Unique Thyroid Carcinoma With Somatic Inactivating TSC2 Mutations, Follicular and Melanocytic Differentiation and Distinct Morphologic Features

Abstract

Thyroid carcinomas with mutations in the TSC2 gene have been described rarely. Here, we report a unique thyroid carcinoma that has somatic mutations in TSC2, follicular and melanocytic differentiation and distinct morphologic features, supported by a comprehensive molecular-genetic analyses. No other known pathogenic driver mutations were identified. The TSC2 mutations are predicted to be truncating and result in constitutive activation of mammalian target of rapamycin (mTOR) signaling that was confirmed by transcriptome expression profiling. This is the first demonstration of a thyroid follicular tumor with melanocytic differentiation and corresponding Melan-A expression may be a biomarker for this thyroid cancer. Its diffuse papillary architecture, macronuclei, multinuclear aggregates and intratumoral T lymphocytes are morphologic features not yet described in other TSC2-mutated thyroid carcinomas. Morphologic overlap with classic thyroid papillary carcinoma and known technical difficulties in sequencing the large TSC2 gene make it likely that TSC2-mutated thyroid carcinomas have been underdiagnosed so far. We suggest that TSC2-mutated thyroid carcinoma should be considered in the differential diagnosis of any thyroid tumor that has unusual morphologic and immunohistochemical features or lacks known driver mutations in thyroid cancer.

The precise diagnosis of cancers arising from thyroid follicular cells often requires integration of morphologic and molecular features in the tumor. In general, most well-differentiated thyroid carcinomas can be diagnosed into one of 3 main morphologic-molecular categories that have characteristic clinical tendencies.^1,2 For example, BRAF mutation-like thyroid carcinomas include classic and infiltrative follicular variant papillary carcinomas that most often have BRAF point or RET fusion mutations^3–9 and a propensity for local-regional lymph node spread. In contrast, RAS mutation-like thyroid carcinomas include invasive encapsulated follicular variant of papillary carcinoma, noninvasive thyroid follicular neoplasm with papillary-like nuclear features and thyroid follicular carcinomas that most often have NRAS point or PPARG fusion mutations^10–18 and a propensity for lower grade and vascular invasion. Some mutations such as the NTRK fusions^19,20 are present in both tumor categories (1), albeit at a lower frequency (2). Third, copy number alteration thyroid carcinomas include oncocytic thyroid carcinomas that most often have large-scale chromosomal losses and mitochondrial DNA mutations^21–26 and a propensity for slightly more aggressive behavior and reduced response to radio-iodine. Most of these well-differentiated thyroid carcinomas have the potential to progress to high-grade/poorly differentiated or anaplastic thyroid carcinomas through the acquisition of additional mutations in genes such as p53, hTERT, AKT1, PIK3CA, EIF1AX, and/or PTEN.²

Less common thyroid tumor categories that have other morphologic-molecular features and clinical tendencies also exist and are important to recognize. Examples in the most recent WHO Classification² include cribriform morular thyroid carcinoma that have APC or CTNNB1 mutations,^27–30 approximately half of which are associated with familial adenomatous polyposis and the high risk of other cancers; secretory thyroid carcinoma with ETV6-NTRK3 mutations^31,32 that are similar to some salivary gland and breast carcinomas and responsive to tyrosine kinase inhibitors; and thyroid hyalinizing trabecular tumors with GLIS alterations^33,34 that are almost always benign.

Another less common thyroid carcinoma group that has morphologic features resembling chromophobe renal cell carcinoma has been described recently. These chromophobe renal cell carcinoma-like thyroid carcinomas exhibit trabecular architecture, clear and eosinophilic cytoplasm, prominent cell membranes, irregular nuclei, binucleation and perinuclear halos^35–38 and are associated with either somatic³⁶ or germ-line³⁷ mutations in the TSC2 gene, the latter associated with tuberous sclerosis complex disease (TSCD).^39–41 We report here a unique and potentially related thyroid carcinoma that has somatic inactivating mutations in TSC2, both follicular and melanocytic differentiation and distinct morphologic features.

MATERIALS AND METHODS

Submission of this case report was approved by the Institutional Review Board of Endeavor Health System, Evanston, IL. Thyroid tumor tissue was dissected, fixed in 10% formalin neutral buffered saline and processed with automation (Tissue-Tec Autotech a120) for paraffin-embedding. In-house immunohistochemistry on paraffin-embedded tissue sections was performed on Ventana or Leica automated immuno-stainers using the following antibodies: pancytokeratins (Ventana, Ae1/Ae3), cytokeratin 7 (Ventana, clone SP532), thyroglobulin (Cell Marque, clone 2H11+6E1), thyroid transcription factor 1 (Ventana, clone 8G7G3/1). PAX8 (Ventana, clone MRQ-50), Melan A (Ventana, clone A-103; Leica, clone A-103), RCC (Cell Marque, clone PN-15), CD8 (Ventana, clone SP239), synaptophysin (Ventana, clone SP11), chromogranin (Ventana, clone LK2H10), calcitonin (Ventana, clone GI191E/A8), Ki67 (Ventana, clone 30.9), p53 (Ventana, clone DO-7), BRAF V600E (Ventana, clone VE1), CK20 (Ventana, clone SP33), CAIX (Ventana, clone EP161), CD117 (Ventana, clone EP10), P504S (Ventana, clone P504S), FH (Santa Cruz, clone J13), and 2SC (Ventana, clone AB). Outside immunohistochemistry at Mayo Clinic Laboratories (Rochester, MN) included Melan A (#MART-1), thyroglobulin (#THYR) and SDHB (#SDHB). RNA in situ hybridization was performed with probes for Trim 63 (ACD, 2.5 Ls). Molecular testing on paraffin-embedded tumor tissue was performed as follows: Tempus (Chicago, IL)—TO tumor of origin assay, xT next generation sequencing (NGS) panel, xR RNA fusion panel and transcriptome expression profiling (the latter research use only); Thyroseq (Sonic Health Care Thyroseq Laboratory, Rye Brook, NY)—NGS panel, RNA-seq fusion panel, copy number alteration and gene expression profiling; Mayo Clinic Laboratories (Rochester, MN)—TFE3 (#TFEF) and TFEB (#TFEBF) fluorescence-in-situ hybridization assays; and Invitae/Labcorp Genetics (San Francisco, CA)—blood-based germ line NGS cancer panel.

RESULTS

A 47-year-old woman presented for surgery with the history of hyperparathyroidism, renal stones and hydronephrosis. Recent laboratory assessments showed an elevated intact parathyroid hormone of 170 pg/mL (reference range 15 to 65 pg/mL), a normal TSH of 0.677 to 0.968 mU/L (reference range 0.300 to 5.600 mU/L) and a normal free T4 of 0.91 (reference range 0.61 to 1.80 ng/dL). Head and neck ultrasound demonstrated a right candidate parathyroid adenoma measuring 1.4×0.5×0.7 cm and a solid, homogeneous, hypoechoic thyroid nodule measuring 1.0×0.7×0.3 cm in the inferior right lobe/isthmus (ACR-TIRAD 4) with no evidence of loco-regional lymph node disease. Preoperative fine needle aspiration biopsy of the right thyroid nodule was interpreted as atypia of uncertain significance (Bethesda System Category III). The patient was counseled on clinical management options, including molecular testing for the thyroid nodule, and elected right hemithyroidectomy with parathyroidectomy.

Parathyroidectomy identified a single parathyroid adenoma (371 mg). Right hemithyroidectomy demonstrated a well-demarcated 0.6 cm tumor with a thin capsule and infiltrative invasion along one edge (Fig. 1A, arrow). There was no extrathyroidal extension, the margins of resection were negative and no regional lymph nodes were sampled. The tumor exhibited diffuse papillary (Fig. 1B) and solid-papillary (Fig. 1C) architecture, patchy clear and eosinophilic cytoplasm (Fig. 1, D-H) and irregular vesicular nuclei with chromatin margination, grooves, and rare pseudo-inclusions (Fig. 1E, arrow). There was frequent binucleation as well as distinctive macronuclei (Fig. 1F, asterisks) and multinuclear aggregates (Fig. 1F-H, arrows) in the tumor. A small lymphocytic tumor infiltrate was also present (Fig. 1D, arrows). One mitosis and no tumor necrosis were identified.

FIGURE 1.

Morphologic features of the thyroid tumor. This thyroid tumor was well-circumscribed with a thin capsule (A) and localized invasion into adjacent normal thyroid tissue along one edge (A, arrow). It exhibited an overall papillary (B) and solid-papillary (C) architecture, clear and eosinophilic tumor cytoplasm (D-F), intratumoral lymphocytes (D, arrows) and irregular vesicular nuclei (E-H) with chromatin margination, nuclear grooves and focal pseudo-inclusions (E, arrow). Binucleation, macronuclei (F, asterixes) and multinuclear aggregates (F-H, arrows) were also present.

The thyroid tumor cells expressed high CK7 (Fig. 2A), high PAX8 (Fig. 2B), very low thyroglobulin (Fig. 2C), low TTF1 (Fig. 2D) and no synaptophysin, chromogranin or calcitonin (images not shown) by in-house immunohistochemistry. The very low thyroglobulin was confirmed by immunohistochemistry at Mayo Clinic Laboratories (image not shown). The small intratumoral lymphocytes were predominantly CD8-positive T cells (Fig. 2E). The Ki-67 proliferation rate was low at 2.5%, BRAF V600E mutant protein was not expressed and p53 was expressed in a wild-type pattern (images not shown).

FIGURE 2.

Immunohistochemical features of the thyroid tumor. The thyroid tumor cells expressed high CK7 (A), high PAX8 (B), very low thyroglobulin (C), and low TTF-1 (D). The intratumoral lymphocytes were CD8-positive T cells (E). The tumor cells also expressed Melan A (F) and RCC (G) compared with normal thyroid tissue. Localized Trim 63 expression was detected in a characteristic dot-like pattern by RNA in-situ hybridization (H). Thyroid tumor and adjacent normal thyroid tissue are present in each panel.

These unusual morphologic features, very low thyroglobulin, low TTF-1 and high PAX-8 led us to at first strongly consider metastatic renal cell carcinoma in the differential diagnoses because it has a proclivity to spread in small deposits to other organs, including the thyroid gland. Relative to normal thyroid tissue, the tumor cells expressed moderate to high levels of RCC (Fig. 2G), a renal cell carcinoma marker with some thyroid cross-reactivity,⁴² and Melan A (Fig. 2F), a specific melanocytic marker that is involved in melanosome biogenesis and maturation^43–45 and often expressed in TFE3-mutated/TFEB-altered renal cell carcinomas.^46–50 The positive Melan A was confirmed by immunohistochemistry at Mayo Clinic Laboratories (image not shown). Trim 63, an additional downstream marker of TFE3-mutated/TFEB-altered renal cell carcinomas,⁵¹ was also expressed in a geographic area of the tumor in characteristic dot-like deposits by RNA in situ hybridization (Fig. 2H). Additional immunohistochemistry on the tumor included CK20 (negative), CAIX (negative), CD117 (patchy positive), P504S (negative), fumarate hydratase (wild-type positive), 2SC (negative), and succinate dehydrogenase B (wild-type positive) (images not shown). Importantly, follow-up high-resolution CT and MRI imaging did not identify a primary renal mass, and we therefore initiated a comprehensive molecular-genetic work-up of the tumor.

DNA fluorescence in situ hybridization performed at Mayo Clinic Laboratories did not identify rearrangements or amplifications of the TFE3 or TFEB genes.^51,52 Thyroseq and Tempus RNA-based tumor testing, including tumor of origin assessment, confirmed with 87% and 99% probability, respectively, that this carcinoma was of thyroid origin based on expression of thyroid follicular markers including the sodium-iodide symporter (NIS). Tempus transcriptome profiling identified expression of thyroglobulin and the melanocytic markers MiTF, Melan A and HMB45/PMEL (highest to very low). Next generation sequencing (NGS) identified 2 somatic mutations in the TSC2 gene, one a deletion (c.2234del; VAF 11.7%) resulting in a frameshift and premature stop codon in the tuberin domain (p.K745R*fs25) and the other a missense mutation (c.3352C>T; VAF 9.2%) resulting in a stop codon gain in the mid-protein domain (p.Q1118*) (Fig. 3, red circles). Both TSC2 mutations are pathogenic and predict biallelic truncation of TSC2 with loss of the carboxyl-terminal Rap-Gap domain, leading to induction of Rheb, mTORC1, and constitutive mTOR signaling. Transcriptome profiling confirmed expression of Rheb, MTOR, ATF4, HIF1α, PGC1α; and eIF4E, all characteristic of mTOR signaling.⁵³ A single variant of unknown significance (c.1039G>A; p.A347T) was identified in the AJUBA gene that encodes a LIM-domain containing protein involved in cell adhesion, differentiation, and proliferation.^54–56 The Tempus and Thyroseq assays assessed more than 648 and 12,000 respective genes/ variants by NGS and 100 and 150 respective fusions by RNA-seq. Copy number alterations and gene expression profiles were also evaluated. Blood-based NGS performed at InVitae/LabCorp did not identify germ-line mutations and there was no clinical evidence of TSCD that is caused by germ-line mutations in TSC2 or TSC1. These overall findings point to somatic mutation of TSC2 as the main oncogenic driver in this thyroid carcinoma.

FIGURE 3.

Schematic depiction of reported mutations and functional domains in TSC2 in thyroid carcinomas. The current papillary TSC2 thyroid carcinoma had 2 somatic truncating mutations in TSC2—one in the tuberin domain and one in the mid-protein domain (red circles). Two previously published chromophobe renal cell carcinoma-like thyroid carcinomas had 2 somatic truncating mutations in the tuberin domain (green circles) or one somatic truncating mutation in the Rap-Gap domain (light blue circle). Another previously published chromophobe renal cell carcinoma-like thyroid carcinoma in a patient with TSCD had a germ-line truncating mutation in the Rap-Gap domain (purple circle). All these truncating TSC2 mutations lead to an incomplete Rap-Gap domain and are predicted to inactivate TSC2 and induce constitutive mTOR signaling. Lastly, one anaplastic thyroid carcinoma was reported to have a somatic truncating mutation in the DUF3384 domain (black circle) and 5 other anaplastic thyroid carcinomas were reported to have somatic missense variants of uncertain significance (yellow circles) in the DUF3384, mid-carboxyl, or Rap-Gap domains of TSC2.

DISCUSSION

We have identified a unique thyroid carcinoma that has both follicular and melanocytic differentiation, somatic inactivating mutations in TSC2 and distinct morphologic features (termed papillary TSC2 thyroid carcinoma for subsequent discussion). Melanocytic differentiation has been observed rarely in medullary thyroid carcinomas,^57–59 but to our knowledge, this is the first report of melanocytic differentiation in a thyroid follicular tumor. This melanocytic differentiation was characterized by immunohistochemical expression of Melan A and transcriptome expression of Melan A and MiTF. Interestingly, melanocytic differentiation is also a major diagnostic feature in PEComa, angiomyolipoma and lymphangioleiomyoma tumors that develop frequently in TSCD patients who have germ-line mutations in TSC2 or less frequently TSC1, with subsequent somatic deletion of the remaining respective TSC2 or TSC1 allele in the tumor. Resulting TSC2 or TSC1 inactivation leads to constitutive mTOR signaling that is believed to induce aberrant melanocytic differentiation, including expression of melanocytic markers and formation of subcellular premelanosome-like structures.^40,41 Thus, somatic inactivation of TSC2 and persistent mTOR signaling provide a plausible pathophysiologic mechanism for the induction of melanocytic differentiation in this papillary TSC2 thyroid carcinoma. Constitutive mTOR signaling may act through eIF4E to upregulate the translation of Melan A,⁶⁰ increase the activity of TFE3 and TFEB^61–64 and/or promote expression of Trim 63,^65,66 although these may be cell context-dependent.⁶⁷ Recent studies have also suggested that the immunophenotypes of TFE3-mutated/TFEB-altered and TSC2-/TSC1-/MTOR-mutated renal cell carcinomas are highly overlapping and that tumors in both groups exhibit increased Melan A, TFE3/TFEB, Trim 63, Cathepsin-K and GPNMB expression/activity.^{61–64,66,68} Related mechanisms could be active to induce Melan A and Trim 63 in this papillary TSC2 thyroid carcinoma, although the papillary TSC2 thyroid carcinoma did not demonstrate expression of Cathepsin-K or membranous GPNMB in our immunohistochemical evaluation (unpublished). Importantly, constitutive mTOR signaling also raises the possibility of clinical use of FDA-approved mTOR inhibitors in patients who have TSC2-mutated thyroid carcinomas, particularly if they become refractory to radio-iodine therapy or are recurrent, as has been shown for some patients with advanced renal cell carcinoma and the benign neoplastic proliferations in TSCD.^69,70

This papillary TSC2 thyroid carcinoma shares some morphologic, immunotypic, and genetic features with the 7 chromophobe renal cell carcinoma-like thyroid carcinomas that have been reported so far,^35–38 including clear and eosinophilic cytoplasm, irregular nuclei, binucleation, high PAX8, very low thyroglobulin, and TSC2 mutations. However, the papillary TSC2 thyroid carcinoma also exhibits some unique features such as papillary architecture, macronuclei, multinuclear aggregates, CD8-positive T lymphocytes, and melanocytic differentiation that may be more specific for it. In our view, all these features should lead to TSC2 mutation testing. In a manner similar to the papillary TSC2 thyroid carcinoma, 2 of the chromophobe renal cell carcinoma-like thyroid carcinomas^36,37 had somatic mutations in TSC2—p.R691Pfs*12 and p.Q785* in the tuberin domain in one tumor (Fig. 3, green circles) and p.Y1650Cfs*4 in the Rap-Gap domain in the other tumor (Fig. 3, light blue circle). Three of the remaining 5 chromophobe renal cell carcinoma-like thyroid carcinomas^35,37,38 arose in patients with TSCD, one with a germ-line TSC2 mutation in the Rap-Gap domain³⁵ (Fig. 3, purple circle), reinforcing an association of the chromophobe renal cell carcinoma-like thyroid carcinoma phenotype with either somatic or germ-line TSC2 mutations. The 2 remaining chromophobe renal cell carcinoma-like thyroid carcinoma patients did not have features of TSCD or apparent TSC2 mutation testing.^37,38 All these TSC2 mutations were truncating, leading to complete or partial loss of the Rap-Gap domain, thereby implicating the GTPase-activating protein activity of TSC2 in its oncogenic function in these thyroid carcinomas. Interestingly, a small number of anaplastic thyroid carcinomas, which are undifferentiated thyroid cancers that tend to have high proliferation rate, necrosis and aggressive behavior, have also been reported⁷¹ with a somatic truncating mutation in the amino-terminal DUF3384 domain (Fig. 3, black circle) or with somatic missense variants of uncertain significance in the DUF3384, midcarboxyl or Rap-Gap domains (Fig. 3, yellow circles) of TSC2. We hypothesize that the mutation type (germ-line, somatic, missense, nonsense, biallelic, truncating, deletion, etc.) and altered functional domains in TSC2 determine, at least in part, the phenotypes observed in TSC2-mutated thyroid carcinomas. Other mutations and/or epigenetic factors may also be involved and we cannot exclude completely that the single missense variant of uncertain significance in the AJUBA gene^54,56 has a functional role in the current papillary TSC2 thyroid carcinoma.

The spectrum of morphologic features present in TSC2-mutated thyroid carcinomas makes their pathologic diagnosis difficult, in particular because of overlap with thyroid papillary carcinoma. For example, 5 of the reported chromophobe renal cell carcinoma-like thyroid carcinomas were diagnosed initially as either the oncocytic variant of papillary thyroid carcinoma (2 cases), the clear cell variant of papillary thyroid carcinoma, widely invasive follicular thyroid carcinoma or poorly differentiated thyroid carcinoma.^35–37 Another was diagnosed initially as metastatic chromophobe renal cell carcinoma.³⁸ We also diagnosed initially the current papillary TSC2 thyroid carcinoma as most consistent with metastatic renal cell carcinoma (favor TFE3-mutated /TFEB-altered subtype) before it became clear by high-resolution imaging that no primary renal mass was present. Thus, only one of 8 reported thyroid carcinomas that had TSC2 mutations or a TSC2 mutant phenotype was diagnosed initially as a TSC2-mutated thyroid cancer. Another complicating factor in the diagnosis of TSC2-mutated thyroid carcinomas, particularly in preoperative fine needle aspiration thyroid biopsies with limited tumor cells, is the known technical difficulty of sequencing the large TSC2 gene that has many small exons, high GC content, repetitive sequences, complex structural variations, and retrotransposon insertions.^72,73 In fact, TSC2 mutations in the current papillary TSC2 thyroid carcinoma were not detected in one of 2 established commercial NGS assays despite the availability of the entire carcinoma in postsurgical tissue. We suggest that TSC2-mutated thyroid carcinoma should be considered in the differential diagnosis of any thyroid tumor that has unusual morphologic and immunohistochemical features or lacks known driver mutations in thyroid cancer.

Although their exact clinical potential remains to be verified, the TSC2-mutated thyroid carcinomas identified so far exhibit clinical patterns similar to other thyroid cancers. Local thyroid invasion, high expression of NIS, low mitotic rate, low Ki-67 proliferation index and lack of necrosis suggests that the current papillary TSC2 thyroid carcinoma is a well-differentiated thyroid cancer. Although angioinvasion was not present in this tumor, 2 of the reported chromophobe renal cell carcinoma-like thyroid carcinomas with somatic TSC2 mutations did have vascular invasion and one had extrathyroidal extension and regional lymph node metastasis.³⁶ Two other chromophobe renal cell carcinoma-like thyroid carcinomas in patients with TSCD had lymph node involvement and one had multiple loco-regional recurrence,^35,37 underscoring their metastatic potential. A role for TSC2 mutations in more advanced high-grade/poorly differentiated or anaplastic thyroid carcinomas requires further investigation.

The main limitation of our report is that it involves a single patient with an apparently unique thyroid cancer. As such, we intended our interpretations to be conservative but also wanted to highlight morphologic and immunohistochemical patterns in the tumor that have not yet been reported in other TSC2-mutated thyroid carcinomas. We anticipate that increased awareness of TSC2-mutated thyroid carcinomas and broader, more accurate sequencing of the TSC2 gene will generate further details on these interesting thyroid cancers.