For all experiments, BALB/c male mice (4 weeks old, 24–25 g) were purchased (Central Laboratory Animals, Seoul, Korea) and kept for 1 week in specific pathogen-free rooms before initiating experiments. All experiments involving animals were approved by the Institutional Animal Care and Use Committee of Seoul National University and complied with governmental and international guidelines on animal experimentation. We utilized the murine NP model following previous reports with some modifications.2223 For induction of NP in mouse nasal cavity, mice were sub-divided into the control group (n = 14) and the experimental group (n = 20), and treated with either phosphate-buffered saline (PBS) or 3% ovalbumin (OVA) plus 10 ng of staphylococcal enterotoxin B (SEB). For treatments with resveratrol (RSV) and dexamethasone, mice were categorized into a control group (group A; n = 6) and three experimental groups (groups B, C, and D; n = 12, 7, and 7, respectively). After treating with 3% OVA for 47 days, mice were administered RSV (0.44 µg per nostril) 3 times a week for 8 consecutive weeks. Dexamethasone (1 mg/kg) was delivered once a week for 8 consecutive weeks. The 3% OVA and SEB were applied before administrating the drugs. Mice were sacrificed 1 day after the last treatment. PBS was applied for both systemic and local stimulation in group A. Mouse sinonasal tissue specimens for various types of analysis were prepared as previously described.24

The skull of each mouse (n = 3 from each group) was subjected to micro-CT (NFR-Polaris-G90; NanoFocusRay, Jeonju, Korea). Cross-sectional images were converted to the Digital Imaging and Communications in Medicine format using PMOD software (PMOD Technologies LLC, Zurich, Switzerland). For neo-osteogenesis evaluation, the same coronal plane showing the true maxillary sinuses and ethmoid labyrinths was selected for each mouse. The nasal bone thickness was measured at 5 different points on both right and left lateral walls and evaluated using Image J software (version 1.52p, National Institutes of Health [NIH] Image processing analysis, http://rsb.info.nih.gov/ij/).

Detailed experimental procedures for processing the sinonasal tissues and immunohistochemistry were previously described.2425 Briefly, the heads of mice were fixed in 4% PFA (Biosesang, Seongnam, Korea), decalcified in 5% nitric acid (Sigma-Aldrich, St. Louis, MO, USA), dehydrated, processed according to standard paraffin-embedding procedures, and coronally sectioned into 3-µm-thick sections. An atlas of normal murine sinonasal anatomy was used to standardize the anatomic locations being examined.26 Several stains were conducted to compare characteristics between groups: hematoxylin and eosin (H&E; Sigma-Aldrich) for general morphology, Sirius red (Polysciences Inc., Warrington, PA, USA) for eosinophils, Giemsa (Sigma-Aldrich) stain for mast cells, periodic Acid-Schiff (Sigma-Aldrich) stain for goblet cells, and Masson's trichrome (Sigma-Aldrich) stain to highlight collagen deposition in the sub-epithelial layer, following the manufacturer's instructions. For immunohistochemistry, sections were stained with RUNX2, osteonectin, and IL-13 antibodies. Detailed information is indicated in Supplementary Table S1. The sections were then incubated with broad antibody enhancer and polymer-horseradish peroxidase (HRP) using polink-2 plus polymerized HRP broad 3,3′-Diaminobenzidine (DAB) Detection System (Golden Bridge International Labs., Bothell, WA, USA) following the manufacturer's recommendations. Five areas from nasal tissue sections were chosen randomly for evaluation under high-powered fields (HPFs; magnification ×1,000) and measured by 2 examiners who were blinded to group assignment. The numbers of NPs were counted following the previous standards,24 quantified microscopically from 5 coronal sections, and presented as total number. The numbers of eosinophils, mast cells, and goblet cells were shown as the average numbers from 5 HPFs. Sub-epithelial fibrosis was evaluated as previously described.27 For the assessment of bone thickness, at least 5 measurements at different points were made in the appropriate area of each HPF (×400 magnification), and measured using Image J software (version 1.52p, NIH Image processing analysis, http://rsb.info.nih.gov/ij/).

For IF staining, the tissue sections were processed as described previously.24 Briefly, mouse nasal tissues were de-paraffinized, rehydrated, boiled in citrate buffer (Dako, Carpinteria, CA, USA) for epitope retrieval, fixed in 4% paraformaldehyde (PFA; Biosesang, Seongnam, Korea), washed 3 × 5 minutes in PBS containing 0.2% Triton X-100 (PBST), blocked in PBST containing 5% bovine serum albumin (BSA) (BSA-PBST) for 3 hours, incubated with primary antibodies diluted in BSA-PBST (Sigma-Aldrich) overnight at 4°C, washed 3 × 5 minutes with PBST, stained with secondary antibodies in BSA-PBST for 1 hour in the dark at room temperature, washed 3 × 5 minutes with PBST, stained with 4′,6-Diamidino-2-phenylindole (Sigma-Aldrich) for 15 minutes at room temperature, mounted with fluorescence-mounting medium (Vector Laboratories Inc., Burlingame, CA, USA). The primary antibodies utilized are summarized in Supplementary Table S1. Images were acquired with a Zeiss fluorescent microscope using AxioVision software (Carl Zeiss Vision, Thornwood, NY, USA). The images were edited using Zen Blue software (Zeiss, Oberkochen, Germany). Immunostaining was not observed in control sections incubated with normal serum instead of the primary antibodies (data not shown).

Total RNA was collected from the tissue samples using QIAshredder and RNeasy Plus Mini kit (both Qiagen, Redwood City, CA, USA) following the manufacturer's recommendations. Total RNA (1 μg) was reverse transcribed to cDNA using the Tetro cDNA-synthesis kit (Bioline, Eveleigh, Australia) following the manufacturer's instructions. Quantitative real-time PCR was performed as previously described.24 Briefly, polymerase chain reaction was conducted with SYBR Green Polymerase Chain Reaction Master Mix (Enzynomics, Daejeon, Korea) and primers that specifically amplify genes for IL13, RUNX2, matrix metallopeptidase 9 (MMP9), collagen type 1 (COL1A1), and osteopontin (SPP1). Additionally, β-actin (ACTB) was used as reference gene for normalization. The relative fold changes were calculated by the comparative 2^−ΔΔCq method. The primer sequences used in this study are described in Supplementary Table S2. Cycling conditions were 95°C for 10 minutes, followed by 40 cycles at 95°C for 10 seconds, 60°C for 1 minutes, and 72°C for 30 seconds. The PCR efficiency in all runs was close to 100%, and all samples were tested in duplicate.

The normalized dataset (GSE23552), which contains 13 normal nasal mucosal tissues and 26 CRS microarray data, was imported from Gene Expression Omnibus (www.ncbi.nih.gov/geo). All tissues (n = 39) were divided into 3 groups: normal tissues (n = 13), chronic rhinosinusitis without nasal polyp (CRSsNP) tissues (n = 15), chronic rhinosinusitis with nasal polyp (CRSwNP) tissues (n = 11). The values of 2828688 (corresponding to IL13), 2908762 (corresponding to RUNX2), 3887210 (corresponding to MMP9), 3762198 (corresponding to COL1A1), and 2735027 (corresponding to SPP1) on each group were calculated and compared between the 3 groups using Spearman's correlation analysis. Enrichment analysis was also performed using Gene Set Enrichment Analysis (GSEA) software v3.0 and a formatted GCT file was used as input for the GSEA algorithm (available from: http://www.broadinstitute.org/gsea). For grouping the GSE23552 dataset, the value of the 2828688 probe was utilized as criteria for high expression group and low expression groups. In the analysis, pre-defined osteogenesis-related gene sets in the Gene Ontology collection (GO) from v6.2 molecular signature database (mSigDB c5 category; http://software.broadinstitute.org/gsea/msigdb) were used, and the number of permutations was set at 1,000. The gene sets with P < 0.05 and false discovery rate (FDR) < 0.25 were considered significantly enriched.

Data are expressed as mean ± standard errors (SEs) from 2 or more distinct samples unless indicated otherwise. We used unpaired Student's t-test (2-sided) or non-parametric Mann-Whitney U test (2-tailed) to compare differences between groups. In addition, correlations of protein or mRNA expression were determined using Spearman's P statistic or Pearson's P statistic. Statistical analyses were performed using SPSS 23.0 software (SPSS, Chicago, IL, USA). Illustrative figures were generated using Prism version 8.2.0 (GraphPad Software Inc., San Diego, CA, USA) and SigmaPlot version 10.0 (Systat Software Inc., San Jose, CA, USA).

To evaluate neo-osteogenesis in mice, we employed an experimental CRSwNP murine model based on previous reports with some modifications (Fig. 1A).242528 Because NPs are associated with osteitis severity and bone remodeling,29 we first checked the number of NPs in mouce nasal cavity. As expected, thickened mucosae with polyp-like lesions were found in the OVA/SEB-treated group, whereas no signs of inflammation were observed in control mice (Supplementary Fig. S1A). When we assessed sub-epithelial fibrosis, eosinophils numbers, mast cells numbers, and goblet cells numbers, following the previous report,30 we observed similar results (Supplementary Fig. S1B-E). In order to find evidence of bone remodeling, we next examined the number of osteoblasts and osteoclasts which are responsible for bone formation and bone resorption, respectively. We detected not only an increased number of osteoblasts and osteoclasts at the rim of the bone but also a woven bone with marked thickening periosteum in OVA/SEB-treated mice (Fig. 1B-D). The micro-CT findings exhibited areas of up-regulated bone thickness and irregular thickening of the sinus walls in OVA/SEB-treated mice. (Fig. 1E and F).

New bone formation is mainly mediated by mature osteoblasts. In this process, RUNX2 serves as master regulator and induces the differentiation of pre-osteoblast.31 Therefore, we inspected the expression of RUNX2 and the number of mature osteoblasts by using immunohistochemistry. We found that RUNX2-immunoreactive cells were significantly elevated in OVA/SEB-treated mice. RUNX2-immunoreactivity was particularly identified in osteoblasts lining the bone (Fig. 2A and B). The mature osteoblasts on the bone surface were evaluated on the basis of the expression of osteonectin (a differentiation marker for bone cells), as shown in Fig. 2A and C. By utilizing IF staining, we observed marked co-localization of RUNX2 and osteonectin mainly lining the bone in mice nasal tissues, as previously reported (Fig. 2C). Because RUNX2-induced differentiated osteoblasts induce osteogenesis, we compared the relationship between RUNX2 level and bone thickness. A significant positive correlation was detected between bone thickness and RUNX2-positive cell numbers (Fig. 2D; Pearson's correlation coefficient, P < 0.01). These results suggest that OVA/SEB-treated mice showed neo-osteogenic characteristics.

We previously found that IL-13, one of the representative Th2 cytokines, induced neo-osteogenesis via RUNX2 expression.32 To verify this in our murine model, we evaluated the expression of IL-13. Immunohistochemistry studies revealed that IL-13 expression, in mucosal tissues overlying the bone, was increased in the OVA/SEB-treated mice compared to the PBS-treated mice (Fig. 3A and B). The number of IL-13-positive cells positively correlated with the number of RUNX2-positive cells (Fig. 3C; Pearson's correlation coefficient, P < 0.01). Moreover, IL-13-positive cell numbers showed positive association with bone thickness (Fig. 3D; Pearson's correlation coefficient, P < 0.05). Furthermore, mRNA expression levels of IL-13, RUNX2, and RUNX2 downstream genes (MMP9, COL1A1, and SPP1) were also increased in OVA/SEB-treated mice (Fig. 3E). This result indicates that IL-13, which was secreted during stimulation with OVA/SEB, could contribute to ossification via RUNX2, suggesting the possibility of osteogenesis in mice.

To verify the role of IL-13 in human nasal tissues, we conducted bioinformatic analysis using a previously published dataset (GSE23552). CRS patient tissues in the GSE23552 data set were categorized into IL13_high and IL13_low groups based on the median IL13 expression value to find gene expression signatures that correlated with IL13 expression (Fig. 4A). Importantly, many osteogenesis-related gene sets were enriched in the IL13_high group (Fig. 4B; P < 0.05, FDR < 0.25). Fig. 4C showed 2 representative enrichment plots of gene sets. Enrichment plots that were identified in Fig. 4B were indicated in Supplementary Fig. S2. We then compared the mRNA levels of IL13, RUNX2, and RUNX2 downstream genes between normal nasal tissues and CRS patient tissues. IL13 expression levels were higher in CRSwNP tissues than in normal nasal tissues. Similarly, transcript expression levels of RUNX2 and RUNX2 downstream target genes (SPP1, MMP9, and COL1A1) were also up-regulated in both CRSsNP and CRSwNP tissues (Fig. 4D). Spearman correlation analysis showed that IL13 expression positively correlated with SPP1 and MMP9 expression (Fig. 4E; Spearman's correlation coefficient, P < 0.05 and P < 0.01, respectively).

Finally, we deployed this eosinophilic murine NP model to evaluate the therapeutic effect of a drug candidate against osteitis (Fig. 5A). For this purpose, we used RSV and dexamethasone which are previously known to reduce NPs.2830 When we applied these drugs to the OVA/SEB-treated mice, we found decreased numbers of osteoblasts and bone thickness comparable to the PBS-treated mice (Fig. 5B). The expression of RUNX2-positive cells was diminished by the treatment of RSV and dexamethasone (Fig. 5C). A similar finding was observed regarding IL-13 (Supplementary Fig. S3). This result implies that our eosinophilic murine NP model might be applicable to evaluating the effects of some type of drugs targeting osteitis in terms of CRS.