Characterizing the mutational landscape of sinonasal squamous cell carcinoma using whole-exome sequencing

Abstract

Background:

Sinonasal squamous cell carcinoma (SNSCC) accounts for less than 3% of all head and neck cancers. The 5-year overall survival rate ranges from 30 to 50%. While whole-exome sequencing (WES) studies have illuminated the mutational landscape of numerous cancer types, few systematic efforts have been conducted for SNSCC, leaving many common driver mutations in this malignancy largely unknown. The objective of this study was to address this gap in knowledge by comprehensively cataloging somatic mutations arising in SNSCC through WES.

Patients and Methods:

This was a retrospective study in which WES was performed on OCT-embedded tumor tissue and paired adjacent normal tissue from 12 patients diagnosed with incident SNSCC at the University of Cincinnati Cancer Center from 2012–2014.

Results:

We identified 263 genes that harbored two or more coding region somatic mutations in multiple SNSCC tumors. Eight genes were significantly mutated (q < 0.1). TP53 was the most frequently mutated gene (6/12; 50%). The other 7 significantly mutated genes included NOTCH1 (4/12; 33.3%), KMT2D (4/12; 33%), VWDE (4/12; 33%), FNBP4 (3/12; 25%), SCAND3 (3/12; 25%), NOD1 (3/12; 25%) and OR5C1 (2/12; 17%). Somatic mutations observed in these genes included truncating or non-synonymous variants in functional domains that may affect regulation of apoptosis, NOTCH signaling, cell cycle, and epithelial cell proliferation, and epigenetic regulation of gene expression.

Conclusion:

This study helps elucidate the mutational landscape of SNSCC, advancing our understanding of potential driver mutations and yielding potential new therapeutic avenues for management of these devastating malignancies.

INTRODUCTION

Sinonasal tract malignancies are relatively rare cancers, accounting for 3–5% of all head and neck cancers in the United States, with an incidence rate of approximately 5.6 new cases per million population [1]. Nearly half of these malignancies are histologically squamous cell carcinoma (SNSCC) [2]. SNSCC affects the nasal cavity and maxillary and paranasal sinuses, and is both aggressive and difficult to diagnose and treat [3]. Risk factors for this malignancy include carcinogen exposure from softwood dust, industrial compounds and chemicals (leather dust, glues, formaldehyde, chrome, nickel, arsenic and welding fumes), smoking, and human papillomavirus (HPV) infection [4]. The 5-year overall survival is estimated to be between 30 – 50% regardless of treatment, with local recurrence being the most common cause of mortality [3]. At present, treatment for SNSCC generally involves a combination of surgery and radiation, with or without chemotherapy. When chemotherapy is employed, it typically involves varying combinations of non-specific cytotoxic drugs, such as cisplatin, anti-metabolites or anti-tubulin agents [5]. However, treatment outcomes remain suboptimal, particularly for more aggressive or advanced tumors. Thus, there is an urgent clinical need to identify patterns of genetic aberrations that can reveal targetable pathways for precision therapies, facilitate discovery of biomarkers for predicting treatment response, and enhance understanding of tumor evolution to inform treatment decisions at various stages of the disease.

Advancement of biomarker-driven targeted therapies can further a precision medicine approach for treating SNSCC and improve overall patient outcomes. Achieving this will necessitate a thorough understanding of the somatic mutations that drive SNSCC development. Currently, there is a relative dearth of knowledge regarding the mutational landscape of SNSCC. To our knowledge, only one study has characterized the mutational landscape of SNSCC using whole genome sequencing (WGS) and WES [6]. Most efforts to date have evaluated mutations using either a limited set of candidate genes [7–13] or targeted sequencing platforms applied to formalin-fixed paraffin embedded (FFPE) tissue, without adjacent-normal or non-pathological control tissue [14–16]. This has not only restricted our understanding of the mutations that drive development of SNSCC but has also limited our ability to identify biomarkers that can predict treatment response, particularly as medicine has increasingly moved towards targeted and personalized therapies [17]. The goal of this study was to reduce this gap in our knowledge of the molecular underpinnings of SNSCC by comprehensively cataloging common somatic mutations in SNSCCs using whole-exome sequencing (WES).

METHODS AND MATERIALS

Patients and tissue samples

OCT-embedded paired tumor and adjacent normal tissue was obtained through the University of Cincinnati cancer center biorepository for 12 cases diagnosed with incident SNSCC from 2012 – 2014. Tissue was collected prior to initiation of chemotherapy or radiation. HPV was assessed using a multi-assay approach that included p16 immunohistochemistry (IHC), RNA in-situ hybridization (ISH), and HPV DNA sequencing [18]. Tumors were considered positive for high-risk HPV if (1) HPV viral oncogenes (E6 or E7) were detected via ISH; or (2) p16 was overexpressed and high-risk HPV DNA was detected. The overall fraction of malignant cells was quantified for each tumor sample by a board-certified anatomic pathologist; histopathologic classification and grading was subsequently confirmed by a second board-certified anatomic pathologist specialized in head and neck cancer. All participants provided written informed consent for use of their tissue samples as approved by the University of Cincinnati Institutional Review Board.

Library construction, cluster generation and sequencing

Genomic DNA was extracted from tumor and adjacent normal samples using the DNeasy Blood & Tissue kit (Qiagen, Valencia, CA) according to the manufacturer’s suggested protocol for purification of total DNA from human tissue. The exome sequencing library was prepared using the SureSelect Human All Exon V6 Target Enrichment System (Agilent, Santa Clara, CA). Briefly, 200 ng of high-quality genomic DNA was sheared by Covaris S2 (Covaris, Woburn, MA) to 150–200 bp and verified via Bioanalyzer (Agilent). The overhang ends of the fragmented DNA were end-repaired, adenylated, and ligated to the sample-unique pre-capture indexing adaptor. Adaptor-tagged DNA libraries were purified by AMPure XP beads (Beckman Coulter, Indianapolis, IN) and amplified by 11 cycles of PCR; quality of the amplified libraries was validated by Bioanalyzer with an expected fragment peak size of approximately 250 bp. Individually prepared libraries were then hybridized to SureSelectXT Human All Exon V6 Capture biotinylated RNA baits targeting exome sequences and captured by magnetic streptavidin beads. Exome libraries were uniquely indexed via 10 cycles of PCR, followed by AMPure XP bead clean up. Quality of the libraries was verified via Bioanalyzer with the expected distribution of the DNA peak size around 300 bp. Quantity of the libraries was measured by KAPA Library Quantification kit (Kapa Biosystems, Wilmington, MA) using an ABI 9700HT real-time PCR system (Thermo Fisher Scientific). Pooled libraries at the concentration of 15 pM were used for clustering in a cBot cluster amplification system (Illumina, San Diego, CA) with the TruSeq PE Cluster Kit v3 (Illumina). Clustered libraries were paired-end sequenced for 2×100 cycles using the TruSeq SBS HS 200-cycles kit v3 on a HiSeq 1000 system (Illumina).

Bioinformatic analysis

NF-core/sarek workflow version 2 [19] was applied to the sequencing reads generated from the SNSCC tumor and adjacent normal samples to detect somatic variants. Briefly, fastqc, and fastp tools were used to evaluate the quality of sequencing reads quality and trimming the sequencing reads (using the ––trim_fastq option). Trimmed sequencing reads were aligned to the human reference genome hg38 using BWA-MEM [20], accompanied by deduplication. GATK Mutect2 variant caller [21] was used to call somatic variants using the tumor with matched normal mode. Putative functional effects annotation for the somatic variants was performed using snpEff and VEP [22, 23]. Workflow output included Variant Call Format (VCF) files containing annotated somatic variant calls, sequences alignment/mapping files, and a MultiQC file containing quality control details of the sequencing reads. MutSig2CV was used to identify significantly mutated genes [24].

Statistical analysis

Mann-Whitney test was used to compare tumor mutation burden (TMB) between categories of clinical and demographic characteristics (HPV status, tumor grade, tumor site, sex and race). The association between TMB and age was evaluated using Spearman correlation. Pair-wise Fisher`s Exact test was used to evaluate co-occurrence or mutual exclusivity of mutated genes. The Benjamin-Hochberg test was used to account for multiple comparison in all statistical analyses [25]. Statistical analyses were conducted using R version 4.4.3. The statistical significance level for all statistical analyses was set at a q-value of 0.1.

RESULTS

Description of study subjects

This study included paired tumor and adjacent normal tissue from 12 patients diagnosed with SNSCC. The median age of the patients was 67.5 years (Interquartile range (IQR): 63.8 – 79.3) with 6 (50%) males and 6 (50%) females. Ten of the 12 cases (83%) were Caucasian, with the other two (17%) identifying as African American. Four of the tumors were positive for high-risk HPV, of which two were positive on RNA ISH; all four were p16 positive with hrHPV DNA detected (three HPV16, one HPV18). The predominant primary tumor sites were maxillary sinus (42%) and nasal cavity (42%), with the remaining 2 cases arising in the paranasal sinus (16%). Clinical and demographic characteristics of the SNSCC cases included in this study are provided in Table 1.

Table 1.

Clinical and demographic characteristics of the sinonasal squamous cell carcinoma cohort (n = 12)

Tumor ID	Age	Sex	Race	HPV Status (subtype)	AJCC TNM Stage	Extracapsular Spread	Perineural Invasion	Primary Tumor Site	ISP-associated SNSCC	Tumor Gradea
HDN1661	64	Male	Caucasian	Positive (HPV16)	T4AN0M0	No	No	Nasal cavity	No	2
HDN0753	60	Female	Caucasian	Positive (HPV18)	T4bN2bM0	Yes	No	Paranasal sinus	No	3
HDN0804	63	Male	Caucasian	Negative	T3N2cM0	Yes	Yes	Nasal cavity	No	1
HDN0466	90	Male	Caucasian	Negative	-	-	Yes	Nasal cavity	No	2
HDN0307	67	Female	Caucasian	Negative	T2N0M0	No	No	Maxillary sinus	No	2
HDN0358	68	Female	Caucasian	Negative	T3N0M0	No	No	Maxillary sinus	No	3
HDN0670	66	Male	Caucasian	Positive (HPV16)	T3N1M0	Yes	No	Nasal cavity	No	2
HND0019	79	Male	Black	Negative	T4aN2bM0	Yes	No	Paranasal sinus	No	1
HDN1261	80	Female	Caucasian	Negative	T2NxMx	-	-	Nasal cavity	No	2
HDN0908	72	Male	Caucasian	Negative	T2N0M0	No	No	Maxillary sinus	No	2
HDN1080	83	Female	Caucasian	Positive (HPV16)	T2N0M0	No	No	Maxillary sinus	No	2
HDN1726	47	Female	Black	Negative	T3N0M0	No	No	Maxillary	Yes	3

^aTumor grade categories: (1) well differentiated (2); moderately differentiated; and (3) poorly differentiated.

All dashes in the table represent missing data.

Abbreviations: HPV = human papillomavirus; AJCC = American Joint Commission on Cancer; TNM = tumor, node, metastasis; ISP = inverted sinonasal papilloma

Somatic mutations in coding regions

The average sequencing coverage of the coding regions in the tumor and adjacent normal samples was 45x, with 92.2% of reads achieving ≥ Q30 quality score (i.e. reads with an error probability ≤ 0.001), which is indicative of high-quality sequencing data. There was an average of 212 somatic mutations per tumor. A total of 263 genes contained ≥ 2 mutations in ≥ 2 tumors in multiple SNSCC tumors. Most variants were missense mutations (81%) dominated by cytosine-to-adenine (C>A) transversion (37%) and cytosine-to-thymine (C >T) transition (34%; (Figure 1). Exons of 11 genes were mutated in ≥ 33% of the tumors (Figure 2). Mutational load varied across tumors, with no statistically significant associations observed with HPV status, tumor grade, age, tumor site, sex or race. The median TMB in our study population was 3.43 mutations per megabase (IQR: 2.35 – 6.67), which is comparable to other head and neck malignancies (Figure 3) [26]. Eight genes were found to be significantly mutated in the SNSCC cohort (q < 0.1). The most frequently mutated gene was TP53 (6/12; 50%). The other significantly mutated genes included NOTCH1 (4/12; 33.3%), KMT2D (4/12; 33%), VWDE (4/12; 33%), FNBP4 (3/12; 25%), SCAND3(3/12; 25%), NOD1 (3/12; 25%) and OR5C1(2/12; 17%).

Figure 1.

Distribution of somatic variant classes, variant type and single nucleotide variant class in 12 sinonasal squamous cell carcinoma (SNSCC) tumors. (A) Total number of variant classes. (B) Total number of variant types. (C) Total number of single nucleotide variants. (D) Number of variant classes per sample. (E) Box plots showing median and interquartile range of variant classes. (F) Frequency of variant classes in 11 most frequently mutated genes.

Figure 2.

Tumor mutation burden, somatic variation classification, single nucleotide variant type and 20 most mutated genes in 12 sinonasal squamous cell carcinoma (SNSCC) tumor samples. The bar chart at the top of the figure indicates tumor mutation burden for each respective tumor. The bar chart on the right-hand side of the figure represents the number of tumors harboring mutations in each respective gene. Abbreviations: hrHPV = high-risk human papillomavirus; TMB = tumor mutational burden.

Figure 3.

Comparison of the median tumor mutation burden per megabase between the 12 sinonasal squamous cell carcinoma (SNSCC) in our study and 33 different cancer types from The Cancer Genome Atlas (TCGA). Median tumor mutation burden in our cohort (SNSCC-ExSeq; highlighted by the red box) is consistent with the TCGA head and neck squamous cell carcinoma HNSC (blue box).

To explore the potential impact of somatic mutations observed in the eight significantly mutated genes, we examined protein changes caused by these variants and how the changes may affect protein function. In TP53, we observed a frame shift deletion located in the proline-rich domain of the protein where valine at position 73 was the first amino acid affected (V73fs; Figure 4A). This V73fs frameshift deletion is a truncating mutation that produces protein forms that are truncated in the C-terminal. We also observed four missense mutations in the DNA binding domain of TP53: 1) amino acid methionine at position 133 changed to threonine (M133T); 2) methionine and alanine at positions 160 and 161 changed to isoleucine and serine, respectively (160_161MA>IS); 3) glycine at position 262 changed to valine (G262V); and 4) arginine at position 280 changes to lysine (R280K; Figure 4A). In Notch Receptor 1 (NOTCH1), a gene that encodes a transmembrane receptor protein and a transcription factor that can function as both an oncogene and a tumor suppressor, we observed 1 deletion, 2 nonsense mutations and 1 missense mutation that may generate truncated or inactive forms of the protein. Protein change analysis of the mutations in this gene identified a deletion of amino acid phenylalanine at position 357 (F357del) in the EGF-like ligand binding domain of the protein. In the ANK domain of NOTCH1, we identified two nonsense mutations where amino acids glutamic acid and glutamine at positions 1665 (E1665*) and 2057 (Q2057*), respectively, were changed to stop codons, and one missense mutation where tyrosine at position 1915 changed to phenylalanine (Y1915F; Figure 4B). We also observed putative driver mutations that could lead to the production of truncated and inactive lysine methyltransferase 2D (KMT2D), an enzyme involved in epigenetic regulation of gene expression. Alterations in this gene include three frame shift deletions (G1910fs, S2431fs, and Q2800fs) and one nonsense mutation (E5444*; Figure 4C). In addition, a nonsense mutation (E83*), a frame shift insertion (R299fs), and a missense mutation (Q518K) were observed in SCAN Domain containing 3 (SCAND3). These mutations could potentially be inactive or alter the function of this protein (Figure 4D). Missense and nonsense mutations and frameshift insertions that could produce truncated and inactive proteins were also identified in the remaining four significantly mutated genes (NOD1, FNBP4, VWDE and OR5C1; Figure 4E – H). Correlative pairwise analysis of mutated genes in our cohort did not show any statistically significant co-occurrence or exclusivity of significantly mutated genes.

Figure 4.

Schematic diagrams show protein domain structures affected by somatic mutations in significantly mutated genes in sinonasal squamous cell carcinoma (SNSCC): (A) TP53, (B) NOTCH1, (C) KMT2D, (D) SCAND3, (E) NOD1, (F) FNBP4, (G) VWDE, and (H) OR5C1.

DISCUSSION

We have identified eight significantly mutated genes in SNSCC, five of which have not been previously reported for this malignancy. To our knowledge, this analysis represents the most comprehensive interrogation to date of the mutational landscape of SNSCC using well-preserved tumor tissue and WES. Uncovering these significantly mutated genes is important for understanding SNSCC tumorigenesis, which will in turn inform the development of targeted therapies for SNSCC. This is particularly significant as the relative rarity of this condition precludes comprehensive clinical trials and current treatment paradigms are often extrapolated from other head and neck subsites. Thus, identification of targetable somatic mutations specific to SNSCC is the first step to advancing precision medicine.

Among the eight significantly mutated genes we identified, TP53, NOTCH1, and KMT2D have been reported to be mutated in SNSCC [6, 8, 15, 16, 27]. In our study, TP53 mutations were observed in half of the tumors, while NOTCH1 and KMT2D mutations each occurred in one-third of the tumors. These mutation frequencies where somewhat higher than those reported in a previous study that employed WGS and WES on FFPE tumor and matched adjacent normal tissue, where TP53, NOTCH1 and KMT2D mutations were observed at rates of 10%, 20% and 23%, respectively [6]. While the frequency of KMT2D mutations in our study is consistent with a small targeted sequencing study involving three inverted sinonasal papilloma (ISP)-associated SNSCC [15], mutation rates for TP53 and NOTCH1 somewhat differed from those observed in another targeted sequencing study by Hieggelke et al. [16], which reported a slightly higher TP53 mutation frequency of 61% and lower NOTCH1 mutation frequency of 9.7% across 49 and 31 SNSCC tumors, respectively. Variations in TP53, NOTCH1, and KMT2D mutation frequencies across studies may reflect differences in sample size, type of tissue samples used, and the proportion of HPV-positive and ISP-associated SNSCC cases.

Two of the other five significantly mutated genes we identified in SNSCC ─ FNBP4 and VWDE ─ have been reported to be mutated in other cancer types, including B cell lymphoma and lung adenocarcinoma (FNBP4) [28, 29] and breast cancer (VWDE) [30]. Another of the significantly mutated genes, OR5C1, encodes an olfactory receptor belonging to the G-protein-coupled receptor family of proteins which are involved in odorant detection within the olfactory epithelium. Perturbation of olfactory receptor pathways have been implicated in multiple other cancer types [31] and is particularly noteworthy in our study given that SNSCC arises in the nasal cavity and surrounding sinuses. To our knowledge, this is the first report identifying OR5C1 gene to be mutated in SNSCC, highlighting a novel tumorigenesis mechanism that warrants further investigation into its biological significance and potential as a therapeutic target. Identification of five significantly mutated genes not previously reported in SNSCC underscores the advantage of using WES or WGS over targeted sequencing approaches, as well as the benefit of fresh-frozen tumor tissue over PPFE when characterizing the mutational landscape of this malignancy. In our study, the use of WES enabled the detection of significant mutations in these five genes, which were not in earlier targeted sequencing panels. Furthermore, by utilizing well-preserved OCT-embedded tissue instead of FFPE tumor tissues, which are known to compromise nucleic acid integrity [32], we ensured greater accuracy and reliability in our analysis.

Many of the observed mutations in the significantly mutated genes appear to generate truncated proteins that could affect protein function. For example, the V73fs frameshift deletion we identified in TP53, a tumor suppressor gene that can also act as an oncogene when it acquires gain of function mutations [33], produces a truncated and inactive stable protein suggesting TP53 loss of function. This mutation is predicted to be inactivating and is associated with poor prognosis in cancer [34]. Experimental studies have shown that this mutation could be involved in promoting cancer cells proliferation, survival, and metastasis [33]. Similarly, a deletion (F357del) in NOTCH1, a gene with both tumor suppressor and oncogene properties [35, 36] has been shown to be inactivating [37]. Additionally, the 2 nonsense mutations in NOTCH1 may generate truncated proteins or proteins without critical functional residues suggesting loss of function leading to dysregulated NOTCH signaling. The nonsense and frameshift insertion mutations in SCAND3, a gene involved in regulating cell cycle and epithelial cell proliferation [38], may produce truncated and inactive proteins that could lead to dysregulation of these critical cell processes leading to the development of SNSCC. Truncating and inactivating frameshift deletions and nonsense mutations observed in KMT2D suggest dysregulated epigenetic regulation of gene expression of tumor suppressor genes that could contribute to the genesis and development of SNSCC. All this evidence points to the involvement of dysregulated apoptosis, NOTCH signaling, cell cycle, and epithelial cell proliferation, and epigenetic regulation of gene expression in SNSCC tumorigenesis.

Previous studies have identified epidermal growth factor receptor (EGFR) mutations as a recurrent feature in ISP, suggesting a potential role in its development [12, 15]. Given that ISP can progress to SNSCC, EGFR mutations should be expected in ISP-associated SCC. In our cohort, which included a single case of SNSCC derived from ISP, EGFR was not significantly mutated. This finding may reflect biological variability or limitations due to small sample size. The interplay between TP53 mutations and high-risk HPV in SNSCC reflects distinct molecular pathways of carcinogenesis. Although previous studies have shown that TP53 and high-risk HPV positivity are largely mutually exclusive in SNSCC [6, 8], we did not observe this exclusivity as expected. Half of the eight high-risk HPV negative tumors harbored TP53 hotspot mutations (three missense mutations and one frameshift deletion), and one of the four high-risk HPV positive tumors also exhibited one TP53 hotspot missense mutation. All TP53 missense mutations were located within the DNA-binding domain, the most critical functional region of p53. The lack of mutual exclusivity between TP53 mutations and high-risk HPV status in our study may be attributed to the small sample size.

Although most of the mutations observed in the significantly mutated genes have never previously been observed in head and neck cancer, all putative driver mutations we observed in TP53 gene and the F357 deletion we observed in NOTCH1 in our SNSCC cohort are also commonly mutated in head and neck squamous cell carcinoma [34, 37]. The presence of these mutations in SNSCC and other malignancies suggests similar tumorigenesis among these malignancies. Moving forward, larger whole-genome sequencing or WES studies are needed to further characterize the mutational landscape of SNSCC to identify new driver mutations and to validate the previously unobserved putative driver mutations we observed in our cohort. Furthermore, experimental studies will be needed to elucidate the functional significance of the putative driver mutations identified in our SNSCC cohort.

Major strengths of the study included availability of tumor tissue from SNSCC cases with well-annotated clinicodemographic data, use of frozen tumor tissue and adjacent normal tissue to call somatic variants, and our application of WES to comprehensively interrogate somatic mutations in coding regions. However, there are also several limitations to our study. The modest sample size limited the statistical power to detect significant somatic mutations in SNSCC or associations of somatic mutation profiles with age, high-risk HPV, tumor grade, tumor site, and race. Regardless, to our knowledge, this is still the largest reported cohort of SNSCC to undergo WES. While the comprehensive nature of WES is viewed as a strength, it is limited in that it is not well-suited for detection of copy number and structural variation and does not cover non-coding regions. Although relatively infrequent, non-coding driver mutations have been reported [39].

CONCLUSION

This study represents the first step towards a comprehensive understanding of the SNSCC mutational landscape. Although further sequencing is required in an expanded cohort to more precisely determine mutation frequencies, the present study identified numerous commonly mutated genes in SNSCC, thus advancing our understanding of its molecular basis and uncovering potential new therapeutic avenues for management of this uncommon, though important and devastating, malignancy.

Highlights.

• Eight genes were significantly mutated in our cohort of sinonasal squamous cell carcinoma (SNSCC).

• Five significant mutated genes have not been previously reported in SNSCC.

• Mutations included truncating and non-synonymous variants in functional domain of the corresponding protein.

• Somatic mutations in SNSCC may affect regulation of apoptosis, NOTCH signaling, cell cycle, and epithelial cell proliferation, and epigenetic regulation.

Data availability

The data generated and/or analyzed during this study are available from the corresponding author on request.