This study was approved by the Ethics Committee of Tongji Hospital and conducted with written informed consent from every participant. Eighty-two control subjects and 375 patients with CRSwNP were enrolled. CRSwNP was diagnosed according to the current guidelines (Fokkens et al., 2020; Orlandi et al., 2021). Control subjects were those undergoing septoplasty because of anatomic variations without any other inflammatory sinonasal disorders. During the surgery, polyp, and inferior turbinate tissue samples were collected from CRSwNP patients and control subjects, respectively. The details of subjects’ characteristics are presented in Supplementary Table SE1. Of note, there was not sufficient tissue from each subject for every analysis. The sample size for each experiment is indicated in the corresponding figures and figure legends.

Among CRSwNP patients, 249 of them were involved in the clinical and immunopathological feature study. Baseline symptoms were scored by visual analog scale (VAS) method as previously described (Fokkens et al., 2020). Endoscopic findings were scored according to the Lund-Kennedy method (Fokkens et al., 2020). Computed tomography (CT) scans were graded based on the Lund-Mackay scoring system (Fokkens et al., 2020). After surgery, patients were routinely given fluticasone propionate nasal spray 200 μg twice daily, and saline irrigation twice daily. Subjects with uncontrolled symptoms were given 1–2 weeks of broad-spectrum antibiotics (usually oral cephalosporin), and/or oral prednisone tablets for 20 days (30 mg daily on days 1–5, 20 mg daily on days 6–10, 10 mg daily on days 11–15, and 5 mg daily on days 16–20) (Liao et al., 2018; Fokkens et al., 2020; Chen et al., 2021). None of patients received macrolides. Patients who did not reach an acceptable level of control (controlled or partially controlled) despite adequate surgery, intranasal corticosteroid treatment, and up to two short courses of antibiotics or systemic corticosteroids in the last year were considered to have difficult-to-treat disease (Fokkens et al., 2020).

To validate the rationality of using inferior turbinate tissue as a control in neutrophil infiltration study, normal ethmoid sinus tissues were obtained from 30 subjects receiving surgery for sinus cyst, nasal tumour or maxillofacial trauma and without rhinosinusitis or rhinitis. The neutrophil infiltration was compared between inferior turbinate and ethmoid sinus tissues. In addition, 46 healthy subjects were recruited for the peripheral blood neutrophil isolation and culture study (Supplementary Table SE2). More information is provided in the Online Supplement.

Hematoxylin and eosin (HE) and immunofluorescence staining, and cell quantification were performed as previously described (Cao et al., 2009; Wang et al., 2018). The 95th percentile of numbers of tissue myeloperoxidase (MPO) positive neutrophils per high-power field (HPF) in control subjects was selected as a cut-off to define neutrophil-high (Neu-high) or neutrophil-low (Neu-low) NPs. More information is provided in the Online Supplement including Supplementary Table SE3 and Supplementary Table SE4.

Surgical tissue samples were weighed and homogenized (Liao et al., 2018). MPO and eosinophilic cationic protein (ECP) levels in tissue homogenates were measured using commercial ELISA kits (Feiya Biological Technology, Jiangsu, China). The protein levels of 34 inflammatory mediators including cytokines, chemokines, and immunoglobulins were quantified with the Bio-Plex suspension chip method (Bio-Rad, Hercules, Calif, United States) (Liao et al., 2018). Concentrations of detected mediators were normalized to total tissue protein levels. More information is provided in this article’s Online Supplement including Supplementary Table SE5.

NPs were classified into T1, T2, and T3 endotypes when the protein levels of IFN-γ, IL-5 and IL-17 A were higher than the 95th percentile of the corresponding cytokine levels in control tissues (cut-off value), as previously described (Tomassen et al., 2016; Chen et al., 2021). When a sample showed levels of two or three of these cytokines above the cut-off values, it was considered as a double or triple mixed type. A sample was defined as all negative when the levels of all three cytokines were below the cut-off values.

Neutrophils were purified from peripheral blood of healthy subjects as previously described (Wilkinson et al., 2018; Grunwell et al., 2019). The purity of isolated CD16⁺Siglec-8^- neutrophils was greater than 97% as determined by flow cytometry (Supplementary Figure SE1). 1×10⁶ neutrophils were incubated with 100 μg/ml homogenates of control tissues, Neu-high NPs, or Neu-low NPs at 37°C for 8 h. Anti-G-CSF (2 μg/ml, R and D Systems, Minneapolis, MN, United States), anti-IL-6 (2 μg/ml, R and D Systems), or isotype control IgG (2 μg/ml, R and D Systems) (Supplementary Table SE6) was added to certain culture experiments with homogenates of same Neu-high NP samples. Apoptotic, live, and necrotic neutrophils were analyzed by flow cytometry after culture. More information is provided in this article’s Online Supplement.

The flow cytometric analysis of blood and cultured neutrophils was performed as previously described (Wilkinson et al., 2018). Please refer to the Online Supplement for more information.

Kolmogorov-Smirnov test or Shapiro-Wilk test was used to test the normality of the data distribution. As all data were non-normally distributed, a Kruskal-Wallis test with the Dunn post hoc test was used to adjust for multiple comparisons, and binary comparisons were carried out with a Mann-Whitney U 2-tailed. Cell culture data are analyzed by Friedman test or Kruskal-Wallis test, multiple comparisons were carried out using Dunn post hoc test. Unless otherwise defined, data are shown as dot plots with horizontal bars representing the medians and error bars showing the interquartile ranges. The correlations were analyzed with a non-parametric Spearman correlation test. For categorical variables, a chi-square test or Fisher’s exact test was used to assess the difference between groups. A logistic regression model was applied to identify risk factors associated with difficult-to-treat disease. SPSS version 22.0 (SPSS Inc., Chicago, IL, United States) software was used to perform the above statistical analyses. In addition, principal component analysis (PCA) and multiple correspondence analysis (MCA) were performed using the R package “devtools” and “MASS,” respectively. Unsupervised hierarchical clustering of variables (Spearman correlation) was performed using R package “pheatmap,” and represented as a dendrogram. A p value of less than 0.05 was considered significant.

We first investigated the tissue neutrophil distribution in control subjects and CRSwNP patients. We found that control inferior turbinate and control ethmoid sinus mucosal samples had comparable numbers of MPO⁺ neutrophils and MPO levels (Supplementary Figures SE2A,B). Given the larger amount of samples generally obtained from inferior turbinates, we used inferior turbinate tissues as a control for further study. We found that the numbers of MPO⁺ neutrophils/HPF were not normally distributed in either control or CRSwNP subjects (Figure 1A). The numbers of neutrophils/HPF in control subjects ranged from 0 to 11.3, with a median of 3.4, 5^th percentile of 0.4, and 95th percentile of 10.5 (Figure 1A). The numbers of neutrophils/HPF in NPs ranged from 0 to 78.0, with a median of 11.0, 5^th percentile of 2.0, and 95th percentile of 45.0 (Figure 1A). We used the 95th percentile of tissue neutrophil numbers in control subjects (>10 MPO⁺ cells/HPF) to define Neu-high NPs. We found that 53% (131/249) of NPs were classified as Neu-high NPs (Figure 1A) and the numbers of MPO⁺ neutrophils were significantly higher in Neu-high NPs than in control tissues and Neu-low NPs (Figure 1B). To further validate this stratification, we analyzed MPO levels in nasal tissues in another set of patients and found that MPO levels were significantly increased in Neu-high NPs in comparison with Neu-low NPs and control tissues (Supplementary Figure SE3A).

Patients with Neu-high NPs had a lower frequency of asthma comorbidity than patients with Neu-low NPs (Table 1). There was no significant difference regarding age, gender, atopy, allergic rhinitis comorbidity, prior surgery, or disease duration between these two patient groups (Table 1). Compared to patients with Neu-low NPs, patients with Neu-high NPs had milder smell dysfunction, lower posterior ethmoid sinus CT scores, and lower blood eosinophil counts and percentages (Table 1). There was no difference in endoscopic scores, including NP scores, between patients with Neu-low and Neu-high NPs.

By evaluating HE-stained tissue sections, we found that Neu-low NPs had higher numbers of eosinophils than Neu-high NPs and control tissues, and there was no difference between Neu-high NPs and control tissues (Supplementary Table SE7). Although both Neu-low and Neu-high NPs had higher numbers of plasma cells and mononuclear cells than in control tissues, there was no difference between Neu-low and Neu-high NPs (Supplementary Table SE7). To confirm the results of eosinophil infiltration, we analyzed the ECP levels in tissues in another set of patients and found that Neu-low NPs had increased ECP levels compared to Neu-high NPs and control tissues (Supplementary Figure SE3B).

Using the Bio-Plex suspension chip method, we measured the protein levels of 34 inflammatory mediators in sinonasal tissues (Supplementary Table SE8). Those having different expression in at least 1 of the 3 subject groups as compared to other groups are shown in the heatmap in Figure 2A, along with tissue neutrophil and eosinophil numbers. The expression levels of selected mediators are shown in Figure 2B. Neu-high NPs displayed increased levels of neutrophil-relevant cytokines and chemokines, including G-CSF, IL-8, IL-6, MCP-1, MIP-1α, IL-1β, and IL-1Ra compared with Neu-low NPs and control tissues (Figures 2A,B). In contrast, the levels of eosinophilic inflammation associated markers including IL-5, IL-13, and IgE were higher in Neu-low NPs than in Neu-high NPs and control tissues (Figures 2A,B). The levels of IFN-γ and IL-17A were increased in both Neu-low and Neu-high NPs compared with control tissues and no significant difference between Neu-high and Neu-low NPs was found (Figures 2A,B).

We next classified Neu-high and Neu-low NPs into T1, T2, and T3 endotypes based on tissue levels of IFN-γ, IL-5, and IL-17A, respectively. We found that the T1, T2, and T3 endotypes accounted for 55.4, 32.4, and 68.0% of Neu-high NPs, respectively, which was distinct from Neu-low NPs (53.7, 66.5, and 46.8% for the T1, T2, and T3 endotypes, respectively) (Figure 2C). Further MCA demonstrated that Neu-high NPs were located near non-T2 and T3 endotypes, whereas the Neu-low NPs were situated in the middle of T1 and T2 endotypes (Figure 2D). We also performed PCA analysis based on the 21 biomarkers listed in the heatmap, and found that control tissues, Neu-high NPs, and Neu-low NPs were mostly separated from each other (Figure 2E). These comprehensive analyses suggest Neu-high and Neu-low NPs have distinct immunopathological features.

To explore the factors driving neutrophilia in NPs, we analyzed the correlations of tissue neutrophils and eosinophils with blood leucocytes and tissue inflammatory mediators. By unsupervised hierarchical cluster analysis, we found that IL-8, G-CSF, IL-1β, bFGF, and tissue neutrophil clustered together in both Neu-high NPs and Neu-low NPs, whereas IL-5, IL-13, IgE, IgG4, tissue eosinophil, and blood eosinophil clustered together in Neu-low NPs (Figures 3A,B). We found that the number of tissue neutrophils were positively correlated with the levels of IL-1β, IL-1Ra, IL-6, IL-8, IL-33, bFGF, G-CSF, MCP-1, and MIP-1α in both Neu-high and Neu-low NPs (Figures 3A,B). In contrast, the number of tissue neutrophils was negatively correlated with the number of tissue eosinophils in Neu-high and Neu-low NPs (Figures 3A,B). The blood neutrophil counts, blood total leukocyte counts, blood lymphocyte percentages, blood neutrophil percentages, tissue MIP-1β levels, and tissue IL-17A levels were positively correlated with tissue neutrophil numbers only in Neu-high NPs (Figure 3A). When combining Neu-high and Neu-low NPs together, we found that tissue neutrophil numbers were positively correlated with tissue levels of IL-8, G-CSF, IL-1β, IL-6, IL-1Ra, MIP-1α, MCP-1, bFGF, IL-33, and IL-17 A (Figure 3C); however, the correlations with IL-1Ra, bFGF, MIP-1α, MCP-1, IL-33, and IL-17A were very weak, with r values less than 0.3. The tissue neutrophil numbers were weakly and negatively correlated with eosinophil numbers in total NPs (Figure 3C). In addition, tissue levels of IgE, IL-5, and IL-13, and blood eosinophil counts and percentages were positively correlated with tissue eosinophil numbers in both Neu-high and Neu-low NPs (Figures 3A,B).

Tissue neutrophilia is mainly determined by the amount of recruitment and survival of neutrophils. Increased expression of neutrophil chemokine IL-8 has also been noted in previous studies. (Van Zele et al., 2006; Wang et al., 2016). In contrast, the apoptotic property of neutrophils in NPs remains poorly understood. The increased expression of survival-promoting factors, such as G-CSF, in Neu-high NPs prompted us to further investigate this finding. We found that the percentage of active caspase-3⁺MPO⁺ apoptotic neutrophils in total neutrophils was significantly lower in Neu-high NPs than in control tissues and Neu-low NPs (Figures 4A,B). The representative photomicrographs showing the isotype control staining are presented in Supplementary Figure SE4. In addition, we found that the percentage of apoptotic neutrophils was negatively correlated with the levels of IL-6 and G-CSF in NPs (Figure 4C). We failed to find any correlation of apoptotic neutrophils with other inflammatory mediators tested in NPs (data not shown).

Given the observation of reduced neutrophil apoptosis in Neu-high NPs, we hypothesized that the inflammatory milieu in Neu-high NPs could promote the survival of neutrophils. To test this possibility, peripheral blood neutrophils isolated from healthy subjects were cultured with tissue homogenates from the different subject groups. Dexamethasone was used as a positive control. Consistent with a previous study (Cox, 1995), we found that the frequency of Annexin-V⁺7-aminoactinomycin D (AAD)^- apoptotic neutrophils was decreased whereas the frequency of Annexin-V^-7⁻AAD^- live neutrophils was increased after dexamethasone treatment when incubated in tissue homogenates of control tissues (Supplementary Figure SE5). Further, we found that the percentage of apoptotic neutrophils was lower whereas the percentage of live neutrophils was higher in neutrophils incubated with tissue homogenates of Neu-high NPs than those incubated with tissue homogenates of control tissues and Neu-low NPs (Figures 5A–E). No difference regarding the frequency of Annexin-V⁺7-AAD⁺ necrotic neutrophils was found among different experimental groups (Figure 5E).

Given the negative correlations between the frequencies of apoptotic neutrophils, G-CSF, and IL-6 levels in NPs, we next explored whether G-CSF or IL-6 was involved in the regulation of neutrophil survival in NPs. We found that the frequency of apoptotic neutrophils was increased, and the frequency of live neutrophils was reduced by anti-G-CSF, but not by anti-IL-6 treatment in neutrophils incubated with tissue homogenates of Neu-high NPs (Figures 6A–E). The frequency of necrotic neutrophils was not affected by either anti-G-CSF or anti-IL-6 treatment (Figure 6E). Next, we detected G-CSF expression in nasal tissues by immunofluorescence staining and found that G-CSF was mainly expressed by epithelial cells. The mean fluorescence intensity of G-CSF staining was significantly higher in Neu-high NPs than that in Neu-low NPs and controls (Supplementary Figure SE6).

Finally, we performed univariate and multivariate regression analyses to identify clinical, cellular, and molecular factors associated with difficult-to-treat disease in patients with CRSwNP 1 year after surgery (Supplementary Table SE9). Both blood and tissue eosinophil and neutrophil abundance were associated with difficult-to-treat disease using univariate regression analysis (Figure 7A). In addition, the characteristics associated with difficult-to-treat NPs are shown in Figure 7A. When these factors were introduced into a multivariate model, only age, gender, asthma comorbidity, prior surgery, bilateral CT scores, tissue neutrophil counts, blood eosinophil counts, and tissue IL-13 levels were significantly and independently associated with difficult-to-treat disease in patients with CRSwNP (Figure 7B). The risk of difficult-to-treat disease increased by 1.036 times for every one unit increase of tissue neutrophil count, and increased by 3.609 times for every 1×10⁹ increase of blood eosinophil counts.