Proto-oncogene mutations in middle ear cholesteatoma contribute to its pathogenesis

Abstract

Background

Chronic inflammation causes bone destruction in middle ear cholesteatomas (MECs). However, the causes of their neoplastic features remain unknown. The present study demonstrated for the first time that neoplastic features of MEC are based on proto-oncogene mutations.

Results

DNA was extracted from MEC and blood samples of five patients to detect somatic mutations using depth-depth exome sequencing. Exons with somatic variants were analyzed using an additional 17 MEC/blood test pairs. Variants detected in MECs but not in blood were considered pathogenic variant candidates. We analyzed the correlation between proto-oncogene (NOTCH1 and MYC) variants and the presence of bone destruction and granulation tissue formation. MYC and NOTCH1 variants were detected in two and five of the 22 samples, respectively. Two of the NOTCH1 variants were located in its specific functional domain, one was truncating and the other was a splice donor site variant. Mutations of the two genes in attic cholesteatomas (n = 14) were significantly related with bone destruction (p = 0.0148) but not with granulation tissue formation (p = 0.399).

Conclusions

This is the first study to demonstrate a relationship between neoplastic features of MEC and proto-oncogene mutations.

Background

Cholesteatomas are benign, squamous epithelial hyperproliferative conditions of the tympanic cavity associated with keratin debris accumulation [1]. They cause gradual destruction of temporal structures, including the ossicles, facial nerve canal, and skull base. This process may be accompanied by severe complications, including facial paralysis, meningitis, and intracranial abscesses [2]. These complications can only be avoided by timely surgical removal of the pathology. Hence, there is a need to establish preventive nonsurgical treatments based on the pathogenesis [2].

Various hypotheses regarding the origin and pathogenesis of cholesteatomas have been proposed [3, 4]. Cholesteatomas are fundamentally non-neoplastic lesions of the temporal bone, but are clinically similar to neoplasms because of their unique epithelial hyperproliferative nature [3]. Interestingly, chronic inflammation may contribute to the pathological aggressiveness by affecting the degree of epithelial migration, cell proliferation, and extracellular matrix deposition [3]. However, it remains unknown whether these neoplasm-like aggressive features are caused by genetic mutations. To develop a fundamental understanding of cholesteatomas without distal metastasis, it is important to investigate underlying genes or mutations, particularly proto-oncogenes, that may control cellular proliferation.

In the present study, we analyzed the pathological genetic variants that may be related to cholesteatomas and investigated whether somatic mutations were associated with bone destruction and inflammatory reactions, such as granulation tissue formation.

Results

As a result of deep-depth WES with 137–212 mean depth of coverage, 24 potential cholesteatoma pathogenic genes (25 somatic variants) were identified (Table 1). The genes with somatic genetic variants detected using WES (samples 02–06) and target capture sequencing (samples 07–19, 21, 24–26) are shown in Tables 1 and 2, respectively. Genes with mutations found in each sample are shown in Fig. 1. MYC and NOTCH1 variants of interest were detected in two and five samples, respectively. Variants in LY75-CD302, EFCAB6, HRASLS, and UBP with approximately 50% variant allele frequencies (VAF) could be somatic variants acquired before cholesteatoma development. All of these variants are registered in the dbSNP database and may not be pathogenic. Meanwhile, the VAF for NOTCH1 was 2–7%, indicating that somatic mutations accounted for a small portion of cholesteatomas. One of the variants (p.I471T) was located in the 11th and 12th epidermal growth factor (EGF) repeat domain, while another (p.C1505R) was located in the LIN-12/NOTCH repeats (LNR) domain. EGF repeat and LNR domains are functionally significant in NOTCH1 [5], and therefore, mutations in these domains could result in the loss of NOTCH1 function (Fig. 2). Variants c2969 + 1 G > T and p.E1102* were a splice alteration and nonsense mutation, respectively, indicating deleterious mutations. NOTCH1 and MYC mutations correlated significantly with bone destruction (p = 0.0148) but not with granulation tissue formation (p = 0.399) (Supplement Fig. 1-a, b) in attic cholesteatomas (n = 14) (Table 3). Bone destruction was defined as at least one of the following clinical or surgical findings: ossicle destruction (> 50%) (Supplement Fig. 1-c, d), dura exposure, facial nerve exposure, and labyrinthine fistulae, described in operative note by surgeons.

Table 1 Somatic genetic variants detected using Exome sequencing 

Fig. 1

Gene variants detected in each cholesteatoma sample

Samples 02–06 were analyzed via whole-exome sequencing, while samples 07–26 were analyzed using target capture sequencing. Genes with somatic genetic variants are listed. MYC variants were detected in two samples, while NOTCH1 variants were detected in five samples. Gray and black boxes indicate genes with variants

Fig. 2

Domain graph for NOTCH1 and MYC with locations of the detected variants

One variant (p.I471T) was located in the 11th and 12th EGF repeat domains, while one (p.C1505R) was located in the LNR domain. EGF repeat and LNR domains were functionally significant in NOTCH1, and therefore, mutations in these domains could result in loss of NOTCH1 function

Table 2 Somatic genetic variants detected using target capture sequencing

Table 3 Patient profiles showing presence or absence of clinical presentations and gene mutations

Discussion

Acquired cholesteatomas are considered non-neoplastic pathologies with epithelial keratinizing lesions, which may lead to invasion and/or destruction of the temporal bone [2,3,4]. Intracranial complications caused by bony destruction can be fatal, therefore, preventive treatment prior to surgical removal of the lesion is necessary. Despite a number of previous studies, the origin of this pathology remains unclear [6]. Alternative molecular strategies, including exploration of genetic alterations, may expand the spectrum of therapeutic choices and lead to the development of nonsurgical preventive options for cholesteatomas. The present study used exome sequencing in five cholesteatoma patients to demonstrate somatic mutations in 24 genes. Gene Ontology (GO) analysis with these candidate genes (Supplement data), including MYC, NOTCH1, JAG1, and PSMC4, demonstrated significant correlations with mesenchymal transition, cell development, cell differentiation, and cellular response to hypoxia, which are clinically assumed to be involved in the pathology of the disease. Exon sequencing of these genes in 17 cholesteatoma/blood test pairs revealed that somatic variants in MYC and NOTCH1 had the highest frequencies among the examined genes. Whole-exome sequencing (WES) analysis was performed on the first five specimens, and the presence of gene mutations in previously known neoplastic variants could be precisely evaluated at the whole-gene level. However, the remaining 17 samples were evaluated with target capture sequencing; these samples may have other mutations in protooncogenes, which are not included in the 24 specific genes already detected in the present study. Moreover, mutations of either gene in attic cholesteatomas were significantly related to bone destruction (p = 0.0148) but not with granulation tissue formation(p = 0.399).

A review of the potential proto-oncogenic modifications in cholesteatomas failed to provide sufficient genetic evidence to support neoplastic features [7]. However, some studies have demonstrated its high proliferative activity using a variety of proliferation markers, including cytokeratin 13/16, Ki67, and proliferating cell nuclear antigen [8,9,10]. Further genetic analysis is required to confirm the highly proliferative nature of the pathology [11]. Interestingly, recent studies using microarray analysis techniques have demonstrated that cholesteatomas express many tumor-related genes, including proto-oncogenes c-MYC and NOTCH1 [11,12,13,14,15]. Based on these studies, abnormalities of proto-oncogene expression in cholesteatomas appear to be linked to their neoplastic features.

In agreement with previous studies, the present study found that bone destruction in cholesteatomas was significantly associated with proto-oncogene mutations. Furthermore, we also detected variants in the 11th and 12th EGF repeat domains and LNR domain, which have previously been shown to be functionally important for NOTCH1 [5]. NOTCH1 is expressed on the cell surface as a heterodimer composed of non-covalently associated extracellular (NEC) and transmembrane subunits. The NEC subunit consists of 36 iterated EGF-like repeats that include the binding region and three LNRs [16] where mutations were found in the present study. Extracellular domains of NOTCH receptors are largely composed of tandemly repeated EGF-1 domains. The 11th and 12th EGF repeat domains, in which the variant was detected in this study, were identified as necessary and sufficient to mediate binding [17]. The importance of these domains in Delta-Serrate-Lag2 ligand binding has been reported in multiple studies [18,19,20]. Our findings suggest an association between NOTCH1 and neoplastic features of cholesteatomas [15]. A previous study reported significantly decreased NOTCH1 expression in cholesteatoma epithelium compared to auditory canal skin epithelium, suggesting that NOTCH1 may alter the balance from cellular differentiation to hyperproliferation and subsequently contribute to neoplastic features of the pathology [15]. In the present study, immunohistochemical analysis of NOTCH1 was performed in a surgical sample (case not shown in Table I). In this specific sample, the levels of expression of NOTCH1 and downstream HES1 in the basement membrane of cholesteatoma epithelium were weaker than in normal skin (Supplement Fig. 2-a and -b). These observations suggest that there is a positive relation between NOTCH1 mutation and protein expression in these lesions. Martincorena et al. [21] reported a high frequency (30–80%) of NOTCH1 mutations in the esophageal epithelia of elderly and middle-aged healthy individuals, suggesting that it is a passenger, not a driver, mutation. That is, high VAFs, including NOTCH1, indicate that clonal expansion in the elderly may result in a predisposition to tumor formation and progression. VAF refers to the proportion of specific variants detected at a particular location in the genome relative to the number of sequencing reads; a value of 50% or 100% is taken to indicate a germline variant, while in the case of somatic mutations VAF varies based on the proportion of a cell population with the mutation within the extracted sample. In this study, VAF ranged from 2 to 8%, indicating a somatic mutation. Protooncogene variants may enhance the pathological features of benign conditions, including cholesteatomas. Therefore, we hypothesized that VAF in NOTCH1 may be useful for identifying the clinical behavior of cholesteatomas, including the degrees of bony destruction.

Finally, we also found that the VAF of LY75-CD302, EFCAB6, HRASLS, and UBP were approximately 50% (Table 2; samples 9 and 10), indicating the presence of heterozygous gene variants. This suggests clonality of cholesteatomas that expanded from a small number of cells. However, this finding requires further investigation using larger samples for exome and targeted mutation analyses.

The present study had some limitations. First, the sample size of 22 was small, so further exome and targeted mutation analyses with larger sample sizes are needed to analyze the roles of genetic changes in cholesteatoma development, inflammation, and neoplastic features. Second, definitive identification of a tissue as neoplastic or whether protooncogene variants induce neoplasia, appropriate validation is required using nearby nonneoplastic tissues or homologous tissue biopsy specimens from the same patient, such as skin from the external auditory canal, in addition to blood samples. Such tissue controls would provide more robust data and confirmation to clarify the correlations between genetic mutations and clinical presentations of cholesteatoma.

Conclusions

Mutations in cholesteatomas, including NOTCH1 and MYC, were significantly correlated with bone destruction. These observations suggest that protooncogene mutations may enhance the pathological features of cholesteatomas.

Methods

Blood and cholesteatoma samples were collected from five Japanese cholesteatoma patients who were treated surgically. These blood-cholesteatoma paired samples were subjected to whole-exome sequencing (WES). DNA was extracted from cholesteatoma using the QIAamp DNA Mini kit (QIAGEN, Hilden, Germany) and from blood using the QIAamp DNA Maxi kit (QIAGEN) according to the manufacturer’s protocols. Coding exons were captured using the SureSelect XT AUTO HUMAN ALL Exon V5 kit (Agilent Technologies Inc., Santa Clara, CA, USA) and sequenced using the HiSeq2500 system (Illumina Inc., San Diego, CA, USA) in 101 bp paired-end reading mode. The reads were aligned to GRCh37/hg19 using Novoalign (Novocraft Technologies, Selangor, Malaysia) and duplicate reads that were excluded from the analysis were marked using the Novosort software (Novocraft Technologies). Local realignment and variant calling were performed using the HaplotypeCaller in the Genome Analysis Toolkit (Broad Institute, Cambridge, MA, USA) to generate variant call format (VCF) files. Variants found in the cholesteatoma but not in the blood of a participant were considered cholesteatoma-specific variants (CSVs). Variants found in any of the five blood samples were excluded as noise, even if they were CSVs in pair comparisons. CSVs were also identified using MuTect2 and were confirmed using MiSeq after specific polymerase chain reaction amplification. CSVs and six additional genes (TP63, RARA, RARB, RARG, BMP4, and TP53) that are frequently mutated in head and neck tumors were sequenced in another 17 cholesteatoma-blood pair test samples. Bait oligos (Sure Design; Agilent Technologies) were used to capture the exons in target genes. Sequencing was performed using the MiSeq platform and the variants were detected using MuTect2. Except for attic cholesteatomas (n = 14), there were less than three cases of all other types of the pathology (Table 1). Therefore, the correlations between gene mutations and clinical severity, including the presence or absence of bone destruction and granulation tissue formation, were examined using Pearson’s chi-square test only in attic cholesteatomas. The presence or absence of bone destruction was evaluated based on surgical records. At our institution, surgeons always describe the following findings in the operative note: ossicle destruction, bone defects in the facial nerve canal, labyrinthine fistulae, dura exposure, and granulation tissue around the surface of cholesteatoma. The first four findings indicate the presence or absence of cholesteatoma-induced bony destruction.