Thyroid carcinoma metastases to central nervous system and vertebrae

Introduction

Metastases from thyroid carcinoma impacting the central nervous system (CNS) are rare, with a 1% incidence of brain metastases for differentiated thyroid carcinoma types and 20% incidence of bone metastases [9]. Not all CNS metastases require biopsy/surgical excision for diagnosis or treatment and thus, specimens for molecular studies are difficult to accrue in large numbers. The most likely metastases to come to surgery are single lesions in patients without a known primary tumor or vertebral bony/epidural masses impacting spinal cord function, necessitating emergency decompression.
Genetic features of thyroid carcinoma at its primary site have been studied, but considerably less is known about molecular features of metastatic lesions, particularly those to specific anatomical regions, including those impacting the CNS. In the primary tumor, the most prevalent reported genetic alterations are recognized to differ by histological subtype [30]. Specifically, papillary carcinoma of the thyroid has been shown to have mutations in BRAF (40-45% of cases), RAS (10-20%), RET/PTC (10-20%), and TRK (< 5%) [30]. Follicular carcinoma demonstrates mutations in RAS (40-50%), PAX8/PPARg (30-35%), PIK3CA (< 10%), and PTEN (< 10%) [30]. As poorly differentiated and anaplastic thyroid carcinomas most often arise from pre-existing differentiated carcinoma types, they often retain the mutations common to differentiated carcinomas, but gain additional mutations in TP53, CTNNB1 and AKT1. Poorly differentiated thyroid carcinomas more frequently have mutations in RAS (20-50%), TP53 (10-35%), BRAF (5-15%), CTNNB1 (0-5%) and PIK3CA (0-15%) [43]. Anaplastic thyroid carcinoma shows TP53 (40-80%), CTNNB1 (0-5%), RAS (10-50%), BRAF (10-50%), PIK3CA (5-25%) and PTEN (10-15%) mutations [44]. Sporadic medullary carcinoma harbors mutations in RET (40-50%) and RAS (25%) [30]. As noted in one recent review, “multiple genes are implicated in the development of thyroid neoplasms, both benign and malignant …in genes which are responsible for cell proliferation, survival and differentiation through different pathways. Approximately 90% of mutations are mutually exclusive activating mutations in oncogenes RAS (~13%) and BRAF (~60%), …whilst the remaining 10% are loss-of-function mutations affecting tumour suppressor genes such as PTEN, PPARg and TP53” [38]. A second recent review emphasizes that “the most frequent driver events can be either point mutations or gene rearrangements, mainly affecting the MAPK pathway and phosphatidylinositol-3 kinase (PI3K)/AKT pathway” [24].
Reactivation of telomerase reverse transcriptase (TERT) occurs in several types of cancer, including thyroid cancers, especially in 2 hotspots in the promotor region c.124C>T and c.146C>T [35]. In a large study of > 200 various histological types of thyroid carcinoma, TERT promotor mutations involving these 2 hotspots were detected in 7% of papillary, 18% of follicular, 25% of Hurthle cell, and 86% of poorly differentiated/anaplastic carcinomas at the primary site [35].
We hypothesized that if targeted therapies become available in the future that molecular assessment of the metastasis itself, rather than the primary tumor, might be optimal for patient management and that metastases might be enriched for mutations seen in a higher percentage of aggressive thyroid carcinoma types, such as TERT promotor mutations. In addition, we especially were interested in TERT promotor mutations, given the fact that their presence is definitional for the highest grade in some tumor types, such as meningiomas and astrocytic lineage tumors that are wildtype for isocitrate dehydrogenase (IDH) [7,39]. Therefore, we interrogated our databases to identify thyroid metastases resected at our institution over the past 15 years from either spinal column or CNS parenchyma. We recorded information regarding the date of first resection of the primary thyroid tumor, histological type of the primary tumor, and assessed the metastases for histological type, following which we tested the metastases for mutations and fusions. We then compared the frequency and types of mutations we identified with well-published data from the literature.

Material and methods

Computer-based database queries of our Surgical Pathology Department files were utilized to identify cases via text word search, 1.01.2006 to 9.08.2021, inclusive, linking the text words “thyroid” and “metastatic” with “brain” or “central nervous system”. Once identified, slides were retrieved from the files and all cases were re-examined microscopically by the expert endocrine pathologist on the paper (CBM) to confirm diagnosis and classify the type of thyroid carcinoma, as most had been originally diagnosed only as “metastatic thyroid carcinoma” without further classification. Metastases were categorized as to papillary, follicular, papillary carcinoma-follicular variant, poorly differentiated, anaplastic, Hurthle cell, and medullary carcinoma subtypes.
After obtaining internal review board approval (COMIRB #21-3458), clinical history was extracted from the medical record, including anatomical location of the metastasis/metastases involving the CNS parenchyma or vertebral bodies, date of diagnosis and histological type of the primary thyroid carcinoma if known, and date of demise or survival as of 9.08.2021. The survival interval from first diagnosis of thyroid carcinoma to development of first metastasis involving the CNS parenchyma or vertebral bodies was calculated.
Unfortunately, given the retrospective nature of the study, the referral nature of patients to our tertiary care center, and the tendency for patients to receive care at multiple different hospitals, the slides or paraffin tissue blocks from the original primary thyroid carcinoma were not available for nearly all our cases. Thus, we were unable to directly compare genetic features between paired primary thyroid and metastatic samples.
The next generation sequencing (NGS) gene fusion and gene mutation assays were validated and performed for 12 cases in the Colorado Molecular Correlates Laboratory in the Department of Pathology at the University of Colorado – Anschutz Medical Campus. Total nucleic acid (TNA) was extracted from formalin fixed paraffin embedded (FFPE) processed material via the Agencourt FormaPure Kit (Beckman Coulter, Brea, CA). TNA was processed for gene fusion analysis via the Archer FusionPlex Solid Tumor library preparation kit and for gene mutation analysis by a customized version of the Archer VariantPlex Solid Tumor library preparation kit (ArcherDx, Boulder, CO). The resulting libraries were sequenced on either the Illumina MiSeq or Illumina NextSeq instruments (Illumina, San Diego, CA). Raw sequence data were analyzed using the ArcherDx Analysis software package (version 4.1.1.7 for fusions, version 5.1.2 for mutations, ArcherDx). Bioinformatically identified fusions and mutations were manually inspected by highly trained personnel. Methods have been previously published [3,13,37,56]. Mutation analysis for three cases was ordered by the oncology team and performed prior to this study at outside national reference laboratories.

Results

Our database search identified 21 resection specimens from 16 patients, with 2 patients each having had 3 surgical resections of spinal bone lesions (cases 2, 3, 4 from a female who was 51 years of age at first resection; cases 6, 7, 8 from a female who was 63 years of age at first resection) (see Table I). Surgical excisions in these 2 women had been undertaken over a 6-year time period for the first and over a 1-year time interval for the second. One man initially presented with a pituitary mass thought preoperatively to be a non-secretory pituitary macroadenoma (case 15), but proven to be thyroid carcinoma after histological examination. He then required surgical decompression of an L1 bone mass the following year (case 16); this was the sole patient with thyroid metastases from both bone and intraparenchymal compartments. All bone lesions were to the spine; none to the skull or other bony sites were identified in our database.
Gender distribution was equal (8 males, 8 females), ages 25-71 years; excluding the youngest patient (age 25) with medullary carcinoma of the thyroid who had MEN2B syndrome, there was a median age of 62 years (Table I). Metastases were of all types, although the most frequent type was follicular thyroid carcinoma (FTC) (n = 9 patients, plus one with FTC-Hurthle cell features, case 11). There was a single case of medullary carcinoma, two cases of classical papillary thyroid carcinoma, and one case of follicular variant of papillary thyroid carcinoma. Four cases were poorly differentiated thyroid carcinoma (PDTC) and three were anaplastic thyroid carcinoma (ATC). In both patients with multiple bone resection specimens, the histologic type changed; from FTC to PDTC (cases 2, 3, 4) and from FTC to ATC (cases 6, 7, 8) (Table I). Three patients had a different histological type for their primary tumor recorded in the medical record than was found in the metastasis (cases 10, 17, 20). However, since the original slides and blocks were not available from the primary tumor, this possible change in type could be neither confirmed nor refuted.
Spinal bone lesions involved all levels in similar numbers, i.e. cervical, thoracic, and lumbar. Intraparenchymal sites constituted 6/21 lesions, with 1 pituitary lesion, 1 in intraventricular/choroid plexus, and 1 each located in frontal, temporal, parieto-occipital and occipital lobes.
Most patients were known to have thyroid carcinoma antecedent to developing their CNS lesion, although in 2 persons, the surgically resected bone metastasis represented their first diagnosis. Intervals for the remaining 14 patients from the time of primary thyroid carcinoma diagnosis to the time of surgical resection of the CNS metastasis ranged from 6 months to 41 years (Table I).
Intervals to demise for those known to have died from disease ranged from < 1 month to over 6 years. A significant number (5/16) are known to be alive with disease at the closure of the study (9.08.2021), with the longest survival interval of 4 years, 10 months, including 2 of the 7 patients with parenchymal brain metastases (both patients are receiving targeted therapy as their tumors have RET mutations). Thus, although the case numbers in our cohort are too limited for a meaningful epidemiological study, we did observe that diagnosis of CNS metastasis did not imply imminent demise (Table I).
For the mutation/fusion testing performed for this study, only cases with FFPE material that had not been subjected to decalcification could be used, as exposure to the acid solution degrades the genetic material. This excluded from analysis 4 vertebral body/bone cases where all material had been decalcified (cases 2, 3, 6, 9). One parenchymal case could not be tested as there was no tumor material remaining in the block (case 16). In total, mutational analysis was completed in 10 of 15 bone metastases and 5 of 6 parenchymal metastases. Fusion analysis was attempted in 13 specimens, with 3 cases (all bone specimens) found to be “uninformative’ for fusion results because the RNA quality needed for fusion testing was determined to be too poor to trust a negative result. The metric that guided this decision was based on a specific sequencing metric utilized in our laboratory. Two cases were found to have a fusion and eight were negative for fusions.
In terms of informative mutation results, the most frequently identified mutations were in TERT promotor, with 80% of all metastases having this mutation (n = 12, only one case with sole TERT mutation, case 11). In our study, of the 10 bony vertebral metastases that could be assessed for the mutation (i.e., not decalcified), 9 of 10 had TERT promotor mutation and of the 5 assessable CNS parenchymal metastases, 3 showed TERT promotor mutation (Table I). The next most frequent were NRAS with 66.6% having this mutation (n = 10), and TP53 with 20% having this mutation (n = 3); fewer ATM, AKT1, PTEN, NOTCH1 mutations were also identified. The latter 4 mutations were all found in concert with other mutations known to be more frequent at the primary site for thyroid carcinoma.
Specifically, papillary thyroid carcinoma (PTC) has been shown to have mutations in BRAF (40-45% of cases), RAS (10-20%), RET/PTC (10-20%), and TRK (< 5%) [30]. In our study, we had only a single papillary carcinoma of the thyroid (case 5, vertebral body metastasis) that unfortunately could not be assessed for mutation. Of the parenchymal metastases, 1 was a follicular variant-PTC but did not show BRAF mutation (case 18). Thus, we did not identify BRAF mutation in the 2 PTCs in our cohort.
Follicular thyroid carcinoma (FTC) demonstrates mutations in RAS (40-50%), PAX8/PPARg (30-35%), PIK3CA (< 10%), and PTEN (< 10%) [30]. We had 9 vertebral body metastases with a FTC component, of which 6 could be assessed; 6/6 had a RAS mutation (cases 7, 8, 12, 13, 14, 15), 2/6 had a PTEN mutation (cases 7, 8), and 1/6 had a PIK3CA mutation (case 15) (Table I). Of the parenchymal metastases, 0 of the 5 assessable cases were FTCs (Table I). Mutations in TP53 were only detected in PDTC or ATC cases.
Two of our patients each had undergone multiple (n = 3) surgical resections of spinal bone lesions (cases 2, 3, 4 from a female who was 51 years of age at first resection; cases 6, 7, 8 from a female who was 63 years of age at first resection) (Table I). Of these 2 patients, only 1 of 3 bony lesions could be assessed in one patient (i.e., cases 2, 3 not informative; case 4 with NRAS, TERT mutation) and 2 of 3 assessable bony lesions from the other patient shared identical mutations in NRAS, PTEN, TERT (cases 7, 8).
The sole medullary carcinoma in the cohort had PDGFRA and RET mutations in the parenchymal brain metastasis (case 21). 5/5 parenchymal cases had more than 1 mutation identified, and 9/10 bone cases had more than 1 mutation identified. Only 2 cases had fusions detected: one NCOA4-RET fusion (mutational testing not performed) and one CCDC6-RET fusion in a tumor that also contained mutations in TERT and ATM (Table I). Thus, only a single metastasis (case 11) had a solo/isolated genetic aberration.

Discussion

Our group has a longstanding interest in metastases impacting the CNS [12,21,22,32,40,41,46,53,57] and in this study, we turned our attention to investigating thyroid metastases involving the CNS parenchyma and vertebral bodies. The most interesting finding in our study was the extremely high percentage of TERT promotor mutations in our metastases, at a rate higher than has been reported for primary tumors in several studies. Specifically, our study identified TERT promotor mutations in 80% overall of thyroid metastases involving the CNS parenchyma and vertebral bodies, suggesting that TERT promotor mutation is enriched in metastasis, regardless of the histological type of the metastasis. In our study, of the 10 bony vertebral metastases that could be assessed for the mutation (i.e., not decalcified), 9 of 10 had TERT promotor mutation and of the 5 assessable CNS parenchymal metastases, 3 showed TERT promotor mutation (Table I).
Previously published studies examining molecular/genetic changes in thyroid carcinoma have focused primarily on the more common diagnosis of PTC [27,49]. Masoodi et al. reported that mutations in the driver genes were preserved between primary PTC and metastases (including bone and brain metastases), while variants of genes involved in DNA methylation and transcriptional repression signaling were restricted to metastases [27]. Osborne et al. undertook molecular assessment on 20 metastases, a number very comparable to our cohort of 21 metastases. However, only 3 of these had been brain metastases; most metastases were from lung or lymph node sites [33]. Overall, 16/20 (80%) showed TERT promotor mutation, 55% were BRAF V600E mutated, 10/20 (50%) had concurrent TERT promotor and BRAF V600E mutation, 4 (20%) had TP53 mutation, 2 had both TERT and TP53, and one had NRAS mutation [27]. Of note, the larger percentage of BRAF mutations in their study compared to ours stems from a larger percentage of PTC (54%) in their study, a histological thyroid carcinoma type known to show a higher percentage of BRAF mutation [30]. Our study only contained a single case of the follicular variant of PTC.
Pozdeyev et al. studied 779 specimens from advanced differentiated (i.e., with distant metastases) and anaplastic thyroid cancers, including metastatic specimens at lung (n = 67), bone (n = 33) and brain (n = 14) sites and found a high percentage with TERT mutations, although results were not detailed as to presence of TERT mutation at a specific metastatic site, such as brain [36]. Song et al. found that while differentiated thyroid carcinoma at the primary site had TERT promoter mutations in 4.5% of all cases, the mutation was enriched in cases where the patient had distant metastases (24%) or died from disease (20%) [47]. Thus, our results are concordant with those from prior limited studies in the literature and our original hypothesis was confirmed.
An acknowledged limitation of our study was that we did not have the paired primary thyroid carcinoma and CNS metastases from the same patient in our cohort. However, it cannot be emphasized strongly enough that paired samples were also limited in the study by Osborne et al. [33]. Indeed, in their study, only 3 of their 20 patients had both the primary thyroid carcinoma and a metastasis for testing and comparison [33]. This reflects the fact that many patients in North America receive care at several different medical facilities, making it significantly less likely that both the primary tumor and the metastasis had been surgically resected at the same hospital. In addition, many thyroid carcinoma patients in our study had had a long time interval between their diagnosis of primary thyroid carcinoma (Table I) and their CNS metastasis, and often the medical record did not have information regarding where the first tumor had been operated, further hampering retrieval of tissue blocks from the primary thyroid tumor for mutational/fusion testing. It is also worth emphasizing that even if the primary tumor had been available for mutation testing, intratumoral heterogeneity mutations occurring in a small subclone of tumor cells might well have made testing inaccurate.
In terms of types of other mutations found in our metastases, all mutations in metastases identified in the study of Osborne et al. were also found in our cohort, except for BRAF V600E. In the study by Nikiforov and Nikiforova, mutations typical of FTCs in primary thyroid carcinomas, including PTEN, NRAS, and PIK3CA, were identified in the FTC metastases in our cohort [30]. The sole medullary carcinoma in our cohort had PDGFRA and RET mutations in the parenchymal brain metastasis (case 21), but not TERT promoter mutation. This is in keeping with prior studies that have not identified TERT promoter mutations in MTC [52]. Thus, as might be expected, driver mutations were preserved, based on the histological type. In our study, even if the metastases were not classifiable as “poor differentiated/anaplastic”, they still had TERT promotor mutation in an overall high percentage of cases (80%). The percentage of TERT promotor mutations detected at the primary site does differ based on this histological subtype, with one of the largest studies showing TERT in 7% of papillary, 18% of follicular, 25% of Hurthle cell, and 86% of poorly differentiated/anaplastic carcinomas [38]. Vinagre et al. report TERT mutations in 3/14 poorly differentiated thyroid (21%), 9/169 follicular (14%), 2/16 anaplastic (13%), 13/169 papillary (8%) and 0/28 medullary carcinomas [52]. Tanaka et al. reported that TERT promoter mutations were associated with transformation from papillary to anaplastic types [49], although in our study anaplastic transformation was not essential to TERT promotor mutation identification.
The possibility that CNS metastases may manifest evolving or acquired features compared to the primary tumor has precedent in many studies, including several by our group. Specifically, we recently investigated prostatic adenocarcinoma CNS parenchymal and dural metastases for ERG, CHD1, and MAP3K7 expression and reported that immunohistochemical markers previously shown to be downregulated in aggressive prostatic carcinomas at their primary site also showed reduced expression in prostatic metastases impacting the CNS [32]. In addition, our group as long ago as 2004 had an interest in assessing possible differences between primary tumors and brain metastases from breast carcinoma [16]. Discordance in epidermal growth factor receptor (EGFR) status in 18% of brain metastases implied that drug therapies should be individualized for patients based on test results of the metastases, not simply extrapolated from the primary tumor [16]. Recent reports comparing primary tumors and brain metastases in breast carcinoma have also shown discordance for various parameters [50].
Groups investigating this topic note that there are “unique biological features of each metastatic site” and emphasize “the need to biopsy metastatic disease in patients with advanced breast cancer” since “understanding the biology of each metastatic site can potentially impact the design of new therapies and ultimately improve patient outcomes” [6]. This same message is appropriate for CNS thyroid metastases.
While clinical features were not our main focus for this study, we did review demographics, tumor location, patient age, and survival prior to conducting molecular testing (Table I). Our cohort included 21 specimens from 16 patients, several of which were located in uncommon, but well-reported, intraparenchymal locations including pituitary [2,4,5,8,10,11,14,23,26,28,29,31,43,48,55,58] and intraventricular/choroid plexus [17,20,25,42,51,53,59]. Bone involvement occurred at cervical, thoracic, and lumbar levels and 2 patients required multiple surgical resections of bone lesions due to symptoms (Table I). None were from the skull.
The numbers in our cohort compare favorably to several prior clinical studies. Slutzky-Shraga et al. reported 10 patients with brain metastases from thyroid carcinoma [45], Hong et al. reported 25 patients with thyroid metastases to CNS and nearby bony sites [19], |and Choi et al. detailed clinical features in 37 patients [9]. Several of these studies differ in several aspects from ours in that they either excluded medullary thyroid cancer [45] or included skull metastases [19]. All were from differing patient populations/geographical areas than ours, namely Israel [45], Taiwan [19,51], and Korea [9] and all contained a slightly different proportion of thyroid carcinoma types. Nevertheless, these 3 studies showed a nearly identical median age for patients with thyroid metastases to CNS to ours of 62 years, with 53.5 years, 63 years, and 63 years, respectively [9,19,45]. Only one included information regarding interval from initial diagnosis of the primary thyroid carcinoma to development of the CNS metastasis, i.e., Slutzky-Shraga et al., who recorded an interval range of 9 months-17 years 3 months [45]. In comparison, 2 of our patients had their first diagnosis of thyroid carcinoma at the time of bone metastasis and one of our patients had a 41-year interval from initial diagnosis of thyroid carcinoma to CNS-impacting metastasis.
In conclusion, we show that that TERT mutation occurs in a higher percentage of metastases (80%) than primary tumors. This study indicates that testing of the metastasis itself will be necessary, should targeted anti-TERT mutation therapies become available in the future. The study also adds to the literature new molecular information on CNS metastases, an area of research focus that is relatively “underserved” compared to that for primary brain tumors. Of note, TERT promotor mutations are also frequent in several types of solid tumors, including hepatocellular carcinoma, in which TERT promotor mutations have been assessed in plasma as a biomarker for prognosis [1,18,34]. In addition, in melanoma, TERT-mutated patients had a significantly worse overall survival than those with wild-type status [15]. Thus, there are already precedents in other solid tumor types for using TERT promotor mutation assessment in prognosis.

Acknowledgements

The authors thank Ms. Jennifer Platte for manuscript preparation, Ms. Lisa Litzenberger for photographic expertise, Dr. Kurt Davies for advice regarding molecular testing, and Dr. Bryan Haugen for review of the paper and critical comments. This work was supported by the Pathology Shared Resource and the University of Colorado Cancer Center Support Grant (P30CA046934).

Disclosure

The authors report no conflict of interest.

References