Stratification of chronic rhinosinusitis with nasal polyps by distinct sinonasal microbial communities and their clinical and pathological correlations

Chronic rhinosinusitis with nasal polyps (CRSwNP) is a heterogeneous inflammatory condition. Dysbiosis is commonly observed. This study identified distinct sinonasal microbial communities in CRSwNP patients and explored their clinical and pathological associations. Two distinct microbial communities were identified: community type 1 (CT1), dominated by Corynebacterium, Propionibacterium, and Peptoniphilus; and community type 2 (CT2), dominated by Streptococcus, Porphyromonas, and Bacteroides. CT1 exhibited significantly reduced microbial diversity and was enriched in pro-inflammatory pathways, whereas CT2 showed higher microbial diversity, with pathways supporting anti-inflammatory. Clinically, CT1 patients demonstrated higher disease severity and increased eosinophil infiltration, compared with CT2. CRSwNP patients can be stratified into two distinct microbial communities, each with specific clinical and pathological profiles. Microbial-based stratification may inform tailored therapeutic strategies to address the diverse clinical and pathological manifestations of CRSwNP. Chronic rhinosinusitis (CRS) is a complex, multifactorial inflammatory disease of the upper airway. Studies have shown that CRS exhibits heterogeneity, with variations in clinical presentation (with or without nasal polyps), immunologic profiles (type 2 or non-type 2 inflammation), and pathologic characteristics (eosinophilic or non-eosinophilic) [1, 2]. The sinonasal microbiome plays a key role in CRS, with reduced microbiota diversity, or “Microbial dysbiosis,” being a common feature among patients [3-5]. It has been proposed that CRS patients could be stratified according to specific sinonasal microbial compositions, illuminating phenotypic heterogeneity. Cope et al. were the first to cluster CRS patients into three subgroups based on microbial composition, noting an association between certain subgroups and nasal polyps or differing immune responses [6]. A smaller study (n = 23) identified two microbial subtypes (DC1 and DC2), finding that DC1 was linked to polyp incidence, while DC2 exhibited lower microbial diversity [7]. Bassiouni et al. further described three main “microbiotypes” in an international multicenter study, but these subtypes largely reflected geographic differences [8]. While these studies highlight the heterogeneity of CRS in relation to microbial communities, several limitations remain. Notably, most conclusions focus on the association of microbiotypes with polyp formation and geographic variation. These factors may also serve as confounding variables, which complicate efforts to link microbial subtypes with clinically or pathologically significant features like disease severity and eosinophil infiltration. Thus, it may be more effective to analyze microbial clusters in specific CRS subtypes, such as CRS with nasal polyps (CRSwNP), and account for regional influences. This study aims to identify sinonasal microbial communities in CRSwNP patients, categorize them based on microbial phenotypes, and explore associations with clinical and pathological characteristics. A total of 86 CRSwNP patients and 15 control subjects were recruited from The First Affiliated Hospital of Sun Yat-sen University. The clinical and pathological characteristics of patients and controls are summarized in Table S1. Sinus swab samples were collected and analyzed using 16S rRNA sequencing. We hypothesized that CRSwNP patients may exhibit distinct local microbial profiles based on the composition of common bacterial genera. Based on bacterial occurrence (present in more than 50% of samples) and relative abundance (greater than 0.1%), seven major bacterial genera were identified in the polyp mucosa of all patients: Alloiococcus, Anaerococcus, Corynebacterium, Haemophilus, Propionibacterium, Staphylococcus, and Streptococcus. Unsupervised clustering approach using the Dirichlet Multinomial Mixtures model revealed two distinct microbial community types (CT1 and CT2). CT1 included 58 patients whose microbiota was dominated by Corynebacterium, Propionibacterium, and Peptoniphilus. CT2, comprising 28 patients, was characterized by Streptococcus, Porphyromonas, and Bacteroides (Figure S1). CT1 showed significantly diminished microbial diversity (Simpson index) and richness compared to CT2 and control subjects, with the control group exhibiting the highest diversity and richness. This suggests that the microbial environment in CT1 patients may be more prone to dysbiosis (Figure 1A–D). The functional potential of the microbiota was predicted using the PICRUSt algorithm. Among the three groups, CT1 showed the lowest functional diversity and the fewest KEGG pathways, while controls exhibited a greater number of pathways than CT1 or CT2. CT1 was significantly enriched in pathways related to bacterial virulence and pro-inflammatory/oxidative stress processes (e.g., lipoarabinomannan biosynthesis, ferroptosis), as well as cellular processes linked to immunosuppressive activities (e.g., morphine and nicotine addiction, neuroactive ligand–receptor interaction) [9, 10]. The enrichment of ferroptosis-related pathways in CT1 may suggest a pro-inflammatory microenvironment, consistent with previous reports linking ferroptosis to epithelial barrier dysfunction [10]. In contrast, CT2 was enriched in metabolic pathways associated with barrier integrity and host defense (e.g., secondary bile acid biosynthesis, teichoic acid biosynthesis) [11] (Figure 1E–G). These findings suggest that CT2 harbors enhanced anti-inflammatory and antimicrobial functions, whereas CT1 microbiota may contribute to a more pathogenic state. We next explored the associations between microbial communities and clinical, demographic, and pathological factors (Table S2). CT1 patients were generally younger than those in the control group. CT1 also had significantly higher Lund-Mackay scores (a CT-based severity scale) and higher polyp scores compared to CT2. Furthermore, CT1 exhibited greater eosinophil infiltration, a hallmark of type 2 inflammation, which is associated with polyp recurrence [12]. Neutrophil counts were similar across both groups. These findings suggest that CT1 patients may exhibit more severe clinical and pathological phenotypes. CT1 patients also demonstrated an increased presence of tertiary lymphoid structures (TLS) and local antibody-producing cells in the sinus mucosa compared to CT2 (Figure S2). Given the relationship between TLS and antigen-driven immune responses [13], TLS in CT1 may represent antigen-driven responses to local microbiota, chronic inflammation markers, or sites of pathogen-focused antibody production (e.g., anti-Corynebacterium IgA). These findings suggest that certain dominant microbiota may influence the local immune environment in CT1 patients. Overall, our results indicate that distinct microbial states exist in CRSwNP subgroups with varying clinical and pathological features. We identified a microbial community (CT1) dominated by Corynebacterium that was associated with high eosinophil infiltration in the nasal mucosa. This finding aligns with previous studies, such as those by Cope et al., who reported a Corynebacterium-dominant subgroup linked to increased nasal polyps and elevated IL-5 transcription [6]. Similarly, Bassiouni et al. found a Corynebacterium-rich microbiotype in patients with aspirin-exacerbated respiratory disease [8]. Both studies, conducted in predominantly Caucasian populations, reported that Corynebacterium was associated with type 2 immunity, consistent with our findings. The Staphylococcus distribution contrasts with several Western population studies. This may be attributable to: (a) Geographic and ethnic variations (East Asian vs. Western populations); (b) Differences in clinical phenotypes (e.g., inclusion of CRSwNP only in our study); (c) Potential variations in sampling protocols despite similar methodologies [6−8]. In addition, a previous multi-center study reported a higher prevalence of Staphylococcus-dominated microbiomes in European compared to Asian centers [8]. The microbial composition of CT2 was similar to Cope's DSI subgroup, which also featured Streptococcus, Porphyromonas, and Bacteroides and resembled healthy controls. Consistently, CT2 exhibited milder clinical severity and lower eosinophil infiltration. Notably, CT1 patients exhibited elevated Lund-Mackay scores and eosinophil infiltration but comparable SNOT-22 symptom scores to CT2. Objective metrics reflect structural and inflammatory status, whereas SNOT-22 captures broader patient-reported outcomes, including quality of life, sleep, and non-nasal symptoms, which may be influenced by individual variability in symptom perception. To validate the microbial community classifications, patients with CRSwNP (n = 32) were recruited from the same institution under identical exclusion criteria. No significant demographic differences (age, sex, comorbidities) existed between primary and validation cohorts (Table S3). Validation cohort samples were clustered independently without prior model training, and then a semi-supervised clustering approach identified two subtypes (Cluster 1: n = 19, 59%; Cluster 2: n = 13, 41%). Cluster 1, dominated by Corynebacterium, Peptoniphilus, and Anaerococcus, resembled CT1, while Cluster 2, enriched in Porphyromonas, Bacteroides, and Streptococcus, mirrored CT2 (Figure 2A). Consistent with our main dataset, Cluster 1 exhibited lower α-diversity (Shannon index) than Cluster 2 (Figure 2B). Additionally, higher CT scores and eosinophil levels were associated with Cluster 1, validating the clinical and pathological characteristics observed in CT1 and CT2 (Figure 2C). This supports the robustness of our microbial community classification and its relevance to clinical and pathological features. Potential clinical application of microbial stratification has been performed in other respiratory mucosal diseases. For example, Zhang et al., who used SourceTracker and DMM modeling to identify two pneumo-types influenced by oral microbes, which correlated with lung function and inflammatory status [14]; another study applying Wald linkage clustering to define four microbiome community types in neutrophilic COPD, in which these microbial subtypes were associated with disease clinical stability and exacerbation frequency [15]. Moreover, Liang et al. used a combination of 15 genera to distinguish patients with eosinophilic NP from healthy control, indicating compositional alterations in the nasal microbiome rather than individual genera could be used to classify disease subgroups [16]. 16S rRNA sequencing restricts strain-level resolution and functional validation. Shotgun metagenomics, transcriptomics and metabolomics approaches are needed to reveal mechanisms linking specific microbial functions to host inflammation in CT1/CT2 subtypes. Cross-sectional design precludes causal inference between dysbiosis and inflammation, and longitudinal studies tracking microbial dynamics and disease progression, as well as experimental models, are needed to infer causation between microbial communities and disease severity. Single-center recruitment limits generalizability; validation in multi-ethnic cohorts is essential. In future study, a larger healthy cohort would also be valuable for establishing a robust baseline (or commensal) nasal microbial profile, which can facilitate to stratify CRSwNP patients by microbial composition patterns. Nonetheless, microbial community-type clustering is a promising approach for classifying patients into subgroups, allowing for the exploration of clinical and pathological correlations. This strategy has the potential to contribute to the development of precise, tailored therapies for CRSwNP patients, addressing the disease's heterogeneity. Bixing Fang: Validation; formal analysis; supervision; writing—original draft; resources; methodology. Yanmin Liu: Conceptualization; methodology; software; data curation; validation. Qing Bao: Investigation; conceptualization. Lin Sun: Conceptualization. Hexin Chen: Supervision; data curation. Qingxiang Meng: Data curation; supervision. Linyi Jiao: Methodology; formal analysis. Xiaoyan Ye: Methodology. Jian Li: Funding acquisition; visualization; project; administration; resources; writing—original draft; writing-review & editing. Tao Ding: Methodology, conceptualization; software; data curation; writing—review & editing. Chunwei Li: Data curation; investigation; writing—review & editing. All authors have read the final manuscript and approved it for publication. This study was supported by grants from the Guangxi Science and Technology Major Program (Guike AB24010081), Guangxi Science and Technology Major Program (Guike AA24206008), National Natural Science Foundation of China (82171112), Zhejiang Provincial Natural Science Foundation of China (ZCLQN25H1302), and Joint TCM Science & Technology Projects of National Demonstration Zones for Comprehensive TCM Reform (GZY-KJS-ZJ-2025-009). We are indebted to the staff and participants involved in this study. We apologize for not being able to cite additional work owing to space limitations. The authors declare no conflicts of interest. This study was conducted in accordance with ethical standards and received approval from the Institutional Review Board of the First Affiliated Hospital of Sun Yat-sen University (approval number [2017] 303). Informed consent was obtained from all participants prior to their inclusion in the study. All data collected were handled in strict confidentiality, ensuring the privacy and anonymity of the participants. The 16S rRNA sequencing data (including Validation dataset for microbial community classifications) from CRSwNP patients and controls are publicly accessible through the National Genomics Data Center (NGDC) GSA repository. The accession number for our dataset in the NGDC GSA repository is GSA: CRA023332 (https://bigd.big.ac.cn/gsa/browse/CRA023332). Intermediate data files associated with the study can be accessed at the following URL: https://github.com/cina0707/CRSwNP. Supplementary materials (methods, figures, tables, graphical abstract, slides, videos, Chinese translated version, and update materials) may be found in the online DOI or iMeta Science http://www.imeta.science/imetaomics/. Table S1: Demographical and clinical pathological characteristics of the subjects. Table S2: Correlation between community types and demographical and clinical pathological variables. Table S3: Comparison of demographic characteristics between the test sets and the verification sets. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.