Introduction

Olfactory neuroblastoma (ONB), also called esthesioneuroblastoma, is a rare, locally aggressive, malignant neoplasm originating in the specialized sensory neuroepithelial olfactory cells found in the upper part of the nasal cavity [1]. Multiple modalities are currently in use to treat ONB, including surgical resection, radiotherapy, and chemotherapy. Numerous studies confirmed that a combination of surgery and radiotherapy is the treatment of choice for the majority of primary-site ONBs [2–6]. Advanced and metastatic ONBs are usually treated with classic chemotherapy, including etoposide, ifosfamide, cisplatin, cyclophosphamide, vincristine, doxorubicin, and nitrogen mustard [3,7]. However, due to the unpredictable biological behavior of the tumor and a lack of consensus on traditional treatment modalities [2,3], the recurrence rate of ONB remains high and effective treatment guidelines for high-grade ONBs are yet to be developed.

In recent years, with the advancement of molecular diagnostic methods, the focus has been on developing individualized targeted therapies for treating different types of cancer [8]. Cytogenetic studies on ONB revealed diverse and complex genomic imbalances in entire chromosomes and chromosome segments [9–11]. Several studies have reported copy number changes in ONB including gains at 7q11 and 20q and deletions at 2q, 5q, 6p, 6q, and 18q [10], as well as novel chromosome aberrations that have not been previously described. In addition, some chromosome regions could be implicated in the tumor progression and metastases formation [1,12]. These results not only indicate complex molecular processes underlying ONB, but also point to the need for a more detailed molecular characterization of ONBs at different stages of tumor progression.

Currently only a few studies have investigated genomic landscape of ONBs, using different sequencing techniques [13–17]. Furthermore, in three of these studies the potential of targeted therapy with specific drugs was explored. Weiss et al. [13] performed whole genome sequencing (WGS) on paired normal and tumor DNA from a patient with metastatic ONB. They detected mutations specific only to the metastatic ONB sample (i.e., in KDR, MYC, SIN3B, and NLRC4 genes) as well as mutations present in both, the metastatic and original surgical resection specimens (i.e., in TP53, TAOK2, and MAP4K2 genes) [1]. Analyzing cancer genomes from seven rare types of metastatic adolescent and young adult cancers (including ONB) using whole exome sequencing [WES], whole-transcriptome sequencing, or OncoScan^™, Cha et al. identified TP53 missense mutation in a metastatic ONB sample, as well as a loss-of-function in CDKN2C gene [14]. Based on these results, they proposed CDK4/6 inhibitors, palbociclib and LY2835219, as potential treatment strategies [16,18]. Similarly, a recent comprehensive genomic study of Gay et al revealed alterations of TP53, PIK3CA, NF1, CDKN2A, and CDKN2C in ONBs [16]. The study of Wang et al. was first to report a case of recurrent ONB treated with a targeted therapy regimen determined after WES. Mutations in EGFR, FGFR2, KDR, and RET genes were detected, therefore the authors utilized a combination of cetuximab and sunitinib [15].

Considering the lack of standardized treatment guidelines, the potential advantages of targeted therapy approaches [8] and the paucity of data exploring the molecular pathogenesis of ONB, we explored potentially targetable biomarkers/pathways in a cohort of recurrent or metastatic ONBs, using multiplatform molecular profiling approach. We identified multiple protein biomarkers of response or resistance to classic chemotherapy and targeted therapy that could be useful in optimizing the cytotoxic chemotherapy and further improving personalized treatment of ONB.

Materials and methods

Gene fusions

Six recent ONB cases were tested for gene fusions using Archer FusionPlex Solid Tumor Kit with Illumina MiSeq (Table 1, S2 Table, and online: http://www.carismolecularintelligence.com/solid_tumors_international).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Results of in situ hybridization, sequencing and gene fusion assays.

https://doi.org/10.1371/journal.pone.0191244.t001

Results

Molecular profiling using IHC, ISH, sequencing and gene fusions

The IHC, ISH, and sequencing results are summarized in Fig 1, Table 1 and S2 Excel File.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Profiling results of 25 biomarkers using immunohistochemistry (IHC) and in situ hybridization (ISH).

AR: androgen receptor; ER: estrogen receptor; PR: progesterone receptor; ALK: anaplastic lymphoma kinase; HER2: Human epidermal growth factor receptor; EGFR: epidermal growth factor receptor; c-Kit: tyrosine protein c-Kit receptor kinase; PDGFR: platelet-derived growth factor receptor; c-MET: MET proto-oncogene, receptor tyrosine kinase; PTEN: phosphatase and tensin homolog; pan-NTRK: tyrosine receptor kinase; PD-1: programmed cell death-1; PD-L1: programmed death ligand-1; ERCC1: excision repair cross-complementation group 1 protein; SPARC-M: Serum protein acidic and rich in cysteine M; RRM1: ribonucleotide reductase M1; TLE3: transducin-like enhancer of split 3; TUBB3: tubulin beta-3 chain; TOPO1 and TOP2A: topoisomerases 1 and 2 alpha; TS: thymidylate synthase; MGMT: O(6)-methylguanine methyltransferase; BCRP: breast cancer resistance protein; MRP1: multidrug resistance associated protein 1; PGP: P-glycoprotein.

https://doi.org/10.1371/journal.pone.0191244.g001

No cases expressed PD-L1 (0/10). Multiple protein biomarkers of response or resistance to classic chemotherapy drugs were identified: PD-1 positive tumor infiltrating lymphocytes in 25% (2/8), low ERCC1 (cisplatin sensitivity [23] in 83% (10/12), high TOPO1 (irinotecan sensitivity [24] in 63% [12/19], high TUBB3 (vincristine resistance [25] in 93% [13/14] (Fig 2F), and high MRP1 (multidrug resistance) in 100% (6/6). Four out of six tested ONBs (67%) were positive for pNTRK (Fig 2E).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2.

(A): Hematoxylin and Eosin (H&E) figure of a case with upregulation of CD24 gene by microarray, confirmed by CD24 protein overexpression in the tumor cells (B); (C) A case of olfactory neuroblastoma (ONB) with CTNNB1 mutation [S33_ H36del] confirmed by the nuclear expression of β-catenin; Another case of ONB with CTNNB1 mutation [S33P] with retained cytoplasmic/membranous expression of β-catenin protein (D); A case of recurrent ONB with pNTRK overexpression (E) and overexpression of TUBB3 (F).

https://doi.org/10.1371/journal.pone.0191244.g002

Mutations (pathogenic and variants of unknown significance [VUS]) were detected in 10/16 (63%) ONBs including tumor suppressor p53 (TP53) [3 cases], beta-catenin 1 gene (CTNNB1), EGFR [2 cases, respectively], while single cases harbored Adenomatous Polyposis Coli (APC), cKIT, cMET, Platelet Derived Growth Factor Receptor Alpha (PDGFRA), CDH1 (E-cadherin), Fumarate Hydratase (FH) and SMAD4 gene mutations (Table 1). CTNNB1 gene alterations were further evaluated by the IHC (Fig 2C and 2D).

None of the cases harbored gene amplifications of EGFR, HER2 and cMET genes. Also, gene fusions were not identified in any of the 6 successfully tested ONBs (the panel of fusion genes is available in S2 Table).

Microarray results

When compared with the control tissue, 21 genes were consistently over-expressed and 19 genes consistently downregulated by an average of 10 fold. Some of the upregulated genes, such as Secretogranin II (SCG2), stem cell marker cluster of differentiation 24 [CD24] (Fig 2A), and insulin-like growth factor binding protein 2 (IGFBP-2), and downregulated genes (ATP-binding cassette transporter 8 [ABCA8] have been described to play a role in different malignancies, and were not hitherto described in ONBs (Table 2). Among the downregulated genes, GHR is a novel observation as most studies associate GHR over-expression as a risk factor for cancer. CD24 gene expression has been confirmed by the IHC (Fig 2B).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Selected genes’ mRNA expression detected in olfactory neuroblastoma samples using Illumina array.

https://doi.org/10.1371/journal.pone.0191244.t002

In order to better understand the pathways perturbed in ONB, we extended the list of over- and under-expressed genes to those that were consistently over- and under-expressed by two fold followed by functional classification of these genes using the Panther website (http://pantherdb.org) and the Reactome database (see S3 Excel File for the list of genes). Overall, 183 genes were found to be downregulated and 146 were over-expressed. At the nominal p-value of 0.05, fifty-nine Reactome pathways were shown to be enriched with the 183 downregulated genes in ONB including 10 genes in the organization of extra cellular matrix and 4 genes in the cell junction organization. Pathway analysis of the 146 upregulated genes identified 100 enriched pathways including 13 genes in the Cell Cycle pathway, 10 genes in the TP53 pathway, 8 genes in Chromatin Modifying Enzyme pathway (The list of all enriched pathways can be found in S3 Excel File).

Discussion

Recent studies demonstrated potential therapeutic benefits of comprehensive molecular profiling for the patients with advanced and/or metastatic cancers [16,26–28]. In this study, we explored a wide range of potentially targetable biomarkers/pathways in recurrent or metastatic ONB samples using multiple molecular profiling platforms, including IHC, ISH, expression microarray and NGS. The sequencing results showed mutations in TP53 (n = 3/16), CTNNB1 (n = 2/16), EGFR (n = 2/16), APC, cKIT, cMET, PDGFRA, CDH1, FH, and SMAD4 genes (n = 1/16, respectively). Multiple genes within the Wnt/β-catenin signaling pathway including CTNNB1, APC and CDH1 exhibited mutations within this cohort. Loss-of-function mutations in these genes lead to deregulated Wnt/β-catenin signaling and excessive stem cell renewal/proliferation, and are associated with metastatic disease [29,30]. However, we found no targetable Wnt pathway enrichment in our cohort using whole-genome expression assay. The potential of several anti-cancer drugs has been explored by targeting different stages or components of Wnt/β-catenin signaling pathway with limited success [31,32]. The role of cKIT and PDGFRA mutations has been, most notably, investigated in gastrointestinal stromal tumors (GIST), where these two mutations are mutually exclusive [33]. Imatinib (tyrosine-kinase inhibitor) response in GIST patients depends not only on the protein expression, but also on the type of mutation in KIT and PDGFRA genes [33,34]. One ONB case in our study had a pathogenic cKIT mutation while another harbored VUS PDGFRA mutation. Both EGFR gene mutations in our study were VUS while EGFR amplification was not observed in any of the tested cases. These results indicate a limited therapeutic benefit of EGFR inhibitors in ONB patients. Mutations in TP53 gene were also detected in other studies that performed DNA sequence analysis in ONB samples [13–17]. Due to the disease progression in those patients, a role of TP53 mutation as an unfavorable prognostic and predictive factor in ONB has been suggested [18]. Of note, tumors harboring TP53 mutations may be sensitive to WEE kinase inhibitors acting against G2-M checkpoint regulators of the cell cycle WEE1 and CHK1 [35].

In addition, our microarray analysis revealed up or downregulation of several genes previously implied in carcinogenesis but not previously described in ONBs, including CD24, SCG2, and IGFBP-2 among the upregulated genes, and ABCA8 and GHR among the downregulated genes. In a recent study by Dvorak et al. a comprehensive expression analysis of all members of the ABC transporter genes across multiple cancers showed that ABCA8 downregulation was more observed in higher grade and its upregulation was associated with lower grade tumors [36]. This is consistent with the data presented in our paper that ABCA8 was consistently downregulated in all four high-grade ONB samples. Interestingly, the study by Dvorak et al. was able to show that out of all 49 ABC transporters that were investigated in various tumors, ABCA8 and four others (ABCC7, ABCC8, ABCA3, and ABCA12) were among the most dysregulated ABC genes [36]. ABCA8 has also been studied by others in relation to cancer and it has been found to be downregulated in multidrug resistant ovarian cancer cell lines [37] as well as in breast and prostate cancer [38,39]. In one study, ABCA8 was found to be upregulated in a subtype of medulloblastoma defined as Sonic hedgehog (SHH) and downregulated in the subtype defined as Wnt signaling [40]. It should be noted that the downregulation of GHR gene in ONB was an unexpected result, as most studies in the literature associate GHR upregulation with increased cancer risk [41].

More recently, immune checkpoint inhibitors (anti-PD-1/PD-L1) have revolutionized the treatment of many tumors with most remarkable benefits in the patients with melanoma, non-small cell lung carcinoma, renal cell carcinoma, bladder carcinoma, and classical Hodgkin lymphoma [42]. Our study is the first to report on the lack of PD-L1 expression in ONB samples, which makes these patients less likely to respond to anti-PD-1/PD-L1 drugs.

Recently, Gay et al. comprehensively profiled 41 samples of ONBs identifying potential targets in the mTOR, CDK and growth factor signaling pathways [16]. Other, small studies on molecular characteristics of ONB showed variable and largely inconsistent results. In the study of Weiss et al. 119 somatically lost genes and 45 gained or amplified genes were reported in a metastatic ONB sample using whole genome sequencing [13]. Seven somatic short nucleotide variants (SNVs) were validated by Sanger sequencing. Specific mutations in KDR, MYC, SIN3B, and NLRC4 genes were present only in the metastatic ONB sample, while mutations in TP53, TAOK2 and MAP4K2 genes were present in both the metastatic and original surgical resection specimens [13]. Our study confirmed some of these mutations (e.g. TP53) but failed other mutations including MYC mutation. In contrast to other genetic alterations (copy number variations, chromosomal translocations, increased enhancer activity), MYC gene mutations are uncommon [43], but have been described in some cancers (e.g. lymphomas) [44]. Furthermore, the described MYC mutation in ONB [13] has not been verified in the COSMIC database (Catalog of Somatic Mutations in Cancer). Notably, none of our ONB cases harbored MYCN gene amplification, a hallmark of pediatric neuroblastomas [45]. Further studies should definitely elucidate the role of MYC gene(s) in ONBs.

Using two different genomics platforms, Cha et al. reported a TP53 missense mutation in a metastatic ONB sample and a loss-of-function in CDKN2C gene [14]. Wang et al. detected mutations in EGFR, FGFR2, KDR, and RET genes in a recurrent ONB sample, using WES. In addition, EGFR and KDR genes were over-expressed in the tumor tissue [15].

Despite over-expression of the tropomyosin receptor kinase receptor family (NTRK) observed in 4/6 tested ONB cases, we did not detect any fusion of either one of the three NTRK genes (NTRK1, NTRK2, and NTRK3). Cancers with overexpression of NTRK driven by gene fusions had been successfully treated with novel NTRK kinase inhibitors [46] but it remains unclear if a “constitutive” overexpression such as this observed in ONB would offer any treatment advantages. Interestingly, one of the cases with NTRK overexpression exhibited TUBB3 positivity and was CD24 positive (Fig 2) indicating a potential therapeutic benefit of retinoid-based therapy [47].

ALK (anaplastic lymphoma receptor tyrosine kinase) gene alterations have been reported in various cancers including pediatric neuroblastomas [48]. We found no ALK gene alterations (mutations or fusions) or over-expression of Alk protein in any of the tested ONB cases, so these patients are unlikely to benefit from the ALK inhibitors.

In several studies, alternative approaches to ONB treatment, especially in metastatic or recurrent cases, have been explored. A combination of cetuximab (a monoclonal antibody to EGFR) and sunitinib (small-molecule inhibitor of receptor tyrosine kinases [RTKs] including kinase insert domain receptor [KDR], fibroblast growth factor receptor 2 [FGFR2], and RET RTK) was selected as a treatment regimen in a case of recurrent ONB, after WES analysis. One month after this treatment, a complete response was observed in the patient [15]. In another study, a significant improvement of clinical symptoms and disease stabilization for 15 months were observed after treatment with sunitinib in a patient with progressive ONB [49]. Furthermore, imatinib mesylate was reported as a potential second-line treatment for inducing long-term remission in heavily pretreated ONB patients [50]. Young et al. went a step further and explored the efficacy of several combinations of targeted drugs in human ONB cell line TC268 [51]. The combinations of AEW541 (insulin-like growth factor 1 [IGF-1] inhibitor) and FS114 (ribosomal protein S6 kinase beta-1 [S6K1] inhibitor) or sunitinib and FS115 (S6K1 inhibitor) were the most effective according to their results [51]. In addition, Sabongi et al. described a case with multiple recurrences of ONB adjuvantly treated with the radiolabeled-somatostatin analogue, ¹⁷⁷Lu-DOTA-TATE. After three cycles of ¹⁷⁷Lu-DOTA-TATE treatment, the stabilization of the disease was reported [52]. Finally, Mao et al. investigated the role of SHH signaling pathway in the development of ONB, by treating ONB cell lines (TC-268 and JFEN) with cyclopamine, a selective inhibitor of the SHH pathway [53]. After the cyclopamine treatment, inhibited ONB cell proliferation and colony formation, induced ONB cell cycle arrest and apoptosis, downregulated expression of SHH signaling components, i.e. PTCH1 and Gli1, and CCND1 (cyclin D1, cycle-related regulator), as well as upregulated p21 expression were observed in vitro [53].

The efficacy of classic chemotherapy in ONBs remains unclear [2,4]. Chemotherapy alone, or in combination with radiotherapy, is often limited to advanced and surgically inoperable ONB cases [54]. Our IHC results indicate that several biomarkers may be used in tailoring the classical cytotoxic drugs including cisplatin and irinotecan sensitivity and vincristine resistance.

Our study have several limitations including a small sample size and lack of clinical (follow-up) data. In addition, all samples were not tested with all methodologies as these have been dynamically changing per molecular testing advances/improvements. This may result in insufficient and biased therapeutic implications.

Although our data indicate limited therapeutic options in patients with advanced and/or metastatic ONBs, several potential biomarkers that could tailor both targeted (e.g., Wnt and cKIT/PDGFRA) and classical therapeutic options merit further research. The therapeutic benefits of immune checkpoint inhibitors are less likely due to the low or lack of PD-1/PD-L1 expression.

Supporting information

Acknowledgments

The preliminary results of the study were presented at the 40^th European Society of Medical Oncology (ESMO) Congress that was held in Copenhagen, Denmark, October 7–11, 2016.

We thank Yvonne Veloso, Peggy Gates, Qing Zhang, Sting Chen and Hongseok Tae (Caris Life Sciences) for their excellent technical and bioinformatics support.