Inflammatory Endotype-Associated Airway Resistome in Chronic Obstructive Pulmonary Disease

ABSTRACT Antimicrobial resistance is a global concern in chronic respiratory diseases, including chronic obstructive pulmonary disease (COPD). The collection of antibiotic resistance genes or resistome in human airways may underlie the resistance. COPD is heterogeneous, and understanding the airway resistome in relation to patient phenotype and endotype may inform precision antibiotic therapy. Here, we characterized the airway resistome for 94 COPD participants at stable disease. Among all demographic and clinical factors, patient inflammatory endotype was associated with the airway resistome. There were distinct resistome profiles between patients with neutrophilic or eosinophilic inflammation, two primary inflammatory endotypes in COPD. For neutrophil-predominant COPD, the resistome was dominated by multidrug resistance genes. For eosinophil-predominant COPD, the resistome was diverse, with an increased portion of patients showing a macrolide-high resistome. The differential antimicrobial resistance pattern was validated by sputum culture and in vitro antimicrobial susceptibility testing. Ralstonia and Pseudomonas were the top contributors to the neutrophil-associated resistome, whereas Campylobacter and Aggregatibacter contributed most to the eosinophil-associated resistome. Multiomic analyses revealed specific host pathways and inflammatory mediators associated with the resistome. The arachidonic acid metabolic pathway and matrix metallopeptidase 8 (MMP-8) exhibited the strongest associations with the neutrophil-associated resistome, whereas the eosinophil chemotaxis pathway and interleukin-13 (IL-13) showed the greatest associations with the eosinophil-associated resistome. These results highlight a previously unrecognized link between inflammation and the airway resistome and suggest the need for considering patient inflammatory subtype in decision-making about antibiotic use in COPD and broader chronic respiratory diseases. IMPORTANCE Antibiotics are commonly prescribed for both acute and long-term prophylactic treatment in chronic airway disorders, such as chronic obstructive pulmonary disease (COPD), and the rapid growth of antibiotic resistance is alarming globally. The airway harbors a diverse collection of microorganisms known as microbiota, which serve as a reservoir for antibiotic resistance genes or the resistome. A comprehensive understanding of the airway resistome in relation to patient clinical and biological factors may help inform decisions to select appropriate antibiotics for clinical therapies. By deep multiomic profiling and in vitro phenotypic testing, we showed that inflammatory endotype, the underlying pattern of airway inflammation, was most strongly associated with the airway resistome in COPD patients. There were distinct resistome profiles between neutrophil-predominant and eosinophil-predominant COPD that were associated with different bacterial species, host pathways, and inflammatory markers, highlighting the need of considering patient inflammatory status in COPD antibiotic management.

The authors provide a very interesting report on COPD. The manuscript is well organized but some issues need to be addressed before it is ready for publication. Line 110: There is no description or ticks on the Y-axis in the bar chart in Figure 1. It should be corrected. Line 111: Procrustes analysis was used here to show the association between resistance and microbial composition, but there are no figures or tables to support this conclusion. The authors should provide further result demonstration, figures or tables. Line 112-114: Clearly, the authors here found that the association between resistance and demographic and clinical features was not as strong as described. In the PERMANOVA results, only the inflammatory endotype was significant, and it can be observed from Figure 1b that this significance may only be a statistically weak association. The authors need to provide further explanations here. Line 117-120: The conclusions in Figure 1d and Figure S1 do not fully conform to the author's description: the results observed in the two hospitals are not consistent. Line 123-128: Among the participants of the four subtypes, the high Multidrug-high subtypes were the most dominant. However, bacterial isolates from the five eosinophilic participants did not have any multiple drug resistance and all of them were macrolides resistance, so the results could not be mutually verified or even contradictory. Line 134: It is not appropriate to use P<0.1 as the criterion of significance. Actually, P<0.1 should not be used as the criterion for all other test, which will make the analysis results questionable. Commonly we use P<0.01 or P<0.05.

Reviewer #3 (Comments for the Author):
This is well written paper which has a refined description and discussion. The results, organization and writing are all good. But after reading this paper, a few remarks and questions arise that require clarification. Generally, please give transition words or sentence. The whole manuscript is too refined, which is good. However, adding transition words or sentence could help readers understand this paper. L101-103: If all participants had not long-term antibiotic use, why remove 5 participants from analysis. L112-114. Pls add the figures of Procrustes analysis; add the corresponding contents into materials and methods. L123 Sputum from 33 participants was subject to bacterial culture. Is there a standard to select these 33 participants? If so, please add. In addition, how many participants from each endotype were selected? L146 is 'sputum transcriptome' same with the terms of 'host transcriptome'? Pls add more content of discussion if possible.

Supplementary document
Four reagent controls were included for sequencing. All of controls are negative controls? Pls give more details. Material: the author measure Sputum Inflammatory Mediators from a subset of patients. So, how many patients were selected? Why did not measure all samples.

Preparing Revision Guidelines
To submit your modified manuscript, log onto the eJP submission site at https://spectrum.msubmit.net/cgi-bin/main.plex. Go to Author Tasks and click the appropriate manuscript title to begin the revision process. The information that you entered when you first submitted the paper will be displayed. Please update the information as necessary. Here are a few examples of required updates that authors must address: • Point-by-point responses to the issues raised by the reviewers in a file named "Response to Reviewers," NOT IN YOUR COVER LETTER. • Upload a compare copy of the manuscript (without figures) as a "Marked-Up Manuscript" file. • Each figure must be uploaded as a separate file, and any multipanel figures must be assembled into one file. For complete guidelines on revision requirements, please see the journal Submission and Review Process requirements at https://journals.asm.org/journal/Spectrum/submission-review-process. Submissions of a paper that does not conform to Microbiology Spectrum guidelines will delay acceptance of your manuscript. " Please return the manuscript within 60 days; if you cannot complete the modification within this time period, please contact me. If you do not wish to modify the manuscript and prefer to submit it to another journal, please notify me of your decision immediately so that the manuscript may be formally withdrawn from consideration by Microbiology Spectrum.
If your manuscript is accepted for publication, you will be contacted separately about payment when the proofs are issued; please follow the instructions in that e-mail. Arrangements for payment must be made before your article is published. For a complete list of Publication Fees, including supplemental material costs, please visit our website.

Response to Reviewer Comments:
Reviewer #1 (Comments for the Author): General comment: This work by Yi et al performed in-depth investigation of respiratory microbiome, in particular its resistomes. Since a number of respiratory diseases are linked to microbiome differences, and that treatment and resistomes are tightly interacting with each other, their finer scale profiling of resistome with regard to endotypes of COPD is of potential interest and clinical relevance.
The authors overall did a good job in metagenomic analysis and ARG profiling, and found intriguing linkage between endotypes and resistomes. I suggest some more details to be refined to reflect the intrinsic questions remaining to be addressed or discussed in this manuscript.

Response to general comment:
We thank the reviewer for the positive remarks along with the very helpful suggestions for us to improve the manuscript. In response, we have performed an additional analysis to demonstrate the capability of sputum transcriptome in reflecting the host immune and inflammatory status. We have also provided additional contexts to discuss the possible cause and effect relationship between the airway resistome and the inflammatory endotype, and the baseline resistome in healthy individuals in comparison with COPD patients. Below please find our point-by-point response to the reviewer's comments. Response 1: We thank the reviewer for this comment. Sputum induction has been the gold standard clinical tool for assessing airway inflammation for chronic airway diseases (1,2). Quality controlled, induced sputum has been widely used to understand the airway gene expression for patients with COPD(3, 4) and asthma(5-7). Sputum gene signatures have been applied to identify patients with acute exacerbations (7), different inflammatory phenotypes(6) and responsiveness to treatment (8). With respect to whether sputum transcriptome reflects host immune and inflammatory status, in a recent study by Peters et al.(5), the authors used sputum transcriptome analysis to define inflammatory patterns (T2-high and T2-low) for asthmatic patients. By using a co-abundance clustering followed by a gene set enrichment analysis, they identified sets of genes enriched for specific immune cell types and showed that the expression levels of these genes were highly correlated with the actually measured sputum immune cell counts, suggesting technical reliability of sputum transcriptome and its capacity in reflecting host inflammatory status.
To demonstrate the same capacity of our transcriptomic data, we performed an analysis similar to Peters et al. (5). Weighted correlation network analysis (WGCNA) on the sputum transcriptomic profile resulted in 497 co-abundance modules, in which the modules of M439, M198 and M226 was most significantly enriched to marker genes for macrophage, neutrophil and eosinophil, respectively (Author Response Figure 1). Similar as the observation in Peters et al. (5), the expression level of these modules were highly correlated with the sputum differential cell count for the corresponding immune cell types in our study, confirming the capability of our sputum transcriptomic data in recapitulating patient inflammatory status and phenotypes.
These data are added to Figure S3 in the revised manuscript and summarized in the main text which reads (Line 156): "For modules significantly enriched for the marker genes for macrophage, neutrophil and eosinophil(16), their expression levels were highly correlated with sputum differential cell count of the corresponding cell type ( Figure S3), suggesting the capability of host transcriptome in sputum in reflecting patient immune and inflammatory status". Figure S3). Enrichment analysis for immune cell marker genes using sputum transcriptomic data. a) The enrichment of transcriptomic co-abundance modules to marker genes for specific cell types, according to Peters et al. (5). The minus log 10 P-value was shown for the enrichment of each module in each cell type. For modules with P<0.01, they were colored according to the color codes of the modules in WGCNA. b) For modules most enriched for the marker genes for macrophage, neutrophil and eosinophil, their expression levels were highly correlated with sputum differential cell count of the corresponding cell type.

Response 2:
We thank the reviewer for this insightful comment. We agree with the reviewer that it is important to understand the cause and effect relationship between the inflammatory endotype and the airway resistome. Although the current study was not designed to tackle causality, two lines of evidence implicate that the observed association between the inflammatory endotype and the resistome could possibly be a result of the microbiome differences, and prior antibiotic history may have limited impact on the endotype-resistome association, based on current data and literature.
1) Procrustes analysis revealed a significant correlation between the airway resistome and microbiome taxonomic composition in our study (M 2 =0.019, P<0.001), suggesting the resistome was likely inherently shaped by the airway microbiota. In addition, the airway microbiota was reported to be significantly associated with COPD inflammatory endotypes both in previous studies (9) and in our cohort (Adonis R 2 =0.092, P=0.013). Therefore, the varied airway microbiota that encode different sets of antibiotic resistance genes could possibly underlie the association between the resistome and inflammatory endotype.
2) All participants were free of antibiotics over the preceding one month. The prior antibiotic history over the last six months was not associated with the endotype (Fisher's exact test P=0.386) or the resistome (PERMANOVA P=0.439) in our study. Furthermore, no association was found between patient antibiotic history and the airway resistome in a previous metagenomic study across a broader range of chronic respiratory disorders including COPD, asthma and bronchiectasis (10), as well as in another study on COPD resistome using qPCR assays (11). Therefore, based on current data and literature, prior antibiotic history may have limited association with the airway resistome.
Together, these evidence suggests that the observed endotype-resistome association was more likely a result of the underlying microbiota differences, whereas antibiotic history seems to be of less relevance. Whether the inflammatory endotype is a cause or a consequence of the microbiota differences needs to be studied further. Nevertheless, we agree with the reviewer that the medical history remains an important factor impacting the resistome, and its effects warrant further investigation in larger cohorts preferably with different demographic backgrounds and antibiotic history, which we have discussed in the original manuscript (Line 194).
We added below statement to the revised manuscript (Line 173): "Our results suggest that the varied resistome across inflammatory endotypes could be a result of the underlying differences in the resident microbiota, whereas prior antibiotic history may be of less relevance, based on current data and literature".
Comment 3: out of scope but important to discuss: baseline in healthy individuals, how much comparison can be performed?

Response 3:
We thank the reviewer for this suggestion. We agree with the reviewer that, although out of scope for the current study, it is important to assess the baseline resistome in healthy individuals and its difference with that in COPD patients. Mac Aogain et al. profiled the airway resistome for 13 healthy individuals and 41 patients with a broad range of chronic respiratory disorders (COPD, asthma, bronchiectasis). Their results revealed a complex airway resistome with the presence of both a core and discriminatory set of ARGs across health and disease status.
We added a few lines of discussion in reference to this study, which reads (Line 196): "In addition, it is important to assess the baseline resistome in healthy individuals in comparison with COPD patients. A previous study identified both shared and unique ARGs in healthy individuals and patients across a broad range of chronic respiratory disorders, suggesting a complex airway resistome with both a core and discriminatory set of elements".

Reviewer #2 (Comments for the Author):
General comment: The authors provide a very interesting report on COPD. The manuscript is well organized but some issues need to be addressed before it is ready for publication.

Response to general comment:
We greatly appreciate the positive remarks from the reviewer along with very helpful suggestions to improve the manuscript. In response, we have added the figure and description for the Procrustes analysis, updated all results using the P<0.05 cutoff, and provided additional clarifications to more properly reflect the statistical associations, the clinical site differences, and differences between the resistome and culture results. Below please find a point-by-point response to the reviewer's comments.

Comment 1:
Line 110: There is no description or ticks on the Y-axis in the bar chart in Figure 1. It should be corrected.

Response 1:
We thank the reviewer for pointing this out. We have now added the description (relative abundance (%)) and ticks to the Y-axis in Figure 1.

Comment 2:
Line 111: Procrustes analysis was used here to show the association between resistance and microbial composition, but there are no figures or tables to support this conclusion. The authors should provide further result demonstration, figures or tables.

Response 2:
At the reviewer's suggestion, we now presented the figure for the Procrustes analysis between resistome and microbiome taxonomic composition as Figure S1 in the revised manuscript (Author Response Figure 2). The Procrustes analysis was performed based on the coordinate matrices generated by PCoA using Bray-Curtis dissimilarity indices on both datasets, using the procrustes function in R vegan package. We have revised the statement in the manuscript which reads (Line 112): "A significant association was found between the resistome and microbiota composition using Procrustes analysis based on Bray-Curtis dissimilarity indices (M 2 =0.19, P<0.001, Figure S1)".
We have also added a statement of methods for Procrustes analysis to the Methods section in the supplementary document which reads (Supplementary document, Line 109): "Procrustes analysis was performed to assess the correlation between the resistome and microbiota taxonomic composition, based on the coordinate matrices of both datasets generated by principal coordinate analysis using Bray-Curtis dissimilarity indices, using the procrustes function in R vegan package".
Author Response Figure 2 (Manuscript Figure S1). Procrustes analysis based on Bray-Curtis dissimilarity matrices showed a significant correlation between the airway resistome and microbiome taxonomic composition.

Comment 3:
Line 112-114: Clearly, the authors here found that the association between resistance and demographic and clinical features was not as strong as described. In the PERMANOVA results, only the inflammatory endotype was significant, and it can be observed from Figure 1b that this significance may only be a statistically weak association. The authors need to provide further explanations here.

Response 3:
We agree with the reviewer that the association between the resistome and the inflammatory endotype was modest and at the borderline of significance (P=0.043). The relatively small sample size could be a factor for the overall lack of significance between the resistome and clinical features. To more properly reflect the statistics, we have revised our statement which reads:  Table S2)".

Comment 4:
Line 117-120: The conclusions in Figure 1d and Figure S1 do not fully conform to the author's description: the results observed in the two hospitals are not consistent.

Response 4:
We thank the reviewer for pointing this out. We agree with the reviewer that there were differences in the resistome profiles between the two clinical sites. Specifically, there was a greater proportion of multidrug-high resistome in Guangzhou versus Shenzhen participants. We have added a distribution barplot to the original Figure S1 (the new Figure S2) to reflect the site differences. Despite the site differences, the distribution pattern of the resistome subtypes in between different inflammatory endotypes (i.e. the high representation of multidrug-high resistome in neutrophilic COPD patients and MLS-high resistome in eosinophilic COPD patients) was generally similar between the two sites.
We have revised our statement which reads (Line 122): "Despite a relatively greater proportion of multidrug-high resistome in Guangzhou versus Shenzhen participants, the high representation of multidrug-high resistome in neutrophilic COPD patients and MLS-high resistome in eosinophilic COPD patients were observed in both sites ( Figure S2)".

Comment 5:
Line 123-128: Among the participants of the four subtypes, the high Multidrug-high subtypes were the most dominant. However, bacterial isolates from the five eosinophilic participants did not have any multiple drug resistance and all of them were macrolides resistance, so the results could not be mutually verified or even contradictory.

Response 5:
We thank the reviewer for pointing this out. For the five eosinophilic participants that had bacterial culture data, bacterial isolates from three of them were of macrolide resistance and the isolates from the other two participants did not show resistance to the tested panel of antibiotics. This pattern was indeed not fully congruent with the overall resistome pattern that multidrug-high resistome was the most highly abundant resistome subtype among all patients. However, it was partially in agreement with the finding regarding the greater representation of MLS (macrolide-lincosamide-streptogramin)-high resistome in eosinophilic COPD patients than the other patient subgroups, and together suggests that macrolide resistance was likely relevant for eosinophilic COPD patients. The small number of patients with culture data is clearly an important caveat in interpreting the results (n=5 for eosinophilic patients).
We have revised the statement which reads (Line 131): "The results were partially in agreement with an endotype-related antimicrobial resistance pattern, although they should be interpreted with caution given the small sample size".
We have deleted the statement "These different patterns were largely consistent with the in vitro antimicrobial susceptibility testing results, suggesting genotype-phenotype translatability" in the original manuscript, to avoid over-interpretation of the data.

Comment 6:
Line 134: It is not appropriate to use P<0.1 as the criterion of significance. Actually, P<0.1 should not be used as the criterion for all other test, which will make the analysis results questionable. Commonly we use P<0.01 or P<0.05.

Response 6:
We thank the reviewer for this suggestion. We agree with the reviewer that P<0.05 should be used to as threshold for statistical significance according to standard practice. We have updated all results using P<0.05 (or FDR<0.05 when applicable) throughout the manuscript.
Reviewer #3 (Comments for the Author): General comment: This is well written paper which has a refined description and discussion. The results, organization and writing are all good. But after reading this paper, a few remarks and questions arise that require clarification. Generally, please give transition words or sentence. The whole manuscript is too refined, which is good. However, adding transition words or sentence could help readers understand this paper.
Response to general comment: We thank the reviewer for the positive remarks. In response, we have added the figure and description for the Procrustes analysis, and provided additional clarifications regarding the definition of sputum transcriptome and reagent controls, and the rationale for the selection of sputum samples for bacterial culture and measurement of inflammatory mediators. Transition words and sentences were also added accordingly. Please find below a point-by-point response to the reviewer's comments.

Response 2:
At the suggestion of this reviewer and the Reviewer 2, we have now presented the figure for the Procrustes analysis between resistome and microbiome taxonomic composition as Figure S1 in the revised manuscript (Author Response Figure 3). The Procrustes analysis was performed based on the coordinate matrices generated by PCoA using Bray-Curtis dissimilarity indices on both datasets, using the procrustes function in R vegan package. We have revised the statement in the manuscript which reads (Line 112): "A significant association was found between the resistome and microbiota composition using Procrustes analysis based on Bray-Curtis dissimilarity indices (M 2 =0.19, P<0.001, Figure S1)".
We have also added a statement of methods for Procrustes analysis to the Methods section in the supplementary document which reads (Supplementary document, Line 109): "Procrustes analysis was performed to assess the correlation between the resistome and microbiota taxonomic composition, based on the coordinate matrices of both datasets generated by principal coordinate analysis using Bray-Curtis dissimilarity indices, using the procrustes function in R vegan package". Figure 3 (Manuscript Figure S1). Procrustes analysis based on Bray-Curtis dissimilarity matrices showed a significant correlation between the airway resistome and microbiome taxonomic composition.

Author Response
Comment 3: L123 Sputum from 33 participants was subject to bacterial culture. Is there a standard to select these 33 participants? If so, please add. In addition, how many participants from each endotype were selected?

Response 3:
The selection of the 33 participants for bacterial culture was based on sample availability. When an extra sputum was available after sample processing for metagenomics and other omics characterization, it was used for bacterial culture, MALDI-TOF identification, and antimicrobial susceptibility testing in the clinical laboratory. We have revised the statement in the manuscript which reads (Line 126): "For 33 participants with extra sputum available, their sputum was subject to bacterial culture, MALDI-TOF identification, and antimicrobial susceptibility testing in the clinical laboratory".
The number of participants from each endotype with culture data was: NEU=13, EOS=5, Mixed=7, Pauci=8. These numbers were indicated in Figure 1d and Table  S3. Your manuscript has been accepted, and I am forwarding it to the ASM Journals Department for publication. You will be notified when your proofs are ready to be viewed.
The ASM Journals program strives for constant improvement in our submission and publication process. Please tell us how we can improve your experience by taking this quick Author Survey.
As an open-access publication, Spectrum receives no financial support from paid subscriptions and depends on authors' prompt payment of publication fees as soon as their articles are accepted. You will be contacted separately about payment when the proofs are issued; please follow the instructions in that e-mail. Arrangements for payment must be made before your article is published. For a complete list of Publication Fees, including supplemental material costs, please visit our website.
ASM policy requires that data be available to the public upon online posting of the article, so please verify all links to sequence records, if present, and make sure that each number retrieves the full record of the data. If a new accession number is not linked or a link is broken, provide production staff with the correct URL for the record. If the accession numbers for new data are not publicly accessible before the expected online posting of the article, publication of your article may be delayed; please contact the ASM production staff immediately with the expected release date.