Documente Academic
Documente Profesional
Documente Cultură
Mahmoud I. Abdel-Aziz, MSc, Paul Brinkman, MSc, Susanne J.H. Vijverberg, PhD,
Anne H. Neerincx, PhD, John H. Riley, PhD, Stewart Bates, PhD, Simone Hashimoto,
MD, PhD, Nazanin Zounemat Kermani, MSc, Kian Fan Chung, MD, DSc, Ratko
Djukanovic, MD, PhD, Sven-Erik Dahlén, MD, PhD, Ian M. Adcock, PhD, Peter H.
Howarth, MD, PhD, Peter J. Sterk, MD, PhD, Aletta D. Kraneveld, PhD, Anke H.
Maitland-van der Zee, PharmD, PhD, on behalf of U-BIOPRED Study Group
PII: S0091-6749(20)30565-0
DOI: https://doi.org/10.1016/j.jaci.2020.04.018
Reference: YMAI 14518
Please cite this article as: Abdel-Aziz MI, Brinkman P, Vijverberg SJH, Neerincx AH, Riley JH, Bates
S, Hashimoto S, Kermani NZ, Chung KF, Djukanovic R, Dahlén S-E, Adcock IM, Howarth PH, Sterk
PJ, Kraneveld AD, Maitland-van der Zee AH, on behalf of the U-BIOPRED Study Group, Sputum
microbiome profiles identify severe asthma phenotypes of relative stability at 12-18 months, Journal of
Allergy and Clinical Immunology (2020), doi: https://doi.org/10.1016/j.jaci.2020.04.018.
This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition
of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of
record. This version will undergo additional copyediting, typesetting and review before it is published
in its final form, but we are providing this version to give early visibility of the article. Please note that,
during the production process, errors may be discovered which could affect the content, and all legal
disclaimers that apply to the journal pertain.
© 2020 Published by Elsevier Inc. on behalf of the American Academy of Allergy, Asthma &
Immunology.
Sputum microbiome profiles identify severe asthma
phenotypes of relative stability at 12-18 months
Airway obstruction
Pathogenic bacteria
More severe
Severe adult asthma patients asthma
Neutrophils
Unsupervised Cluster 2
Macrophages
microbiome-driven
clustering
Baseline 12-18 month Cluster 1
follow-up
Less severe
asthma
Commensal bacteria
3 Mahmoud I. Abdel-Aziz MSc1,2, Paul Brinkman MSc1, Susanne J. H. Vijverberg PhD1, Anne H.
4 Neerincx PhD1, John H. Riley PhD3, Stewart Bates PhD3, Simone Hashimoto MD, PhD1,4, Nazanin
5 Zounemat Kermani MSc5, Kian Fan Chung MD, DSc6, Ratko Djukanovic MD, PhD7, Sven-Erik
6 Dahlén MD, PhD8, Ian M. Adcock PhD6, Peter H. Howarth MD, PhD7, Peter J. Sterk MD, PhD1,
7 Aletta D. Kraneveld PhD9,10, Anke H. Maitland-van der Zee PharmD, PhD1,4, on behalf of the U-
8 BIOPRED Study Group
45 Key messages:
46 • Using unbiased clustering based on sputum microbiome profiles we show that, severe
47 asthma can be stratified into two phenotypic clusters.
48 • These clusters significantly differed by age of asthma onset, patient residential locations,
49 smoking history, percentage of blood and/or sputum neutrophils and macrophages,
50 spirometry and used asthma medications.
51
52
53 Capsule summary
54 Two distinct sputum-microbiome driven clusters of severe asthmatics were revealed, exhibiting
55 relative stability after 12-18 months. An interplay between bacteria and innate immunity was
56 suggested, which may help to design urgently needed precision medicine approaches.
57
58 Keywords:
101 U-BIOPRED has received funding from the Innovative Medicines Initiative (IMI) Joint
102 Undertaking under grant agreement no. 115010, resources of which are composed of financial
103 contributions from the European Union’s Seventh Framework Programme (FP7/2007–2013) and
104 European Federation of Pharmaceutical Industries and Associations (EFPIA) companies’ in-kind
105 contributions (www.imi.europa.eu). The salary of MIA was sponsored by the Egyptian
106 Government PhD Research Scholarships.
108 MIA, PB, SJHV, AHN, SH, NZK, and AK have no conflicts of interest to disclose. JHR, SB and PH are
109 employees by and share-holders of GlaxoSmithKline. RD has received fees for lectures at
110 symposia organized by Novartis, AstraZeneca and TEVA, consultation for TEVA and Novartis as
111 member of advisory boards, and participation in a scientific discussion about asthma organized
112 by GlaxoSmithKline. RD is a co-founder and current consultant, and has shares in Synairgen, a
113 University of Southampton spin out company. SED reports personal fees from AstraZeneca,
114 GlaxoSmithKline, Merck & Co, Novartis, Regeneron, Sanofi & Teva , outside the submitted work.
115 IMA, FC and PJS have received grants from Innovative Medicines Initiative (IMI), during the
116 conduct of the study. AHM has been reimbursed for visiting the ATS by Chiesi, received a fee for
117 participating in advisory boards for Boehringer lngelheim and Astra Zeneca, and received
118 unrestricted research grants from GSK, Chiesi and Boehringer Ingelheim.
119 Abstract
120 Background
123 Objectives
124 To identify severe asthma phenotypes based on sputum microbiome profiles and assess their
125 stability after 12-18 months. Furthermore, to evaluate clusters’ robustness after inclusion of an
127 Methods
128 In this longitudinal multicenter cohort study, sputum samples were collected for microbiome
129 profiling from a subset of the U-BIOPRED adult patient cohort at baseline and after 12-18
130 months of follow-up. Unsupervised hierarchical clustering was performed using the Bray-Curtis
131 β-diversity measure of microbial profiles. For internal validation, partitioning around medoids,
132 consensus cluster distribution, bootstrapping and topological data analysis were applied.
133 Follow-up samples were studied to evaluate within-patient clustering stability in severe
134 asthmatics. Cluster robustness was evaluated by an independent mild-moderate asthma cohort.
135 Results
136 Data were available for 100 severe asthma subjects (median age: 55 yrs, 42% males). Two
1
138 smoking status, residential locations, percentage of blood and/or sputum neutrophils and
139 macrophages, lung spirometry, and concurrent asthma medications (all p-values <.05). Cluster 2
140 patients displayed a commensal-deficient bacterial profile which was associated with worse
141 asthma outcomes compared to cluster 1. Longitudinal clusters revealed high relative stability
142 after 12-18 months in the severe asthmatics. Further inclusion of 24 independent mild-to-
144 Conclusion
145 Unbiased microbiome-driven clustering revealed two distinct robust severe asthma phenotypes,
146 which exhibited relative overtime stability. This suggests that the sputum microbiome may
2
148 Introduction
149 Severe asthma patients represent approximately 5% of the total asthma population.1 Severe
150 asthma places substantial health and cost burdens on patients and healthcare communities.2 It
151 is a heterogeneous disease consisting of multiple phenotypes that show differences in clinical
153 requirements. Better characterization of the severe asthma patient population should
154 eventually lead to more effective tailoring of therapeutic decisions to meet patients’ needs and
156 Asthma phenotyping aims to classify the asthmatic population into subgroups based on clinical
158 mechanisms that drive these phenotypes. Omics technologies utilize high throughput advanced
159 analytical and computational tools to elucidate biological pathways and/or highlighting novel
160 biomarkers that can improve diagnosis and therapeutic decisions.4 Classifying cohorts of
161 asthmatics by omics methods can be done in a supervised or an unsupervised approach. The
162 latter can be considered as “unbiased” since it does not involve any priori assumptions and is,
163 therefore, the preferred option. Applying this principle, the Unbiased BIOmarkers in PREDiction
164 of respiratory disease outcomes (U-BIOPRED) project has published asthma phenotypes driven
166 Over the past decade several studies have investigated the airway microbial dysbiosis in asthma
167 patients8-12 with some focusing on airway microbiome profiles in asthmatics with different
168 inflammatory phenotypes.13-17 Inconsistencies in the reported results between the studies have
3
169 been observed which might hinder direct clinical applicability.18 In addition, most of the studies
170 conducted have used amplicon sequencing of bacterial ribosomal RNA (rRNA) which has a
171 limited ability to identify bacteria at species level, and therefore limits its clinical relevance.
172 Studying the induced sputum microbiome within the context of a large scale multicenter asthma
173 cohort study, such as the U-BIOPRED, using both 16s rRNA sequencing and metagenomics could
174 lead to more conclusive results. In this study, we hypothesized that the sputum microbiome,
175 through its host and environment interaction, can reveal distinct severe asthma phenotypes.
176 Specifically, we aimed to (1) identify severe asthma phenotypes through unsupervised unbiased
177 clustering of sputum microbiome profiles of severe asthmatics, (2) assess within-patient
178 longitudinal stability of the identified clinical clusters after 12-18 months and (3) evaluate the
179 robustness of the clinical clusters by subsequently analyzing the microbiome of patients from an
4
181 Methods
183 The U-BIOPRED project is a multi-center prospective observational pan-European cohort study,
184 comprising 3 asthma sub-cohorts defined by standard clinical criteria: severe non-smoking
185 asthmatics (cohort A), severe previous or current smoking asthmatics (cohort B), mild-to-
186 moderate non-smoking asthmatics (cohort C), described in detail previously.19 All recruited
187 participants provided written informed consents and each study center obtained local medical
188 ethics committee approval. The study was registered under identifier NCT01976767 at
189 ClinicalTrials.gov.
190 The study involved 2 research visits: screening and baseline for the mild-moderate and severe
191 asthma patients and an additional longitudinal visit (12-18 months after baseline) for the severe
192 asthmatics.19 At the baseline and longitudinal visits, several questionnaires and biological
193 measurements were obtained from the recruited participants as described in detail elsewhere.19
194 Participants
195 One hundred severe and 24 mild-moderate U-BIOPRED adult asthmatics from 13 study centers
196 spanning 9 different European countries, provided induced sputum samples at baseline that
197 passed quality-control, with 46 severe asthmatics providing single additional follow-up samples
198 after 12-18 months (Figure E1) . Mild-to-moderate and severe asthmatics were defined
199 according to criteria of the Innovative Medicines Initiative and Global Initiative for Asthma
200 (GINA) guidelines.20, 21 They completed standard asthma control and quality of life
5
201 questionnaires,22, 23 underwent spirometry24 and were assessed for inflammatory biomarkers25,
26
202 and for atopy27 (for details see online supplement). All participants were considered as non-
203 smokers if they had not smoked for at least 12 past months and had <5 pack-year smoking
204 history.
205 Sputum induction, 16s rRNA amplicon sequencing and shotgun metagenomics processing
206 Sputum at baseline and longitudinal visits was induced by inhaling hypertonic (0.9% to 4.5%)
207 saline according to standardized protocols.26, 28 Induced sputum samples were prepared for
208 microbiome and metagenomics profiling, as described in details in the online supplement and
209 elsewhere.29
211 The general data analysis workflow is shown in Figure E2 and is further described below:
213 Cluster benchmarking was based on the analysis performed by Brinkman et al.7 with
214 modifications to suit the microbiome data. To assess the patient variability in the microbiome
215 profiles, the Bray-Curtis β-diversity dissimilarity measure was computed separately on
216 numerical count data of 16s rRNA microbiome operational taxonomic units (OTUs) and
217 metagenomics species. Clustering was then performed using hierarchical ward2 agglomerative
218 clustering on the Bray-Curtis measure.30 The optimum number of clusters was determined
219 based on several indices such as; optimum average silhouettes width,31 total within-cluster sum
220 of square (WCSS),32 and Calinski-Harabasz33 indices. Cluster assignment of the patients was
6
221 internally validated by using partition around the medoids (PAM).34 Agreement in the clustering
222 of patients’ assignments between hierarchical Ward, and PAM clustering was quantified by
223 means of Pearson Chi-Square test or Fisher’s Exact test and Rand index.35 This was also
224 performed to assess if the clustering would differ in 16s rRNA sequencing as compared to
225 metagenomics approach. Clustering was further validated visually using topological data
226 analysis (TDA) with the Ayasdi workbench (version 7.15.0; Ayasdi, Menlo Park, Calif) as reported
227 previously.6, 7 In TDA, visual depiction of the shape of patients’ metagenomics data was
228 performed, where nodes represent patients’ points that are highly similar and connected by
229 edges (lines) to nodes that have data in common. The neighborhood lenses (filter functions) 1
230 and 2 were used that generate a graph of patients’ metagenomics data into two-dimensional
231 space by k-nearest neighbors algorithm. The TDA graph, thus, shows connections of each
232 patient point to its nearest neighbors by only information driven from the metagenome.
233 Cluster-wise stability was evaluated by consensus cluster distribution36 and by resampling the
234 data (1000 iterations) using bootstrapping, jittering and replacement of points by noise with
236 Metagenomics data was used to reveal the bacterial profiles of the identified clusters up to
239 The same clustering protocol was re-performed for the longitudinal samples. Longitudinal
240 cluster migration and stability was assessed by cross-tabulating baseline and longitudinal
7
242 Cluster robustness
243 Robustness of the clustering was evaluated by including mild-moderate asthmatics (U-BIOPRED
244 cohort C) to test whether their inclusion would fit the previous clustering solution or result in
247 Patient cluster distribution according to the inclusion country and season of sample collection
248 was tested using Chi-Square test with Monte Carlo simulation (2000 permutations) and later
250 Differences in patients’ demographic and clinical characteristics between the baseline and
251 longitudinal visits and between the revealed clusters were compared using Wilcoxon signed-
252 rank and McNemar's tests for paired data, and Pearson Chi-Square, Fisher’s exact or Mann-
253 Whitney U tests for independent data as appropriate. Results are considered as significant at
255 Differences in microbiome profiles between the clusters and between baseline and longitudinal
256 visits were compared using a Mann-Whitney U test and Wilcoxon signed-rank test, respectively.
257 P-values for metagenomics species differences were adjusted for multiple testing using
258 Benjamini-Hochberg false discovery rate correction (FDR).38 Results are considered significant at
8
260 A correlation heatmap (spearman and point-biserial correlations) was depicted between
261 bacterial species and asthma clinical characteristics that were found to be statistically significant
263 All analyses were performed using R studio (version 1.1.453) with R software (version 3.5.1)
264 supported with the following packages; phyloseq, vegan, stats, cluster, factoextra,
9
266 Results
267 The baseline and follow-up characteristics for the included participants are summarized in Table
268 1.
271 The 16s rRNA microbiome sequencing identified a total of 2777 OTUs, while the metagenomics
272 approach identified a total of 251 bacterial species. Bray-Curtis beta microbiome diversity
273 suggested 2 optimum clusters as evaluated by multiple indices (Figure E3). Applying hierarchical
274 ward clustering revealed two severe asthmatic groups which were driven only by the
275 microbiome profiles at baseline visit (Figure 1A). Partition around medoids (PAM) clustering also
276 revealed two isolated clusters (Figure 1B). Quantitative assessment of similarity in patients’
277 assignment between hierarchical ward and PAM clustering was performed by means of Pearson
278 Chi-Square test (χ2= 51.85, p < 1 x 10-11) and Rand index (RI= 0.82) suggesting great similarity.
279 The 16s rRNA microbiome generated clusters were highly concordant with the metagenomics
280 generated clusters as indicated by Pearson Chi-Square test (χ2= 63.659, p < 1.48 x 10-15) and
281 Rand index (RI= 0.85). Visual representation by the TDA analysis shows that two patient groups
282 can be driven by the microbiome profiles (Figure 1C) when the metagenomics species were
284 There were no statistically significant associations between the patients’ clusters assignments
285 and the 9 countries from which samples were collected (p-values were 0.110 and 0.229 for
286 Hierarchical and PAM clustering, respectively) or the season of sample collection (p-values were
10
287 0.633 and 0.702 for hierarchical and PAM clustering, respectively). This was visually confirmed
288 by running PCoA on Bray-Curtis dissimilarity measure showing random patient allocations
291 Table 2 shows the demographic and clinical characteristics of the two baseline severe asthma
292 clusters. Cluster 1 represent 75 % (n=75) of the severe asthmatics. More than half of them lived
293 in urban areas. As compared to cluster 2, cluster 1 patients had significantly lower percentages
294 of sputum neutrophils, higher percentages of sputum macrophages, and higher values for FEV1
295 and FEV1/FVC % predicted pre- and post- salbutamol (Figure E5A). In contrast, cluster 2 patients
296 had significantly younger age of asthma onset, were mostly non-smokers (84%), and were more
297 likely to live in suburban areas. In addition, a higher percentage of cluster 2 patients were
300 Cluster 2 patients had lower microbial richness and alpha-diversity indices when compared to
301 cluster 1 (Figure 2). Figure E6 shows the bacterial phylogenetic map of the two baseline clusters.
302 Statistical testing showed that a total of 28 species remained significantly different between the
303 two clusters after FDR correction. All of them were more abundant in cluster 1 as compared to
304 cluster 2 patients (Figure 3 and Figure 4). Those species were related to three dominant phyla;
305 Firmicutes, Bacteriodetes and Actinobacteria, and to Proteobacteria to a lesser extent which
306 comprise the following main genera; Veillonella, Prevotella, Alloprevotella, Streptococcus,
307 Porphyromonas, Rothia, Haemophilus, Neisseria, Megasphaera, and few others. In contrast,
11
308 there was a trend in increased relative abundances of few pathogenic species, such as
310 compared to cluster 1 patients, however these results were not statistically significant (Figure
311 E7). Correlations between individual bacterial species and asthma characteristics were of weak
312 or moderate (r <0.50) strength (Figure E8) and reflected the findings revealed by the clustering.
315 Microbiome data were available for 46 (out of 100) severe asthma patients after 12-18 months
316 from baseline inclusion. Similar to the baseline visit, different indices suggested that the
317 microbiome profiles of severe asthma patients at the longitudinal visit allocated the patients
318 into two main clusters (Figure E9). Hierarchical ward clustering had a great similarity with PAM
319 (Figure E10) as quantified by Fisher’s Exact test (p < 1 x 10-6) and Rand index (RI= 0.875).
320 Demographic and clinical characteristics of the longitudinal sputum microbiome clusters
321 Table 3 shows the demographic and clinical characteristics of the two longitudinal severe
322 asthma clusters. Longitudinal cluster 2 patients were more likely to live in rural areas (55.6%) as
323 compared to cluster 1 patients who were more likely to live in urban areas (70.3%). Similar to
324 the baseline analysis, longitudinal cluster 2 patients had higher percentages of sputum and
325 blood neutrophils, lower percentages of sputum macrophages, and lower values for FEV1 %
326 predicted pre- and post- salbutamol (Figure E5B). In addition, higher percentage of cluster 2
328
12
329 Microbial profiles of the longitudinal severe asthma clusters
330 Longitudinal cluster 2 patients had lower microbial richness and diversity compared to
331 longitudinal cluster 1 patients estimated by multiple indices as shown in Figure E11. Figure E12
332 shows the bacterial phylogenetic map of the two longitudinal clusters. A total of 13 species
333 remained significantly different between the two clusters after FDR correction. All of them were
334 more abundant in cluster 1 compared to cluster 2 patients (Figure E13, and Figure E14). Those
335 species were related to the following main genera; Veillonella, Prevotella, Streptococcus, Rothia,
338 Patients who had both baseline and longitudinal microbiome samples were cross-tabulated to
339 check similarity in clusters assignment between the two visits (Figure 5). Out of 46 patients, 39
341 significant Fisher’s Exact test (p <.01) and relatively high Rand index (RI = 0.74) suggesting
342 relative cluster stability after 12-18 months. No significant differences in microbial richness and
343 diversity (Figure E15) or in species relative abundances after FDR correction between baseline
344 and longitudinal visits have been observed. Figure E16 shows a bar chart of mean percentage in
347 Repeating the baseline clustering including both the data from the severe asthmatics (n=100)
348 and those with mild-moderate asthma (n=24) also resulted in two main clusters (Figure E17); 23
13
349 of 24 patients (95.8%) of the mild-moderate asthmatics were assigned to cluster 1, with only 1
350 mild-moderate asthmatic being assigned to cluster 2 using hierarchical clustering. This patient
351 displayed clinical characteristics like those of severe asthmatics in cluster 2 in respect of young
352 age of asthma onset, sputum neutrophilia, and decreased percentage of sputum macrophages
353 and FEV1 values (Table E1). A cross-table to asses if the severe asthma patients had changed
354 their initial cluster assignment after inclusion of the mild-moderate asthma group is shown in
355 Table E2. High cluster stability is indicated by Pearson Chi-Square test (χ2= 73.397, p < 1 x 10-15)
356 and Rand index (RI= 0.904), suggesting robustness of the clustering model.
14
357 Discussion
358 This is the first study to investigate microbiome-driven phenotypes and their stability overtime
359 in severe asthmatics. Using unbiased clustering based on microbiome profiles we show that,
360 severe asthma can be stratified into two phenotypic clusters. These clusters significantly
361 differed by age of asthma onset, patient residential locations, smoking history, percentage of
362 blood and/or sputum neutrophils and macrophages, spirometry and used asthma medications.
363 At both baseline and longitudinal visits, the severe asthmatics in cluster 2 had worse lung
364 function, with associated blood and/or sputum neutrophilia and decreased sputum
365 macrophages compared to cluster 1 patients. In addition, they were more likely to receive add-
366 on asthma medications, such as theophylline or long acting anticholinergics, possibly indicating
368 The two phenotypic clusters were associated with markedly distinct microbiome profiles; cluster
369 1 had higher bacterial richness and diversity compared to cluster 2, both at baseline and
370 longitudinal visits. Cluster 2 was characterized by a clear deficiency of several bacterial species
372 Haemophilus, Neisseria, and Megasphaera genera. Most of these species are considered as
373 commensals inhabiting the oropharyngeal region and the airways. This microbial dysbiosis was
374 associated with blood and/or sputum neutrophilia, deceased sputum macrophages and worse
375 lung function outcomes at baseline and longitudinal visits. A study by the Severe Asthma
376 Research Program (SARP) has shown that sputum neutrophilia (with or without eosinophilia) is a
377 characteristic of more severe asthma phenotypes.39 In our study, the neutrophilia in cluster 2
378 could be attributed to the presence of either “subclinical infection” or modulation of the airway
15
379 “immune tone” by the microbiota. Thus, deficiency of commensal bacteria leads to increased
380 risk of infections with pathogenic ones, as manifested by increased abundance of Haemophilus
382 bacteria could be responsible for the observed blood and/or sputum neutrophilia. Our results
383 are complementary to another U-BIOPRED study which showed that adult severe asthmatics
384 had lower sputum microbiome α-diversity compared to mild-moderate asthmatics and healthy
385 controls, and this diversity was inversely correlated with sputum neutrophils.29 This was in-line
386 with previously reported studies showing that neutrophilic asthma is characterised by low
387 airway bacterial diversity and high abundances of the Proteobacteria phylum, especially the
388 Moraxella and Haemophilus genera.14-16 In addition, a decrease in the percentage of sputum
389 macrophages may suggest a defective innate immune response with impaired macrophage
390 phagocytosis of these pathogenic bacteria. This finding was in agreement with a previous study
391 that showed severe asthmatics have a reduced macrophages phagocytic capacity for certain
392 pathogenic bacteria such as H. influenzae, compared to non-severe asthmatics and healthy
393 subjects.40
394 A study in COPD patients showed that survivors of 1-year mortality had higher relative
395 abundances of Veillonella compared to non-survivors.41 This was partly in agreement with the
396 finding that the less severe cluster 1 patients in our study had higher relative abundances of
397 Veillonella bacteria among others as compared to cluster 2 patients, which may suggest these
398 bacteria might have protective role in chronic respiratory diseases. Although few species, such
399 as Haemophilus parainfluenzae, can become opportunistic pathogens in some situations, their
400 increased abundances in cluster 1 patients was not associated with asthma severity
16
401 characteristics such as exacerbation frequency. These findings imply that the crosstalk of several
402 species within the airway microbial community and their interplay with innate immunity may
403 provide a better clinical relevance instead of looking at the roles of single/few bacterial species.
404 A striking finding in our study is that approximately 85% of the patients remained cluster stable
405 after 12-18 months. In addition, the bacterial dysbiosis was not related to either current or
406 previous antibiotic intake (ever), suggesting that the microbial profiles of these patients were
407 not a main consequence of short-term intake of antibiotics. Rather, they were probably shaped
408 over a long-time period and might have a genetic background42 and/or resulted from a life-long
410 Tobacco smoking has been reported previously to be associated with sputum microbiome
411 diversity of asthmatic patients, 44 in addition, it has been reported to induce neutrophilia in
412 chronic airway disease including asthma.45, 46 However, in our study, the more severe cluster 2
413 patients were more likely to be non-smokers (84.0%). Since microbial dysbiosis and neutrophilia
414 were more often observed in cluster 2 patients, it seems unlikely that this is mainly attributed to
415 smoking and possibly denoting the interplay between microbial dysbiosis and innate immunity
416 altering the airway immune tone. In another study investigating the airway microbiome in
417 asthmatics, neutrophilia was the strongest predictor for microbiota variance, while smoking was
418 not a predictor14 supporting the findings observed in cluster 2. This warrants further
419 investigation to gain more insight on the trajectories of neutrophilic asthma and possible
421 Cluster 2 patients were more likely to live in suburban or rural areas, in contrast to cluster 1
422 patients who were more likely to live in urban areas. We could speculate that there is relatively
17
423 more traffic in or around suburbs and longer driving times due to commuting to work.
424 Therefore, more exposure to car gases/particulate air pollution may influence the airway
425 microbiome profiles.47 In addition, exposure to fine particulate air pollution can increase
426 neutrophils,48 which may contribute to the neutrophilia we see in cluster 2 patients.
427 The relative cluster stability and robustness we found in our study is an ideal criterion for
428 personalized diagnostic or therapeutic decisions in asthmatics. In addition, the inclusion of mild-
429 moderate asthma group fitted well to our previous clustering scheme where most of this
430 patient group (> 95 %) were assigned to the “milder” asthma cluster except for one patient who
431 also showed close characteristics to the severe asthma cluster 2 patients regarding young age of
432 asthma onset, neutrophilia and low values of FEV1. This suggests that using the criteria of the
433 observed microbiome profiles may enable better diagnosis and/or prediction of more severe
434 phenotypes and hence tailor therapeutic decisions. In addition, the relative stability of the
435 clusters after 12-18 months indicates that management approaches aiming at correcting the
436 pathophysiology of these patients should be sought instead of just symptoms control which is
437 not sufficient. This may include approaches that aim to reshape the lung microbiome, for
438 example using long term pre-, pro- or synbiotics. Moreover, therapeutic strategies directed
439 towards the neutrophilic corticosteroid-resistant asthma phenotype and innate immunity may
440 be tried in these patients, such as macrolide antibiotics,49, 50 low dose theophylline,51 and other
442 Our study has many strengths. First, the analysis protocol is unbiased and its clinical significance
443 is driven only by the microbiome. Second, the pan-European nature of this study makes our
444 results more generalizable and valid than previously reported single country/center studies with
18
445 lower sample sizes. Third, the utilization of both 16s rRNA sequencing and metagenomics
446 further increase the reliability and generalizability to other studies. In addition, the limited, but
447 affordable 16s rRNA method provided highly concordant clustering results to the metagenomics
448 approach, which further extend it potential applicability in clinical practice in sittings where
449 metagenomics could not be performed. Fourth, we internally validated our finding by applying
450 different clustering algorithms that were highly consistent. Although adding an independent
451 mild-moderate asthma cohort is not considered as a validation, its perfect harmonization with
452 the clustering solution prove the robustness and clinical relevance of the clustering technique
453 used. Fifth, shotgun metagenomics allowed us to reveal bacterial associations up to species
454 levels and hence increasing the clinical relevance, unlike most previous studies that used 16s
455 rRNA based-methods (limited in identifying bacterial species). Finally, this study is regarded as
456 one of the first attempts to uncover microbiome-driven phenotypes and elaborate targets for
458 However, there are also limitations. First, we sampled only one airway compartment (induced
459 sputum). Whether other sampling compartments would provide additional microbiome-related
460 information needs to be determined. Second, only two time points were measured over a
461 follow-up design of 12-18 months, which may not be sufficient to adequately assess longitudinal
462 shifts of microbiome clusters in asthmatics. Third, some patients were lost to follow-up which
463 may create further bias in the assessment. However, we are the first study to investigate the
464 microbiome clusters in severe asthmatics which might serve as a basis for future investigations.
465 Finally, using an external cohort outside the U-BIOPRED for validation is still needed to confirm
19
467 In conclusion, our study has shown that microbiome-driven clustering can be used as an
468 unbiased way to detect phenotypes in severe asthmatics. Furthermore, they are relatively
469 stable after 12-18 months of follow-up and demonstrated robustness after inclusion of
470 independent asthma cohort. Asthma patients with the more severe, microbiome-driven
471 phenotype, comprised approximately 25% of the severe asthmatics and need to be considered
472 targets for personalized medicine decisions targeting the airway microbiome.
473 Acknowledgment
474 Some of the drawn objects in the graphical abstract were adapted from Servier Medical Art
20
477 References
478 1. Larsson K, Stallberg B, Lisspers K, Telg G, Johansson G, Thuresson M, et al. Prevalence and
479 management of severe asthma in primary care: an observational cohort study in Sweden
480 (PACEHR). Respir Res 2018; 19:12.
481 2. O'Neill S, Sweeney J, Patterson CC, Menzies-Gow A, Niven R, Mansur AH, et al. The cost of
482 treating severe refractory asthma in the UK: an economic analysis from the British Thoracic
483 Society Difficult Asthma Registry. Thorax 2015; 70:376-8.
484 3. Corren J. Asthma phenotypes and endotypes: an evolving paradigm for classification. Discov Med
485 2013; 15:243-9.
486 4. Wheelock CE, Goss VM, Balgoma D, Nicholas B, Brandsma J, Skipp PJ, et al. Application of 'omics
487 technologies to biomarker discovery in inflammatory lung diseases. Eur Respir J 2013; 42:802-25.
488 5. Kuo CS, Pavlidis S, Loza M, Baribaud F, Rowe A, Pandis I, et al. T-helper cell type 2 (Th2) and non-
489 Th2 molecular phenotypes of asthma using sputum transcriptomics in U-BIOPRED. Eur Respir J
490 2017; 49.
491 6. Schofield JPR, Burg D, Nicholas B, Strazzeri F, Brandsma J, Staykova D, et al. Stratification of
492 asthma phenotypes by airway proteomic signatures. J Allergy Clin Immunol 2019.
493 7. Brinkman P, Wagener AH, Hekking PP, Bansal AT, Maitland-van der Zee AH, Wang Y, et al.
494 Identification and prospective stability of electronic nose (eNose)-derived inflammatory
495 phenotypes in patients with severe asthma. J Allergy Clin Immunol 2018.
496 8. Hilty M, Burke C, Pedro H, Cardenas P, Bush A, Bossley C, et al. Disordered microbial
497 communities in asthmatic airways. PLoS One 2010; 5:e8578.
498 9. Marri PR, Stern DA, Wright AL, Billheimer D, Martinez FD. Asthma-associated differences in
499 microbial composition of induced sputum. J Allergy Clin Immunol 2013; 131:346-52.e1-3.
500 10. Huang YJ, Nariya S, Harris JM, Lynch SV, Choy DF, Arron JR, et al. The airway microbiome in
501 patients with severe asthma: Associations with disease features and severity. J Allergy Clin
502 Immunol 2015; 136:874-84.
503 11. Zhang Q, Cox M, Liang Z, Brinkmann F, Cardenas PA, Duff R, et al. Airway Microbiota in Severe
504 Asthma and Relationship to Asthma Severity and Phenotypes. PLoS ONE 2016; 11:e0152724.
505 12. Lee JJ, Kim SH, Lee MJ, Kim BK, Song WJ, Park HW, et al. Different Upper Airway Microbiome and
506 Their Functional Genes Associated with Asthma in Young Adults and Elderly Individuals. Allergy
507 2018.
508 13. Pang Z, Wang G, Gibson P, Guan X, Zhang W, Zheng R, et al. Airway Microbiome in Different
509 Inflammatory Phenotypes of Asthma: A Cross-Sectional Study in Northeast China. Int. J. Med. Sci.
510 2019; 16:477-85.
511 14. Taylor SL, Leong LEX, Choo JM, Wesselingh S, Yang IA, Upham JW, et al. Inflammatory
512 phenotypes in patients with severe asthma are associated with distinct airway microbiology. J
513 Allergy Clin Immunol 2018; 141:94-103.e15.
514 15. Yang X, Li H, Ma Q, Zhang Q, Wang C. Neutrophilic Asthma Is Associated with Increased Airway
515 Bacterial Burden and Disordered Community Composition. Biomed Res Int 2018; 2018:9230234.
516 16. Simpson JL, Daly J, Baines KJ, Yang IA, Upham JW, Reynolds PN, et al. Airway dysbiosis:
517 Haemophilus influenzae and Tropheryma in poorly controlled asthma. Eur Respir J 2016; 47:792-
518 800.
519 17. Li N, Qiu R, Yang Z, Li J, Chung KF, Zhong N, et al. Sputum microbiota in severe asthma patients:
520 Relationship to eosinophilic inflammation. Respir Med 2017; 131:192-8.
21
521 18. Abdel-Aziz MI, Vijverberg SJH, Neerincx AH, Kraneveld AD, Maitland-van der Zee AH. The
522 crosstalk between microbiome and asthma: Exploring associations and challenges. Clin Exp
523 Allergy 2019; 49:1067-86.
524 19. Shaw DE, Sousa AR, Fowler SJ, Fleming LJ, Roberts G, Corfield J, et al. Clinical and inflammatory
525 characteristics of the European U-BIOPRED adult severe asthma cohort. Eur Respir J 2015;
526 46:1308-21.
527 20. Bel EH, Sousa A, Fleming L, Bush A, Chung KF, Versnel J, et al. Diagnosis and definition of severe
528 refractory asthma: an international consensus statement from the Innovative Medicine Initiative
529 (IMI). Thorax 2011; 66:910-7.
530 21. Bousquet. Global Initiative for Asthma (GINA) and its objectives. Clin Exp Allergy 2000; 30:2-5.
531 22. Juniper EF, Svensson K, Mork AC, Stahl E. Measurement properties and interpretation of three
532 shortened versions of the asthma control questionnaire. Respir Med 2005; 99:553-8.
533 23. Juniper EF, Guyatt GH, Ferrie PJ, Griffith LE. Measuring quality of life in asthma. Am Rev Respir
534 Dis 1993; 147:832-8.
535 24. Miller MR, Hankinson J, Brusasco V, Burgos F, Casaburi R, Coates A, et al. Standardisation of
536 spirometry. Eur Respir J 2005; 26:319-38.
537 25. Weiszhar Z, Horvath I. Induced sputum analysis: step by step. Breathe 2013; 9:300-6.
538 26. Pavord ID, Pizzichini MM, Pizzichini E, Hargreave FE. The use of induced sputum to investigate
539 airway inflammation. Thorax 1997; 52:498-501.
540 27. Siles RI, Hsieh FH. Allergy blood testing: A practical guide for clinicians. Cleve Clin J Med 2011;
541 78:585-92.
542 28. Paggiaro PL, Chanez P, Holz O, Ind PW, Djukanovic R, Maestrelli P, et al. Sputum induction. Eur
543 Respir J Suppl 2002; 37:3s-8s.
544 29. Howarth P, Bates S, Versi A, Riley JH, Loza M, Baribaud F, et al. Sputum Metagenomic Profiling
545 Identifies Altered Airway Microbiome in Severe Asthma Linked to Neutrophilic Airways. In: D92.
546 THE MICROBIOME AND LUNG DISEASE; 2019. p. A7023-A.
547 30. Ward JH. Hierarchical Grouping to Optimize an Objective Function. J Am Stat Assoc 1963; 58:236-
548 44.
549 31. Rousseeuw PJ. Silhouettes: A graphical aid to the interpretation and validation of cluster
550 analysis. J Comput Appl Math 1987; 20:53-65.
551 32. Hartigan JA. Clustering Algorithms: John Wiley \& Sons, Inc.; 1975.
552 33. Caliński T, Harabasz J. A dendrite method for cluster analysis. Commun Stat 1974; 3:1-27.
553 34. Kaufman L, Leonard Kaufman PJR, Rousseeuw PJ. Finding Groups in Data: An Introduction to
554 Cluster Analysis: Wiley; 1990.
555 35. Rand WM. Objective Criteria for the Evaluation of Clustering Methods. J Am Stat Assoc 1971;
556 66:846-50.
557 36. Monti S, Tamayo P, Mesirov J, Golub TJML. Consensus Clustering: A Resampling-Based Method
558 for Class Discovery and Visualization of Gene Expression Microarray Data. 2003; 52:91-118.
559 37. Hennig C. Cluster-wise assessment of cluster stability. Comput Stat Data Anal 2007; 52:258-71.
560 38. Benjamini Y, Hochberg Y. Controlling the False Discovery Rate: A Practical and Powerful
561 Approach to Multiple Testing. J R Stat Soc Series B Stat Methodol 1995; 57:289-300.
562 39. Moore WC, Hastie AT, Li X, Li H, Busse WW, Jarjour NN, et al. Sputum neutrophil counts are
563 associated with more severe asthma phenotypes using cluster analysis. J Allergy Clin Immunol
564 2014; 133:1557-63.e5.
565 40. Liang Z, Zhang Q, Thomas CM, Chana KK, Gibeon D, Barnes PJ, et al. Impaired macrophage
566 phagocytosis of bacteria in severe asthma. Respir Res 2014; 15:72.
22
567 41. Filho FSL, Alotaibi NM, Ngan D, Tam S, Yang J, Hollander Z, et al. Sputum Microbiome Is
568 Associated with 1-Year Mortality after Chronic Obstructive Pulmonary Disease Hospitalizations.
569 Am J Respir Crit Care Med 2019; 199:1205-13.
570 42. Patarcic I, Gelemanovic A, Kirin M, Kolcic I, Theodoratou E, Baillie KJ, et al. The role of host
571 genetic factors in respiratory tract infectious diseases: systematic review, meta-analyses and
572 field synopsis. Sci Rep 2015; 5:16119.
573 43. Renz H, Holt PG, Inouye M, Logan AC, Prescott SL, Sly PD. An exposome perspective: Early-life
574 events and immune development in a changing world. J Allergy Clin Immunol 2017; 140:24-40.
575 44. Munck C, Helby J, Westergaard CG, Porsbjerg C, Backer V, Hansen LH. Smoking Cessation and the
576 Microbiome in Induced Sputum Samples from Cigarette Smoking Asthma Patients. PLoS One
577 2016; 11:e0158622.
578 45. Barnes PJ. Medicine. Neutrophils find smoke attractive. Science 2010; 330:40-1.
579 46. Kobayashi Y, Bossley C, Gupta A, Akashi K, Tsartsali L, Mercado N, et al. Passive smoking impairs
580 histone deacetylase-2 in children with severe asthma. Chest 2014; 145:305-12.
581 47. Mariani J, Favero C, Spinazze A, Cavallo DM, Carugno M, Motta V, et al. Short-term particulate
582 matter exposure influences nasal microbiota in a population of healthy subjects. Environ Res
583 2018; 162:119-26.
584 48. Xu X, Jiang SY, Wang TY, Bai Y, Zhong M, Wang A, et al. Inflammatory response to fine particulate
585 air pollution exposure: neutrophil versus monocyte. PLoS One 2013; 8:e71414.
586 49. Simpson JL, Powell H, Boyle MJ, Scott RJ, Gibson PG. Clarithromycin targets neutrophilic airway
587 inflammation in refractory asthma. Am J Respir Crit Care Med 2008; 177:148-55.
588 50. Brusselle GG, Vanderstichele C, Jordens P, Deman R, Slabbynck H, Ringoet V, et al. Azithromycin
589 for prevention of exacerbations in severe asthma (AZISAST): a multicentre randomised double-
590 blind placebo-controlled trial. Thorax 2013; 68:322-9.
591 51. Barnes PJ. Theophylline. Am J Respir Crit Care Med 2013; 188:901-6.
592 52. Gao H, Ying S, Dai Y. Pathological Roles of Neutrophil-Mediated Inflammation in Asthma and Its
593 Potential for Therapy as a Target. J Immunol Res 2017; 2017:3743048.
594 53. Chung KF. Neutrophilic asthma: a distinct target for treatment? Lancet Respir Med 2016; 4:765-
595 7.
23
596 Tables
597
598 Table 1: Baseline and longitudinal patient characteristics for the severe and mild-moderate asthmatics.
Severe asthmatics Mild-moderate
Characteristic asthmatics
Baseline Longitudinal p-value* Baseline
(n=100) (n=46) (n=24)
Age (years), median (IQR) 55 (46.0- 57 (51.3-63.5) 4.3 x 10-11 40.50 (25.75-
62.0) 51.00)
Age of onset (years), median (IQR) 27 (7-46) 37 (14-49) NA 9 (3-22)
Females (%) 58 (58.0%) 25 (54.3%) NA 11 (45.8%)
BMI (kg/m2), median (IQR) 27.72 (24.67- 27.71 (24.37- NA 24.02 (21.80-
32.45) 31.04) 30.18)
Race (white Caucasian, %) 92 (92.0%) 43 (93.5%) NA 23 (95.8%)
Residential Location (%) NA
• Rural 26 (26.0%) 10 (21.7%) 6 (25%)
• Suburban 25 (25.0%) 8 (17.4%) 8 (33.3%)
• Urban 49 (49.0%) 28 (60.9%) 10 (41.7%)
Atopy (%) 70 (70.0%) 31 (67.4%) NA 23 (95.8%)
Non-smoking patients (%) 66 (66.0%) 30 (65.2%) NA 24 (100%)
Eosinophils % in sputum, median (IQR) 2.75 (0.37- 2.19 (0.76- .838 0.72 (0.21-1.81)
19.27) 17.12)
Neutrophils % in sputum, median 57.98 (39.59- 62.95 (51.19- .096 42.17 (26.16-
(IQR) 81.98) 78.24) 75.18)
Macrophages % in sputum, median 22.82 (10.15- 21.57 (10.01- .068 42.97 (21.62-
(IQR) 39.55) 37.84) 66.69)
Eosinophils % in blood, median (IQR) 3.43 (1.58- 3.50 (1.86- .024 3.55 (1.81-4.46)
6.51) 4.97)
Neutrophils % in blood, median (IQR) 58.48 (53.96- 60.55 (54.88- .190 58.75 (53.35-
67.45) 76.83) 64.82)
FEV1 % predicted pre salbutamol, 62.78 (45.31- 61.42 (49.03- .642 92.00 (87.34-
median (IQR) 74.16) 74.75) 104.45)
FEV1 % predicted post salbutamol, 73.08 (52.77- 72.79 (55.24- .468 103.13 (88.51-
median (IQR) 86.92) 85.61) 114.07)
FEV1/FVC % predicted pre salbutamol, 73.46 (61.03- 75.21 (61.35- .282 89.99 (79.80-
median (IQR) 83.66) 83.33) 96.54)
FEV1/FVC % predicted post 76.49 (65.03- 78.70 (64.03- .461 96.91 (87.42-
salbutamol, median (IQR) 87.28) 85.90) 100.83)
FENO in ppb, median (IQR) 23.50 (13.25- 27.50 (16.90- .708 30.25 (19.50-
45.00) 50.00) 58.13)
Number of exacerbations per year, 2 (1-3) 2 (0-3) .737 0 (0-1)
median (IQR)
ACQ5 score average, median (IQR) 2.3 (1.60- 1.9 (0.95- .589 1.00 (0.45-1.55)
3.20) 3.00)
AQLQ score average, median (IQR) 4.46 (3.50- 4.59 (3.67- .245 5.59 (4.84-6.56)
5.50) 5.60)
24
Current asthma medication used (%):
• ICS 100 (100%) 42 (91.3%) .125 24 (100%)
• SABA 63 (63.0%) 27 (58.7%) .999 19 (79.2%)
• LABA 99 (99.0%) 43 (93.5%) .250 1 (4.2%)
• OCS 87 (87.0%) 14 (30.4%) .070 0 (0.0%)
• Short-acting Anticholinergics 9 (9.0%) 3 (6.5%) .999 1 (4.2%)
• Long-acting Anticholinergics 29 (29.0%) 15 (32.6%) .999 0 (0.0%)
• Leukotriene antagonists 43 (43.0%) 19 (41.3%) .999 0 (0.0%)
• Theophylline 21 (21.0%) 7 (15.2%) .999 0 (0.0%)
Antibiotic usage (%):
• Current intake 18 (18.0%) 7 (15.2%) .500 0 (0.0%)
• Current and previous (ever) 22 (22.0%) 26 (56.5%) <.001 1 (4.2%)
intake
599 Variables described as n (% of n) unless specified, IQR: interquartile range, BMI: body mass index, FEV1:
600 forced expiratory volume in 1 Second, FVC: forced vital capacity, FENO: fraction of exhaled nitric oxide,
601 ACQ5: 5-item asthma control questionnaire, AQLQ: asthma quality of life questionnaire, ICS: inhaled
602 corticosteroids, SABA: short acting beta agonist, LABA: long acting beta agonist, OCS: oral
603 corticosteroids. *p-values were computed for paired differences for only patients who have both
604 baseline and longitudinal sputum samples.
25
605 Table 2: Demographic and clinical characteristics of the baseline clusters.
Characteristics Baseline P-
Cluster 1 (n=75) Cluster 2 (n=25) value
Age (years), median (IQR) 55 (46-62) 57 (49-63) .327
Age of onset (years), median (IQR) 30.50 (14.00- 16.00 (5.00-33.50) .012
48.25)
Females (%) 44 (58.7%) 14 (44.0%) .815
BMI (kg/m2), median (IQR) 27.73 (24.60- 27.47 (24.48- .591
32.61) 30.82)
Race (white Caucasian, %) 70 (93.3%) 22 (88.0%) .409
Residential Location (%) .009
• Rural 21 (28.0%) 5 (20.0%)
• Suburban 13 (17.3%) 12 (48.0%)
• Urban 41 (54.7%) 8 (32.0%)
Atopy (%) 54 (72.0%) 16 (64.0%) .359
Non-smokers (%) 45 (60.0%) 21 (84.0%) .028
Eosinophils % in sputum, median (IQR) 3.81 (0.19-21.92) 2.28 (0.39-7.08) .446
Neutrophils % in sputum, median (IQR) 53.40 (32.40- 86.90 (57.32- <.0001
70.74) 92.73)
Macrophages % in sputum, median (IQR) 26.69 (14.60- 9.21 (3.89-19.49) <.0001
47.30)
Eosinophils % in blood, median (IQR) 3.50 (1.52-6.53) 3.28 (1.74-6.42) .987
Neutrophils % in blood, median (IQR) 58.16 (53.93- 58.80 (53.48- .678
66.54) 70.57)
FEV1 % predicted pre salbutamol, median (IQR) 65.24 (52.38- 47.66 (38.36- .035
74.15) 74.18)
FEV1 % predicted post salbutamol, median (IQR) 74.88 (61.52- 51.44 (42.90- .009
86.94) 87.62)
FEV1/FVC % predicted pre salbutamol, median 74.62 (65.60- 65.99 (51.82- .030
(IQR) 83.97) 77.86)
FEV1/FVC % predicted post salbutamol, median 79.25 (67.02- 68.19 (52.31- .012
(IQR) 89.11) 83.86)
FENO in ppb, median (IQR) 26.25 (12.63- 22.00 (14.00- .328
53.00) 26.75)
Number of exacerbations per year, median (IQR) 2 (1-3) 2 (1-4) .416
ACQ5 score average, median (IQR) 2.40 (1.40-3.20) 2.20 (1.65-3.15) .783
AQLQ score average, median (IQR) 4.53 (3.52-5.53) 4.09 (3.33-5.14) .417
Current asthma medication use (%):
• ICS 75 (100%) 25 (100%) NA
• SABA 45 (60.0%) 18 (72.0%) .282
• LABA 74 (98.7%) 25 (100%) .999
• OCS 33 (44.0%) 14 (56.0%) .298
• Short-acting Anticholinergics 8 (10.7%) 1 (4.0%) .444
• Long-acting Anticholinergics 20 (26.7%) 9 (36.0%) .373
• Leukotriene antagonists 32 (42.7%) 11 (44.0%) .907
• Theophylline 11 (14.7%) 10 (40.0%) .007
26
OCS normalized dose (mg) 10.00 (8.44-16.25) 10.00 (5.00-14.38) .363
(n=33) (n=14)
Antibiotic usage (%):
• Current intake 14 (18.7%) 4 (16.0%) .999
• Current and previous (ever) intake 18 (24%) 4 (16.0%) .403
606 Variables described as n (% of n) unless specified, IQR: interquartile range, BMI: body mass index, FEV1:
607 forced expiratory volume in 1 Second, FVC: forced vital capacity, FENO: fraction of exhaled nitric oxide,
608 ACQ5: 5-item asthma control questionnaire, AQLQ: asthma quality of life questionnaire, ICS: inhaled
609 corticosteroids, SABA: short acting beta agonist, LABA: long acting beta agonist, OCS: oral
610 corticosteroids. P-values for testing statistical significance between the two longitudinal clusters were
611 calculated using Pearson Chi-Square or Fisher’s Exact tests as appropriate for categorical variables and
612 Mann-Whitney U test for continuous variables.
27
613 Table 3: Demographic and clinical characteristics of the longitudinal clusters.
Characteristics Longitudinal P-
Cluster 1 (n=37) Cluster 2 (n=9) value
Age (years), median (IQR) 57 (48-64) 55 (53-65) .957
Age of onset (years), median (IQR) 37.00 (17.50- 29.50 (3.50-49.75) .716
48.50)
Females (%) 22 (59.5%) 3 (33.3%) .264
BMI (kg/m2), median (IQR) 27.96 (24.25- 29.56 (24.83- .744
31.87) 30.36)
Race (white Caucasian, %) 35 (94.6%) 8 (88.9%) .488
Residential Location (%) .012
• Rural 5 (13.5%) 5 (55.6%)
• Suburban 6 (16.2%) 2 (22.2%)
• Urban 26 (70.3%) 2 (22.2%)
Atopy (%) 25 (67.6%) 6 (66.7%) .872
Non-smokers (%) 23 (62.2%) 7 (77.8%) .463
Eosinophils % in sputum, median (IQR) 2.65 (0.88-16.45) 0.99 (0.51-18.69) .651
Neutrophils % in sputum, median (IQR) 59.40 (45.03- 83.76 (72.58- .002
72.64) 95.66)
Macrophages % in sputum, median (IQR) 27.24 (15.23- 5.68 (1.37-12.45) <.0001
38.99)
Eosinophils % in blood, median (IQR) 3.65 (2.07-5.29) 2.07 (0.81-4.13) .103
Neutrophils % in blood, median (IQR) 59.81 (53.72- 76.61 (65.31- .023
65.82) 82.98)
FEV1 % predicted pre salbutamol, median (IQR) 63.76 (50.55- 56.17 (40.00- .036
77.23) 62.28)
FEV1 % predicted post salbutamol, median (IQR) 74.33 (63.23- 64.22 (44.64- .036
86.27) 78.08)
FEV1/FVC % predicted pre salbutamol, median 77.35 (62.08- 65.50 (51.22- .116
(IQR) 83.38) 76.02)
FEV1/FVC % predicted post salbutamol, median 80.40 (67.68- 65.04 (50.82- .081
(IQR) 86.22) 78.93)
FENO in ppb, median (IQR) 25.00 (15.90- 28.25 (19.00- .999
54.50) 35.88)
Number of exacerbations per year, median (IQR) 2 (0-3) 2 (0-3) .663
ACQ5 score average, median (IQR) 2.0 (0.9-3.1) 1.8 (1.0-2.7) .957
AQLQ score average, median (IQR) 4.59 (3.86-5.53) 4.48 (3.44-6.09) .807
Current medication use (%):
• ICS 33 (89.2%) 9 (100%) .571
• SABA 20 (54.1%) 7 (77.8%) .270
• LABA 34 (91.9%) 9 (100%) .999
• OCS 10 (27.0%) 4 (44.4%) .423
• Short-acting Anticholinergics 2 (5.4%) 1 (11.1%) .488
• Long-acting Anticholinergics 7 (18.9%) 8 (88.9%) <.001
• Leukotriene antagonists 15 (40.5%) 4 (44.4%) .999
• Theophylline 4 (10.8%) 3 (33.3%) .124
28
OCS normalized dose (mg) 10.00 (8.13-12.50) 13.75 (5.63-20.00) .999
(n=10) (n=4)
Antibiotics usage (%):
• Current intake 4 (10.8%) 3 (33.3%) .124
• Current and previous (ever) intake 20 (54.1%) 6 (66.7%) .711
614 Variables described as n (% of n) unless specified, IQR: interquartile range, BMI: body mass index, FEV1:
615 forced expiratory volume in 1 Second, FVC: forced vital capacity, FENO: fraction of exhaled nitric oxide,
616 ACQ5: 5-item asthma control questionnaire, AQLQ: asthma quality of life questionnaire, ICS: inhaled
617 corticosteroids, SABA: short acting beta agonist, LABA: long acting beta agonist, OCS: oral
618 corticosteroids. P-values for testing statistical significance between the two longitudinal clusters were
619 calculated using Pearson Chi-Square or Fisher’s Exact tests as appropriate for categorical variables and
620 Mann-Whitney U test for continuous variables.
29
621 Figure Legends
622 Figure 1: A; Hierarchical cluster dendogram in a tree-like structure, where patients’ nodes (leaves at the
623 bottom of the dendogram) that are statistically closely connected are joined together by edges (small
624 branches). The small branches are further joined up by larger branches (bottom-up), to the upper part of
625 the dendogram representing the two main branches (clusters) originating from the severe asthma
626 cohort. B; partition around the medoids (PAM) clusering show two relatively detached ellipses. Similarity
627 in patients’ assignment between the two clustering algorithms was assessed by Pearson Chi-Square test
628 (χ2= 51.85, p << 1 x 10-11) and Rand index (RI= 0.802) suggesting great similarity. Bootstrapping, jittering
629 and replacement of points by noise schemes (1000 iterations) resulted in Jaccard similarity indices
630 ranged from 0.82-1 for both clusters by either hierarchical clustering or PAM suggesting highly stable
631 clusters. C; Topological data analysis (TDA) graph, where nodes were colored in accordance with baseline
632 hierarchical clustering of patients. Two distinct patient clusters based on the metagenome profile were
633 observed, where blue nodes represent cluster 1 patients, while red nodes represent cluster 2 patients.
634 Yellow nodes represent less matched patient cluster assignment by TDA compared to the hierarchical
635 clustering.
636 Figure 2: Upper panel; Venn diagram represents the metagenomics species distribution between the two
637 clusters. Lower panel; different alpha diversity measures reveal that cluster 2 patients had much lower
638 microbial diversity compared to cluster 1 patients (all p-values < .0001).
639 Figure 3: Metagenomics phylogenetic map shows statistically significant differential bacterial taxa
640 between the two baseline clusters after false discovery rate (FDR) correction. Nodes color corresponds
641 to the median difference in relative abundances of the bacterial taxa. The darker the color of the
642 phylogenetic branches, the higher median differences, while grey nodes and branches indicate no
643 significant differences. The magenta color indicates that all significant taxa were more abundant in
644 cluster 1 compared to cluster 2, while absence of cyan color in the phylogenetic map indicate that no
645 significant taxa were more abundant in cluster 2 as compared to cluster 1.
646 Figure 4: Statistically significant differentially abundant species after false discovery rate (FDR) correction
647 between baseline cluster 1 and cluster 2 patients by metagenomics. All of them were more abundant in
648 cluster 1 compared to cluster 2 patients.
649 Figure 5: Cross-tabulated and Sankey diagram of patients’ clusters assignments among 46 patients with
650 both baseline and follow-up visits: 39 patients are cluster stable (in boldface), and 7 patients migrate
651 between clusters. Cells described as n (% of baseline clusters n). p-value was generated using Fisher’s
652 exact test.
653
30
A Hierarchical cluster dendogram C Topological data analysis (TDA)
Cluster 2
Baseline clusters
B Partition around the medoids (PAM)
Cluster 1
Cluster 1
Cluster 2
4 53 144
Alpha diversity measures
Baseline clusters
Neisseria_unclassified
Neisseria_sicca Haemophilus_influenzae
Neisseria_elongata Haemophilus_sputorum
Neisseria_polysaccharea Haemophilus_parainfluenzae
Neisseria_flavescens
Kingella_oralis Haemophilus_paraphrohaemolyticus
Neisseria_bacilliformis Neisseria_sp_oral_taxon_014
Haemophilus
Polaromonas_unclassified Kingella_unclassified Neisseria_cinerea Haemophilus_parahaemolyticus
Fusobacterium_necrophorum Neisseria Haemophilus_haemolyticus Aggregatibacter_unclassified
Aggregatibacter_segnis
Neisseria_meningitidis Neisseria_lactamica
Fusobacterium_nucleatum Kingella
Fusobacterium_periodonticum Variovorax_unclassified Aggregatibacter
Polaromonas Neisseria_subflava Haemophilus_pittmaniae Actinobacillus_unclassified
Veillonella_dispar Aggregatibacter_aphrophilus
Neisseria_macacae Neisseria_gonorrhoeae Pantoea_agglomerans
Veillonella_unclassified Fusobacterium Kingella_denitrificans Actinobacillus
Veillonella_atypica Eggerthia_catenaformis Variovorax Pantoea_unclassified
Alicycliphilus_denitrificans Pasteurellaceae
Neisseriaceae Simonsiella_muelleri
Moraxella_catarrhalis Cardiobacterium_hominis
Veillonella Eggerthia Leptotrichia_unclassified
Alicycliphilus
Simonsiella Acinetobacter_lwoffii Pantoea
Veillonella_parvula
Solobacterium_moorei Comamonadaceae Acinetobacter_unclassified
Comamonas Pseudomonas_unclassified
Moraxella Escherichia_unclassified
Selenomonas_infelix Eikenella
Solobacterium Lautropia_mirabilis Acinetobacter Cardiobacterium_valvarum Agrobacterium_unclassified
Leptotrichia_shahii
Selenomonas_flueggei Megasphaera_micronuciformis Leptotrichia Comamonas_unclassified Pseudomonas_aeruginosa
Selenomonas_artemidis Lautropia unclassified Eikenella_corrodens
Pseudomonas Cardiobacterium Escherichia
Mitsuokella_unclassified
Erysipelotrichaceae Fusobacteriaceae Moraxellaceae Enterobacteriaceae
Agrobacterium_tumefaciens
Centipeda_periodontii Gallionellaceae_unclassified Agrobacterium
Megasphaera_unclassified Leptotrichia_wadei
Selenomonas_sputigena Pseudomonadaceae
Selenomonas Megasphaera
Burkholderiaceae Stenotrophomonas
Mitsuokella
Centipeda
Neisseriales unclassified
Brevundimonas_unclassified
Bulleidia
Burkholderiales Pasteurellales Cardiobacteriaceae_unclassified
Cardiobacteriaceae Shinella_zoogloeoides
Eubacterium_yurii Selenomonas_noxia Gallionellaceae Stenotrophomonas_maltophilia
Leptotrichiaceae Pseudomonadales Shewanella_baltica Hyphomicrobiaceae_unclassified Shinella
Veillonellaceae Xanthomonadaceae unclassified
Peptostreptococcaceae_noname
Brevundimonas_diminuta
Dialister_micraerophilus Fusobacteriales Rhizobiaceae
Bulleidia_extructa Gallionellales Hyphomicrobiaceae
Sinorhizobium_unclassified
Brevundimonas
Dialister Ochrobactrum_anthropi
unclassified Peptostreptococcus_unclassified Erysipelotrichales Xanthomonadales Cardiobacteriales Shewanella
Leptotrichiaceae_unclassified Sinorhizobium
Enterobacteriales
Peptostreptococcus Anaeroglobus Betaproteobacteria Ochrobactrum Campylobacter_showae
Campylobacter_rectus
Clostridiales_Family_XIII_Incertae_Sedis_unclassified Dialister_invisus Brucellaceae
Gammaproteobacteria Caulobacteraceae
Eubacterium_infirmum Peptostreptococcaceae unclassified
Peptostreptococcus_stomatis Anaeroglobus_geminatus Shewanellaceae Campylobacter Treponema_denticola
Filifactor
Clostridiales_Family_XIII_Incertae_Sedis_noname Parvimonas_micra Selenomonadales Alteromonadales RhizobialesCaulobacterales Campylobacter_concisus
Campylobacteraceae Treponema_maltophilum
Campylobacter_curvus
Clostridiales_Family_XIII_Incertae_Sedis Filifactor_alocis Fusobacteriia
Eubacteriaceae_bacterium_CM5 Alphaproteobacteria Campylobacter_gracilis
Treponema_socranskii
Parvimonas Treponema
Eubacteriaceae_bacterium_OBRC8
Eubacteriaceae_noname
Clostridiales_Family_XI_Incertae_Sedis
Erysipelotrichia Proteobacteria Campylobacterales Desulfovibrio_desulfuricans
Desulfovibrio
Treponema_lecithinolyticum
Eubacteriaceae Epsilonproteobacteria
Eubacteriaceae_bacterium_ACC19a Parvimonas_unclassified
Spirochaetaceae Treponema_medium
Clostridiales Treponema_vincentii
Eubacteriaceae_noname_unclassified Eubacterium DesulfovibrionaceaeBilophila_unclassified
Eubacterium_saphenum
Catonella_morbi Negativicutes Deltaproteobacteria
Desulfovibrionales Bilophila
Fusobacteria candidate_division_TM7_single_cell_isolate_TM7c
Catonella Clostridia Candidatus_Saccharibacteria_noname Jonquetella_unclassified unclassified
Eubacterium_brachy Spirochaetales Candidatus_Saccharibacteria_noname_unclassified
Lachnospiraceae_bacterium_oral_taxon_082
Spirochaetia candidate_division_TM7_single_cell_isolate_TM7b
Lachnospiraceae
Lachnospiraceae_noname Stomatobaculum
Stomatobaculum_longum
Firmicutes Spirochaetes
Candidatus_Saccharibacteria_noname
Candidatus_Saccharibacteria_noname
Synergistales
Jonquetella
Candidatus_Saccharibacteria Jonquetella_anthropi
Shuttleworthia
Synergistaceae Actinomyces_cardiffensis
Bacteria
Lachnoanaerobaculum
Lachnospiraceae_bacterium_ICM7 Oribacterium Synergistia Candidatus_Saccharibacteria_noname
Oribacterium_sinus Synergistetes Actinomyces_odontolyticus
Shuttleworthia_satelles Pyramidobacter
Fretibacterium Actinomyces_viscosus
Oribacterium_sp_oral_taxon_078 Abiotrophia_defectiva
Lachnoanaerobaculum_saburreum Tenericutes Fretibacterium_fastidiosum
Actinomyces Actinomyces_graevenitzii
Pyramidobacter_piscolens
Granulicatella_adiacens Abiotrophia
Mollicutes Mycoplasma
Granulicatella_unclassified Granulicatella Mycoplasma_pneumoniae Actinomycetaceae Actinomyces_oris
Mycoplasmatales Mycoplasmataceae Actinomyces_naeslundii Actinomyces_sp_ICM47
Lactococcus_lactis Aerococcaceae
Lactococcus Carnobacteriaceae Actinobacteria Gordonia_terrae
Granulicatella_elegans
Bacilli Gordonia
Gordoniaceae Rothia_aeria
Streptococcus_tigurinus Lactobacillales
Streptococcus_vestibularis
Streptococcaceae Bacillus_cereus_thuringiensis Actinobacteria Actinomycetales Micrococcaceae
Rothia_mucilaginosa
Streptococcus_salivarius Streptococcus_constellatus Atopobium_parvulum Rothia
Actinomycetales_noname
Streptococcus_australis Streptococcus_agalactiae Atopobium_vaginae
Streptococcus_thermophilus Bacillus Bacteroidetes_noname unclassified
Rothia_dentocariosa
Streptococcus_intermedius
Bacteroidetes Atopobium_rimae
Nocardioidaceae Propionibacteriaceae Rothia_unclassified
Nodes
Streptococcus_peroris
Bacteroidetes_noname Tropheryma Propionibacteriaceae_unclassified
Streptococcus Bacillaceae
Atopobium
Streptococcus_oligofermentans
Streptococcus_sp_BS35b
Streptococcus_pseudopneumoniae Coriobacteriales Corynebacteriaceae Propionibacterium_acidipropionici
Streptococcus_sanguinis
Lactobacillaceae Flavobacteriia Bifidobacteriales
Streptococcus_cristatus Bacillales Bacteroidetes_noname
Tropheryma_whipplei Corynebacterium_durum
Nocardioides Propionibacterium_granulosum
Streptococcus_mitis_oralis_pneumoniae
1.00
Streptococcus_gordonii
−0.3500
%DFWHULDOVSHFLHV count
Streptococcus_anginosus Bacillales_noname Cryptobacterium_curtum
Corynebacterium Propionibacterium
Bacteroidetes_oral_taxon_274
Streptococcus_parasanguinis
Gemella_morbillorum Coriobacteriaceae Corynebacterium_tuberculostearicum
Bacteroidetes_noname Cryptobacterium Nocardioides_unclassified Corynebacterium_matruchotii
Streptococcus_infantis Propionibacterium_propionicum
Flavobacteriales Bacteroidia
median difference
Scardovia_wiggsiae
Corynebacterium_pseudodiphtheriticum Propionibacterium_acnes
Gemella_haemolysans
Gemella Bacteroidetes_bacterium_oral_taxon_272 Slackia
Lactobacillus
Lactobacillus_fermentum
Gemella_unclassified
Staphylococcaceae
Gemella_sanguinis
Chryseobacterium_gleum
Scardovia Bifidobacteriaceae
Scardovia_unclassified Alloscardovia
Slackia_exigua
Slackia_unclassified
−0.1560 7.94
Lactobacillus_salivarius Chryseobacterium Alloscardovia_omnicolens
Lactobacillus_ultunensis Chryseobacterium_unclassified Olsenella
Flavobacteriaceae Parascardovia
Staphylococcus_epidermidis
Staphylococcus_saprophyticus Riemerella
Parascardovia_denticolens Bifidobacterium
Olsenella_unclassified
Olsenella_uli
−0.0389 28.80
Staphylococcus Tannerella_forsythia Bacteroidales Prevotella_sp_C561
Prevotella_dentalis
Tannerella Prevotella_oris
Riemerella_unclassified
Staphylococcus_aureus Bifidobacterium_dentium
Staphylococcus_haemolyticus
Staphylococcus_hominis
Capnocytophaga_ochracea
Porphyromonadaceae
Prevotella_oulorum
Prevotella_maculosa
Prevotella_pallens
Prevotella_salivae
Prevotella_marshii
Prevotella_denticola
Bifidobacterium_longum
0.0000 63.50
Capnocytophaga_granulosa Prevotella_loescheii Prevotella_histicola
Porphyromonas_asaccharolytica Prevotella_multiformis
Capnocytophaga Prevotella_buccae
Capnocytophaga_unclassified
Porphyromonas_uenonis
Capnocytophaga_sp_oral_taxon_329
Porphyromonas_sp_oral_taxon_279
Porphyromonas
Prevotellaceae
Prevotella_micans
Prevotella_intermedia
Prevotella Prevotella_bivia
Prevotella_veroralis
Prevotella_melaninogenica
0.0389 112.00
Capnocytophaga_gingivalis Porphyromonas_gingivalis Prevotella_multisaccharivorax Candidatus_Prevotella_conceptionensis
Prevotella_sp_oral_taxon_473
Porphyromonas_endodontalis Prevotella_baroniae
Porphyromonas_gulae
Porphyromonas_catoniae
Alloprevotella
Alloprevotella_tannerae
Prevotella_nigrescens Prevotella_disiens
Prevotella_nanceiensis
Prevotella_pleuritidis
0.1560 175.00
Porphyromonas_sp_oral_taxon_278 Alloprevotella_rava Prevotella_saccharolytica
Prevotella_oralis
Alloprevotella_unclassified
0.3500 251.00
Differentially abundant features Significance Fold change Prevalence shift
Veillonella atypica
Haemophilus parainfluenzae
Prevotella melaninogenica
Rothia mucilaginosa
Veillonella unclassified
Veillonella parvula
Prevotella histicola
Veillonella dispar
Megasphaera micronuciformis
Streptococcus parasanguinis
Prevotella salivae
Prevotella pallens
Alloprevotella tannerae
Streptococcus salivarius
Baseline cluster 1
Baseline cluster 2
Porphyromonas endodontalis
Prevotella nigrescens
Granulicatella unclassified
Prevotella veroralis
Prevotella oris
Neisseria flavescens
Prevotella nanceiensis
Campylobacter concisus
Fusobacterium nucleatum
Actinomyces graevenitzii
Capnocytophaga unclassified
Streptococcus infantis
Alloprevotella rava
Haemophilus haemolyticus
(í (í (í (í (í (í ( 2 4 í í
Abundance ORJíVFDOH íORJDGMS YDOXH Generalized fold change Prevalence [%]
Cluster 1 Baseline Cluster 2 Baseline p-value Rand index
Cluster 1 Longitudinal 33 (91.7%) 4 (40%) .0014 0.736
Cluster 2 Longitudinal 3 (8.3%) 6 (60%)
'( '+
"* 0
•••••• !)!#$•••%&• •••••• !)!,-&.%••/%&$•
Subjects screened
Screen failure n= 94
n=730 Adverse events n= 4
Lost to follow up n=2
Withdrew consent n=7
Participated in another study
n=1
Subjects enrolled Other reasons n=12
n=610
Baseline Baseline
n=100 n=24
Follow-up
n=46
Hierarchical cluster dendogram
Longitudinal clusters
Partition around the medoids (PAM)
Alpha diversity measure
Longitudinal clusters
Longitudinal cluster 1
Longitudinal cluster 2
Bacterial species count Bacterial species count
Neisseria_unclassified
Neisseria_sicca Haemophilus_influenzae
Neisseria_elongata Haemophilus_sputorum
Neisseria_polysaccharea Haemophilus_parainfluenzae
Neisseria_flavescens
Kingella_oralis Haemophilus_paraphrohaemolyticus
Neisseria_bacilliformis Neisseria_sp_oral_taxon_014
Haemophilus
Polaromonas_unclassified Kingella_unclassified Neisseria_cinerea Haemophilus_parahaemolyticus
Fusobacterium_necrophorum Neisseria Haemophilus_haemolyticus Aggregatibacter_unclassified
Aggregatibacter_segnis
Neisseria_meningitidis Neisseria_lactamica
Fusobacterium_nucleatum Kingella
Fusobacterium_periodonticum Variovorax_unclassified Aggregatibacter
Polaromonas Neisseria_subflava Haemophilus_pittmaniae Actinobacillus_unclassified
Veillonella_dispar Aggregatibacter_aphrophilus
Neisseria_macacae Neisseria_gonorrhoeae Pantoea_agglomerans
Veillonella_unclassified Fusobacterium Kingella_denitrificans Actinobacillus
Veillonella_atypica Eggerthia_catenaformis Variovorax Pantoea_unclassified
Alicycliphilus_denitrificans Pasteurellaceae
Neisseriaceae Simonsiella_muelleri
Moraxella_catarrhalis Cardiobacterium_hominis
Veillonella Eggerthia Leptotrichia_unclassified
Alicycliphilus
Simonsiella Acinetobacter_lwoffii Pantoea
Veillonella_parvula
Solobacterium_moorei Comamonadaceae Acinetobacter_unclassified
Comamonas Pseudomonas_unclassified
Moraxella Escherichia_unclassified
Selenomonas_infelix Eikenella
Solobacterium Lautropia_mirabilis Acinetobacter Cardiobacterium_valvarum Agrobacterium_unclassified
Leptotrichia_shahii
Selenomonas_flueggei Megasphaera_micronuciformis Leptotrichia Comamonas_unclassified Pseudomonas_aeruginosa
Selenomonas_artemidis Lautropia unclassified Eikenella_corrodens
Pseudomonas Cardiobacterium Escherichia
Mitsuokella_unclassified
Erysipelotrichaceae Fusobacteriaceae Moraxellaceae Enterobacteriaceae
Agrobacterium_tumefaciens
Centipeda_periodontii Gallionellaceae_unclassified Agrobacterium
Megasphaera_unclassified Leptotrichia_wadei
Selenomonas_sputigena Pseudomonadaceae
Selenomonas Megasphaera
Burkholderiaceae Stenotrophomonas
Mitsuokella
Centipeda
Neisseriales unclassified
Brevundimonas_unclassified
Bulleidia
Burkholderiales Pasteurellales Cardiobacteriaceae_unclassified
Cardiobacteriaceae Shinella_zoogloeoides
Eubacterium_yurii Selenomonas_noxia Gallionellaceae Stenotrophomonas_maltophilia
Leptotrichiaceae Pseudomonadales Shewanella_baltica Hyphomicrobiaceae_unclassified Shinella
Veillonellaceae Xanthomonadaceae unclassified
Peptostreptococcaceae_noname
Brevundimonas_diminuta
Dialister_micraerophilus Fusobacteriales Rhizobiaceae
Bulleidia_extructa Gallionellales Hyphomicrobiaceae
Sinorhizobium_unclassified
Brevundimonas
Dialister Ochrobactrum_anthropi
unclassified Peptostreptococcus_unclassified Erysipelotrichales Xanthomonadales Cardiobacteriales Shewanella
Leptotrichiaceae_unclassified Sinorhizobium
Enterobacteriales
Peptostreptococcus Anaeroglobus Betaproteobacteria Ochrobactrum Campylobacter_showae
Campylobacter_rectus
Clostridiales_Family_XIII_Incertae_Sedis_unclassified Dialister_invisus Brucellaceae
Gammaproteobacteria Caulobacteraceae
Eubacterium_infirmum Peptostreptococcaceae unclassified
Peptostreptococcus_stomatis Anaeroglobus_geminatus Shewanellaceae Campylobacter Treponema_denticola
Filifactor
Clostridiales_Family_XIII_Incertae_Sedis_noname Parvimonas_micra Selenomonadales Alteromonadales RhizobialesCaulobacterales Campylobacter_concisus
Campylobacteraceae Treponema_maltophilum
Campylobacter_curvus
Clostridiales_Family_XIII_Incertae_Sedis Filifactor_alocis Fusobacteriia
Eubacteriaceae_bacterium_CM5 Alphaproteobacteria Campylobacter_gracilis
Treponema_socranskii
Parvimonas Treponema
Eubacteriaceae_bacterium_OBRC8
Eubacteriaceae_noname
Clostridiales_Family_XI_Incertae_Sedis
Erysipelotrichia Proteobacteria Campylobacterales Desulfovibrio_desulfuricans
Desulfovibrio
Treponema_lecithinolyticum
Eubacteriaceae Epsilonproteobacteria
Eubacteriaceae_bacterium_ACC19a Parvimonas_unclassified
Spirochaetaceae Treponema_medium
Clostridiales Treponema_vincentii
Eubacteriaceae_noname_unclassified Eubacterium DesulfovibrionaceaeBilophila_unclassified
Eubacterium_saphenum
Catonella_morbi Negativicutes Deltaproteobacteria
Desulfovibrionales Bilophila
Fusobacteria candidate_division_TM7_single_cell_isolate_TM7c
Catonella Clostridia Candidatus_Saccharibacteria_noname Jonquetella_unclassified unclassified
Eubacterium_brachy Spirochaetales Candidatus_Saccharibacteria_noname_unclassified
Lachnospiraceae_bacterium_oral_taxon_082
Spirochaetia candidate_division_TM7_single_cell_isolate_TM7b
Lachnospiraceae
Lachnospiraceae_noname Stomatobaculum
Stomatobaculum_longum
Firmicutes Spirochaetes
Candidatus_Saccharibacteria_noname
Candidatus_Saccharibacteria_noname
Synergistales
Jonquetella
Candidatus_Saccharibacteria Jonquetella_anthropi
Shuttleworthia
Synergistaceae Actinomyces_cardiffensis
Bacteria
Lachnoanaerobaculum
Lachnospiraceae_bacterium_ICM7 Oribacterium Synergistia Candidatus_Saccharibacteria_noname
Oribacterium_sinus Synergistetes Actinomyces_odontolyticus
Shuttleworthia_satelles Pyramidobacter
Fretibacterium Actinomyces_viscosus
Oribacterium_sp_oral_taxon_078 Abiotrophia_defectiva
Lachnoanaerobaculum_saburreum Tenericutes Fretibacterium_fastidiosum
Actinomyces Actinomyces_graevenitzii
Pyramidobacter_piscolens
Granulicatella_adiacens Abiotrophia
Mollicutes Mycoplasma
Granulicatella_unclassified Granulicatella Mycoplasma_pneumoniae Actinomycetaceae Actinomyces_oris
Mycoplasmatales Mycoplasmataceae Actinomyces_naeslundii Actinomyces_sp_ICM47
Lactococcus_lactis Aerococcaceae
Lactococcus Carnobacteriaceae Actinobacteria Gordonia_terrae
Granulicatella_elegans
Bacilli Gordonia
Gordoniaceae Rothia_aeria
Streptococcus_tigurinus Lactobacillales
Streptococcus_vestibularis
Streptococcaceae Bacillus_cereus_thuringiensis Actinobacteria Actinomycetales Micrococcaceae
Rothia_mucilaginosa
Streptococcus_salivarius Streptococcus_constellatus Atopobium_parvulum Rothia
Actinomycetales_noname
Streptococcus_australis Streptococcus_agalactiae Atopobium_vaginae
Streptococcus_thermophilus Bacillus Bacteroidetes_noname unclassified
Rothia_dentocariosa
Streptococcus_intermedius
Bacteroidetes Atopobium_rimae
Nocardioidaceae Propionibacteriaceae Rothia_unclassified
Nodes
Streptococcus_peroris
Bacteroidetes_noname Tropheryma Propionibacteriaceae_unclassified
Streptococcus Bacillaceae
Atopobium
Streptococcus_oligofermentans
Streptococcus_sp_BS35b
Streptococcus_pseudopneumoniae Coriobacteriales Corynebacteriaceae Propionibacterium_acidipropionici
Streptococcus_sanguinis
Lactobacillaceae Flavobacteriia Bifidobacteriales
Streptococcus_cristatus Bacillales Bacteroidetes_noname
Tropheryma_whipplei Corynebacterium_durum
Nocardioides Propionibacterium_granulosum
Streptococcus_mitis_oralis_pneumoniae
1.00
Streptococcus_gordonii
−0.3500
%DFWHULDOVSHFLHV count
Streptococcus_anginosus Bacillales_noname Cryptobacterium_curtum
Corynebacterium Propionibacterium
Bacteroidetes_oral_taxon_274
Streptococcus_parasanguinis
Gemella_morbillorum Coriobacteriaceae Corynebacterium_tuberculostearicum
Bacteroidetes_noname Cryptobacterium Nocardioides_unclassified Corynebacterium_matruchotii
Streptococcus_infantis Propionibacterium_propionicum
Flavobacteriales Bacteroidia
median difference
Scardovia_wiggsiae
Corynebacterium_pseudodiphtheriticum Propionibacterium_acnes
Gemella_haemolysans
Gemella Bacteroidetes_bacterium_oral_taxon_272 Slackia
Lactobacillus
Lactobacillus_fermentum
Gemella_unclassified
Staphylococcaceae
Gemella_sanguinis
Chryseobacterium_gleum
Scardovia Bifidobacteriaceae
Scardovia_unclassified Alloscardovia
Slackia_exigua
Slackia_unclassified
−0.1560 7.94
Lactobacillus_salivarius Chryseobacterium Alloscardovia_omnicolens
Lactobacillus_ultunensis Chryseobacterium_unclassified Olsenella
Flavobacteriaceae Parascardovia
Staphylococcus_epidermidis
Staphylococcus_saprophyticus Riemerella
Parascardovia_denticolens Bifidobacterium
Olsenella_unclassified
Olsenella_uli
−0.0389 28.80
Staphylococcus Tannerella_forsythia Bacteroidales Prevotella_sp_C561
Prevotella_dentalis
Tannerella Prevotella_oris
Riemerella_unclassified
Staphylococcus_aureus Bifidobacterium_dentium
Staphylococcus_haemolyticus
Staphylococcus_hominis
Capnocytophaga_ochracea
Porphyromonadaceae
Prevotella_oulorum
Prevotella_maculosa
Prevotella_pallens
Prevotella_salivae
Prevotella_marshii
Prevotella_denticola
Bifidobacterium_longum
0.0000 63.50
Capnocytophaga_granulosa Prevotella_loescheii Prevotella_histicola
Porphyromonas_asaccharolytica Prevotella_multiformis
Capnocytophaga Prevotella_buccae
Capnocytophaga_unclassified
Porphyromonas_uenonis
Capnocytophaga_sp_oral_taxon_329
Porphyromonas_sp_oral_taxon_279
Porphyromonas
Prevotellaceae
Prevotella_micans
Prevotella_intermedia
Prevotella Prevotella_bivia
Prevotella_veroralis
Prevotella_melaninogenica
0.0389 112.00
Capnocytophaga_gingivalis Porphyromonas_gingivalis Prevotella_multisaccharivorax Candidatus_Prevotella_conceptionensis
Prevotella_sp_oral_taxon_473
Porphyromonas_endodontalis Prevotella_baroniae
Porphyromonas_gulae
Porphyromonas_catoniae
Alloprevotella
Alloprevotella_tannerae
Prevotella_nigrescens Prevotella_disiens
Prevotella_nanceiensis
Prevotella_pleuritidis
0.1560 175.00
Porphyromonas_sp_oral_taxon_278 Alloprevotella_rava Prevotella_saccharolytica
Prevotella_oralis
Alloprevotella_unclassified
0.3500 251.00
Differentially abundant features Significance Fold change Prevalence shift
Veillonella atypica
Prevotella melaninogenica
Haemophilus parainfluenzae
Rothia mucilaginosa
Veillonella parvula
Longitudinal cluster 1
Veillonella unclassified Longitudinal cluster 2
Neisseria unclassified
Veillonella dispar
Prevotella salivae
Streptococcus parasanguinis
Streptococcus salivarius
Rothia dentocariosa
(í (í (í4 (í3 (í (í ( í í
0
5
10
15
0
10
20
30
op
h ilu
s
in
flu
H en
ae za
m e
op
h ilu
s
pa Ac
ra tin
i nf ob
lu ac
en ter
z ia
ae
N
ei
ss
er
ia
un
cl
a ss
i fie
d
Pr
ev
ot
el
la
hi
st
ic
Pr ol Ba
ev a cte
ot roi
el d ete
la s
m
el
an
i no
ge
ni
ca
R
ot
hi
St a
re m
pt uc
oc
Phylum
ila
Species
oc gi
cu no
s sa
m
iti
s
or
al
is Fir
pn mi
eu cu
m tes
on
ia
e
Tr
op
h er
ym
a
w
hi
pp
l ei
Ve
i llo
ne
lla
at Pr
yp ote
i ca ob
ac
Ve ter
ill ia
on
el
la
un
cl
a ss
i fie
d
Timepoint
Baseline
Longitudinal
Hierarchical cluster dendogram
og
Baseline clusters
Partition around the medoids (PAM)
Mild-
-mo
Mild-moderate group
Severe
asthmatics
Induced
sputum
ɴ-diversity ɴ-diversity
Unsupervised approach
Clustering validation
50 ● ●●●●●●
● ●
●
●
●
●●
●
●
●●●●●
●
●
●●
● ●●
●●
● ●●● ●●
●● ●
● ●
● ● ●● ● ● ●● ● ●●● ●● ●● ●
●● ●● ●● ●●
● ● ● ● ●●●
● ●● ●● ●
● ● ●●
25
●
●●
●
●
● ●●
● ●● 50 ● ●●
●
● ●
●
● 50 ●●
● ●●
● ● ●●
Clinical characteristics
● ● ● ● ●● ●
● ●● ● ● ● ●● ● ●●
25 ●
●●● ●
● ●
●
●●●●● ● ●●●
●● ● ●
●
●● ●
●●● ●
●
● ●
● ●
●●
● ●
● ●●
●
● ● ●● ●● ● ● ● ● ●
● ● ●● ●
●●●
●● ●●● ●●
●
● ●
●●
●●
●●
●
●●
● 25 ●●
●
25 ● ●
●
0 ● ● ● ● ●
FEV1/FVC pre−salbutamol (% pred) FEV1/FVC post−salbutamol (% pred) Age of asthma onset (years)
0.03 ●
● ●
0.012 ● ●●
● ● 0.012
● ● ● ● ●
100 ● 100 60 ●●
● ●● ●
●● ● ● ●●
● ●●
●
● ● ● ●●● ● ● ●
●●
●
● ● ● ●
● ●
●●●● ●●● ●● ●● ●●●● ●● ●
● ● ● ● ●
●● ●● ● ● ●● ●●●
●● ● ●● ● ●
80 ●●
●●●●●
● ●
80 ●●●● ●●● ●
●
●
● 40 ●● ●
●● ●
● ● ●
●
●
●
●
●
●● ● ● ●●
● ● ●●
● ● ● ●● ●
● ● ● ●
● ●● ●● ● ● ● ● ● ●●
●
●
● ●
● ●
● ●● ● ● ● ● ●
● ●●●
●● ● ● ●● ● ●
●
● ●
● ● ●●
●●
● ●● ● ●● ●
●
●● ●
● ● ●● ●
60 ●●●
●●● ●
●● ●● ● ●
60 ●
● ● ●
20 ●●
●
●
●
●
●
● ●
● ●
●
● ● ● ● ●● ● ●
● ●
● ● ● ● ● ●
● ●
● ● ● ● ● ● ●● ●
●
● ● ●●
40 ●
●● ● ● 40 ●●
●
0
●● ●● ●●
●
● ● ●●
●
75 ●
●
●●
●
● 60 ●
●
● 80 ● ● ●
●
● ●
● ● 80 ●
●
●
● ●
●●
●● ● ● ● ●● ● ●
● ● ● ● ● ●● ●
●●
●
●
● ●
40
● ● 70 ● ● ●
● ●
●
50
●● ●
●● ● ● ● ●● ●
● ●●
●
●
● ●
●
60 ● ●
● ●●●● ●● ●
● ● ●●
● ●
●●
●● ● 60 ● ● ●
●
● ● ● ● ●● ● ● ● ●
●● ● ●
●●● ● ●●
20 ●
Clinical characteristics
● ● ● ● ● ●
25 ●
●
●
●● ● ●
● ● ●
●
●●
50 ●
●
●
● 40 ●
● ● ●● ● ●
●
●● ● ● ● ●
●● 0 ● ● ●●
40
● ● ●
●● ●
50 ●● ●
● ●
●
●
●
●
Cluster 1 Cluster 2
Microbiome−driven clustersDWEDVHOLQHDQGIROORZXSYLVLWV
Baseline cluster 1
Baseline cluster 2
Bacterial species count Bacterial species count
0.26 0.24 0.25 0.23 Veillonella atypica
* * * *
−0.20 0.22 0.25 0.27 0.25 0.22 Prevotella histicola
* * * ** * *
−0.23 0.21 0.24 0.26 0.24 0.22 0.28 Megasphaera micronuciformis
* * * ** * * **
−0.29 0.42 0.26 Streptococcus parasanguinis
** **** **
−0.38 0.27 0.22 0.30 0.26 −0.36 Prevotella melaninogenica
*** ** * ** ** **
−0.30 0.38 0.25 0.27 0.20 0.23 Veillonella parvula
** **** * ** * *
−0.21 0.24 0.22 0.24 0.25 0.27 Actinomyces graevenitzii
* * * * * **
−0.26 0.34 0.20 0.25 0.20 0.27 Prevotella salivae
** *** * * * **
−0.27 0.37 0.23 0.25 Fusobacterium nucleatum
** *** * *
0.33 0.21 Campylobacter concisus
*** *
−0.25 0.22 0.27 Veillonella dispar
* * **
−0.39 0.30 0.21 0.23 Granulicatella unclassified
**** ** * *
−0.32 0.26 0.32 0.25 Veillonella unclassified
** ** ** *
0.20 0.23 Prevotella veroralis
* *
0.22 0.25 Alloprevotella rava
* *
−0.25 0.30 0.21 0.25 0.29 Prevotella pallens
* ** * * **
−0.33 0.26 0.26 0.23 Haemophilus parainfluenzae
*** ** ** *
−0.43 0.44 0.23 0.25 0.28 −0.20 Rothia mucilaginosa
**** **** * * ** *
−0.37 0.43 0.21 0.22 Prevotella nigrescens
*** **** * *
−0.33 0.37 Prevotella nanceiensis
*** ***
−0.38 0.38 0.20 Porphyromonas endodontalis
*** **** *
−0.41 0.43 Alloprevotella tannerae
**** ****
−0.24 0.25 −0.25 Streptococcus infantis
* * *
0.25 −0.24 Prevotella oris
* *
0.30 0.30 Streptococcus salivarius
** **
−0.21 0.31 Neisseria flavescens
* **
0.27 Haemophilus haemolyticus
**
−0.34 0.35 Capnocytophaga unclassified
*** ***
0.23 −0.25 −0.29 Haemophilus influenzae
* * **
0.27 Moraxella catarrhalis
**
)
%
s%
ol
ol
ol
e
ng
mo
nse
ban
tak
tam
am
tam
oki
hils
uta
age
e in
ao
bur
but
Sm
lbu
lbu
&orrelationV (rVDQGUSE)
rop
alb
thm
oph
(su
yllin
sal
−sa
−sa
1
eut
t−s
re−
f as
acr
ion
oph
0.5
ost
pre
mn
pos
Cp
mm
cat
0
eo
Cp
The
V1
utu
V1
−0.5
l lo
/FV
Ag
/FV
utu
FE
Sp
FE
−1
nta
V1
Sp
V1
e
FE
sid
FE
Re
Within-cluster sum of squares Average silhouettes width Calinski-Harabasz index
Use of list:
This list is to be used for all non-core clinical papers.
Instructions
Follow up clinical papers should re-use the baseline cohort description paper lists, in order to recognize
the clinical staff involved in the study.
Contributors
Partner organisations
Novartis Pharma AG University of Southampton, Southampton, UK
Università Cattolica del Sacro Cuore, Rome, University Hospital, Copenhagen, Denmark
Italy
European. Fed. of Allergy and Airways Lega Italiano Anti Fumo, Catania, Italy
Diseases Patients’ Associations, Brussels,
Belgium
Chiesi GlaxoSmithKline
Roche UCB
UK
David Supple (chair)
The Netherlands
Dominique Hamerlijnck
Jenny Negus UK
Lehanne Sergison UK