Non-steroidal anti-inflammatory drug (NSAID)-exacerbated respiratory disease (N-ERD) affects 8-10% of patients suffering from chronic rhinosinusitis with nasal polyps (CRSwNP) (1). In addition to nasal polyps, N-ERD patients suffer from asthma and hypersensitivity to NSAIDs. The disease burden is evidenced by a high incidence of polyp recurrence and severely impaired quality of life (2, 3). In the US and Europe, patients with nasal polyps show predominantly a type 2 inflammation profile, marked by high local T helper 2 cytokine levels and a strong eosinophilic influx in the nasal polyp tissue (4). This pattern is even more pronounced in N-ERD patients, where additionally increased levels of nasal IL-5 and eosinophilic cationic protein (ECP) are observed as compared to aspirin-tolerant CRSwNP patients (5, 6). A recent study employing global Transcriptomics in combination with genome-wide methylomics and targeted metabolomics has demonstrated that these cells display an overall reduction in DNA methylation, combined with aberrant metabolic profiles and increased chemokine expression in N-ERD patients (7). Therefore the persistent metabolic and epigenetic reprogramming in macrophages could play an important role in the pathomechanism of N-ERD.

In the last decade, monoclonal antibodies targeting mediators of the type 2 inflammation pathway have become available and are a therapeutic add-on for patients with severe, refractory disease (8, 9). For example, omalizumab, a monoclonal antibody targeting free IgE, reduces the polyp burden (10) and induces complete aspirin tolerance in more than 60% of patients (11, 12). The IL-4 receptor α chain-blocking antibody dupilumab, which targets both the IL-4 and IL-13 receptors, also improves aspirin tolerance (13) and leads to significant clinical improvement in N-ERD patients (14, 15). Furthermore, it leads to a significant reduction of type 2 inflammation associated biomarkers such as Eotaxin-3, IgE and ECP in nasal secretions (13, 16). In tissue homogenates, ECP, eotaxins and pulmonary and activation-regulated chemokine (PARC) are also significantly reduced after dupilumab therapy (16). Thus, it is thought that dupilumab reduces type 2 inflammation, IgE production, drivers of and eosinophilic inflammation locally in nasal tissue.

CRSwNP, N-ERD and other chronic type 2 inflammatory diseases such as type 2 asthma or atopic dermatitis (AD) are associated with epithelial barrier dysfunction. This epithelial impairment can lead to colonization with opportunistic bacteria such as Moraxella, Haemophilus, or Staphylococcus aureus ( 17), the latter being considered a major contributor to symptom burden in all above-mentioned diseases (18). In this context, recent findings have shown that treatment with dupilumab does not only lead to a significant clinical improvement in AD patients, but also to improved skin barrier function accompanied by reduced colonization with Staphylococcus aureus in both lesional and non-lesional skin (19–21). This reduction was shown to occur as early as 3 days after the beginning of treatment (22). However, whether and to what extent blockage of IL-4 and IL-13 may alter the respiratory microbiome in CRS or asthma patients is currently unknown (23). A recent cultivation based study assessing microbial profiles in CRSwNP patients after surgery or dupilumab indicates that Staphylococcus aureus prevalence remained stable under IL-4Rα blockage, while it increased in patients after surgery (24).

In this study, we longitudinally assessed the effect of 24 weeks of dupilumab treatment on the composition of the nasal microbiome in patients suffering from N-ERD. Patients with N-ERD have very high levels of type 2 inflammation (6) and the cohort used for this study has been carefully selected and well characterized (13). Furthermore, we previously showed that N-ERD patients on average display a higher abundance of Staphyloccus aureus than CRSwNP or CRSsNP patients (25). As Staphylococcus aureus is thought to play an important role in driving the disease, this group is very well suited to investigate changes in this amplicon sequence variant (ASV) under dupiluma therapy. As the nasal microbiome is a low (microbial) biomass habitat, making samples particularly susceptible to sampling, handling, and sequencing contamination (26–28), we included numerous sampling and DNA extraction controls and applied stringent contamination filtering workflows to ensure the validity of our data.

Here we report on the nasal microbiome composition and longitudinal changes therein in an initial cohort of 31 N-ERD patients undergoing dupilumab treatment for 24 weeks. After stringent decontamination and filtering, samples from 8 patients with full microbiome datasets were included for downstream analysis. We found significant clinical and nasal biomarkers improvement, although no major consistent changes in microbiome composition and diversity during dupilumab therapy were observed. Selected, predominantly relatively low abundant ASVs, showed significant changes in pairwise abundance testing. While patients showed a general trend toward increased abundance of Staphylococcus ASVs, two patients displayed a severe drop in Staphylococcus abundance independently development of aspirin tolerance after 24 weeks of therapy.

Out of the samples from 28 patients subjected to 16S rRNA gene amplicon sequencing, full datasets of only 8 patients (29%) remained for the final analysis. This considerable loss in data is due to the known issue of very low initial microbial biomass of nasal samples (27). In another study involving nasopharyngeal swabs of 82 patients in the context of tuberculosis, a recovery rate of only 53% of samples using comparable filtering criteria was achieved (39). Including all samples in the downstream analysis, regardless of read counts and presence or absence of air and reagent control contaminations, bears the substantial risk of non-proportional target amplification of contaminant DNA deriving from environmental or reagent sources, which can severely impact the outcome (40). In this context, Salter et al. demonstrated in a longitudinal study of children that the differences in nasal microbial composition were not - as initially thought - age-dependent, but could be attributed solely to the usage of four extraction kits bearing different contaminations during the study (41). Thus, to avoid potential sampling or experimental bias, we employed in the present study stringent criteria as proposed in the “RIDE” checklist (26, 42) thereby accepting the exclusion of more than 70% of the samples. This high loss shows that there is an urgent need to test and develop alternative methods for nasal sampling, ideally yielding higher biomass (43), and to implement mandatory guidelines for stringent quality control during sampling allowing for comparison between published results (27). Methodologic studies in saliva microbiome, have shown that stimulated (i.e. paraffin-stimulation) and unstimulated saliva as well as simple mouth wash show comparable results with regards to microbiome diversity (44, 45). Similar studies are warranted in the field of nasal sampling comparing i.e. brushes, scrapes or nasal washes with currently used swabs using appropriate controls.

The impact of CRS on the overall diversity of the nasal microbiome has yet not fully been elucidated: While some studies observed a reduced microbial diversity in CRS-affected patients (46–49), we and others could not detect any statistically significant difference in well-established diversity indices (25, 50, 51). However, there is clear evidence for the role of Staphylococcus aureus and corynebacteria in the pathogenesis of CRS and their abundance is associated with type 2 mediator levels (25, 52–54). Due to the tight interaction between the microbiome, epithelium, and type 2 cytokines, it is conceivable that dupilumab may also have an influence on barrier function and, as a consequence, on microbial composition. In AD, a significant reduction of Staphylococcus aureus colonization of the skin was demonstrated during dupilumab treatment regardless of whether samples were taken from healthy skin or lesions (19–22, 55). Interestingly, this reduction was observed as early as 3 days after the start of treatment and preceded clinical improvements or changes in serum CCL17, a marker of type 2 disease, by 11 days (22). These results suggest an important role of the microbiome in shaping the epithelial barrier. However, data regarding the effect of dupilumab on nasal microbiome are scarce: to date only one study performed in AD patients also collected swabs from the anterior nares before and 16 weeks after dupilumab therapy (20). They initially observed a significant increase of S. hominis in patients achieving significant disease improvement, but subgroup analysis showed that this result was mainly driven by a few patients. This is in accordance with our observation that there is a strong inter-individual difference between patients with regards to relative abundance changes of Staphylococcus associated ASVs during dupilumab therapy.

In contrast to the reported decrease in Staphylococcus aureus colonization of skin in AD patients during dupilumab therapy, we observed no consistent significant changes in the relative abundance of staphylococci in this cohort. This is in line, a recent study showing that dupilumab as opposed to surgery stabilizes Staphylococcus aureus prevalence using a cultivation-based approach (24). However, our longitudinal study design does show that large, individuum-specific fluctuations in the relative abundance of Staphylococcus-related ASVs occur during dupilumab therapy. One explanation for the observed discrepancy between observations made in nasal and oral samples may be the differential microbial composition between nose and skin: In the nasal cavity, staphylococci usually shows (depending on the underlying inflammatory disease) a lower abundance of 2-30% within the local microbiome (25, 56, 57) as compared to AD, where staphylococci make up 75% or more of the skin microbiome (22, 58). Thus, as relative Staphylococcus aureus abundance drops in the skin are on average around 10% in non-lesional skin during dupilumab therapy (21, 22), such relatively small changes may not be detectable even if present in our cohort due to the small sample size, large intra-individual variability and initially low S. aureus abundance in many patients. Based on the results reported here, including the lack of any association of changes in Staphylococcus relative abundance with clinical parameters such as the development of aspirin tolerance, we propose that the link between S. aureus colonization and disease burden might not be as strong or omnipresent in type 2 diseases affecting respiratory epithelia as it is for those affecting skin, like AD.

The quest for biomarkers for evaluation of treatment success of biological treatment is ongoing. Here, we observed a significant reduction of type 2-associated biomarkers IgE and eotaxin during therapy. In this line, it would have been interesting to assess also non-typical biomarkers such as IL-24, which is tightly associated with IL-4 stimulation (59). Interestingly the effect of IL-4 can be directly counteracted by IFN-γ stimulation and this could be exploited as a novel treatment option for severe CRS. In this line, exogenous IFN-gamma application has been shown to lead to a better symptom control in a small scale study in ten treatment-resistant CRS patients with dysregulated IFN-γ production (60).

Limitations of the current study include the small initial participant number due to the design as a pilot trial and that the initial samples size calculation was based on the clinical primary endpoint of maximally tolerated aspirin levels (13). To calculate the number of patients required for detecting significant changes in the total microbial community in future larger trials, we conducted a power calculation based on Mattiello et al (61). The distribution from the included “anterior nares” dataset was applied; the estimated number of taxa was set to 50 (based on assuming a maximally double as high diversity as observed in our previous study (25)); and 5% of the most abundant taxa were assumed to be either 50% more or 50% less abundant each. At a sample size of n=90 and n=30 for case vs control this resulted in a power of 0.87, 0.95, and 0.51 for the significant detection of changes in the total microbial community, in the abundance of at least one taxon, and in the abundance of all taxa, respectively. However, the sample size of 90 patients with N-ERD and treated with dupilumab is beyond the scope of a pilot trial. This was further reduced by stringent quality control. A confounding factor might be the continuation of standard CRS therapy during the clinical trial which includes topical corticosteroid therapy. But due to the severe burden in N-ERD patients, mainstay therapy had to be continued following current guidelines (9). As in the current sub study we included clinical data of only 8 patients of the initial trial involving 31 subjects, the higher tolerance development toward aspirin observed as compared to our previous observation (13) is most likely due to the small sample size. Another limitation is that patients also received standard nasal corticosteroids for the duration of the study. This may have partly confounded our findings however due to the severe symptom burden of N-ERD patients it was not possible to stop nasal corticosteroids for the duration of the trial. Importantly, none of the patients received oral steroids or antibiotics for the duration of the study.

In summary, we showed in this pilot trial, that dupilumab does not affect the overall diversity of the nasal microbiome, but may lead to alterations in microbiome composition, including changes in Staphylococcus relative abundance in selected individuals. The fact that more than 70% of the samples did not yield enough biomass to be processed warrants further refinement of nasal microbiome sampling techniques and mandatory guidelines on stringent quality control for analysis of this low biomass data.