Comparison of subtyping approaches and the underlying drivers of microbial signatures for chronic rhinosinusitis

Chronic rhinosinusitis (CRS) is a heterogeneous condition characterised by persistent sinus inflammation and microbial dysbiosis. This study aimed to identify clinically relevant sub-groups of CRS patients based on distinct microbial signatures, with a comparison to the commonly used phenotypic subgrouping approach. The underlying drivers of these distinct microbial clusters were also investigated, together with associations with epithelial barrier integrity. Sinus biopsies were collected from CRS patients (n=23), and disease controls (n=8). Expression of 42 tight junction genes was evaluated using quantitative PCR, together with microbiota analysis and immunohistochemistry for measuring mucosal integrity and inflammation. CRS patients clustered into two distinct microbial sub-groups using probabilistic modelling Dirichlet (DC) multinomial mixtures. DC1 exhibited significantly reduced bacterial diversity, increased dispersion, and was dominated by Pseudomonas, Haemophilus, and Achromobacter. DC2 had significantly elevated B-cells, incidence of nasal polyps, and higher numbers of Anaerococcus, Megasphaera, Prevotella, Atopobium, and Propionibacterium. In addition, each DC exhibited distinct tight junction gene and protein expression profiles compared with controls. Stratifying CRS patients based on clinical phenotypic subtypes (absence or presence of nasal polyps (CRSsNP or CRSwNP respectively) or with cystic fibrosis (CRSwCF)) did account for a larger proportion of the variation in the microbial dataset compared with DC groupings. However, no significant differences between CRSsNP and CRSwNP cohorts were observed for inflammatory markers, beta-dispersion and alpha diversity measures. In conclusion, both stratification approaches used had benefits and pitfalls, but DC clustering did provide greater resolution when studying tight junction impairment. Future studies in CRS should give careful consideration into the patient subtyping approach used.