This study aimed to characterize the microbiome signature of sinonasal tissues from healthy controls, subjects with CRSsNP, and those with CRSwNP using different collection methods (i.e. nasal swab vs. tissue biopsy) and tissue types (UT vs. NP). We demonstrated that tissue samples may be more optimal specimens for assessing the microbiome of CRS than are nasal swabs. Tissues provided less variation between samples; moreover, in UTs different microbial compositions were associated with disease subtype. In addition, NP tissues had a unique microbiome that was clearly differentiated from UT and revealed a strong association with clinical severity of CRSwNP. To our knowledge, this is the first study to use UT to assess the microbiome of patients with CRS.
An advantage of using UT is the relative simplicity of sample collection without causing significant morbidity. Therefore UT can be easily acquired regardless of disease status. The immunological profile of UT has been demonstrated to differ according to disease status. For instance, the eosinophil cationic protein (ECP) level in UT is strongly correlated with overall disease severity, comorbid asthma, and the risk of polyp recurrence(25, 26). Furthermore, specific immunological characteristics of UT in different disease subtypes (6, 27) seem to be associated with the distinguishing microbiome profile.
Recent studies of CRS have found dysbiosis of nasal microbiota. Despite an overall decrease of microbial richness, evenness, and diversity in CRS, inconsistencies or biases in previous studies limited the identification of clinically relevant genera or species (14, 28, 29). Furthermore, an optimal sampling method of the nasal microbiome was not yet determined. To date, most studies have utilized nasal swabs, primarily collecting from the middle meatus. The middle meatus is an intersectional area that shares a common drainage pathway with other paranasal sinuses. Therefore, samples from this area may reflect multiple sites, which could produce increased heterogeneity. However, another study comparing microbial compositions from swab samples collected at different anatomical sites from patients with CRS still demonstrated a high interpersonal variation, which outweighed location-specific differences.11 Our study found similar results, where a high interpersonal variability of swab samples may obscure the differences among disease subtypes.
Tissue samples from the UT and NP were superior to swabs and demonstrated decreased interpersonal differences within subgroups. This allowed for the identification of discrete tissue-associated microbiomes. The high variation in swab samples could be a function of the vulnerability of the nasal surface microbiome to environmental changes.
Our results are further supported by a previous study that found significantly different bacterial compositions in tissue specimens as compared to swab samples (30). Tissue specimens are more inclusive than swabs, as they incorporate bacterial biofilms that grow on the surface and bacteria that penetrate the mucosal epithelium. This is supported by the observation that tissue samples have a greater biomass than swabs, although non-microbial DNA is present (31). Thus, tissue samples should be given greater attention in the future as a collection method.
In CRS, the airway microbiome composition is associated with inflammatory profiles. For example, CRSwNP accompanied by asthma presented with significantly higher levels of Th2-related cytokines. Simultaneously, there was a higher abundance of Proteobacteria than in patients with CRSwNP without asthma. Furthermore, Corynebacterium, which are members of the phylum Actinobacteria, were more abundant in CRSwNP without asthma (32). There is an upregulation of several important cytokines, such as IL-5, in NP compared to UT regardless of disease subtype (6, 27). Therefore, the distinct microbial composition of NP compared to that of UT in our study could be associated with differences in inflammatory profiles.
In the NP, the relative abundance of the phylum Firmicutes was remarkably lower than in the UT, whereas the abundance of the phylum Proteobacteria was higher. Despite the use of different sampling methods, the higher abundance of Proteobacteria than that in controls has been consistently reported. At the genus level, Haemophilus, Escherichia, and Moraxella were highly enhanced in CRSwNP as compared to controls (33–35). Furthermore, in our study, Sphingomonas, a genus in the phylum Proteobacteria, was significantly enriched in the NP as compared to in the UT. In a murine asthma model, glycoproteins from Sphingomonas induced type 2 inflammation via natural killer T cells in an IL-4-, IL-13-, and IL-33-dependent manner (36, 37). Interestingly, Sphingomonas was found to be more abundant in bronchoalveolar lavage fluid from patients with eosinophil-high asthma than in those with eosinophil-low asthma; this is also related to increased airway hyperresponsiveness to methacholine (38, 39).
We also evaluated the severity of CRS using LM scores, a widely used radiological parameter assessed by CT. The score increases with certain markers of disease severity, such as the increasing grade of polyposis, nature of surgery offered (i.e. more extensive surgery), and treatment outcome (16). Correlation between genera and disease severity differed by both tissue and disease subtype. This indicates unique interactions of microbiome according to the different microenvironmental conditions.
In the NP microbiome, the genus Prevotella was significantly and inversely correlated with disease severity. Similar findings have been observed in other inflammatory diseases including multiple sclerosis (MS). In that study, the abundance of gut Prevotella was reduced in untreated patients with MS, and treatment with disease-modifying therapy was associated with an increased relative abundance of Prevotella (40). This protective effect of Prevotella appears to be mediated through the induction of CD4+ FoxP3+ regulatory T cells (41).
There were several limitations of our study. First, the number of patients was rather small, considering the heterogeneity of CRS. Second, we did not analyze the differences between eosinophilic NP and non-eosinophilic NP, which are known to exhibit immunological differences (27). Third, 16S rRNA analysis only reveals the presence of bacteria, but it does not analyze their metabolic activity. Fourth, tissue samples dominated by non-microbial DNA can introduce contamination during amplicon sequencing. Therefore, further validation studies with larger sample sizes are necessary. Future studies should also identify interactions between the tissue microenvironment and microbes in association with immunological profiles.