyla observed were similar to the analysis done with MetaPhlAn3, with Firmicutes dominating (Fig. S7).Consistent with the MetaPhlAn3 analysis, Streptococcus, Veillonella, Actinomyces, Rothia, and Prevotella were the 5 most abundant genera observed by mOTUs2.0 (Fig. S1).When we carried out a LEfSe analysis at LDA ≥ 3 (p-value < 0.01) the output derived from MetaPhlAn3 and mOTU2 was highly similar, with Veillonella (as a genus) and V. atypica, B. longum, and L. gastricus found to be enriched in the control group (Fig. S8).mOTUs2 also highlighted Atopobium parvum and V. tobetsuensis as enriched in the control group (Fig. S8).The genus Granulicatella was found by both tools as enriched in the bronchial obstruction group (Fig. S8).
We also attempted to discriminate the control group from the bronchial obstruction group by investigating MetaPhlAn3, mOTUs2 and HUMAnN3 outputs with a series of machine learning (ML) algorithms and feature selection methods (see "Material and methods" and Table S4).Our highest classification accuracy (76%) was achieved, after hyperparameter tuning, by the model using mOTUs2 species relative abundance as input, Chi-squared test as feature selection method and decision tree as classification algorithm (Table S4).We next examined the features retained by the best model for each microbiome data type (Supplementary Data File).Interestingly, many of those features were also highlighted in the LEfSe analysis.For instance, the best model using MetaPhlAn3 genus relative abundance as input selected Veillonella, Lactobacillus, Actinomyces, Bifidobacterium and Megasphaera as negative predictors of bronchial obstruction, and Granulicatella as positive predictor of bronchial obstruction (Fig. 7) consistent with our LEfSe analysis (Fig. 5).

Discussion
We tallied a large respiratory health survey of the Inuit population of Nunavik 6 by a comprehensive OP microbiome study.The survey indicated that airway obstruction (17% vs. 12%) and diagnosed COPD (14% vs. 9%) but not diagnosed asthma (3% vs. 15%) were more frequent in Nunavik compared to the rest of Canada (numbers age-adjusted based on the Nunavik population) 6 .The results derived from this recent survey are in line with previous studies 2,5 .These prevalence rates must be considered in the context that participants had to travel to an icebreaker in order to be evaluated for the health survey, possibly reducing the possibility for individuals with severe bronchial obstruction to participate.Yet we had a large group (n = 198) of Nunavimmiut OP swabs with linked spirometry values that provide indication about respiratory health and the presence of bronchial obstruction.
The overall Inuit OP microbiome composition was consistent with other studies where the four predominant phyla found here (Firmicutes, Actinobacteria, Bacteroidetes and Protobacteria) were also reported either in healthy individuals or in people with various respiratory diseases 13,16,23,32 .Streptococcus was the most abundant genus in our samples, a finding consistent with its predominance in oral, bronchial and lung tissues 13 .In addition to Streptococcus, we found that Veillonella, Actinomyces, Rothia and Prevotella made up the 5 most abundant genera.These results are reproducible as they were replicated by two distinct taxonomy profilers (MetaPhlAn3 and mOTUs2).These five genera are indeed considered to be part of the core pulmonary microbiome 33,34 and are thought to be pivotal for a healthy respiratory flora 16,35 .While the most abundant classes and genera in our Inuit group were alike other OP microbiomes, an analysis of similarities revealed that they nonetheless had distinctive features.The Inuit were indeed richer with the phyla Firmicutes and Actinobacteria, while US/German groups www.nature.com/scientificreports/were richer in Bacteroidetes and Proteobacteria.Firmicutes and Actinobacteria are predominantly associated with respiratory health (in contrast to Proteobacteria) and their abundance in the Inuit may reflect the relatively good respiratory health of this population despite high smoking rates 6 .Diet and lifestyle have been shown to influence the gut microbiome 15,19 which is distinct between industrialized, non-industrialized and Nunavik populations 36 .Traditional Inuit foods, which mostly includes meat from hunted species and fish, may affect to some extent the Inuit gut microbiome, which itself may influence the OP microbiome through the gut-lung axis 37 .
It is important to acknowledge that a possible limitation of our comparison of Inuit and non-Inuit OP microbiomes pertains to the sample preparation methods of the different studies involved (Table S3), as sample preparation may introduce biases in taxon levels.However, sample preparation methods most likely had no impact here and should not affect our conclusions.Indeed, samples from different studies are intertwined with Inuit samples in an NMDS plot colored by sample prep kits (Fig. S3B).More importantly, the separation of Inuit samples in the NMDS plot remains when the analysis is restricted to external samples prepared with the same kit as we used (Fig. S3C).The sequencing platform used by the different studies also had no impact (Fig. S3D).
Alterations in microbial richness or composition are usually observed with severe COPD 12,17,26,33 or bronchiectasis 38 .In our analysis we did not find a difference in α-diversity based on the respiratory health of our participants.One possible explanation is that samples from our bronchial obstruction group were derived from people with only mild or moderate bronchial obstruction whose OP microbiomes do not display a significantly different α-diversity 33 .While not the norm, some studies reported no differences in α-diversity between people with and without respiratory disease 8,39,40 .However, we were able to find differences in α-diversity metrics when we stratified groups according to sex, women having a less diverse microbial flora.Diseases of the respiratory systems were found to be higher in Inuit females 2,41,42 .This difference in α-diversity in the OP microbiome between men and women was significant and consistent with studies in other populations 43,44 .The OP microbiome of mid-aged adult and elders were found to be different 45 , a finding consistent with our observation that α-diversity decreases with age.
While no difference in microbial α-diversity could be detected between our two spirometry groups, their microbial composition was significantly different according to β-diversity metrics.Such difference in microbial composition was also observed when considering sex and age.Overall, a variety of genera and species part of the Firmicutes phylum were increased in the control group.Veillonella, a member of the core pulmonary bacterial microbiome 16,33 was also increased in the control group, including all its taxonomic affiliations (family, order, class, phylum).This genus, notably the species V. dispar and V. atypica, stands out as being consistently associated with healthy oral microbiota 35 and preserved lung function 18,34 .Other Firmicutes of the Bacilli (e.g. the Lactobacillus genus) and Negativicutes (e.g. the Megasphaera genus) classes were also increased in the control group.Several Lactobacillus, including L. gastricus, are potential probiotic candidates 46 and M. micronuciformis seems beneficial to health 47 and is often associated with Veillonella 48 .The other phylum increased in the control group was the Actinobacteria (e.g. the genera Actinomyces and Bifidobacterium).Many reports have shown, at least in murine models, that B. longum can protect against lung inflammation and can decrease the severity of airway disorders 49 .These data are meaningful as a similar enrichment was observed when we used the mOTUs2 algorithm instead of MetaPhlAn3.Moreover, through a logistic regression model of machine learning, the same species were predictive of respiratory health.Species found in the control group are also more likely to be correlated together.Neisseria elongata was the exception but it had the lowest LDA score and may thus have been a false positive in the LefSe analysis.
In the bronchial obstruction group, diverse genera and species part of the Proteobacteria phylum (e.g.Kingella, Haemophilus, Neisseria) and the Bacilli class (e.g.Granulicatella) were enriched.Two Haemophilus species were enriched in participants from this group.Our data were also in line with frequent bacterial genera described in COPD patients, for example the Haemophilus, Streptococcus and Neisseria genera associated here to participants with impaired lung function 9,13,14,50 .Similarly, increased levels of Carnobacteriaceae have been associated with respiratory infections in a pneumonia mouse model 51 .Granulicatella elegans, which belongs to this bacterial family, was increased (p-value < 0.01) in our bronchial obstruction group and while part of the normal flora, this bacteria has been shown to induce mediators of inflammation 52 .Notably, the latter was also found to be increased by mOTUs2, a feature associated with bronchial obstruction in our logistic regression machine learning model.Haemophilus spp.and Neisseria spp.were also part of the main features of our machine learning model with the best score.
A number of biochemical pathways related to vitamin metabolism were enriched in the analysis of the OP microbiome of the control group (menaquinone, biotin and the pantothenate containing coenzyme A).The vitamin biosynthetic pathways are in general depleted across groups with various chronic diseases in comparison to healthy controls 53 .Menaquinone (vitamin K2) is proposed to play a distinct role in the prevention of multiple diseases, while biotin is linked to microbial and host metabolism as well as to inflammation 54 .Reduced inflammation mediated by bariatric surgery was associated with increased bacterial biotin 55 .Coenzyme A is an enzymatic co-factor involved in a plethora of metabolic reactions and the increase of one of those enzymes involved in butyrate metabolism is linked to asthma protection 56 .Fermentation pathways (ANAEROFRUCAT-PWY, P461-PWY) were also increased in our healthy controls, with the possibility of producing short chain fatty acids that are linked to anti-inflammatory effects 54 .In the bronchial obstruction group, we detected distinct enriched pathways that are involved in long chain fatty acid biosynthesis and in complex sugar catabolism that could impact negatively on lung function 57 , increasing the risk of respiratory diseases 58 .
In summary, taking advantage of a large survey of respiratory health, we have carried out a detailed OP microbiome analysis of the Inuit population of Nunavik.We found that the Inuit OP microbiome was distinct in comparison to other populations.In the context of respiratory health, we also found differences in bacterial taxonomic groups between healthy controls and the group with mild to moderate bronchial obstruction.Through two different algorithms and further supported by a machine learning strategy, we found that Veillonella and

Ethic statement and sampling
This study was approved by the Research Ethical Committee of Laval University Hospital center, Québec, with project number MP-20-2019-4110 pertaining to the 2017 Qanuilirpitaa? survey.All methods were performed in accordance with the relevant guidelines and regulations.This survey was a broad health survey which include a comprehensive respiratory part into it.Participants were enrolled and sampled in the period between August 19, 2017 and October 5, 2017 in 14 villages from three regions of Nunavik-Hudson Bay, Ungava Bay and Hudson strait.All participants were recruited with informed consent for the use of their anonymized data in research.OP samples were obtained by research staff that rubbed a sterile flocked swab at the back of the subject's throat in accordance with the guidelines from the medical microbiology laboratory of the Centre Hospitalier Universitaire de Québec.The swabs were put in Universal Transport Medium immediately upon sampling and kept at 4 °C for no longer than 30 min prior to be stored at − 80 °C.Swabs were then kept frozen at − 80 °C until further processing.Also, spirometry was performed by experienced respiratory therapists using an EasyOneTM Spirometer (New Diagnostic Design, Andover, MA, USA), and following the protocol from the CanCOLD study 59 .Two lung volumes were measured: forced vital capacity (FVC) and forced expiratory volume in one second (FEV 1 ).All spirometric tracings were reviewed for validity by a respirologist.All participants with at least three acceptable spirometry curves were included in the analysis 60 .Bronchial obstruction was defined by a FEV 1 /FVC ratio below 0.7, a standard value used to compare the lung function of many ethnic groups in different countries 24 .

Extraction and sequencing
Throat swabs were thawed on ice and vortexed for 10 s.DNA extraction from the swabs was performed using the QIAamp DNA Microbiome kit (QIAGEN) according to the manufacturer's instructions.We used a Mini-Beadbeater-24 (BioSpec Products) for mechanical lysis, applying 3500 oscillations per minute 3 times for 2 min with an iced-storage 5-min interval.Lysates were incubated at 56 °C for 30 min.For elution, we used 50 µl of buffer AVE which was eluted twice per column.The extracted DNAs were quantified using a Quantus™ Fluorometer (Promega) and stored at − 20 °C.Next generation sequencing libraries were made using the Nextera DNA Flex Library Prep kit (Illumina) following the manufacturer's instructions.Libraries were quantified using a Quantus™ Fluorometer, and quality was checked with an Agilent 2100 Bioanalyzer with High Sensitivity DNA chips (Agilent).As a control we used mocked swabs and Universal Transport Medium spiked with 10 5 CFU of Pseudomonas aeruginosa (ATCC 27853) and Acinetobacter baumanii (ATTC 19606) that were processed similarly as above.Sequencing of the libraries was done with an Illumina NovaSeq 6000 system, generating approximately 100-million paired-end reads of 150 nucleotides per sample.Sequencing of the mocked library revealed only the spiked bacteria, suggesting that contamination was marginal.

Bioinformatics
Trimming of paired-end sequencing reads was done using Trimmomatic.To quantify the percentage of human DNA contamination in our sample, we aligned the reads to the Genome Reference Consortium Human Build 38 patch release 13 (https:// www.ncbi.nlm.nih.gov/ assem bly/ GCF_ 00000 1405.39/) using Bowtie2 61 .For taxonomic profiling, sequence reads from each sample were mapped against the MetaPhlAn3 clade-specific marker gene database mpa_v30_CHOCOPhlAn_20190 using MetaPhlAn3 30 .For metabolic profiling, we used the HUMAnN3 (v3.0.0.α.1) 30 and the databases uniref90 and mpa_v30_CHOCOPhlAn_201901 30 .All pathways assigned as "unclassified", "unintegrated" or "unidentified" were excluded from subsequent analyses.Taxonomic profiles were also generated using the mOTUs2 profiler version 2.0.0 62 at the species, genus, family, and class taxonomic levels using default parameters.For comparison between Inuit and US/German populations, reads were taken from three databases: 1.Sequence Read Archive-SRA (ncbi.nlm.nih.gov/sra); 2. European Nucleotide Archive-ENA (ebi.ac.uk/ena); 3. Human Microbiome Project (hmpdacc.org).Criteria selection were shotgun sequencing and OP swabs from participants over 16 years of age.These were analyzed using the same MetaPhlAn3 pipeline as for the Inuit samples.Given that we observed a correlation between the number of reads per sample and the number of genera identified, samples with less than 5 genera and with ≥ 90% of unassigned reads were excluded from downstream analyses, to avoid possible biases coming from low sequencing coverage.The correlation between low read counts and the number of genera was no longer observed upon sample exclusion.A total of 327 US/ German samples were kept for analysis.All Inuit samples satisfied the exclusion criteria.

Statistical analysis
Species relative abundances obtained from MetaPhlAn3 were imported into R v4.0.4 and α-diversity analyses were performed with the vegan 2.5-7 (Shannon and Simpson), fossil 0.4.0 (ACE) and breakaway 4.7.3 (Chao1) packages [63][64][65] .We assessed statistical differences of these four measures between more than two groups by Analysis of Variance (ANOVA) and by student's t-test for two-group comparisons.Post-hoc pairwise group comparisons for more than two groups were performed with Tukey-HSD (Honest Significant Differences) tests.For β-diversity, Bray-Curtis dissimilarity and Jaccard index were calculated using the vegan package and PCoA plots were drawn using the ggplot 2 3.3.5 package.Analysis of similarities (ANOSIM) and Permutational ANOVA (PERMANOVA) was used to compare β-diversity measures.Post-hoc comparisons were made with the SIMPER function from the vegan package in R 66 .Significant differences between groups, for both taxonomic affiliation and metabolic pathways, were depicted by LEfSe analysis using Kruskal-Wallis rank sum and Wilcoxon tests

Figure 1 .
Figure 1.Relative abundance of bacterial classes identified with MetaPhlAn3 in the bronchial obstruction and control groups.Classes on the left are in bronchial obstruction group and classes on the right are in the control group.Only the bacterial classes with a relative abundance higher than 1% are depicted.The abundances for the 25 remaining bacterial classes were summed and shown by the 'Others' label.***, p-value < 0.001 with Student's t-test.

Figure 2 .
Figure 2. Heat tree for the prevalence and abundance of taxa in the group of 198 participants.The whole group includes 105 participant's samples with normal spirometry and 93 with evidence of bronchial obstruction.The color code depicts the mean relative abundance of taxa within samples.The prevalence of taxa within the samples group is represented by the size of the nodes.

Figure 3 .
Figure 3. β-diversity metrics and signature taxa of the oropharyngeal microbiome from different populations.(A) Clustering of samples from the Inuit (red) and US/German (green) population based on genus-level taxonomic assignments.Clustering is displayed as the non-metric multidimensional scaling (NMDS) plot of all samples, in which the dissimilarity between samples is calculated as the Bray-Curtis distance.The statistical significance of the clustering pattern in the ordination plot was evaluated using the Permutational ANOVA (PERMANOVA) and Analysis of group Similarities (ANOSIM) tests.(B) The most significantly discriminant taxa between participants from the Inuit (red) and US/German (green) population were identified by linear discriminant analysis effect size (LEfSe).Taxa with a logarithmic LDA score ≥ 3.8 and p-value < 0.001 are shown.p phylum, o order, c class, f family, g genus, s species.

Figure 4 .
Figure 4. α-diversity and β-diversity metrics of the oropharyngeal microbiome.(A) α-diversity measured by the Shannon index for the oropharyngeal microbiomes at the genus level from the bronchial obstruction and the control groups.Each dot denotes the Shannon diversity of a sample.The boxes show inter-quartile ranges for each group with the group's median denoted by a line.No significant difference was observed between the groups by t-test.(B) Clustering of samples from the bronchial obstruction (khaki) and control (blue) groups based on genus-level taxonomic assignments.Clustering is displayed as the non-metric multidimensional scaling (NMDS) plot of all samples, in which the dissimilarity between samples is calculated as the Bray-Curtis distance.The statistical significance of the clustering pattern in the ordination plot was evaluated using the Permutational ANOVA (PERMANOVA) and Analysis of group Similarities (ANOSIM) tests.

Figure 5 .
Figure 5. Signature taxa associated with the bronchial obstruction and control groups.(A) The most significantly discriminant taxa between participants from the bronchial obstruction (red) and control groups (green) were identified by linear discriminant analysis effect size (LEfSe).Taxa with a logarithmic LDA score ≥ 2.0 and p-value < 0.05 (*) or p-value < 0.01 (**) are shown.p phylum, o order, c class, f family, g genus, s species (B) Heatmap of Pearson's correlations for the relative abundance of the most discriminant species highlighted by LEfSe analysis in panel A. The orange shaded palette indicates negative correlation while the purple shaded palette indicates positive correlation between species relative abundance.

Figure 6 .
Figure 6.Signature metabolic pathways associated with the bronchial obstruction and control groups.The most significantly discriminant metabolic pathways between participants from the bronchial obstruction (red) and control groups (green) as identified by linear discriminant analysis effect size (LEfSe).

Figure 7 .
Figure 7. Machine learning models to predict healthy and bronchial obstruction groups.The best model using MetaPhlAn3 genus relative abundance as input was obtained with the features selected by the Wilcoxon-test and a Logistic regression model with an accuracy score of 63.1%.Since the bronchial obstruction group was considered as the positive class, a positive coefficient means an increase and a negative coefficient a decrease in the probability of having obstructive syndrome.The strength of the effect is represented by the magnitude of the coefficient.
other Firmicutes were more abundant in healthy controls, while Proteobacteria and Graniculatella were enriched in the bronchial obstruction group.These associations, which are consistent with other studies in completely different populations, suggest potential microbiological ecological strategies for improving respiratory health.