Association between the sinus microbiota with eosinophilic inflammation and prognosis in chronic rhinosinusitis with nasal polyps

Dysbiosis of the sinus microbiome affects the pathophysiology of chronic rhinosinusitis with nasal polyps (CRSwNPs). We investigated whether the sinus microbiota in CRSwNPs is associated with eosinophilic inflammation, especially in relation to innate lymphoid cells (ILCs), prognosis, and serum extracellular vesicles (EVs). Middle meatal swabs and serum from 31 CRSwNPs patients and six healthy controls were analyzed by 16S ribosomal RNA sequencing. ILC2s and cytokines from sinonasal tissues were measured by flow cytometry and ELISA, respectively. The relative abundances (RAs) of bacteria were compared based on eosinophilic inflammation and surgical outcome. The correlations between sinus bacteria and ILC2s, cytokines, and serum EVs were analyzed. The compositions of sinus bacteria were different between groups at the genus level. In eosinophilic CRSwNPs patients, the RA of Anaerococcus was significantly decreased (P = 0.010), whereas that of Lachnoclostridium was significantly increased (P = 0.038) compared with that in controls. The RA of Lachnoclostridium showed a significant positive correlation with interleukin (IL)-5-producing ILC2 populations (R = 0.340, P = 0.049), whereas the RA of Anaerococcus showed a negative correlation with IL-5-producing ILC2 populations (R = −0.332, P = 0.055). The RAs of Corynebacterium, Anaerococcus, and Tepidimonas were significantly decreased in patients with suboptimal outcomes compared with those in patients with optimal outcomes and control subjects. Some sinus bacteria and serum EVs showed positive correlations. CRSwNPs patients showed distinct microbiota compositions based on eosinophilic inflammation in relation to ILC2s and surgical outcome. These findings support a relationship between the microbiota and the host immune response in CRSwNPs.


Introduction
Chronic rhinosinusitis (CRS) is a multifactorial chronic upper airway inflammatory disease. Putative pathological factors include changes in the microbiota, an imbalance of the local or systemic immune system, allergens, toxins, and genetic predisposition [1][2][3] . Several studies have reported that the microbiota may affect the pathophysiology of CRS. In general, a decrease in microbial diversity and an increase in the total number of microorganisms have been reported in patients with CRS 4,5 .
Therefore, dysbiosis of the microbiota could be the driving force of CRS.
Immunologic features associated with the nature or composition of the sinus microbiota could partially explain the heterogeneity of CRS. CRS with nasal polyps (CRSwNPs) is generally classified as eosinophilic or noneosinophilic, which show distinct immunological features. Eosinophilic polyps are associated with the type 2 immune response, while noneosinophilic polyps are associated with type 1 and 17 immune responses 6 . When a CRS patient cluster was divided into distinct subgroups according to the specific pattern of bacterial cocolonization, each subgroup was associated with a host immune response. Some subgroups exhibited increased expression of the interleukin 6, tumor necrosis factor, IL8, and IL10 genes, whereas other subgroups were associated with induced IL5 gene expression 7 . Innate lymphoid cells (ILCs) are a recently identified innate lymphocyte population. Group 2 ILCs (ILC2s) secrete type 2 cytokines, such as IL-5 and IL-13, and contribute to the pathogenesis of CRSwNPs, particularly in eosinophilic tissue inflammation 8 . ILCs interact with bacteria and coordinate host-bacteria relationships that are associated with the pathogenesis and progression of numerous chronic human infectious, inflammatory, and metabolic diseases 9 . We speculated that in CRSwNPs, ILCs could interact with the microbiota and influence eosinophil inflammation.
Bacteria secrete extracellular vesicles (EVs), including transmembrane proteins, cytosolic proteins, lipids, and nucleic acids. EVs play a role in immune and airway epithelial cell communication and can induce immune responses in allergic diseases. EVs can be detected in the blood circulation far from the local disease site [10][11][12][13] . In nasal lavage fluid from CRS patients, the composition of EVs secreted from microbiota has been positively correlated with bacterial composition 14 . We hypothesized that paranasal sinus bacteria-derived EVs can be detected in the blood. However, the relationship between sinus bacteria and serum EVs is unknown.
In the present study, the sinus microbiota in specimens from middle meatus swabs from CRSwNPs patients was analyzed based on eosinophilic vs. noneosinophilic inflammation. In addition, correlations were analyzed between sinus bacteria and ILCs and cytokines from tissues and between sinus bacteria and serum EVs.

Study populations
This study was conducted in accordance with the Declaration of Helsinki and approved by the Institutional Review Board of Asan Medical Center (2018-0121). All participants provided written informed consent.
Thirty-one adult CRSwNP patients, including 21 eosinophilic CRSwNPs patients and 10 noneosinophilic CRSwNPs patients, and 6 healthy controls were recruited from the Department of Otorhinolaryngology-Head and Neck Surgery, Asan Medical Center, between February 2018 and December 2018. CRSwNPs was diagnosed based on the established guideline criteria of the European Position Paper on Rhinosinusitis and Nasal Polyps (2012), considering the history, nasal endoscopy findings, and computed tomography (CT) images of the paranasal sinuses 2 . Eosinophilic and noneosinophilic CRSwNPs were defined based on the proportion of eosinophils >10% and <10%, respectively, of the total inflammatory cells in the NP tissue 15,16 . Subjects with antrochoanal polyps, fungal sinusitis, allergic fungal sinusitis, aspirin-exacerbated respiratory disease, immunodeficiency, pregnancy, cystic fibrosis, or primary ciliary dyskinesia were not included. All patients refrained from the use of decongestants, antibiotics, and systemic corticosteroids for at least 4 weeks prior to surgery, did not have a history of acute respiratory infection for 4 weeks before surgery, and did not have medical comorbidities that included diabetes mellitus, hypertension, or hyperlipidemia. Demographic and clinical variables recorded for each subject were age, sex, ethnicity, smoking history, allergy, asthma, antibiotic use in the past 6 months, saline rinse use, and topical steroid use in the preceding 4 weeks.
Atopic status was evaluated using a skin prick test (Allergopharma; Hermann-Körner-Straße, Reinbeck, Germany), and specific IgE antibodies against various common inhalant allergens were detected with ImmunoCAP™ tests (Thermo-Fisher Scientific, Waltham, MA, USA). Asthma and aspirin intolerance were diagnosed based on medical history. Preoperative Lund-Kennedy endoscopy score, polyp grading according to the Davos classification, Lund-Mackay CT scores, and sinonasal outcome test (SNOT-22) scores were evaluated 2 . Control subjects >18 years of age underwent septoplasty for nasal obstruction without any other inflammatory nasal diseases, such as allergic rhinitis or sinusitis, evident on CT scan.
After at least 6 months, CRSwNP patients were examined to determine changes in Lund-Kennedy endoscopy score, the need for revision of sinus procedure(s), and the use of additional courses of antibiotics or systemic steroids after routine standard perioperative administration. The optimal outcome was defined as a reduction in endoscopy scores >50%, no requirement for revision of surgical treatment, and no further use of antibiotics and systemic steroids before the 6-month follow-up visit 17 .
Swab samples (Isohelix DNA swab, Cell Projects Ltd., Maidstone, Kent, UK) for DNA extraction were obtained using an endoscopic guide from the middle meatus and rotated at least five full turns until saturated. The samples were placed on ice upon collection and frozen within an hour at -80°C until DNA extraction. Care was taken to avoid contamination from the anterior nasal cavity during swabbing. NP tissues from CRSwNP patients and middle turbinate tissues from controls were harvested after swabbing. Blood samples were collected from the median cubital vein after overnight fasting and were transferred into serum separation tubes (BD, Franklin Lakes, NJ, USA). Serum was separated from the blood and frozen at −80°C until DNA extraction.

DNA extraction from samples
Human swab samples were vortexed after dilution in 800 µl of phosphate-buffered saline (PBS, Welgene, Gyeongsan, South Korea). After centrifugation at 10,000 × g for 1 min at 4°C, the pellets of swab samples containing bacterial cells were heated for 40 min at 100°C to extract DNA from the bacterial cells. To eliminate remaining floating particles and debris, the supernatants were collected after centrifugation at 13,000 rpm for 30 min at 4°C. For serum samples, floating particles were precipitated by centrifugation at 10,000 × g for 1 min at 4°C, after which the supernatants were mixed with PBS. After centrifugation, bacteria and foreign particles were thoroughly eliminated by sterilizing the supernatants through a 0.22-μm filter. Next, the supernatants were boiled for 40 min at 100°C to extract DNA from EVs and centrifuged at 13,000 rpm for 30 min at 4°C. The supernatants were collected 18,19 . After this preprocessing, DNA from sinus bacteria and serum bacterial EVs was extracted using a MO BIO PowerSoil DNA Isolation Kit (QIAGEN, Hilden, Germany) and quantified using the QIAxpert system (QIAGEN).

Bacterial metagenomic analysis using DNA from human samples
Bacterial genomic DNA was amplified with the 16S_V3_F (5′-TCGTCGGCAGCGTCAGATGTGTATAAGAGACA GCCTACGGGNGGCWGCAG-3′) and 16S_V4_R (5′-GT CTCGTGGGCTCGGAGATGTGTATAAGAGACAGGA CTACHVGGGTATCTAATCC-3′) primers specific for the V3-V4 hypervariable regions of the 16S ribosomal RNA gene. The libraries were prepared using PCR products according to the MiSeq system guide (Illumina, San Diego, CA, USA) and quantified using the QIAxpert system (QIAGEN). Each amplicon was then quantified, the equimolar ratio was obtained, amplicons were pooled, and they were sequenced on a MiSeq device (Illumina) according to the manufacturer's recommendations.

Analysis of the bacterial microbiota composition
Paired-end reads that matched the adapter sequences were trimmed using cutadapt version 1.1.6 20 . The resulting FASTQ files containing paired-end reads were merged with The Context-Aware Scheme for Paired-End Read tool, and they were quality filtered using the Phred (Q) score based on previously described criteria 21,22 . Reads <350 and >550 bp after merging were discarded. To identify chimeric sequences, a reference-based chimera detection step was conducted using the VSEARCH tool against the SILVA gold database 23,24 . Next, the sequence reads were clustered into operational taxonomic units (OTUs) using VSEARCH with a de novo clustering algorithm, with 97% sequence similarity as the threshold. The representative sequences of the OTUs were finally classified using the SILVA 128 database with UCLUST (parallel_assign_taxonomy_uclust.py script on QIIME version 1.9.1) and default parameters 25 .

Tissue processing and flow cytometric analysis for ILC2s
Sinonasal tissues were minced using sterile scissors. The tissues were then strained through a 70-μm nylon mesh cell strainer into 50-ml polypropylene tubes containing PBS and 1% fetal bovine serum (FBS; MP Biomedicals, Aurora, OH, USA). Tissue pellets were obtained by centrifugation for 5 min at 600 × g.

Statistical analyses
The demographic and clinical variables were compared using the Kruskal-Wallis test or Chi-square test. Standard ecological diversity indices (e.g., Chao1 and Shannon diversity) were assessed using the Kruskal-Wallis test.
Beta diversity (distance between samples based on differences in OTUs present in each sample) was measured using principal coordinate analysis (PCoA) based on Bray-Curtis dissimilarity. A heat map was used to visualize ordination. Comparison of the relative abundance (RA) of OTUs between groups was performed using the Wilcoxon rank sum test.
Linear discriminant analysis (LDA) effect size (LEfSe) was used to determine genera that best characterized each group (control vs. eosinophilic CRSwNPs vs. noneosinophilic CRSwNPs). LEfSe uses LDA to estimate the effect size of each differentially rich feature 27 . An LDA score >3 and P < 0.05 were considered significant. Correlations were evaluated using Pearson's correlation analysis. Statistical analyses were performed with the R software package. Statistical significance was evident as P < 0.05.

Results
Eosinophilic CRSwNPs patients were older and had higher asthma incidence, polyp grade, Lund-Kennedy endoscopic score, and Lund-Mackay CT score than control and noneosinophilic CRSwNPs patients. Noneosinophilic CRSwNPs patients had a higher prevalence of purulent discharge and saline irrigation in the 6 months before surgery than eosinophilic CRSwNPs patients (Table 1).
Sinus bacterial alpha diversity did not differ significantly among control, eosinophilic CRSwNPs, and noneosinophilic CRSwNPs subjects (Fig. E1a). At the phylum level, analysis of OTUs demonstrated a rich set of bacterial taxa in both control subjects and patients dominated by characteristic bacterial phyla that included Firmicutes (mainly Staphylococci), Actinobacteria (mainly Corynebacterium, Bifidobacterium, and Propionibacterium species), and Proteobacteria (mainly Moraxella, Pseudomonas, Enterobacter, and Aggregatibacter). Sinus bacteria did not differ at the phylum level between control subjects, eosinophilic CRSwNPs patients, and noneosinophilic CRSwNPs patients (Fig. E1b).
However, sinus bacterial compositions were different at the genus level among control subjects, eosinophilic CRSwNPs patients, and noneosinophilic CRSwNPs patients (Fig. 1a). In eosinophilic CRSwNPs patients, the RAs of Anaerococcus and Tepidimonas were significantly decreased (P = 0.010 and P = 0.009, respectively), while that of Lachnoclostridium was significantly increased (P = 0.038) when compared with those in controls. The RA of Stenotrophomonas was significantly increased in eosinophilic CRSwNPs patients compared with that in noneosinophilic CRSwNPs patients (P = 0.038), the RA of Lachnospiraceae NK4A136 was increased in noneosinophilic CRSwNPs patients compared with that in controls (P = 0.039), and the RA of Mesorhizobium was significantly decreased in eosinophilic CRSwNPs patients compared with that in controls (P = 0.012) (Fig. 1b). When LEfSe was used to compare bacterial abundance at the genus level, Anaerococcus and Tepidimonas were significantly enriched in control subjects (Fig. 2), similar to the results of the comparison between groups. Correlations between immunological features, such as ILCs, cytokines, and chemokines, and distinct bacteria (Anaerococcus, Tepidimonas, Lachnoclostridium, Stenotrophomonas, and Lachnospiraceae NK4A136) were analyzed. The RA of Lachnoclostridium was significantly positively correlated with IL-5-producing ILC2 populations (R = 0.340, P = 0.049; Fig. E2) without a significant correlation with IL-5 levels (R = 0.188, P = 0.266). Furthermore, IL-5-producing ILC2 populations correlated with IL-5 (R = 0.351, P = 0.042) and CCL24 (R = 0.351, P = 0.042) levels in tissues. The RA of Anaerococcus showed a negative correlation with IL-5-producing ILC2 populations (R = −0.332, P = 0.055) and a significant negative correlation with IL-5 levels in tissues (R = −0.420, P = 0.010; Table 2). There were significant correlations between the levels of other cytokines and chemokines (IL-8, IL-13, IFN-γ, CCL11, and CCL24) and the RAs of Anaerococcus and Lachnoclostridium. The RAs of Tepidimonas, Stenotrophomonas, and Lachnospiraceae NK4A136 were not correlated with ILCs, cytokines, and chemokines (all P > 0.05; Table E1).
Patients with suboptimal outcomes exhibited significantly decreased abundances of Corynebacterium at the time of surgery compared with those of patients with optimal outcomes and controls (6.9% RA vs. 15.3% RA vs. 19.0% RA, P = 0.049). The RAs of Anaerococcus and Tepidimonas were also significantly decreased in patients Fig. 1 Significantly differentially abundant bacterial genera in the sinus among control subjects, patients with eosinophilic chronic rhinosinusitis with nasal polyps, and patients with noneosinophilic chronic rhinosinusitis with nasal polyps. a Differences in the composition of sinus bacteria between the three groups were identified by a heat map, although there was distinct clustering on the PCoA plot. b In patients with eosinophilic chronic rhinosinusitis with nasal polyps, the RAs of Anaerococcus, Tepidimonas, and Mesorhizobium were significantly decreased, whereas that of Lachnoclostridium was significantly increased compared with those in controls. *P < 0.05, **P < 0.01.

Discussion
Various microbes, including bacteria, viruses, and fungi, thrive in the human body in a symbiotic relationship 28 . Many microbiota-related studies have been performed with the knowledge that the microbiota strongly influences human health and diseases 29 . These studies have been performed not only for the gut but also for the respiratory tract, which is one of the entry routes of bacteria into the body. Bacteria have been implicated in the pathogenesis of CRS, and staphylococcal superantigen, immune barrier, and biofilm hypotheses have been proposed [30][31][32] . However, the exact roles of bacteria in CRS remain controversial.
In the present study, eosinophilic CRSwNPs patients showed significantly lower RAs of Anaerococcus and Tepidimonas and a higher RA of Lachnoclostridium at the genus level than controls and noneosinophilic CRSwNPs patients. LEfSe also showed Anaerococcus and Tepidimonas enrichment in controls. The clinical and demographic variables were different among controls, eosinophilic CRSwNPs patients, and noneosinophilic CRSwNPs patients. These variables might affect bacterial composition. Eosinophilic CRSwNPs patients had higher asthma incidence and clinical severity scores but lower purulence incidence and saline irrigation than noneosinophilic CRSwNPs patients. These findings indicate that asthma, clinical severity, and purulence may be associated with significant alterations in bacterial composition, similar to the findings of a previous study 17 , although a direct relationship was not analyzed. The three groups showed similar rates of smoking, systemic antibiotic administration before surgery, and saline irrigation. Given that these factors were not associated with distinct alterations in sinonasal bacteria [33][34][35] , their effects on the microbiome might not be significant.  In one study, when a CRS patient cluster was divided into distinct subgroups, each subgroup was associated with a distinct metabolism and host immune response 7 . The immunologic features associated with the nature of the sinus microbiota can partially explain CRS heterogeneity and can be used as a basis for establishing a customized treatment strategy. In our study, the IL-5producing ILC2 population showed a significant positive correlation with the RA of Lachnoclostridium but a negative correlation with the RA of Anaerococcus. The Fig. 3 Significantly differentially abundant bacterial genera in the sinus according to the surgical outcome. Corynebacterium, Anaerococcus, and Tepidimonas were significantly less abundant in patients with suboptimal outcomes than in those with optimal outcomes and control subjects. *P < 0.05. observation that IL-5-producing ILC2 populations correlated with the mediators of eosinophilic inflammation, such as IL-5 and CCL24 levels in tissues, might explain the RA of Lachnoclostridium and the reduction in Anaerococcus abundance in eosinophilic CRSwNPs patients. Increased Lachnoclostridium abundance with type 2 responses and gut dysbiosis was suggested as a mechanism of food allergy induction during consumption of a high-fat diet 36 . Knowledge about the association between Anaerococcus and the immune response is limited. Given that A. prevotii is frequently identified from vaginal discharges and ovarian, peritoneal, sacral, or lung abscesses 37 , Anaerococcus may be related to the type 1 response.
CD4 + T helper (Th) cells play an important role in coordinating adaptive immune responses. Endotypes of CRS are differentiated mainly on the basis of Th cells other than ILCs, which produce cytokines and chemokines, leading to the accumulation of eosinophils and neutrophils 38 . The bacteria are associated with the endotypes of CRS in relation to cytokines produced by Th cells. Gram-negative bacteria were isolated more frequently in IL-5/IL-17/IFN-γ-negative NPs patients, whereas gram-positive bacteria were more associated with IL-5-positive NPs patients 39 . Moreover, depletion of commensal bacteria using oral broad-spectrum antibiotics increases serum IgE levels and exacerbates basophilmediated Th2 cell responses and allergic inflammation in the airway 40 .
Bacterial diversity and composition were previously suggested as predictors of surgical outcome. In 27 patients with CRS, more diverse bacterial communities and enriched Actinobacteria at the phylum level and Corynebacterium at the genus level showed optimal outcomes than suboptimal outcomes 17 . Similarly, in our study, patients with optimal outcomes were more enriched for the genera Corynebacterium, Anaerococcus, and Tepidimonas at the time of surgery than those with suboptimal outcomes and less enriched than controls.
EVs circulate through the blood and directly and/or indirectly communicate with tissues and organs 10 . Blood bacterial EVs were suggested to be useful for metagenome analysis of the bodily microbiota in preterm birth and a mouse model of Alzheimer's disease 41,42 . In this study, Deinococcus, Sphingomonas, and Lactobacillus among the sinus bacteria were positively correlated with serum EVs. Individual bacteria in the middle meatus and serum were also correlated. These correlations suggested that sinus bacteria might communicate with blood bacterial EVs. Although direct causality between sinus bacteria and blood bacterial EVs and the mechanism of the latter's formation have not been proven, Sphingomonas may be potentially meaningful in terms of sinus bacteria and blood bacteria communication. The Sphingomonadaceae family was found to be abundant in the urine-derived EVs of children with asthma and chronic rhinitis compared with that in EVs of controls 43 . Sphingomonas species in the airway were associated with increased airway hyperresponsiveness to methacholine, positively correlated with eosinophil cationic protein levels, and strongly and positively correlated with serum IgE levels and blood and sputum eosinophil counts in patients with asthma [44][45][46] . This association may be explained by glycosphingolipid, a cell membrane component of Sphingomonas that activates natural killer T cells producing IL-4 and IL-13 [47][48][49] . Further research is needed on the association of Sphingomonas between the sinus and serum in relation to type 2 immune responses and eosinophilic CRSwNPs. Another potentially interesting bacterium is Lactobacillus; its RA in the sinus is less than that of other bacteria, but its sinus RA is positively correlated with its RA in serum EVs. Considering that Lactobacillus has been suggested as a potential protective species against allergens and virus infection 50,51 , deficiency in Lactobacillus in the sinus and blood may be associated with the pathophysiology of CRS.
The current study included a small number of subjects. However, the observed relationship between immunological characteristics in terms of ILCs and bacteria in the sinonasal cavity in CRSwNPs patients is a novel finding. Because there was large variation in the bacterial composition of the sinonasal cavity between individuals, similar to that in a previous study 7 , clustering based on the clinical characteristics was not distinct. Bacterial metabolites were not investigated in this study. A previous study demonstrated that the Streptococcaceae dominant group, Pseudomonadaceae dominant group, Corynebacteriaceae dominant group, and Staphylococcaceae dominant group encoded enriched gene pathways, including ansamycin biosynthesis, tryptophan metabolism, the two-component response system virulence pathway, and the peroxisome proliferator-activated receptor gamma signaling pathway, respectively. These metabolites may be involved in the immune responses of nasal polyp tissues 7 . Therefore, the effect of the metabolites of Lachnoclostridium and Anaerococcus on ILC2s in NP tissues should be further investigated.
Finally, this study showed that the sinus microbiota, particularly Anaerococcus and Lachnoclostridium in CRSwNPs patients, was associated with eosinophilic inflammation and IL-5-producing ILC2s. The abundances of the genera Corynebacterium, Anaerococcus, and Tepidimonas were related to surgical outcome. Furthermore, some sinus bacteria were correlated with blood bacterial EVs. Therefore, understanding the sinus microbiota of individuals could be the basis for effective treatment of CRS and for predicting prognosis.