Transcriptomic and immunophenotypic characterization of two cases of adamantinoma‐like Ewing sarcoma of the thyroid gland

Adamantinoma‐like Ewing sarcoma (ALES) is a rare aggressive malignancy occasionally diagnosed in the thyroid gland. ALES shows basaloid cytomorphology, expresses keratins, p63, p40, frequently CD99, and harbours the t(11;22) EWSR1::FLI1 translocation. There is debate on whether ALES resembles more sarcoma or carcinoma.


Introduction
Adamantinoma-like Ewing sarcoma (ALES) is a rare aggressive malignancy, sharing morphologic features with Ewing's sarcoma (ES) of the bone and soft tissue (BST) and harbouring the t(11;22) EWSR1::FLI1 translocation. 1 Initially considered a variant of ES or adamantinoma, 2 ALES frequently shows squamous differentiation and expression of high molecular weight keratins, p63 and p40, findings more suggestive of basaloid epithelial or myoepithelial origin rather than genuine sarcoma. 1 There is ongoing debate regarding the true nature of this entity, and whether it is best classified as a sarcoma or carcinoma. 3 Originally described as a lesion of the bones, 2 ALES occasionally arises in the head and neck, particularly the salivary glands, sinonasal tract, and thyroid gland. 1 The true nature of ALES of the thyroid has been debated, with Oliveira et al. using the term "carcinoma with Ewing family tumor elements (CEFTE)" and hypothesizing an origin from papillary thyroid carcinoma or solid cell nests. 4 We performed an investigative analysis aiming at describing the transcriptomic profile of ALES, also in comparison with skeletal ES, in an attempt to better characterize this interesting entity and elucidate its origin and pathogenesis.

C A S E S E L E C T I O N A N D P A T H O L O G Y R E V I E W
The study was approved by the Institutional Review Board (IRB# 18-005928, 8/6/2018). The institutional and consultation files were searched for patients diagnosed with thyroid ALES and skeletal ES, between 2013 and 2018. Archival surgical pathology material was retrieved and reviewed for adequacy and morphology. Nonneoplastic thyroid tissue from patients with recent thyroidectomy for papillary thyroid carcinoma was obtained to serve as a baseline control. Unstained sections (4 lm) were obtained from formalin-fixed paraffinembedded (FFPE) tissue blocks and submitted for further testing. Clinicopathologic data were obtained through chart review.
The mRNA libraries for the tumour samples were prepared using Agilent Exome V7 capture baits with respect to human genome reference hg19 genomic positions. The libraries were sequenced as 101 9 2 paired-end reads on an Illumina HiSeq 2500 instrument (Illumina, San Diego, CA, USA). The RNA-Seq data were analysed using the clinical Mayo Bioinformatics RNA-Seq workflow, MAPRSeq v.1.2.1.5, 5 which includes alignment with TopHat 2.0.6 6 and quantification of gene expression using the HTSeq software. 7 Gene fusions were identified using the Tophat-Fusion alogorithm. 8 Annotation of fusions in the sarcoma samples was performed using an in-house-developed and clinically validated bioinformatics module. Normalized gene counts were also obtained from MAP-RSeq, where expression for each gene were normalized to 1 million reads and corrected for gene length (fragments per kilobase pair per million mapped reads, FPKM). The FPKM values were used to perform unsupervised clustering using principal component analysis (PCA) to inspect differences between the various sarcoma and normal groups. Differential expression analysis was performed on the raw gene counts, using the bioinformatics package edgeR 2.6.2. 9 Genes found significantly different between the various sarcoma and normal groups were reported with magnitude of change (log2 scale) and level of significance (False Discovery Rate, FDR <5%). Canonical pathway analysis was performed using the Ingenuity pathway analysis software IPA (Ingenuity Systems, Qiagen, Redwood City, CA, USA) to identify pathways associated with differentially expressed genes in the sarcoma samples. Gene set overlap analysis was performed by using the Molecular Signatures Database (MSigDB) 10,11 Queries for tissue-specific gene expression were submitted to the Human Protein Atlas (http://www.proteinatlas.org). 12

C L I N I C O P A T H O L O G I C C H A R A C T E R I S T I C S
A summary of patient clinicopathologic characteristics is presented in Table 1. Two patients were diagnosed with ALES in our consultation practice, between 2013 and 2018. The patients were a 49-year-old-man and a 31-year-old woman with large thyroid masses (4.5 and 6 cm, respectively). Microscopic examination ( Figure 1) revealed high-grade malignancies with nested, solid, and trabecular architecture ( Figure 1A, B), areas of geographic necrosis ( Figure 1C), and increased mitoses. No low-grade component or readily identifiable area of preexisting papillary thyroid carcinoma was identified in either of the tumours. In a few areas, entrapped thyroid follicles were present, highlighted with immunostain for TTF1 ( Figure 2C). Immunohistochemically, the tumour cells expressed keratins (AE1/AE3 and CAM5.2), particularly of high molecular weight (keratin 5) ( Figure 3B), as well as p40 and p63 ( Figure 3C) and focally CD99 ( Figure 2A). We saw "block-like" reactivity for p16 ( Figure 2B), which prompted further evaluation with high-risk HPV DNA-ISH, which was negative. Expression of INI1 was retained in both tumours. No expression of keratin 7, keratin 20, CD45, CD20, CD5, chromogranin, synaptophysin, calcitonin, thyroglobulin, PAX8, TTF1, S100, NUT, desmin, ER, FLI1, desmin, or myogenin was seen in tumour cells. A summary of the immunohistochemical profile, along with the corresponding RNA expression findings is presented in Table 2.
A search for patients with a diagnosis of BST ES yielded 19 cases, of which four were selected for further evaluation on the basis of tissue availability and suitability (nondecalcified tissue). All four patients were young (age range 6-20 years) and were diagnosed either with ES of the chest wall or the skull. Morphologically, all four lesions showed classic morphologic features of ES.
We randomly selected two patients who underwent thyroidectomy for papillary thyroid carcinoma and macrodissected nonneoplastic thyroid tissue to serve as a control.
We identified EWSR1::FLI1 fusion transcripts with retained EWSR1 exon 8 in both ALES cases ( Figure 4B, C). All four BST ES cases had a EWSR1::FLI1 fusion transcript with break points between EWSR1 exon 7 and FLI1 exon 6, as determined by RNA sequencing ( Figure S1). We also found that splicing regulators of the pre-mRNA of EWSR1::FLI1 fusion transcript HNRNPH1, 13 SUPT6H, and SF3B1 14 were overexpressed in all ALES and BST ES cases. The complete gene fusion dataset is presented in Table S2.
Three hundred and forty-one genes were overexpressed in ALES only in relation to thyroid without significant expression difference from BST ES. Further query at the Molecular Signature Database 10,11 showed that 53 (15.5%) of those 341 genes overlapped with a gene set overexpressed in mesenchymal stem cells with induced expression of the EWSR1::FLI1 fusion  Figure 4A). Of those, NKX2-2 has been extensively studied in BST ES 16 and TNNT1 seems to be a promising biomarker of adverse behaviour. 17 In ALES 461 genes where differentially expressed from BST ES, of which 395 (85.7%) overexpressed and 66 (14.3%) underexpressed. In relation to nonneoplastic thyroid tissue ALES had differential expression of 1346 genes, of which 682 (50.7%) were upregulated and 664 (49.3%) were underexpressed.
Unique to ALES was the differential expression of 137 genes (Table S3), which were further divided into four groups as follows: Group 1: 86 (62.8%) genes were overexpressed relative both to ES and nonneoplastic thyroid; Group 2: 22 (16.1%) genes were underexpressed in ALES relative to both ES and nonneoplastic thyroid; Group 3: 25 (18.2%) genes were overexpressed relative to ES, but underexpressed in nonneoplastic thyroid; and Group 4: The following four (2.9%) genes were underexpressed relative to ES, but overexpressed relative to nonneoplastic thyroid: CSMD2; HOXD9; ENPP3, and SERPINB13.
Further analysis of the above gene sets with the Molecular Signatures Database (MSigDB) 10,11 showed a significant overlap of subset of Group 1 genes with gene sets related to development of the skin and epidermis, 18,19 keratinocyte differentiation, 20 and keratin intermediate filaments 21 (Figure 3A). Organ-specific gene expression queries (http://www.proteinatlas.org) 12  showed that Group 2 genes are normally expressed in skeletal muscle and thyroid or in various tissues with nonspecific distributions. Group 3 genes show thyroidspecific expression and can be attributed to the background thyroid included in the ALES samples. Group 4 genes seem to play a role in head and neck neoplasia, as further discussed in detail below.
Immunohistochemistry correlated well with the RNA expression findings (Table 2). This was most prominent in the case of KRT5 and strongly positive keratin 5 immunostain, as well as TP63 with nuclear expression of p63 and its isoform p40. 22 Notably, the lack of TTF1 and thyroglobulin expression in ALES was associated with underexpression of the corresponding genes NKX2-1 (TTF1) and TG (thyroglobulin).
Detailed results of the RNA expression profiling, particularly regarding genes with statistically significant differential expression between ALES and BST ES, can be found in Table S3.

Discussion
ALES of the thyroid is a rare aggressive malignancy with broad differential diagnosis, including Ewing's sarcoma, poorly differentiated carcinoma of thyroid, medullary carcinoma, and spindle cell epithelial tumour with thymus-like elements. Immunohistochemical expression of keratins, p63, p40, and CD99 in ALES is particularly useful for distinguishing it from those morphologically similar lesions. 23 In organs of the head and neck other than the thyroid, the differential diagnosis of ALES is even broader. Bishop et al. highlight ALES as a diagnostic pitfall in the differential diagnosis of the much more common basaloid and myoepithelial neoplasms. Strong keratin reactivity can lead to the misdiagnosis of carcinoma with basaloid features, further confounded by the reactivity for p16 that can occasionally be observed. 1 Analogous findings are reported in the recent study by Bal et al., who also praise the role of NKX2.2 and FLI1 immunohistochemistry to help differentiate ALES from its close mimics. 24 In the salivary glands, ALES has to be distinguished from basaloid neoplasms such as adenoid cystic carcinoma, myoepithelial carcinoma, 25 basal cell adenocarcinoma, 26 and in paediatric patients sialoblastoma. 27 ALES of the salivary glands can show less palisading and keratinization, as well as more frequent synaptophysin expression than in nonsalivary, clues that can be diagnostically useful. 25 In the sinonasal tract ALES has to be differentiated from sinonasal undifferentiated carcinoma (SNUC), 28 as well as SMARCB1 (INI1) deficient 26 or SMARCA4 (BRG1) deficient carcinomas. 26,27,29 Immunohistochemistry helps in the latter two, but SNUC remains a diagnosis of exclusion, although immunohistochemistry for IDH1/2 may prove to be useful. 30 ALES is a high-grade thyroid malignancy with EWSR1::FLI1 fusion, 1 long considered as pathognomonic for the diagnosis of ES. We therefore focused our analysis in comparing the molecular profile of ALES to that of BST ES. At the same time, we chose to compare the ALES molecular profile with nonneoplastic thyroid to assess findings that could elucidate a possible cellular origin of ALES from thyroid follicular cells. The large number of differentially expressed genes between ALES and nonneoplastic thyroid, along with a lack of thyroglobulin expression by tumour cells, does not support such a hypothesis. Although the exact histogenesis of ALES remains unclear, a possible origin from solid cell nests has been suggested. 31 The possibility of ALES arising in a background of papillary thyroid carcinoma needs also to be taken into consideration by future studies, especially given the fact that EWSR1 rearrangements, as well as the presence of EWSR1::FLI1 fusions have been detected in papillary thyroid microcarcinomas. 32 In our study we did not observe any well-differentiated component in any of the tumours that could suggest development of ALES as a progression of low-grade disease. Despite its overall high-grade features and aggressive biologic behaviour, occasional cases of long-term ALES survivors have been reported. 33 We identified the presence of EWSR1::FLI1 fusion transcript with retained EWSR1 exon 8 in both cases of ALES in our study. The most common EWSR1:: FLI1 transcripts found in about 70% of BST Ewing's sarcoma, and also known as "type 1" transcripts, have breakpoints in EWSR1 exons 1-7 and FLI1 exons 6-10. 34,35 Apparently, a retained EWSR1 exon 8 leads to a nonfunctional out-of-frame EWSR1::FLI1 transcript that needs to be spliced out in order to be translated into a functional oncogenic protein. 36 The splicing process demands expression of several genes, of which the most important is HNRNPH1, encoding heterogeneous nuclear ribonucleoprotein H1. 13 Also, SUPT6H and SF3B1 need to be expressed for nondisrupted splicing of EWSR1 exon 8. 14 Our RNA sequencing results showed overexpression of all three genes in ALES, which allows us to hypothesize that the retained EWSR1 exon 8 was effectively spliced out in both of our cases. The expression upregulation of genes known to be downstream to the EWSR1::FLI1 fusion transcript cascade (TNNT1, NKX2.2) is also in keeping with this hypothesis, which nevertheless warrants further experimental validation. More specifically, TNNT1 encodes slow skeletal muscle troponin T and is overexpressed in various malignancies, 37 including breast 38 and uterine leiomyosarcoma, 39 and has been associated with enhancement of cellular motility and invasiveness. 37 In Ewing's sarcoma, a high expression level of TNNT1 is a promising prognostic biomarker of adverse outcomes. 17 NKX2.2 is a downstream target of EWSR1::FLI1 which is necessary for its oncogenic function and is consistently expressed in Ewing's sarcoma, as well as in ALES. 16 Uniquely overexpressed in relationship to nonneoplastic thyroid tissue, but expressed less than in BST Ewing's sarcoma, were four genes related to head and neck neoplasia: CSMD2 encodes a synaptic transmembrane protein essential for neuronal maturation, 40 which has also been reported as a fusion partner for EWSR1 in Ewing's sarcoma, 41 and its expression is upregulated in carcinomas of the head and neck. 42 Yang et al. also reported somatic driver mutations of CSMD2 in papillary thyroid carcinoma, 43 which could be supportive of the hypothesis previously mentioned by Oliveira et al., according to which ALES, or "carcinoma of the thyroid with Ewing family tumor elements", could arise in a background of preexisting papillary thyroid carcinoma. 4 However, we did not observe any other morphologic, immunophenotypic, or molecular findings supportive of this hypothesis. ENPP3 (also known as CD203c) encodes a marker of plasmacytoid dendritic cells that are frequently abundant in squamous cell carcinomas of the head and neck, as well as their cervical nodal metastases. 44 SERPINB13 expression is modulated in squamous cell carcinomas of the head and neck, possibly in relationship with exposure to ultraviolet radiation. 45 Finally, HOXD9 is overexpressed in oesophageal squamous cell carcinoma. 46 The present study was designed as a phenotype-genotype correlation study with the goal of assessing the molecular relationship of thyroid ALES, as compared to that of classical BST ES. Since cases were retrieved from institutional consultation files, comprehensive data regarding treatment and patient outcome are not available. Moreover, the small number of cases would not allow for unbiased outcome analysis. 47 This limitation hinders the ability to further elucidate possible similarities or differences between thyroid ALES and BST ES regarding patient treatment and clinical outcomes.
In conclusion, ALES is a high-grade malignancy harbouring the EWSR1::FLI1 fusion and showing overlapping features of basaloid squamous cell carcinoma and skeletal Ewing's sarcoma. Our study sheds more light into a common classification dilemma that has been expressed by several authors 3,23 regarding whether ALES constitutes a genuine sarcoma, or a carcinoma with the EWSR1::FLI1 fusion transcript. Our findings showed strong expression of genes related to squamous epithelial differentiation, with coexpression of genes expressed downstream of the EWSR1::FLI1 fusion transcript activation cascade. This supports the hypothesis that ALES is a unique malignancy with concurrent, overlapping features of both carcinoma and sarcoma.

Supporting Information
Additional Supporting Information may be found in the online version of this article: Table S1. Antibodies used for immunohistochemical characterization of adamantinoma-like Ewing's sarcoma of the thyroid gland Table S2. Gene fusion dataset. Table S3. Complete RNA expression profiling dataset. Figure S1. Breakpoints for the EWSR1::FLI1 fusion transcripts identified in skeletal Ewing's sarcoma by RNA sequencing: A. Case 1; B. Case 2; C. Case 3; D. Case 4.