Genetic mutation analysis of the malignant transformation of sinonasal inverted papilloma by targeted amplicon sequencing

The mechanism underlying the malignant transformation of inverted papilloma (IP) has not yet been elucidated. To clarify the genes responsible for the malignant transformation, we analyzed 10 cases of IP, 8 of IP with dysplasia, and 11 of squamous cell carcinoma (SCC) by targeted amplicon sequencing. The number of mutant genes increased in the order of IP < dysplasia < SCC. Significant differences were observed in the mutation rates of three genes (KRAS, APC and STK11) in particular. TP53 was altered frequently in each group and might be involved in malignant transformation based on to the site of the mutation. A comparison of the genetic variants by region of IP tissue among patients with IP alone, and those with dysplasia or SCC revealed significant differences in the mutation rate of the KRAS gene. Identification of genetic mutations in KRAS is effective for predicting the malignant transformation of IP.


Introduction
Sinonasal papilloma accounts for 0.4-4.7% of all nasal tumors and is histopathologically classified into three types: inverted papilloma (IP), exophytic papilloma and oncocytic papilloma. 1 IP represents the most commonly diagnosed subtype of sinonasal papilloma with an incidence of 0.74-1.5 cases per 100.000 inhabitants per year. 2 IP is pathologically regarded as a benign tumor, but its malignant transformation is known to occur at a rate of 3.8-10%. [3][4][5][6] Therefore, the standard treatment for IP is complete resection. However, it is impossible to completely diagnose the malignant transformation of IP by macroscopic findings and computed tomography (CT) or magnetic response imaging (MRI) before surgical resection. Most of the malignant tissues present as squamous cell carcinoma (SCC). SCC within IP has been reported to occupy 10 to 95% of the total tumor volume. 7 Therefore, a preoperative biopsy may not include areas of the IP showing malignant change and, therefore, is not always effective in assisting the clinician in reaching a correct diagnosis. Therefore, multiple preoperative biopsies from different sites within the tumor tissue may be necessary to prevent misdiagnosis.
Recently, next-generation sequencer (NGS) has been widely used at the forefront of research, affording considerable progress in the identification of the responsible genes in several diseases. 8 The clinical applications of whole genome sequencing and whole exosome sequencing were limited in terms of the cost and time required; however, it has recently become possible to efficiently retrieve the target genes by amplicon sequencing. 9 These technological innovations mean that NGS can now be clinically applied for some types of tumors, including head and neck tumors. 10 Although many investigations into the etiology and behavior of IP have been undertaken, the exact mechanism of its malignant transformation has not yet been fully elucidated. To clarify the genetic variants involved in the malignant transformation of IP, we analyzed the genetic variants in the tissues associated with IP, dysplasia and SCC by target amplicon sequencing with NGS and compared the differences in variants between the tissues. If the genes responsible for the malignant transformation of IP can be identified, it will become possible to predict malignant change in IP through examining alterations in those genes.  Service, Tokyo, Tokyo, Japan) was used. Kruskal-Wallis test was used to compare the number of mutant genes among the 3 groups, and chi-square test was used to compare the mutation rate of each gene. A p value of less than 0.05 was judged to be statistically significant.

Correlations between genetic variants and histological type
We could detect one or more genetic mutations in each of the 29 cases, with a total of 129 mutations with amino acid substitutions detected. In the 26 genes analyzed using the sequence panel, the average number of mutant genes was 2.9 (range 1-5) in the IP tissue from patients with IP alone, 4.4 (range 3-6) in the dysplastic tissue from patients with IP with dysplasia, and 5.7 (range 3-9) in the carcinoma tissue from patients with SCC with or without IP ( Figure   2). The number of mutant genes increased in the order of IP<dysplasia<SCC, and there was a significant difference in the number of mutated genes between the IP and SCC tissue (P = 0.042, Figure 3). When dysplasia was divided into two groups based on the grading of dysplasia, the average number of mutant genes was 4.7 (range 3-6) in the low-grade dysplastic tissues (case number 11, 12, 13, 14, 17 and 18) and 3.5 (range 3-4) in the highgrade dysplastic tissues (case number 15 and 16). There was no significant difference between the two groups in terms of the number or pattern of mutant genes. When SCC was further divided into two groups, the average number of mutant genes was 5.6 (range 4-7) in the carcinoma tissue from patients with SCC with IP and 5.8 (range 3-9) in the carcinoma tissue from patients with SCC without IP, which we regarded as de novo SCC. There was no significant difference between the two groups in terms of the number or pattern of mutant genes.
Analysis of each gene revealed significant differences in the mutation rates of the following 3 genes among the IP, dysplasia and SCC groups ( Table   1). KRAS and APC were found to be altered less frequently in the IP group (10 and 0%), while they were more frequently mutated in the dysplasia (75 and 50%) and SCC (73 and 64%) groups. No alterations in STK11 were found in the IP and dysplasia groups, while it was more frequently mutated in the SCC group (36%). TP53 was found to be altered frequently in each group (IP, 70%; dysplasia, 75%; SCC, 91%), with no significant differences observed among the three groups.

Differences in genetic variants in the same patient
We compared the differences in genetic variants between the regions of IP and dysplastic tissue in the same patient with IP with coexistent dysplasia.
A total 39 mutant genes were found in either the IP or dysplastic tissue in 8 cases, while 4 mutant genes (10%) were found in the IP alone, and 6 mutant genes (15%) in the dysplastic tissue alone (Figure 4). Similarly, we compared the differences in genetic variants between the IP and SCC regions in the same patient with IP with coexistent SCC. A total 29 mutant genes were found in either the IP or SCC tissue in 5 cases, while 1 mutant gene (3%) was found in the IP alone, and 2 mutant genes (7%) in the SCC tissue alone ( Figure 5). In both IP with dysplasia and IP with SCC, little difference was observed in the genetic variants between the IP tissue and coexisting dysplasia or SCC tissue.
Genetic variants in the IP tissue among patients with IP alone, IP with dysplasia and IP with SCC Furthermore, we compared genetic variants in the IP tissue among patients with IP alone, IP with dysplasia and IP with SCC ( Figure 6). The number of mutant genes increased in the order of IP<IP with dysplasia<IP with SCC, and there was a significant difference in the number of mutated genes in the IP tissue between patients with IP alone and those with IP with SCC (P = 0.030, Figure 7). The analysis of each gene revealed that there was significant difference in the genetic mutation of KRAS in the IP tissue between patients with IP and those with IP with dysplasia/SCC (P = 0.005, Table 3). Therefore, we calculated the sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) for patients with coexistent dysplasia or SCC.
Results showed that the sensitivity of the detection of dysplasia or SCC using the KRAS mutation was 85%, the specificity was 90%, the PPV was 91% and the NPV was 75%.

Discussion
Recently, several investigators have reported on candidate biomarkers for the malignant transformation of sinonasal IP. Cyclooxygenase-2, which is induced by mitogenic and inflammatory stimuli and promotes tumorigenesis, has been reported to be overexpressed significantly in IP with SCC compared to the level observed in benign IP. 11 Human papilloma virus (HPV) shows higher viral loads in patients with IP/SCC than in those with IP. 12 A recent systemic review with meta-analysis found a statistically significant association between HPV infection and malignant transformation of IP. 13 Furthermore, proapoptotic factors such as p53, p21, p16, p27, tissue factor pathway inhibitor-2, p63, bcl-2 family, Ki-67, proliferating cell nuclear antigen and intercellular adhesion molecules have all been evaluated as possible contributors to IP malignant transformation. 14 Therefore, we undertook a comprehensive analysis of genetic alterations in 26 genes closely related to solid tumors by target amplicon sequencing with NGS. As a result, the number of mutant genes was found to increase in the order of IP<dysplasia<SCC (Figure 3). In particular, there were significant differences in the mutation rates of 3 genes: KRAS, APC, and STK11 (Table 1). Furthermore, there was a significant difference in the genetic mutation of KRAS in the IP tissue between patients with IP alone and those with IP with dysplasia/SCC (Table 3) However, they also reported that there were no KRAS mutations in either IP or IP-associated SCC.
On the other hand, TP53 was found to be altered frequently in all groups (IP, 70%; dysplasia, 75%; SCC, 91%), but there were no significant differences among the three groups in this study (Table 1). However, in the analysis of gene mutation site, the g. 7570312C>A mutation was detected in 7% of all cases, with all cases limited to the SCC group (18%). In addition, the g. 7577150T>A mutation was detected in 10% of all cases, with all cases found in patients with either dysplasia or SCC (13% and 18%, respectively) ( Table   2). Lin et al. showed that tissue from IP with carcinoma is positively stained for p53 at more than twice the frequency than that observed in tissue from IP alone based on an immunohistochemical approach (62% versus 30%), and suggested that alterations in p53 are important in the progression of IP to malignant disease. 16 These results suggest that p53 protein dysfunction, dependent on the site of the TP53 mutation, could be involved in the malignant transformation of IP.
Currently, the diagnosis of SCC derived from IP requires the pathological existence of IP. However, if the IP is completely replaced by carcinoma, or if carcinoma alone occurs after complete IP resection, a differential diagnosis between SCC derived from IP and de novo SCC would be impossible. However, if the IP and SCC occur independently and the SCC invades the nearby IP, de novo SCC might be misdiagnosed as SCC derived from IP. In this study, we compared genetic variants between SCC with IP and de novo SCC, and no significant differences were observed between the two groups in terms of the number and pattern of mutant genes. Therefore, this study indicated that a differential diagnosis between SCC derived from IP and de novo SCC is impossible based on genetic alterations.
As IP occasionally progresses to malignant transformation, its standard treatment is complete resection. The nasal and paranasal sinuses, where IP arises, are anatomically adjacent to important organs such as the orbit and skull base; therefore, extended surgery for carcinoma derived from papilloma could have a great influence on postoperative complications and dysfunction. Accordingly, if we can predict the incidence of carcinoma within IP tissue prior to surgery, we could set an appropriate resection line and avoid unnecessarily extensive surgery. However, it is currently impossible to completely diagnose the malignant transformation of IP by macroscopic findings and CT or MRI before surgical resection. Furthermore, even if a preoperative biopsy is performed, carcinoma tissue might not be contained in the biopsied specimen. Therefore, it would be clinically very beneficial to develop a new preoperative method for the diagnosis of the malignant transformation of IP.
We examined the differences in genetic variants between the IP and dysplasia or SCC tissues in the same patient. As a result, a total of 39 mutant genes were found in either the IP or dysplastic tissue, while 4 mutant genes (10%) were found in the IP tissues alone and 6 mutant genes (15%) were found in the dysplastic tissue alone (Figure 4). Similarly, a total of 29 mutant genes were found in either the IP or SCC tissue, while 1 mutant gene (3%) was found in the IP tissue alone and 2 mutant genes (7%) were found in the SCC tissue alone ( Figure 5). Accordingly, there was little difference in the genetic variants between IP and the coexisting dysplasia or SCC. These results indicated that it might be possible to predict genetic variants in carcinoma within IP by genetic mutation analysis of the biopsied IP tissue. Furthermore, we compared genetic variants in the IP tissue among patients with IP alone, IP with dysplasia and IP with SCC. As a result, significant differences were observed in the mutation rates of the KRAS gene in the IP tissue between patients with IP and those with IP with dysplasia/SCC (Table 3). Regarding other genes, no genetic mutations of APC, NRAS and STK11 were found in patients with IP alone, while such mutations were frequently found in the IP tissue from patients with coexistent dysplasia or SCC. These results suggested that genetic variants of KRAS, APC, NRAS and STK11 in preoperative biopsy tissue could be predictive of the existence of carcinoma within the IP tissues and the risk of future malignant transformation. If we can predict the existence of carcinoma within IP tissue prior to surgery, it may be possible to select the following treatment strategy: minimally invasive endoscopic sinus surgery for IP with low risk of malignant transformation, while external approach or advanced endoscopic sinus surgery including the bone resection of orbit or skull base for IP with high risk of malignant transformation.
One limitation to this study is that only a limited number of cases were analyzed; therefore, it is necessary to pursue further analysis of a greater number of cases. Further research will allow for the identification of mutant gene combinations and their mutation sites that can predict malignant transformation of IP more accurately and efficiently.