The remodeling patterns in different types of chronic rhinosinusitis (CRS) have rarely been compared, particularly the difference between eosinophilic and noneosinophilic CRS with nasal polyps (CRSwNP). Moreover, whether there is a link between remodeling and inflammation remains controversial.
To directly compare the remodeling features of different CRS and to explore their relationship with inflammation in Chinese patients.
Histologic characteristics of surgical samples were analyzed in 33 controls, 72 eosinophilic and 76 noneosinophilic CRSwNP, and 72 CRS without nasal polyps (CRSsNP) patients. Tissue samples from 38 controls, 26 eosinophilic and 26 noneosinophilic CRSwNP, and 32 CRSsNP patients were measured for mRNA and/or protein expression of profibrotic growth factors, metalloproteinases (MMPs), tissue inhibitor of metalloproteinases (TIMPs), hypoxia-inducible factor (HIF)-1α, interleukin (IL)-8, eosinophil cationic protein (ECP), and myeloperoxidase (MPO).
The amount of collagen decreased, whereas the edema scores increased, from CRSsNP to noneosinophilic CRSwNP and to eosinophilic CRSwNP. Transforming growth factor (TGF)-β2 protein levels were enhanced in CRSsNP compared with CRSwNP. TIMP-4 protein levels decreased in eosinophilic CRSwNP compared with noneosinophilic CRSwNP and CRSsNP. The number of neutrophils decreased from CRSsNP to noneosinophilic CRSwNP and to eosinophilic CRSwNP. ECP levels were only up-regulated in eosinophilic CRSwNP. ECP levels and neutrophil number correlated positively with the severity of edema and fibrosis, respectively. Neutrophils were the major sources of TGF-β2 in CRSsNP and noneosinophilic CRSwNP.
Distinct remodeling patterns are revealed for different types of CRS, particularly for eosinophilic and noneosinophilic CRSwNP. Tissue remodeling associates with inflammation in CRS.
Besides the persistent inflammation of the sinonasal mucosa, chronic rhinosinusitis (CRS) is also characterized by marked tissue remodeling that includes epithelial damage, basement membrane thickening, fibrosis, and/or edema . Currently, CRS is considered as a heterogeneous group of sinus disorders and divided into two types: CRS without nasal polyps (NPs) (CRSsNP) and CRS with NPs (CRSwNP), primarily on the basis of the absence or presence of NPs [2, 3]. Studies from white patients have showed that CRSsNP and CRSwNP express distinct inflammation and remodeling patterns [4, 5]. CRSsNP is characterized by a Th1 milieu and an elevated expression of transforming growth factor (TGF)-β1 and collagen, whereas CRSwNP is featured by a Th2-biased eosinophilic inflammation and a prominent edema formation with a down-regulation of TGF-β1 expression [4, 5]. Distinct from CRSwNP in Caucasians, we found that more than 50% of CRSwNP in Chinese demonstrates noneosinophilic inflammation with milder Th1/Th2/Th17 mixed responses . Recently, without stratification analysis of CRSwNP based on the extent of eosinophil infiltration, Li et al. found that Chinese CRSwNP is also characterized by a significant edema formation and a lack of TGF-β1 expression in comparison with CRSsNP . However, our and others' preliminary data indicated that eosinophilic and noneosinophilic CRSwNP may have distinct remodeling patterns, for example, compared with eosinophilic CRSwNP, noneosinophilic CRSwNP is distinguished by vigorous glandular hypertrophy [6, 8]. Therefore, it is necessary to further define the difference between eosinophilic and noneosinophilic CRSwNP. Until now, the link between remodeling and inflammation is subject to discussion. Although Chinese and white patients with CRS demonstrate distinct inflammatory patterns [6, 9], Li et al. found similar remodeling patterns in Chinese as compared with those reported in Caucasians [5, 7]; therefore, they proposed that there may be a dissociation of inflammation and remodeling in CRS . However, Li et al. did not directly investigate the relationship between remodeling and inflammation in their study. Through studying on ‘early stage’ CRSsNP, Van Bruaene et al. found that TGF-β1 and collagen production are up-regulated in sinonasal mucosa, despite the absence of obvious inflammation , challenging the paradigm that remodeling is dependent on the prior development of inflammation. Nevertheless, Chen et al. found that interleukin (IL)-5 levels correlated with tissue inhibitors of metalloproteinase (TIMP)-1 levels in CRSwNP, suggesting an association between inflammation and remodeling .
The aim of this study was to compare the difference in remodeling patterns between CRSsNP and eosinophilic and noneosinophilic CRSwNP and to define the relationship between remodeling and inflammation patterns in Chinese patients.
This study was approved by the Ethics Committee of Tongji Medical College of Huazhong University of Science and Technology, Wuhan, China, and conducted with written informed consent from patients. Seventy-two CRSsNP, 72 eosinophilic CRSwNP, 76 noneosinophilic CRSwNP, and 33 control subjects were enrolled for histological study, and additional 32 CRSsNP, 26 eosinophilic CRSwNP, 26 noneosinophilic CRSwNP, and 38 control subjects were recruited for immunohistochemistry and RT-PCR study. CRSwNP was classified as eosinophilic when percent tissue eosinophils exceeded 10% of total infiltrating cells . Diseased ethmoid sinus mucosa or NP tissues were collected during surgery. Control subjects were patients undergoing septoplasty because of anatomic variations and did not have sinus diseases, and the biopsy was taken from the inferior turbinate mucosa during surgery. Details of subjects' characteristics are described in the online supplement and Table S1.
Paraffin sections (5 μm) were stained with hematoxylin and eosin (HE) and Masson dye. The stained sections were observed by 2 independent physicians who were blind to the clinical data as previously described . Sections stained with HE were used to determine the general pathologic features including glandular hyperplasia and edema at ×200 magnification. Edema was semiquantitatively scored on a 3-point scale, with 0 representing the lowest and 2 representing the highest scores as mentioned elsewhere . The number of eosinophils was counted at high-power (HP) magnification (×400) . Sections stained with Masson dye were employed to assess the collagen deposition. The percent of positive area was determined at ×200 magnification by using pro-plus 6.0 software (Media Cybernetics, Inc. MD, USA). Ten fields per sample were randomly selected for analysis. According to Hellquist's classification , each of the tissues was classified into edematous, glandular hyperplastic, fibrotic, or atypical type based on remodeling features.
For immunohistochemical staining, the paraffin sections were stained with specific primary antibodies (Table S2 in the online supplement). The protein expression was detected with streptavidin–biotin–peroxidase complex method as previously described . As the expression intensity of fibronectin varied distinctly across different areas, the results were presented as integral optical density (IOD) values that were calculated by multiplying mean intensity of positive staining and area with positive staining per field at ×200 magnification by using pro-plus 6.0 software . For the analysis of other protein expression, the positive cells were counted at HP magnification, and the vascular expression of α-smooth muscle actin (α-SMA) was excluded from measurement . Ten fields per sample were randomly selected for analysis . To further identify the relationship between neutrophils [myeloperoxidase (MPO) positive] and TGF-β2 and hypoxia-induced factor-1α (HIF-1α) expression, immunohistochemical staining was conducted on consecutive serial sections (3 μm).
Total RNA was extracted and reverse-transcribed to cDNA with random hexamer primer as mentioned previously . PCR was performed on the Light-Cycler system (Roche Diagnostics, Mannheim, Germany) using the SYBR Premix Ex Taq kit (TaKaRa Biotechnology, Dalian, China) with the appropriate primers (Table S3 in the online supplement) as stated elsewhere . Relative gene expression was calculated by using the comparative CT method . An inferior turbinate sample from a control was used as a calibrator [16, 17]. GAPDH was used as a housekeeping gene for normalization, and a ‘no template’ sample was used as a negative control.
For continuous variables, results are expressed as medians and interquartile ranges, or in box-and-whisker plots. In continuous variables, when comparisons were made between groups, the Kruskal–Wallis H-test was used to assess significant intergroup variability. The Mann–Whitney U 2-tailed test was used for between-group comparison. Differences in proportions between groups were tested by chi-square test. Bonferroni correction was used to adjust significance levels for multiple comparisons between different groups using an α = 0.05/7 = 0.0071 for each comparison. The Spearman test was used to determine correlations, and significance was accepted for P < 0.05.
General remodeling patterns in different types of CRS
In the present study, CRS was subgrouped based on remodeling features according to Hellquist's classification . Atypical type characterized by large and pleomorphic histiocytes was rarely found . Edematous type was more frequently found in eosinophilic CRSwNP [70.8% (51/72)] compared with noneosinophilic CRSwNP [43.4% (33/76)] (P = 0.001) and CRSsNP [20.8% (15/72)] (P < 0.0001) and in noneosinophilic CRSwNP compared with CRSsNP (P = 0.005). On the contrary, fibrotic type was less commonly found in eosinophilic CRSwNP [12.5% (9/72)] than in noneosinophilic CRSwNP [36.8% (28/76)] (P = 0.001) and CRSsNP [48.6% (35/72)] (P < 0.0001). Although the incidence of glandular hyperplastic type tended to increase from eosinophilic CRSwNP [16.7% (12/72)] to noneosinophilic CRSwNP [19.7% (15/76)] and to CRSsNP [30.6% (22/72)], there was no significant difference among them.
Furthermore, we found that the total collagen amount was significantly lower in eosinophilic and noneosinophilic CRSwNP and significantly higher in CRSsNP compared with controls (Fig. 1). Although both eosinophilic and noneosinophilic CRSwNP had a reduced deposition of collagen compared with CRSsNP and controls, noneosinophilic CRSwNP had a relatively higher amount of collagen compared with eosinophilic CRSwNP (Fig. 1). In contrast, the edema scores were significantly higher in eosinophilic and noneosinophilic CRSwNP than in CRSsNP and controls and were further increased in eosinophilic CRSwNP compared with noneosinophilic CRSwNP (Fig. 1).
The expression of α-SMA and fibronectin
The number of α-SMA-positive myofibroblasts was significantly higher in all types of CRS compared with controls (Fig. S1). Although there was a trend toward increased number of myofibroblasts from eosinophilic CRSwNP to noneosinophilic CRSwNP and to CRSsNP, only the difference between eosinophilic CRSwNP and CRSsNP reached the statistical significance (Fig. S1). The expression pattern of fibronectin in different types of CRS was consistent with that of α-SMA (Fig. S1).
The expression of growth factors, proteases, and protease inhibitors
We investigated expression of TGF-β isoforms including TGF-β1, TGF-β2, and TGF-β3. As illustrated in Fig. 2A, there was a significant lower level of TGF-β1 mRNA expression in all types of CRS compared with controls, which is consistent with our previous finding . Interestingly, the mRNA level of TGF-β2 was significantly up-regulated in CRSsNP compared with controls and CRSwNP. No significant difference in TGF-β3 mRNA expression was observed among different study groups. In addition, we found that compared with CRSwNP and controls, heparin-binding epithelial growth factor (HB-EGF) levels were increased in CRSsNP (Fig. 2A). Moreover, we measured mRNA expression of metalloproteinase (MMP)-2, MMP-7, TIMP-1, and TIMP-4, whose expression has been shown to be altered in CRS by previous studies  (Fig. 2A). We found that the MMP-2 mRNA levels were significantly down-regulated in eosinophilic and noneosinophilic CRSwNP, but not in CRSsNP, compared with controls, whereas the MMP-7 mRNA levels were up-regulated in all kinds of CRS compared with controls. TIMP-4 mRNA expression was significantly down-regulated in eosinophilic CRSwNP compared with controls, and CRSsNP had a higher level of TIMP-4 mRNA expression than both eosinophilic and noneosinophilic CRSwNP. No significant difference in TIMP-1 mRNA expression was observed among different study groups.
We further studied the protein expression of TGF-β2 and TIMP-4 in sinonasal mucosa. We found TGF-β2- and TIMP-4-positive staining in epithelium and lamina propria (Fig. 2B). TGF-β2-positive cells were extensively distributed in lamina propria in CRS, and most TIMP-4-positive staining was found in glandular cells and cells around the vessels and glands (Fig. 2B). The number of the TGF-β2-positive cells in lamina propria was significantly increased in all kinds of CRS compared with controls and in CRSsNP compared with CRSwNP (Fig. 2B). The number of TIMP-4-positive cells in lamina propria was significantly increased in CRSsNP compared with CRSwNP and controls, but decreased in eosinophilic CRSwNP compared with other study groups (Fig. 2B). There was no significant difference in TIMP-4 and TGF-β2 expression in epithelium between these study groups.
The relationship between inflammation and remodeling
We evaluated the inflammation patterns in different kinds of CRS. Compared with controls, CRSsNP, and noneosinophilic CRSwNP, the number of eosinophils and eosinophil cationic protein (ECP) mRNA expression levels were significantly enhanced in eosinophilic CRSwNP (Fig. 3A). On the contrary, the number of neutrophils was significantly higher in all kinds of CRS compared with controls, and the number of neutrophils increased significantly from eosinophilic CRSwNP to noneosinophilic CRSwNP and to CRSsNP (Fig. 3A). IL-8 mRNA expression levels were significantly increased in CRSsNP compared with other study groups (Fig. 3A). In addition, we evaluated the HIF-1α expression which reflected the hypoxic status in diseased sinus tissue. HIF-1α mRNA and protein expression was enhanced in noneosinophilic CRSwNP and CRSsNP compared with controls (Fig. 3B). HIF-1α was mainly expressed in nucleus of inflammatory cells in lamina propria (Fig. 3B).
Analyzing the relationship between inflammation and remodeling mediators, we found that the number of neutrophils correlated positively, whereas ECP mRNA levels correlated negatively, with the expression of profibrotic factors (Table 1). In contrast, the ECP levels had a positive correlation, but the neutrophil number had a negative correlation with edema severity (Table 1). The ECP levels correlated negatively with TIMP-4 expression (Table 1). The number of TGF-β2-positive cells correlated positively with the number of myofibroblasts (n = 45, r = 0.723, P < 0.05) and the expression level of fibronectin (n = 46, r = 0.679, P < 0.05). Moreover, there was a positive correlation between HB-EGF and TGF-β2 expression (n = 48, r = 0.653, P < 0.05). The HIF-1α mRNA level was positively correlated with the number of neutrophils (Table 1), IL-8 (n = 50, r = 0.585, P < 0.05), and TGF-β2 (n = 53, r = 0.668, P < 0.05) mRNA levels.
Table 1. Correlation between tissue remodeling and inflammation parameters
To further identify the relationship between neutrophils and TGF-β2 and HIF-1α expression, immunohistochemical staining was conducted on consecutive serial sections (Fig. 4A and B). Interestingly, we found that the percentage of TGF-β2-positive neutrophils as a proportion of total TGF-β2-positive cells was significantly higher in all types of CRS than in controls and significantly higher in noneosinophilic CRSwNP and CRSsNP than in eosinophilic CRSwNP (Fig. 4C). The percentage of neutrophils positive for TGF-β2 did not differ significantly among these study groups (Fig. 4C). As to HIF-1α, we found that the percentage of HIF-1α-positive neutrophils as a proportion of total HIF-1α-positive cells was significantly higher in noneosinophilic CRSwNP and CRSsNP than in controls and eosinophilic CRSwNP (Fig. 4C). On the other hand, no significant difference in the percentage of neutrophils positive for HIF-1α among these study groups was found (Fig. 4C).
In this study, on the basis of stratifying CRSwNP into eosinophilic and noneosinophilic type, we directly compared the features of tissue remodeling in different types of CRS in Chinese patients and demonstrated that different types of CRS presented distinct patterns of tissue remodeling. CRSsNP was confirmed to be characterized by prominent fibrosis . Importantly, we further found that, although both eosinophilic and noneosinophilic CRSwNP presented an edema formation, noneosinophilic CRSwNP was relatively less edematous and more fibrotic than eosinophilic CRSwNP.
Tissue remodeling is a dynamic process that involves growth factor-induced extracellular matrix (ECM) production and proteases and protease inhibitor-controlled degradation . TGF-β family members are important profibrotic growth factors, and three isoforms (TGF-β1, 2, and 3) have been identified in human beings . Consistent with our previous reports , we found that the TGF-β1 mRNA expression was down-regulated in all types of CRS. However, Li et al. and Van Bruaene et al. found that TGF-β1 protein levels were significantly increased in CRSsNP in Chinese and Caucasians, respectively [5, 7]. The factors contributing to these discrepancies are not very clear; nevertheless, it is well known that the regulation of TGF-β1 mainly occurs at the post-transcriptional level . TGF-β2 has been reported to be the predominant isoform expressed in severe asthma and associated with local fibrosis . Van Bruaene found that TGF-β2 protein levels increased similarly in CRSsNP and CRSwNP compared with controls in Caucasians . In this study, however, we found that TGF-β2 protein levels were significantly up-regulated in CRSsNP in comparison with CRSwNP and controls in Chinese. Moreover, we found that TGF-β2 protein levels were marginally, but insignificantly, increased in noneosinophilic CRSwNP in comparison with eosinophilic CRSwNP. We further discovered that TGF-β2 protein levels correlated positively with the number of myofibroblasts and the expression of fibronectin, underscoring the potential role of TGF-β2 instead of TGF-β1 in local fibrosis in CRS. As to TGF-β3, our present data did not favor its role in CRS remodeling. A part of effect of TGF-β on tissue remodeling may be mediated through the induction of expression of HB-EGF , which is a potent mitogen and chemotactic factor for fibroblasts . Here, we demonstrated that HB-EGF mRNA expression correlated with TGF-β2 expression, indicating a potential role of HB-EGF in TGF-β2-mediated tissue fibrosis in CRS.
The balance in proteases and their inhibitors plays a crucial role in ECM degradation. Li et al. found that TIMP-1 and TIMP-4 protein expression was inhibited in CRSwNP compared with CRSsNP in Chinese . However, in the current study, no difference in TIMP-1 mRNA expression among different study groups was discovered. Moreover, compared with controls, we found that the protein expression of TIMP-4 was only inhibited in eosinophilic CRSwNP but not in noneosinophilic CRSwNP. The lower expression of TIMP-4 may lead to the loss of inhibition of MMPs, such as MMP7, the degrading enzymes toward ECM proteins, and result in more extensive edema formation in eosinophilic CRSwNP.
The relationship between airway inflammation and structural changes is still an open, crucial question. In this study, we explored the relationship between inflammation and remodeling across different types of CRS directly. We found that the eosinophil infiltration and activation were much more prominent in eosinophilic CRSwNP than in other types of CRS, whereas the neutrophil infiltration increased from eosinophilic CRSwNP to noneosinophilic CRSwNP and to CRSsNP, reflecting a distinguished eosinophil and neutrophil granulocyte activation bias in different kinds of CRS. Furthermore, we found a positive correlation between edema and eosinophilic inflammation and between fibrosis and neutrophilic inflammation in CRS, respectively. Eosinophils have the potential to cause epithelium and ECM damage through the release of toxic proteins, lipid mediators, and reactive oxygen species and lead to edema formation . Neutrophils are likely the sources of many important remodeling mediators, such as MMPs and growth factors [23, 24]. Chu et al. have reported that compared with controls, asthmatic subjects had greater numbers of tissue neutrophils positive for TGF-β, which may contribute to the airway remodeling process . Interestingly, in the present study, for the first time, we found that neutrophils were the main sources of TGF-β2 in noneosinophilic CRSwNP and CRSsNP, highlighting an important role of neutrophil in tissue remodeling in CRS. In this study, we also found an increase in HIF-1α expression in CRSsNP and noneosinophilic CRSwNP compared with controls, indicating tissue hypoxia in these two kinds of CRS. In contrast, Shin et al. recently found that HIF-1α protein expression was up-regulated similarly in eosinophilic and noneosinophilic CRSwNP, but not in CRSsNP, and HIF-1α was mainly expressed in nucleus of epithelial cells . The factors contributing to the discrepancies between their and our results are not very clear; nevertheless, our HIF-1α mRNA expression data are consistent with others . The hypoxia can induce the expression of IL-8 in nasal tissue, thus promoting the infiltration of neutrophils and tissue remodeling . In the current study, a positive correlation between HIF-1α and IL-8, neutrophil infiltration, and TGF-β2 was found.
In conclusion, our current study reveals distinct remodeling patterns for different types of CRS, particularly for eosinophilic and noneosinophilic CRSwNP. The remodeling features are related to inflammation patterns in CRS. Eosinophilic inflammation may contribute to the marked edema in eosinophilic CRSwNP, whereas hypoxia, which is possibly secondary to obstruction of sinus ostia, may lead to neutrophilic inflammation with overproduction of TGF-β2 and subsequent fibrosis in CRSsNP and noneosinophilic CRSwNP (Fig. 5).
The authors thank Prof. Guang-Lian Xiong in the Department of Statistics, College of Public Health, Huazhong University of Science and Technology, for assisting in statistical analysis of our data.
Grant support: This study was supported by National Nature Science Foundation of China (NSFC) grant 81020108018 to ZL, NSFC grant 30901660 to XL, and a grant from Ministry of Health of China (201202005).
Disclosure of conflict of interest: None for every author.
Author contribution: LLS performed PCR and immunohistochemistry experiments and data analysis and manuscript preparation. PX performed cell counting, data analysis, and manuscript preparation. PPC and BL performed some histology experiments. LZ, XL, and YHC participated in tissue sample collection and manuscript preparation. ZL designed the study and prepared the manuscript.