Comprehensive Genetic Analysis of Follicular Thyroid Carcinoma Predicts Prognosis Independent of Histology

Comprehensive Genetic Analysis of Follicular Thyroid Carcinoma Predicts Prognosis Independent of... Abstract Context Follicular thyroid carcinoma (FTC) is classified into minimally invasive (miFTC), encapsulated angioinvasive (eaFTC), and widely invasive (wiFTC) subtypes, according to the 2017 World Health Organization guidelines. The genetic signatures of these subtypes may be crucial for diagnosis, prognosis, and treatment but have not been described. Objective Identify and describe the genetic underpinnings of subtypes of FTC. Methods Thirty-nine tumors, comprising 12 miFTCs, 17 eaFTCs, and 10 wiFTCs, were whole-exome sequenced and analyzed. Somatic mutations, constitutional sequence variants, somatic copy number alterations, and mutational signatures were described. Clinicopathologic parameters and mutational profiles were assessed for associations with patient outcomes. Results Total mutation burden was consistent across FTC subtypes, with a median of 10 (range 1 to 44) nonsynonymous somatic mutations per tumor. Overall, 20.5% of specimens had a mutation in the RAS subfamily (HRAS, KRAS, or NRAS), with no notable difference between subtypes. Mutations in TSHR, DICER1, EIF1AX, KDM5C, NF1, PTEN, and TP53 were also noted to be recurrent across the cohort. Clonality analysis demonstrated more subclones in wiFTC. Survival analysis demonstrated worse disease-specific survival in the eaFTC and wiFTC cohorts, with no recurrences or deaths for patients with miFTC. Mutation burden was associated with worse prognosis, independent of histopathological classification. Conclusions Though the number and variety of somatic variants are similar in the different histopathological subtypes of FTC in our study, mutational burden was an independent predictor of mortality and recurrence. Follicular thyroid carcinoma (FTC) is a well-differentiated endocrine malignancy that accounts for 10% of all thyroid cancers (1–3). FTC is ∼2.5 times more common in women, similar to papillary thyroid carcinoma (PTC) (2, 4). The 5-year survival of FTC is ∼88%, but drops to 78% at 10 years (5). FTCs have traditionally been classified as minimally invasive (miFTC) or widely invasive FTC (wiFTC) based on the presence of vascular and/or capsular infiltration (6). In 2017, the World Health Organization (WHO) unveiled new guidelines that include an intermediate histopathologic type: encapsulated angioinvasive FTC (eaFTC) (7). Although determination of invasive status is critical for prognostication, even the new classification relies on subjective pathological evaluation of the degree of invasion. The earlier system, despite its limitations, has been demonstrated to predict considerable differences in tumor recurrence, metastatic potential, and mortality, though the 2017 criteria have not yet been similarly validated (8–11). Although the genomic landscape and driver events in PTC have been well described (12), the molecular etiology of FTC is less well known, particularly wiFTC and eaFTC. Recurrent somatic mutations in the RAS family (NRAS, KRAS, and HRAS) have been reported in FTC, usually in the 61st codon (13–15). Most common is NRAS, mutated in 15% to 40% of FTCs (16, 17). RAS family mutations hold clinical significance, as they have been shown to increase metastatic potential and disease-specific mortality (18). The PAX8-PPARγ fusion gene is identified in about one-third of FTCs, with estimates ranging from 12% to 56% (19, 20). Although PAX8-PPARγ likely contributes to follicular tumorigenesis, it does not appear to impact prognosis (20). The increased availability and accuracy of next-generation sequencing technology has allowed recent advances in understanding the mutational landscape of FTC and the major differences between FTC and follicular thyroid adenoma (FTA) (13–15). However, nearly all prior sequencing studies in FTC have been performed on miFTC, with limited applicability to the rarer but much deadlier wiFTC. Moreover, there have been few data characterizing the molecular underpinnings of invasion or the genetic distinctions between the different categories of FTC, particularly since the recent introduction of eaFTC in the WHO 2017 guidelines. The lack of genetic markers associated with each category limits the use of the histopathological classification of follicular tumors to surgical specimens rather than fine-needle aspiration (FNA) biopsies. Our study uses next-generation sequencing techniques and bioinformatics tools to investigate the genomic landscape of FTC, with particular attention to the recently described WHO 2017 histopathological categories of invasiveness. We investigate whether the divergent behavior of these tumors in clinical practice is associated with distinct molecular profiles. Materials and Methods Patient cohort and sample acquisition The 39 patients recruited for this study received surgical treatment at Yale New Haven Hospital (n = 24; FTC600 series) or Karolinska University Hospital (n = 15; FTC1 series) between 2002 and 2013 (Supplemental Table 1). All samples were independently reviewed by a minimum of two experienced endocrine pathologists for histopathological confirmation, and poorly differentiated thyroid cancer was ruled out by the Turin criteria, in accordance with the 2017 recommendations from the WHO. Matched normal samples for each tumor were obtained from adjacent histologically normal thyroid or blood leukocyte DNA. The diagnosis and degree of invasion were confirmed according to the 2017 guidelines established by the WHO (7). Although the 2017 WHO guidelines distinguish Hürthle cell (oxyphilic) carcinomas from FTCs, both Hürthle cell and conventional FTCs were included in this study, as both exhibit invasive behavior. Informed consent was obtained from all patients involved in this study. The acquisition and use of protected health information and tissue specimens were performed as specified by the Health Insurance Portability and Accountability Act (Yale) or Swedish Act on Biobanks (Karolinska). The study was approved by the Yale University and Karolinska Institutet Institutional Review Boards. Whole-exome sequencing Genomic DNA was extracted from formalin-fixed paraffin-embedded (FFPE) or fresh-frozen tissue. Three 1-mm–thick tissue cores were obtained per block for FFPE samples, with paraffin enzymatically removed and genomic DNA prepared using a proprietary in-house method at the Yale Center for Genome Analysis. Before and after coring, FFPE blocks were sectioned, stained, and analyzed via light microscopy to ensure tissue identity. For fresh-frozen samples, after tissue disruption via sonication, genomic DNA was isolated using the DNeasy Blood & Tissue Kit (Qiagen) in accordance with the manufacturer’s instructions. The quality and quantity of isolated DNA were then measured via spectrophotometry (NanoDrop Technologies). Tumor and normal genomic DNA pairs were then subjected to exome capture and sequencing as previously described (21). Adaptors of known sequence were ligated to genomic DNA fragments that were amplified by ligation-mediated PCR. Specimens were then subjected to capture using the NimbleGen 2.1M human exome array. Exome-specific DNA was eluted and then underwent 74-base paired-end sequencing on the HiSeq 2000 instrument per the specifications of the manufacturer (Illumina). Removal of PCR duplicates was performed with Picard Tools (http://broadinstitute.github.io/picard). Though 42 tumor-normal pairs originally generated libraries for whole-exome sequencing (WES), 39 of these passed quality metrics and were included in subsequent analyses. The Burrow-Wheeler Aligner-MEM program was used to map reads to the human reference genome GRCh37/hg19 (22). Somatic and germline single nucleotide variants and short insertions or deletions were called by MuTect2 using the Bayesian classifier followed by in-house filtering scripts to increase variant calling specificity (23). Likely false-positives were excluded using the D-ToxoG filter, and variant allele frequencies <0.10 were manually censored for variant calling (24, 25). Tumor purity was calculated with a nested approach, taking into account B-allele frequency in loss-of-heterozygosity regions, followed by tumor driver allele frequency and average overall variant allele frequency (26). Known variants in annotated databases [1000 Genomes (27, 28); and NHLBI ESP6500 (Exome Variant Server, NHLBI GO Exome Sequencing Project, Seattle, WA; http://evs.gs.washington.edu/EVS/) and 2577 noncancer exomes sequenced at Yale] were excluded, and novel exonic variants were evaluated for impact on transcriptional and/or translational processing. Copy-number variation detection EXCAVATOR was used to examine somatic copy number variation (CNV) in tumors (29). Briefly, EXCAVATOR normalizes the nonuniform WES read depth taking guanine/cytosine content into account and thereby calculates the ratios of normalized read depth between tumor and normal for the exome capture intervals and performs segmentation of copy number (CN) data using a novel heterogeneous hidden Markov model algorithm. Somatic CN gain and loss were defined relative to the matched normal samples. The significance of individual CNV peaks was evaluated using GISTIC (30). Fusion gene detection in FTC Fresh-frozen tumor samples (n = 13) were assayed for the presence of three chromosomal translocations common to thyroid carcinoma: PAX8-PPARγ, RET-PTC1, or RET-PTC3. The remaining samples from the WES cohort were excluded due to low RNA yield or quality or were in FFPE and not amenable to reliable RNA gene fusion detection. After isolating RNA using an RNeasy Plus Mini Kit (Qiagen), spectrophotometry was used to determine the concentration and purity of RNA (NanoDrop Technologies). We then assayed the samples for the defined chromosomal translocations with the Thyroid Cancer Fusion Gene Detection Kit (THRNA-RT32; EntroGen, Inc.). In this protocol, cDNA synthesis and gene amplification are performed simultaneously. The products are exposed to dual-labeled hydrolysis probes designed to detect the amplification products. Fusion gene amplification was detected via a FAM-BHQ–labeled probe in a CFX96 Real-Time System thermocycler (Bio-Rad). Additionally, exome sequencing data were investigated for evidence of somatic gene fusions using the DELLY2 and Manta packages and filtered for recurrent or previously described fusion events (31, 32). Mutational signatures, clonality, and pathway analyses Mutational signatures, as described by Alexandrov et al. (33), were also analyzed. Signatures of the individual samples and across each type of tumor were constructed through the compilation and quantification of somatic variant patterns using the deconstructSigs algorithm (34). SciClone was used to detect clonal populations of tumor cells in each sample (35). In brief, this algorithm clusters mutations by variant allele frequency to predict subclonal populations within the tumor, focusing on CN neutral regions. For clonality analysis, mutations at a more generous variant allele frequency cutoff of 0.05 were included. Additionally, using the network and pathway analysis platforms from STRING (https://string-db.org), GeNets (https://apps.broadinstitute.org/genets#), and Kyoto Encyclopedia of Genes and Genomes (KEGG; http://www.genome.jp/kegg/mapper.html), all called somatic alterations were evaluated to determine if genetic events were enriched in specific biologic pathways. Networks were mapped with any disconnected nodes excluded for visual clarity. The STRING database maps connections between proteins based on experimental data as well as text mining of major biomedical databases (36). GeNets analysis uses a meta-network of several other protein interaction tools and databases to map connections between genes of interest, whereas KEGG lists all pathways enriched in each set of genes, ranked by number of genes altered per pathway (37). Statistical analysis Clinicopathologic parameters including patient age, sex, tumor size (largest diameter), and American Joint Committee on Cancer (AJCC) stage (7th and 8th editions) (38) and genetic parameters including total count of somatic variants, clonality, and number of cancer-associated driver events were assessed for association with invasive status. All variables were assumed to be nonparametrically distributed due to small sample size. Continuous variables were analyzed using the Kruskal-Wallis test and/or two-tailed Mann-Whitney test, as appropriate, with Dunn correction for multiple testing. Fisher exact test with Freeman-Halton extension was used for categorical variables. Disease-specific survival (i.e., time to death due to FTC) was illustrated with Kaplan-Meier curves, and Cox proportional hazards regression was used for univariate and multivariate survival analysis. All statistical analyses were performed using Prism 7 (GraphPad Software), except for three-category Fisher exact tests, which were performed using VassarStats (vassarstats.net) and Cox proportional hazards analyses, which were performed in R (39). Results Patient and tumor characteristics Twenty-six of the 39 patients included in the exome analysis were female (67%), and 13 were male (33%). The mean age of the population was 55.2 years old (median 54; range 14 to 86 years). The overall median tumor diameter was 3.6 cm, and the majority of patients had stage I or II disease (92%; 36 out of 39) based on AJCC 8th edition. Following surgery, patients were clinically monitored for disease progression or recurrence for a median duration of 5.8 years. The overall recurrence rate and disease-specific mortality rate were both 15% (6 out of 39), as all patients in our cohort who had disease recurrence ultimately died of FTC; these figures are concordant with previously reported trends (5). All patients had surgery, and most had radioactive iodine treatment, whereas only one had cytotoxic or targeted chemotherapy (Supplemental Table 1). On average, miFTCs were smaller and found in younger patients, though these differences were not statistically significant. Relative to miFTC and eaFTC, wiFTC had a higher AJCC stage at time of surgery (P = 0.013) and were more likely to recur or be the cause of death [P = 0.0017; Table 1; Fig. 1(a)]. Table 1. Clinicopathologic Features of Patients With miFTC, eaFTC, and wiFTC Variable All (N = 39) miFTC (n = 12) eaFTC (n = 17) wiFTC (n = 10) P Value a Age at diagnosis, y, mean ± SD 55.2 ± 16.1 48.1 ± 15.6 58.0 ± 15.9 58.9 ± 15.7 0.24 Female sex, n/total n (%) 26/39 (67) 8/12 (67) 12/17 (71) 6/10 (60) 0.91 Tumor diameter, cm, mean ± SD 3.6 ± 1.5 3.0 ± 0.9 3.9 ± 2.0 3.8 ± 0.9 0.29 AJCC stage greater than II, n/total n (%) 3/39 (8) 0/12 (0) 0/17 (0) 3/10 (30) 0.013 (Pmi-ea > 0.99; Pmi-wi = 0.078; Pea-wi = 0.041) Disease recurrence, n/total n (%) 6/39 (15) 0/12 (0) 1/17 (6) 5/9 (56) 0.0017 (Pmi-ea > 0.99; Pmi-wi = 0.0062; Pea-wi = 0.0097) Death due to FTC, n/total n (%) 6/39 (15) 0/12 (0) 1/17 (6) 5/9 (56) 0.0017 (Pmi-ea > 0.99; Pmi-wi = 0.0062; Pea-wi = 0.0097) Disease-specific survival, 5-year, n/total n (%) 22/24 (92) 7/7 (100) 9/9 (100) 7/9 (78) 0.31 Disease-specific survival, 10-year, n/total n (%) 8/14 (57) 4/4 (100) 3/4 (75) 1/6 (17) 0.052 Variable All (N = 39) miFTC (n = 12) eaFTC (n = 17) wiFTC (n = 10) P Value a Age at diagnosis, y, mean ± SD 55.2 ± 16.1 48.1 ± 15.6 58.0 ± 15.9 58.9 ± 15.7 0.24 Female sex, n/total n (%) 26/39 (67) 8/12 (67) 12/17 (71) 6/10 (60) 0.91 Tumor diameter, cm, mean ± SD 3.6 ± 1.5 3.0 ± 0.9 3.9 ± 2.0 3.8 ± 0.9 0.29 AJCC stage greater than II, n/total n (%) 3/39 (8) 0/12 (0) 0/17 (0) 3/10 (30) 0.013 (Pmi-ea > 0.99; Pmi-wi = 0.078; Pea-wi = 0.041) Disease recurrence, n/total n (%) 6/39 (15) 0/12 (0) 1/17 (6) 5/9 (56) 0.0017 (Pmi-ea > 0.99; Pmi-wi = 0.0062; Pea-wi = 0.0097) Death due to FTC, n/total n (%) 6/39 (15) 0/12 (0) 1/17 (6) 5/9 (56) 0.0017 (Pmi-ea > 0.99; Pmi-wi = 0.0062; Pea-wi = 0.0097) Disease-specific survival, 5-year, n/total n (%) 22/24 (92) 7/7 (100) 9/9 (100) 7/9 (78) 0.31 Disease-specific survival, 10-year, n/total n (%) 8/14 (57) 4/4 (100) 3/4 (75) 1/6 (17) 0.052 P values <0.05 shown in bold. a Kruskal-Wallis followed by corrected Dunn test for continuous variables and Fisher exact test (with Freeman-Halton extension) for categorical variables. P values in parentheses are pairwise comparisons between miFTC, eaFTC, and wiFTC when significance was found in the overall three-way analysis. View Large Table 1. Clinicopathologic Features of Patients With miFTC, eaFTC, and wiFTC Variable All (N = 39) miFTC (n = 12) eaFTC (n = 17) wiFTC (n = 10) P Value a Age at diagnosis, y, mean ± SD 55.2 ± 16.1 48.1 ± 15.6 58.0 ± 15.9 58.9 ± 15.7 0.24 Female sex, n/total n (%) 26/39 (67) 8/12 (67) 12/17 (71) 6/10 (60) 0.91 Tumor diameter, cm, mean ± SD 3.6 ± 1.5 3.0 ± 0.9 3.9 ± 2.0 3.8 ± 0.9 0.29 AJCC stage greater than II, n/total n (%) 3/39 (8) 0/12 (0) 0/17 (0) 3/10 (30) 0.013 (Pmi-ea > 0.99; Pmi-wi = 0.078; Pea-wi = 0.041) Disease recurrence, n/total n (%) 6/39 (15) 0/12 (0) 1/17 (6) 5/9 (56) 0.0017 (Pmi-ea > 0.99; Pmi-wi = 0.0062; Pea-wi = 0.0097) Death due to FTC, n/total n (%) 6/39 (15) 0/12 (0) 1/17 (6) 5/9 (56) 0.0017 (Pmi-ea > 0.99; Pmi-wi = 0.0062; Pea-wi = 0.0097) Disease-specific survival, 5-year, n/total n (%) 22/24 (92) 7/7 (100) 9/9 (100) 7/9 (78) 0.31 Disease-specific survival, 10-year, n/total n (%) 8/14 (57) 4/4 (100) 3/4 (75) 1/6 (17) 0.052 Variable All (N = 39) miFTC (n = 12) eaFTC (n = 17) wiFTC (n = 10) P Value a Age at diagnosis, y, mean ± SD 55.2 ± 16.1 48.1 ± 15.6 58.0 ± 15.9 58.9 ± 15.7 0.24 Female sex, n/total n (%) 26/39 (67) 8/12 (67) 12/17 (71) 6/10 (60) 0.91 Tumor diameter, cm, mean ± SD 3.6 ± 1.5 3.0 ± 0.9 3.9 ± 2.0 3.8 ± 0.9 0.29 AJCC stage greater than II, n/total n (%) 3/39 (8) 0/12 (0) 0/17 (0) 3/10 (30) 0.013 (Pmi-ea > 0.99; Pmi-wi = 0.078; Pea-wi = 0.041) Disease recurrence, n/total n (%) 6/39 (15) 0/12 (0) 1/17 (6) 5/9 (56) 0.0017 (Pmi-ea > 0.99; Pmi-wi = 0.0062; Pea-wi = 0.0097) Death due to FTC, n/total n (%) 6/39 (15) 0/12 (0) 1/17 (6) 5/9 (56) 0.0017 (Pmi-ea > 0.99; Pmi-wi = 0.0062; Pea-wi = 0.0097) Disease-specific survival, 5-year, n/total n (%) 22/24 (92) 7/7 (100) 9/9 (100) 7/9 (78) 0.31 Disease-specific survival, 10-year, n/total n (%) 8/14 (57) 4/4 (100) 3/4 (75) 1/6 (17) 0.052 P values <0.05 shown in bold. a Kruskal-Wallis followed by corrected Dunn test for continuous variables and Fisher exact test (with Freeman-Halton extension) for categorical variables. P values in parentheses are pairwise comparisons between miFTC, eaFTC, and wiFTC when significance was found in the overall three-way analysis. View Large Figure 1. View largeDownload slide Clinicopathological and mutational characteristics of FTC. (a) Clinical and pathological parameters of FTCs, grouped by histologic invasion subtype. Disease-specific survival is shown; cases lost to follow-up or who died for another reason were censored. (b) Somatic mutation profiles of recurrent and driver mutations as generated by WES. The top row contains the total number of nonsynonymous variants per tumor, including single nucleotide variants and short insertion/deletion events. Each subsequent row contains the mutations in a specific gene. Bars to the right illustrate prevalence in our cohort. Driver genes were selected based on Vogelstein et al. (41) and supplemented with purported FTC drivers from the three most recent sequencing studies in FTC: BRAF, BRIP1, CNOT1, DICER1, EIF1AX, EZH1, HRAS, IDH1, IGF2BP3, KDM5C, KMT2C, KRAS–MAP4K3, NF1, NRAS, PTEN, SOS1, SPOP, STAG2, TCF12, TP53, and TSHR (13–15). Recurrent mutations were noted in at least two samples, whereas private mutations were identified in only one sample in our cohort. (c) Recurrent arm-level CN events. Cases unable to have CNV analysis performed due to low DNA quality are marked with an “X.” F, female; M, male. Figure 1. View largeDownload slide Clinicopathological and mutational characteristics of FTC. (a) Clinical and pathological parameters of FTCs, grouped by histologic invasion subtype. Disease-specific survival is shown; cases lost to follow-up or who died for another reason were censored. (b) Somatic mutation profiles of recurrent and driver mutations as generated by WES. The top row contains the total number of nonsynonymous variants per tumor, including single nucleotide variants and short insertion/deletion events. Each subsequent row contains the mutations in a specific gene. Bars to the right illustrate prevalence in our cohort. Driver genes were selected based on Vogelstein et al. (41) and supplemented with purported FTC drivers from the three most recent sequencing studies in FTC: BRAF, BRIP1, CNOT1, DICER1, EIF1AX, EZH1, HRAS, IDH1, IGF2BP3, KDM5C, KMT2C, KRAS–MAP4K3, NF1, NRAS, PTEN, SOS1, SPOP, STAG2, TCF12, TP53, and TSHR (13–15). Recurrent mutations were noted in at least two samples, whereas private mutations were identified in only one sample in our cohort. (c) Recurrent arm-level CN events. Cases unable to have CNV analysis performed due to low DNA quality are marked with an “X.” F, female; M, male. WES Thirty-nine FTC samples and their respective matched normal samples passed quality metrics after WES. Mean depth of coverage for tumor and normal samples was 258.6 and 115.4, respectively. The proportion of bases with a minimum of 20 times coverage was 96.9% and 93.1% for tumor and normal samples, respectively, with associated error rates of 0.3% and 0.3%. Mean tumor purity was calculated at 70% and not associated with the number of called somatic mutations (Supplemental Fig. 1). Overall, there was no considerable difference in total or cancer-specific somatic mutational burden across histologic subtypes, though there was a trend toward more mutations in eaFTC and wiFTC (Table 2). Mutations in RAS family genes were noted in 20.5% of samples (8 out of 39), all in codon 61. NRAS mutations were found in 5 out of 39 samples, whereas KRAS and HRAS variants were found in 2 and 1 specimens, respectively, with no difference across histologic categories [Fig. 1(b)]. TSHR mutations were identified in four tumors. Among known cancer- or FTC-specific driver genes, DICER1, EIF1AX, KDM5C, NF1, PRDM1, PTEN, and TP53 were recurrently mutated in two samples each. Although these were all too rare to be distributed in a statistically significant manner, TP53 mutations were only noted in wiFTC (P = 0.06). There was a trend toward more driver gene mutations in the wiFTC cohort. An additional set of recurrently mutated genes in our cohort have not been previously described as drivers, and though two of these (CAMTA1 and SFPQ) are in the Catalogue of Somatic Mutations in Cancer (COSMIC) cancer gene census, the specific mutations identified in our study are not listed in COSMIC (http://cancer.sanger.ac.uk/cosmic). Table 2. Genetic Features of miFTC, eaFTC, and wiFTC Variable All (n = 39) miFTC (n = 12) eaFTC (n = 17) wiFTC (n = 10) P Value a Nonsynonymous somatic variants, mean ± SD 12.5 ± 8.9 8.2 ± 5.3 15.4 ± 10.7 12.8 ± 7.6 0.08 Transition/transversion ratio per sample, mean ± SD 1.3 ± 1.2 1.4 ± 2.0 1.1 ± 0.7 1.5 ± 1.0 0.42 RAS family mutation, n/total n (%) 8/39 (21) 1/12 (8) 5/17 (29) 2/10 (20) 0.37 Cancer driver gene nonsynonymous mutations, mean ± SD 0.8 ± 1.0 0.5 ± 0.5 0.8 ± 0.8 1.2 ± 1.5 0.62 Number of clones, mean ± SD 1.9 ± 0.7 1.8 ± 0.6 1.8 ± 0.6 2.4 ± 0.7 <0.05 (Pmi-ea > 0.99; Pmi-wi = 0.15; Pea-wi = 0.05) Variable All (n = 39) miFTC (n = 12) eaFTC (n = 17) wiFTC (n = 10) P Value a Nonsynonymous somatic variants, mean ± SD 12.5 ± 8.9 8.2 ± 5.3 15.4 ± 10.7 12.8 ± 7.6 0.08 Transition/transversion ratio per sample, mean ± SD 1.3 ± 1.2 1.4 ± 2.0 1.1 ± 0.7 1.5 ± 1.0 0.42 RAS family mutation, n/total n (%) 8/39 (21) 1/12 (8) 5/17 (29) 2/10 (20) 0.37 Cancer driver gene nonsynonymous mutations, mean ± SD 0.8 ± 1.0 0.5 ± 0.5 0.8 ± 0.8 1.2 ± 1.5 0.62 Number of clones, mean ± SD 1.9 ± 0.7 1.8 ± 0.6 1.8 ± 0.6 2.4 ± 0.7 <0.05 (Pmi-ea > 0.99; Pmi-wi = 0.15; Pea-wi = 0.05) P values <0.05 shown in bold. a Kruskal-Wallis test followed by corrected Dunn test for continuous variables or Fisher exact test (with Freeman-Halton extension) for categorical variables. P values in parentheses are pairwise comparisons between miFTC, eaFTC, and wiFTC when significance was found in the overall three-way analysis. View Large Table 2. Genetic Features of miFTC, eaFTC, and wiFTC Variable All (n = 39) miFTC (n = 12) eaFTC (n = 17) wiFTC (n = 10) P Value a Nonsynonymous somatic variants, mean ± SD 12.5 ± 8.9 8.2 ± 5.3 15.4 ± 10.7 12.8 ± 7.6 0.08 Transition/transversion ratio per sample, mean ± SD 1.3 ± 1.2 1.4 ± 2.0 1.1 ± 0.7 1.5 ± 1.0 0.42 RAS family mutation, n/total n (%) 8/39 (21) 1/12 (8) 5/17 (29) 2/10 (20) 0.37 Cancer driver gene nonsynonymous mutations, mean ± SD 0.8 ± 1.0 0.5 ± 0.5 0.8 ± 0.8 1.2 ± 1.5 0.62 Number of clones, mean ± SD 1.9 ± 0.7 1.8 ± 0.6 1.8 ± 0.6 2.4 ± 0.7 <0.05 (Pmi-ea > 0.99; Pmi-wi = 0.15; Pea-wi = 0.05) Variable All (n = 39) miFTC (n = 12) eaFTC (n = 17) wiFTC (n = 10) P Value a Nonsynonymous somatic variants, mean ± SD 12.5 ± 8.9 8.2 ± 5.3 15.4 ± 10.7 12.8 ± 7.6 0.08 Transition/transversion ratio per sample, mean ± SD 1.3 ± 1.2 1.4 ± 2.0 1.1 ± 0.7 1.5 ± 1.0 0.42 RAS family mutation, n/total n (%) 8/39 (21) 1/12 (8) 5/17 (29) 2/10 (20) 0.37 Cancer driver gene nonsynonymous mutations, mean ± SD 0.8 ± 1.0 0.5 ± 0.5 0.8 ± 0.8 1.2 ± 1.5 0.62 Number of clones, mean ± SD 1.9 ± 0.7 1.8 ± 0.6 1.8 ± 0.6 2.4 ± 0.7 <0.05 (Pmi-ea > 0.99; Pmi-wi = 0.15; Pea-wi = 0.05) P values <0.05 shown in bold. a Kruskal-Wallis test followed by corrected Dunn test for continuous variables or Fisher exact test (with Freeman-Halton extension) for categorical variables. P values in parentheses are pairwise comparisons between miFTC, eaFTC, and wiFTC when significance was found in the overall three-way analysis. View Large The range of somatic mutational burden across the cohort was 1 to 44 nonsynonymous variants per tumor (mean 12.5; median 10), with no marked difference in mutational signature between subtypes (Fig. 2; Supplemental Table 2). Although 55 constitutional variants were noted in potential cancer-associated genes, none of these was in BROCA-curated thyroid cancer susceptibility genes (tests.labmed.washington.edu/BROCA) (Supplemental Table 3). There was a higher number of subclones per tumor in the wiFTC category (P = 0.05), though the number of somatic variants in our cohort is low, limiting the reliability of the clonality analysis (Supplemental Fig. 2). Figure 2. View largeDownload slide Somatic mutational burden and signature of FTC. (a) Burden of somatic variants per tumor, displayed according to histological subtype. (b) Base-change spectra of individual tumors. (c) Mutational signatures [as originally described in Alexandrov et al. (33) and curated now by COSMIC] in each subtype of FTC. Signatures were generated for each tumor individually using deconstructSigs and summed to generate the distribution shown for each subtype. “Unknown” represents base changes that could not be mapped to a known mutational signature. Figure 2. View largeDownload slide Somatic mutational burden and signature of FTC. (a) Burden of somatic variants per tumor, displayed according to histological subtype. (b) Base-change spectra of individual tumors. (c) Mutational signatures [as originally described in Alexandrov et al. (33) and curated now by COSMIC] in each subtype of FTC. Signatures were generated for each tumor individually using deconstructSigs and summed to generate the distribution shown for each subtype. “Unknown” represents base changes that could not be mapped to a known mutational signature. CN analysis Overall, FTCs were defined in our study by a general pattern of CN gain, with differences at some loci between the different tumor subtypes. Arm-level CN events were identified across the genome [Fig. 1(c); Supplemental Fig. 3A and 3B]. Although the most common arm-level CN events were gains of 5q, 7p, and 12q, recurrent CN losses at 22q were also identified. The regions with the most substantial CN gains were 7q22.3-36.3 and 12q24.33, whereas 1p36.11 demonstrated CN loss (Supplemental Fig. 3C–3E). The 7q22.3-36.3 region contains 306 genes, including the oncogene PRKAR2B (40). Interestingly, the eaFTC cohort carried recurrent CN gains in region 5q31.3-q35.3, which were not seen in the other tumor types; this region carries the genes CSF1R and NPM1, both of which are implicated as cancer drivers in other studies (41). Survival analysis Invasive status and AJCC stage were significantly associated with disease-specific survival, whereas, surprisingly, tumor size had no statistically noteworthy effect on survival in our cohort [Fig. 3(a)–3(d)]. Higher-than-mean total nonsynonymous mutation burden was associated with worse survival, as was burden of cancer driver gene mutations (41) [Fig. 3(e) and 3(f)], but not mutational burden of COSMIC census genes or clones per tumor [Fig. 3(g) and 3(h)]. Univariate Cox proportional hazards analysis demonstrated that patient age, histologic subtype, AJCC stage, total nonsynonymous mutational burden (as a continuous variable), and number of driver events were all associated with disease-specific survival. Although the number of events in our cohort was too low to support multivariate analysis of all of these factors at once, total mutational burden was found to predict mortality in a limited multivariate analysis that controlled for stage and histologic subtype (Table 3). Figure 3. View largeDownload slide Kaplan-Meier survival curves. Disease-specific survival is plotted against time to event, out to 10 years of follow-up. Tick marks along survival curves indicate cases censored at that time point. Cases were stratified on the basis of: (a) invasion (n = 12 miFTC, 17 eaFTC, and 10 wiFTC); (b) tumor size (n = 30 tumors ≤4 cm and 9 tumors >4 cm); AJCC stage: (c) 7th edition (n = 25 stage I or II, 10 stage III, and 4 stage IV) or (d) 8th edition (n = 36 stage I or II and 3 stage IV); (e) total nonsynonymous mutational burden (n = 24 with <13 variants and 15 with ≥13 variants); (f) cancer driver gene mutations as proposed by Vogelstein et al. (41) (n = 19 with 0 drivers, 17 with 1 driver, and 3 with 2 or more drivers); (g) COSMIC census gene mutations (n = 9 with 0 genes mutated, 11 with 1 gene mutated, 10 with 2 genes mutated, and 9 with 3 or more genes mutated); and (h) subclones per tumor as determined by SciClone (35) (n = 8 with 1 clone, 25 with 2 clones, and 5 with 3 or more clones). Figure 3. View largeDownload slide Kaplan-Meier survival curves. Disease-specific survival is plotted against time to event, out to 10 years of follow-up. Tick marks along survival curves indicate cases censored at that time point. Cases were stratified on the basis of: (a) invasion (n = 12 miFTC, 17 eaFTC, and 10 wiFTC); (b) tumor size (n = 30 tumors ≤4 cm and 9 tumors >4 cm); AJCC stage: (c) 7th edition (n = 25 stage I or II, 10 stage III, and 4 stage IV) or (d) 8th edition (n = 36 stage I or II and 3 stage IV); (e) total nonsynonymous mutational burden (n = 24 with <13 variants and 15 with ≥13 variants); (f) cancer driver gene mutations as proposed by Vogelstein et al. (41) (n = 19 with 0 drivers, 17 with 1 driver, and 3 with 2 or more drivers); (g) COSMIC census gene mutations (n = 9 with 0 genes mutated, 11 with 1 gene mutated, 10 with 2 genes mutated, and 9 with 3 or more genes mutated); and (h) subclones per tumor as determined by SciClone (35) (n = 8 with 1 clone, 25 with 2 clones, and 5 with 3 or more clones). Table 3. Predictors of Disease-Specific Mortality Variable Hazard Ratio 95% CI P Value a Univariate analysis  Age 1.1 1.0–1.2 0.043  Sex, male 1.3 0.24–7.2 0.76  Size, cm 1.7 0.74–3.9 0.21  Subtypeb 8.2 1.1–58.6 0.036  AJCC 8th edition stageb 2.3 1.3–4.0 0.0081  Total nonsynonymous variants 1.1 1.0–1.2 0.013  Clones 1.8 0.67–4.7 0.28  Cancer driver gene mutationsc 2.8 1.4–5.4 0.0018  FTC driver gene mutationsd 3.3 1.2–9.3 0.018  RAS gene mutation 0.53 0.06–4.5 0.54  CN loss of arm 22q 0.97 0.09–11 0.98 Multivariate analysis  Subtypeb 174 0.66–45,546 0.07  AJCC 8th edition stageb 1.8 0.64–5.0 0.27  Total nonsynonymous variants 1.4 1.06–2.0 0.02 Variable Hazard Ratio 95% CI P Value a Univariate analysis  Age 1.1 1.0–1.2 0.043  Sex, male 1.3 0.24–7.2 0.76  Size, cm 1.7 0.74–3.9 0.21  Subtypeb 8.2 1.1–58.6 0.036  AJCC 8th edition stageb 2.3 1.3–4.0 0.0081  Total nonsynonymous variants 1.1 1.0–1.2 0.013  Clones 1.8 0.67–4.7 0.28  Cancer driver gene mutationsc 2.8 1.4–5.4 0.0018  FTC driver gene mutationsd 3.3 1.2–9.3 0.018  RAS gene mutation 0.53 0.06–4.5 0.54  CN loss of arm 22q 0.97 0.09–11 0.98 Multivariate analysis  Subtypeb 174 0.66–45,546 0.07  AJCC 8th edition stageb 1.8 0.64–5.0 0.27  Total nonsynonymous variants 1.4 1.06–2.0 0.02 P values <0.05 shown in bold. a Likelihood ratio test for univariate analysis or Wald test for multivariate analysis. b Per each increment of subtype (miFTC, eaFTC, or wiFTC) or stage (I–IV). c Cancer driver gene mutations as curated by Vogelstein et al. (41). d Drawn from recent exome sequencing studies in FTC: BRAF, BRIP1, CNOT1, DICER1, EIF1AX, EZH1, HRAS, IDH1, IGF2BP3, KDM5C, KMT2C, KRAS, MAP4K3, NF1, NRAS, PTEN, SOS1, SPOP, STAG2, TCF12, TP53, and TSHR (13–15). View Large Table 3. Predictors of Disease-Specific Mortality Variable Hazard Ratio 95% CI P Value a Univariate analysis  Age 1.1 1.0–1.2 0.043  Sex, male 1.3 0.24–7.2 0.76  Size, cm 1.7 0.74–3.9 0.21  Subtypeb 8.2 1.1–58.6 0.036  AJCC 8th edition stageb 2.3 1.3–4.0 0.0081  Total nonsynonymous variants 1.1 1.0–1.2 0.013  Clones 1.8 0.67–4.7 0.28  Cancer driver gene mutationsc 2.8 1.4–5.4 0.0018  FTC driver gene mutationsd 3.3 1.2–9.3 0.018  RAS gene mutation 0.53 0.06–4.5 0.54  CN loss of arm 22q 0.97 0.09–11 0.98 Multivariate analysis  Subtypeb 174 0.66–45,546 0.07  AJCC 8th edition stageb 1.8 0.64–5.0 0.27  Total nonsynonymous variants 1.4 1.06–2.0 0.02 Variable Hazard Ratio 95% CI P Value a Univariate analysis  Age 1.1 1.0–1.2 0.043  Sex, male 1.3 0.24–7.2 0.76  Size, cm 1.7 0.74–3.9 0.21  Subtypeb 8.2 1.1–58.6 0.036  AJCC 8th edition stageb 2.3 1.3–4.0 0.0081  Total nonsynonymous variants 1.1 1.0–1.2 0.013  Clones 1.8 0.67–4.7 0.28  Cancer driver gene mutationsc 2.8 1.4–5.4 0.0018  FTC driver gene mutationsd 3.3 1.2–9.3 0.018  RAS gene mutation 0.53 0.06–4.5 0.54  CN loss of arm 22q 0.97 0.09–11 0.98 Multivariate analysis  Subtypeb 174 0.66–45,546 0.07  AJCC 8th edition stageb 1.8 0.64–5.0 0.27  Total nonsynonymous variants 1.4 1.06–2.0 0.02 P values <0.05 shown in bold. a Likelihood ratio test for univariate analysis or Wald test for multivariate analysis. b Per each increment of subtype (miFTC, eaFTC, or wiFTC) or stage (I–IV). c Cancer driver gene mutations as curated by Vogelstein et al. (41). d Drawn from recent exome sequencing studies in FTC: BRAF, BRIP1, CNOT1, DICER1, EIF1AX, EZH1, HRAS, IDH1, IGF2BP3, KDM5C, KMT2C, KRAS, MAP4K3, NF1, NRAS, PTEN, SOS1, SPOP, STAG2, TCF12, TP53, and TSHR (13–15). View Large Fusion gene analysis Of the available samples tested directly for fusion genes, only one tumor was found to have the PAX8-PPARγ fusion gene (7.7%; 1 out of 13). This widely invasive, stage II tumor was successfully resected, and the patient had 10 years of recurrence-free survival prior to the conclusion of this study. This patient had no other known cancer drivers identified in our study. The observed rate of PAX8-PPARγ fusion in this study is considerably lower than previously reported (20), most likely due to the smaller cohort sizes in each FTC category. As expected for FTC, no RET-PTC1 or RET-PTC3 fusion events were identified in our cohort. The algorithmic approach to fusion gene identification from exome sequencing data did not reveal any previously described FTC fusions or any recurrent fusion events across our cohort. Pathway analyses KEGG pathway analysis demonstrated enrichment of mutations in the MAPK signaling pathway, as expected, and several generic malignancy-associated pathways, but these were not specific to particular categories of invasiveness (Supplemental Table 4). The wiFTC cohort did carry more mutations in several pathways known to be associated with more aggressive malignancies, including cAMP signaling, mitophagy, and longevity regulation. Similarly, protein-protein interaction analysis demonstrated networks of mutations in Ras signaling, cell cycle regulation, and RNA processing, which were not specific to any of the FTC subtypes (Supplemental Fig. 4). Interestingly, each subtype of tumor often carried mutations in different genes within similar networks, hinting at a final common pathway of FTC tumorigenesis. Discussion Follicular thyroid tumors present a unique dilemma for clinicians and patients. Minimally invasive FTCs carry an excellent prognosis with low risk of recurrence or disease-specific mortality, whereas widely invasive tumors are much more aggressive even with maximally intensive treatment (8–10). The new category of eaFTC provides additional granularity in histologic diagnosis, but its effect on prognosis remains unclear. Additionally, the histologic classification requires a surgical specimen in which the capsule can be examined thoroughly, which limits the utility of FNA as a diagnostic tool. Unfortunately, even with the advent of newer molecular markers from thyroid FNA specimens, many patients must still undergo partial or total thyroidectomy to arrive at a definitive diagnosis when faced with a follicular thyroid neoplasm. Although wiFTCs have a distinctly aggressive clinical phenotype, the molecular evolution leading from a normal thyroid follicular cell to wiFTC is incompletely understood. wiFTC is relatively rare, and most next-generation sequencing studies in FTC have analyzed few widely invasive tumors. Furthermore, it is not clear based on currently available evidence whether follicular tumors progress from FTA to wiFTC in a stepwise fashion through miFTC or eaFTC intermediates or whether these subtypes represent distinct entities arising from disparate genetic backgrounds. To determine whether genetic distinctions determine the different histologic categories of FTC as defined in the 2017 WHO guidelines, our group preformed WES analysis of 39 FTC specimens across a spectrum of miFTC, eaFTC, and wiFTC cases. Our cohort is characterized by relatively modest mutational burden [mean 12.5 nonsynonymous variants per tumor or 0.38 per megabase (Mb)], concordant with other recent studies of well-differentiated thyroid cancer (mean 0.41/Mb for PTC or 0.31/Mb in FTC); although the wiFTCs behave aggressively in many cases, they do not seem to carry the large mutational burden described in anaplastic thyroid cancers (mean 2.7/Mb in anaplastic thyroid cancer or 0.72/Mb excluding hypermutators) (12, 13, 42). Mutations within RAS genes were noted in 20.5% of our cohort overall. This frequency of RAS mutations is slightly lower than generally reported in FTC, but within the range reported across prior studies. Mutations in the TSH receptor gene TSHR were common in our cohort, found in 10.3% of samples. TSHR mutations occurred in both the miFTC and wiFTC groups. Mutations in TSHR and the RAS family were mutually exclusive and did not appear to occur at significantly different rates based on invasive status in our cohort. Cancer driver genes EIF1AX and DICER1 carried mutations in two samples each; recurrent mutations in these genes were demonstrated in the recent genomic analyses of FTC (13, 15), and recurrent EIF1AX mutations were noted in a recent WES analysis of anaplastic thyroid carcinoma, the most aggressive thyroid carcinoma (42). In this cohort, one wiFTC sample was found to have a BRAFK601E mutation, without any other FTC driver events. Although somatic BRAFV600E mutations are found in ∼60% of PTCs, they are not historically associated with follicular thyroid tumors (12). This FTC had no nuclear features of PTC, confirming the diagnosis of FTC rather than follicular variant of PTC. Indeed, recent reports have identified BRAF mutations in a small proportion of FTAs (14) and specifically one BRAFK601E mutation in FTC (13). Although only a single driver gene fusion event was identified in this study, WES-indiscernible fusion events such as fusions involving introns playing a driving role in some FTCs cannot be ruled out. Future comprehensive analyses in the field will hopefully identify additional gene fusions in FTC in addition to the well-described PAX8-PPARγ fusion gene and other less common fusion events such as THADA (19, 43). Our cohort was characterized by numerous somatic arm-level copy changes, including recurrent CN loss of 22q. Losses of 22q have also been identified in recent next-generation sequencing studies of FTCs (13, 15). The functional consequences of 22q deletion have not been fully elucidated, though one study has shown that these tumors behave similarly to RAS-family mutated tumors on a transcriptional level (15). The recurrent CN gains identified in 5q in eaFTC and 7q in miFTC and wiFTC may be tumor-driving events in some cases, though the molecular mechanisms have not yet been definitively identified. Survival analysis demonstrated considerable differences in outcome based on AJCC stage and invasion subtype. Interestingly, the AJCC 7th edition staging system seemed to perform better in our cohort than the 8th edition system, likely due to our cohort’s enrichment in more aggressive widely invasive cases, which were downstaged under the new system. The genetics of the tumors also informed the survival analysis, with total mutation burden, cancer driver burden, and FTC driver burden all associated with worse prognosis. Most patients who had recurrence or mortality due to FTC had a more advanced genetic profile as well as more invasive histopathological phenotype, raising the question of which factor is more important. Our multivariate survival analysis suggests that total mutational burden may be a strong prognostic indicator independent of histopathology. Total mutational burden is already in clinical use as a biomarker to guide immunomodulatory treatment in melanoma and other cancers (44). Together with histological classification, the genetic profile of invasive follicular thyroid tumors can be used to make informed treatment decisions for patients facing this diverse group of cancers. In this study, the genetic profile seems to predict survival as a complement to the traditional histological approach, and next-generation sequencing is fast becoming affordable and practical for clinical use, particularly in high-volume centers. Whether similar results could be obtained using less-invasive biopsy specimens remains to be seen, but the approach shows promise of obviating the need for diagnostic surgery for some patients with follicular lesions. Conclusion Overall, our data demonstrate that the 2017 WHO subtypes of invasive FTC arise from comparable genetic backgrounds, in spite of the dramatic differences in clinical outcome. The differences in tumor behavior may be due to time elapsed since starting on a shared pathway toward malignancy, as tumors have time to accumulate more mutations and become more aggressive over time, without a signature genetic event defining invasiveness. Alternatively, the difference may lie in the epigenetics or noncoding genetic profile of the tumor or a background thyroid milieu that encourages the more invasive and aggressive behavior even with similar somatic mutational landscape. Some tumors have an aggressive phenotype in spite of relatively bland mutational profiles, lending further credence to the idea of nongenetic driver events in some cases. In the absence of specific genetic markers to guide prognosis in FTC, clinicians will need to rely on an integrated approach to this disease, incorporating clinical, pathological, and genetic factors to guide treatment decisions. Abbreviations: Abbreviations: AJCC American Joint Committee on Cancer CN copy number CNV copy number variation COSMIC Catalogue of Somatic Mutations in Cancer eaFTC encapsulated angioinvasive follicular thyroid carcinoma FFPE formalin-fixed paraffin-embedded FNA fine-needle aspiration FTA follicular thyroid adenoma FTC follicular thyroid carcinoma KEGG Kyoto Encyclopedia of Genes and Genomes Mb megabase miFTC minimally invasive follicular thyroid carcinoma PTC papillary thyroid carcinoma WES whole-exome sequencing WHO World Health Organization wiFTC widely invasive follicular thyroid carcinoma Acknowledgments The authors thank the Ohse Research Foundation at the Yale School of Medicine, the Stockholm County Council, the Swedish Cancer Society, and the Swedish Society for Medical Research for support of this research. Financial Support: This study was supported by the Yale School of Medicine (to T.C.), Stockholms Läns Landsting (to C.C.J.), Svenska Sällskapet för Medicinsk Forskning (to C.C.J.), Cancerfonden (to C.C.J.), and the Damon Runyon Cancer Research Foundation (to T.C.). Disclosure Summary: The authors have nothing to disclose. References 1. Aschebrook-Kilfoy B , Ward MH , Sabra MM , Devesa SS . Thyroid cancer incidence patterns in the United States by histologic type, 1992-2006 . Thyroid . 2011 ; 21 ( 2 ): 125 – 134 . 2. Enewold L , Zhu K , Ron E , Marrogi AJ , Stojadinovic A , Peoples GE , Devesa SS . Rising thyroid cancer incidence in the United States by demographic and tumor characteristics, 1980-2005 . Cancer Epidemiol Biomarkers Prev . 2009 ; 18 ( 3 ): 784 – 791 . 3. Lim H , Devesa SS , Sosa JA , Check D , Kitahara CM . Trends in thyroid cancer incidence and mortality in the United States, 1974-2013 . JAMA . 2017 ; 317 ( 13 ): 1338 – 1348 . 4. Aschebrook-Kilfoy B , Grogan RH , Ward MH , Kaplan E , Devesa SS . Follicular thyroid cancer incidence patterns in the United States, 1980-2009 . Thyroid . 2013 ; 23 ( 8 ): 1015 – 1021 . 5. James BC , Aschebrook-Kilfoy B , Cipriani N , Kaplan EL , Angelos P , Grogan RH . The incidence and survival of rare cancers of the thyroid, parathyroid, adrenal, and pancreas . Ann Surg Oncol . 2016 ; 23 ( 2 ): 424 – 433 . 6. DeLellis RA , Lloyd RV , Heitz PU , Eng C , eds. Pathology & Genetics of Tumours of Endocrine Organs. World Health Organization Classification of Tumours . Vol 8. 3rd ed . Lyon : IARC Press ; 2004 . 7. Lloyd RV , Osamura RY , Klöppel G , Rosai J , eds. WHO Classification of Tumours of Endocrine Organs . Vol 10. Geneva : WHO Press ; 2017 . 8. D’Avanzo A , Treseler P , Ituarte PH , Wong M , Streja L , Greenspan FS , Siperstein AE , Duh QY , Clark OH . Follicular thyroid carcinoma: histology and prognosis . Cancer . 2004 ; 100 ( 6 ): 1123 – 1129 . 9. Sugino K , Ito K , Nagahama M , Kitagawa W , Shibuya H , Ohkuwa K , Yano Y , Uruno T , Akaishi J , Kameyama K , Ito K . Prognosis and prognostic factors for distant metastases and tumor mortality in follicular thyroid carcinoma . Thyroid . 2011 ; 21 ( 7 ): 751 – 757 . 10. Asari R , Koperek O , Scheuba C , Riss P , Kaserer K , Hoffmann M , Niederle B . Follicular thyroid carcinoma in an iodine-replete endemic goiter region: a prospectively collected, retrospectively analyzed clinical trial . Ann Surg . 2009 ; 249 ( 6 ): 1023 – 1031 . 11. Cipriani NA , Nagar S , Kaplan SP , White MG , Antic T , Sadow PM , Aschebrook-Kilfoy B , Angelos P , Kaplan EL , Grogan RH . Follicular thyroid carcinoma: how have histologic diagnoses changed in the last half-century and what are the prognostic implications ? Thyroid . 2015 ; 25 ( 11 ): 1209 – 1216 . 12. Cancer Genome Atlas Research Network . Integrated genomic characterization of papillary thyroid carcinoma . Cell . 2014 ; 159 ( 3 ): 676 – 690 . 13. Jung SH , Kim MS , Jung CK , Park HC , Kim SY , Liu J , Bae JS , Lee SH , Kim TM , Lee SH , Chung YJ . Mutational burdens and evolutionary ages of thyroid follicular adenoma are comparable to those of follicular carcinoma . Oncotarget . 2016 ; 7 ( 43 ): 69638 – 69648 . 14. Swierniak M , Pfeifer A , Stokowy T , Rusinek D , Chekan M , Lange D , Krajewska J , Oczko-Wojciechowska M , Czarniecka A , Jarzab M , Jarzab B , Wojtas B . Somatic mutation profiling of follicular thyroid cancer by next generation sequencing . Mol Cell Endocrinol . 2016 ; 433 : 130 – 137 . 15. Yoo SK , Lee S , Kim SJ , Jee HG , Kim BA , Cho H , Song YS , Cho SW , Won JK , Shin JY , Park J , Kim JI , Lee KE , Park YJ , Seo JS . Comprehensive analysis of the transcriptional and mutational landscape of follicular and papillary thyroid cancers . PLoS Genet . 2016 ; 12 ( 8 ): e1006239 . 16. Fukahori M , Yoshida A , Hayashi H , Yoshihara M , Matsukuma S , Sakuma Y , Koizume S , Okamoto N , Kondo T , Masuda M , Miyagi Y . The associations between RAS mutations and clinical characteristics in follicular thyroid tumors: new insights from a single center and a large patient cohort . Thyroid . 2012 ; 22 ( 7 ): 683 – 689 . 17. Vuong HG , Kondo T , Oishi N , Nakazawa T , Mochizuki K , Inoue T , Tahara I , Kasai K , Hirokawa M , Tran TM , Katoh R . Genetic alterations of differentiated thyroid carcinoma in iodine-rich and iodine-deficient countries . Cancer Med . 2016 ; 5 ( 8 ): 1883 – 1889 . 18. Garcia-Rostan G , Zhao H , Camp RL , Pollan M , Herrero A , Pardo J , Wu R , Carcangiu ML , Costa J , Tallini G . ras mutations are associated with aggressive tumor phenotypes and poor prognosis in thyroid cancer . J Clin Oncol . 2003 ; 21 ( 17 ): 3226 – 3235 . 19. Nikiforova MN , Biddinger PW , Caudill CM , Kroll TG , Nikiforov YE . PAX8-PPARgamma rearrangement in thyroid tumors: RT-PCR and immunohistochemical analyses . Am J Surg Pathol . 2002 ; 26 ( 8 ): 1016 – 1023 . 20. Boos LA , Dettmer M , Schmitt A , Rudolph T , Steinert H , Moch H , Sobrinho-Simões M , Komminoth P , Perren A . Diagnostic and prognostic implications of the PAX8-PPARγ translocation in thyroid carcinomas-a TMA-based study of 226 cases . Histopathology . 2013 ; 63 ( 2 ): 234 – 241 . 21. Choi M , Scholl UI , Ji W , Liu T , Tikhonova IR , Zumbo P , Nayir A , Bakkaloğlu A , Ozen S , Sanjad S , Nelson-Williams C , Farhi A , Mane S , Lifton RP . Genetic diagnosis by whole exome capture and massively parallel DNA sequencing . Proc Natl Acad Sci USA . 2009 ; 106 ( 45 ): 19096 – 19101 . 22. Li H , Durbin R . Fast and accurate short read alignment with Burrows-Wheeler transform . Bioinformatics . 2009 ; 25 ( 14 ): 1754 – 1760 . 23. Cibulskis K , Lawrence MS , Carter SL , Sivachenko A , Jaffe D , Sougnez C , Gabriel S , Meyerson M , Lander ES , Getz G . Sensitive detection of somatic point mutations in impure and heterogeneous cancer samples . Nat Biotechnol . 2013 ; 31 ( 3 ): 213 – 219 . 24. Costello M , Pugh TJ , Fennell TJ , Stewart C , Lichtenstein L , Meldrim JC , Fostel JL , Friedrich DC , Perrin D , Dionne D , Kim S , Gabriel SB , Lander ES , Fisher S , Getz G . Discovery and characterization of artifactual mutations in deep coverage targeted capture sequencing data due to oxidative DNA damage during sample preparation . Nucleic Acids Res . 2013 ; 41 ( 6 ): e67 . 25. Chen G , Mosier S , Gocke CD , Lin MT , Eshleman JR . Cytosine deamination is a major cause of baseline noise in next-generation sequencing . Mol Diagn Ther . 2014 ; 18 ( 5 ): 587 – 593 . 26. Zhao S , Choi M , Overton JD , Bellone S , Roque DM , Cocco E , Guzzo F , English DP , Varughese J , Gasparrini S , Bortolomai I , Buza N , Hui P , Abu-Khalaf M , Ravaggi A , Bignotti E , Bandiera E , Romani C , Todeschini P , Tassi R , Zanotti L , Carrara L , Pecorelli S , Silasi DA , Ratner E , Azodi M , Schwartz PE , Rutherford TJ , Stiegler AL , Mane S , Boggon TJ , Schlessinger J , Lifton RP , Santin AD . Landscape of somatic single-nucleotide and copy-number mutations in uterine serous carcinoma . Proc Natl Acad Sci USA . 2013 ; 110 ( 8 ): 2916 – 2921 . 27. Sudmant PH , Rausch T , Gardner EJ , Handsaker RE , Abyzov A , Huddleston J , Zhang Y , Ye K , Jun G , Fritz MH , Konkel MK , Malhotra A , Stütz AM , Shi X , Casale FP , Chen J , Hormozdiari F , Dayama G , Chen K , Malig M , Chaisson MJP , Walter K , Meiers S , Kashin S , Garrison E , Auton A , Lam HYK , Mu XJ , Alkan C , Antaki D , Bae T , Cerveira E , Chines P , Chong Z , Clarke L , Dal E , Ding L , Emery S , Fan X , Gujral M , Kahveci F , Kidd JM , Kong Y , Lameijer EW , McCarthy S , Flicek P , Gibbs RA , Marth G , Mason CE , Menelaou A , Muzny DM , Nelson BJ , Noor A , Parrish NF , Pendleton M , Quitadamo A , Raeder B , Schadt EE , Romanovitch M , Schlattl A , Sebra R , Shabalin AA , Untergasser A , Walker JA , Wang M , Yu F , Zhang C , Zhang J , Zheng-Bradley X , Zhou W , Zichner T , Sebat J , Batzer MA , McCarroll SA , Mills RE , Gerstein MB , Bashir A , Stegle O , Devine SE , Lee C , Eichler EE , Korbel JO ; 1000 Genomes Project Consortium . An integrated map of structural variation in 2,504 human genomes . Nature . 2015 ; 526 ( 7571 ): 75 – 81 . 28. Auton A , Brooks LD , Durbin RM , Garrison EP , Kang HM , Korbel JO , Marchini JL , McCarthy S , McVean GA , Abecasis GR ; 1000 Genomes Project Consortium . A global reference for human genetic variation . Nature . 2015 ; 526 ( 7571 ): 68 – 74 . 29. Magi A , Tattini L , Cifola I , D’Aurizio R , Benelli M , Mangano E , Battaglia C , Bonora E , Kurg A , Seri M , Magini P , Giusti B , Romeo G , Pippucci T , De Bellis G , Abbate R , Gensini GF . EXCAVATOR: detecting copy number variants from whole-exome sequencing data . Genome Biol . 2013 ; 14 ( 10 ): R120 . 30. Mermel CH , Schumacher SE , Hill B , Meyerson ML , Beroukhim R , Getz G . GISTIC2.0 facilitates sensitive and confident localization of the targets of focal somatic copy-number alteration in human cancers . Genome Biol . 2011 ; 12 ( 4 ): R41 . 31. Chen X , Schulz-Trieglaff O , Shaw R , Barnes B , Schlesinger F , Källberg M , Cox AJ , Kruglyak S , Saunders CT . Manta: rapid detection of structural variants and indels for germline and cancer sequencing applications . Bioinformatics . 2016 ; 32 ( 8 ): 1220 – 1222 . 32. Rausch T , Zichner T , Schlattl A , Stütz AM , Benes V , Korbel JO . DELLY: structural variant discovery by integrated paired-end and split-read analysis . Bioinformatics . 2012 ; 28 ( 18 ): i333 – i339 . 33. Alexandrov LB , Nik-Zainal S , Wedge DC , Aparicio SA , Behjati S , Biankin AV , Bignell GR , Bolli N , Borg A , Børresen-Dale AL , Boyault S , Burkhardt B , Butler AP , Caldas C , Davies HR , Desmedt C , Eils R , Eyfjörd JE , Foekens JA , Greaves M , Hosoda F , Hutter B , Ilicic T , Imbeaud S , Imielinski M , Jäger N , Jones DT , Jones D , Knappskog S , Kool M , Lakhani SR , López-Otín C , Martin S , Munshi NC , Nakamura H , Northcott PA , Pajic M , Papaemmanuil E , Paradiso A , Pearson JV , Puente XS , Raine K , Ramakrishna M , Richardson AL , Richter J , Rosenstiel P , Schlesner M , Schumacher TN , Span PN , Teague JW , Totoki Y , Tutt AN , Valdés-Mas R , van Buuren MM , van ’t Veer L , Vincent-Salomon A , Waddell N , Yates LR , Zucman-Rossi J , Futreal PA , McDermott U , Lichter P , Meyerson M , Grimmond SM , Siebert R , Campo E , Shibata T , Pfister SM , Campbell PJ , Stratton MR ; Australian Pancreatic Cancer Genome Initiative; ICGC Breast Cancer Consortium; ICGC MMML-Seq Consortium; ICGC PedBrain . Signatures of mutational processes in human cancer [published correction appears in Nature. 2013;502(7470):258]. Nature . 2013 ; 500 ( 7463 ): 415 – 421 . 34. Rosenthal R , McGranahan N , Herrero J , Taylor BS , Swanton C . DeconstructSigs: delineating mutational processes in single tumors distinguishes DNA repair deficiencies and patterns of carcinoma evolution . Genome Biol . 2016 ; 17 ( 1 ): 31 . 35. Miller CA , White BS , Dees ND , Griffith M , Welch JS , Griffith OL , Vij R , Tomasson MH , Graubert TA , Walter MJ , Ellis MJ , Schierding W , DiPersio JF , Ley TJ , Mardis ER , Wilson RK , Ding L . SciClone: inferring clonal architecture and tracking the spatial and temporal patterns of tumor evolution . PLOS Comput Biol . 2014 ; 10 ( 8 ): e1003665 . 36. Szklarczyk D , Morris JH , Cook H , Kuhn M , Wyder S , Simonovic M , Santos A , Doncheva NT , Roth A , Bork P , Jensen LJ , von Mering C . The STRING database in 2017: quality-controlled protein-protein association networks, made broadly accessible . Nucleic Acids Res . 2017 ; 45 ( D1 ): D362 – D368 . 37. Kanehisa M , Goto S , Sato Y , Furumichi M , Tanabe M . KEGG for integration and interpretation of large-scale molecular data sets . Nucleic Acids Res . 2012 ; 40 ( Database issue ): D109 – D114 . 38. Tuttle RM , Haugen B , Perrier ND . Updated American Joint Committee on Cancer/tumor-node-metastasis staging system for differentiated and anaplastic thyroid cancer (eighth edition): what changed and why ? Thyroid 2017 ; 27 ( 6 ): 751 – 756 39. R: A language and environment for statistical computing [computer program]. Vienna, Austria: R Foundation for Statistical Computing: 2017. 40. Sha J , Xue W , Dong B , Pan J , Wu X , Li D , Liu D , Huang Y . PRKAR2B plays an oncogenic role in the castration-resistant prostate cancer . Oncotarget . 2017 ; 8 ( 4 ): 6114 – 6129 . 41. Vogelstein B , Papadopoulos N , Velculescu VE , Zhou S , Diaz LA Jr , Kinzler KW . Cancer genome landscapes . Science . 2013 ; 339 ( 6127 ): 1546 – 1558 . 42. Kunstman JW , Juhlin CC , Goh G , Brown TC , Stenman A , Healy JM , Rubinstein JC , Choi M , Kiss N , Nelson-Williams C , Mane S , Rimm DL , Prasad ML , Höög A , Zedenius J , Larsson C , Korah R , Lifton RP , Carling T . Characterization of the mutational landscape of anaplastic thyroid cancer via whole-exome sequencing . Hum Mol Genet . 2015 ; 24 ( 8 ): 2318 – 2329 . 43. Panebianco F , Kelly LM , Liu P , Zhong S , Dacic S , Wang X , Singhi AD , Dhir R , Chiosea SI , Kuan SF , Bhargava R , Dabbs D , Trivedi S , Gandhi M , Diaz R , Wald AI , Carty SE , Ferris RL , Lee AV , Nikiforova MN , Nikiforov YE . THADA fusion is a mechanism of IGF2BP3 activation and IGF1R signaling in thyroid cancer . Proc Natl Acad Sci USA . 2017 ; 114 ( 9 ): 2307 – 2312 . 44. Khagi Y , Goodman AM , Daniels GA , Patel SP , Sacco AG , Randall JM , Bazhenova LA , Kurzrock R . Hypermutated circulating tumor DNA: correlation with response to checkpoint inhibitor-based immunotherapy . Clin Cancer Res . 2017 ; 23 ( 19 ): 5729 – 5736 . Copyright © 2018 Endocrine Society http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Journal of Clinical Endocrinology and Metabolism Oxford University Press

Loading next page...
 
/lp/ou_press/comprehensive-genetic-analysis-of-follicular-thyroid-carcinoma-Dd8coFEdJz
Publisher
Oxford University Press
Copyright
Copyright © 2018 Endocrine Society
ISSN
0021-972X
eISSN
1945-7197
D.O.I.
10.1210/jc.2018-00277
Publisher site
See Article on Publisher Site

Abstract

Abstract Context Follicular thyroid carcinoma (FTC) is classified into minimally invasive (miFTC), encapsulated angioinvasive (eaFTC), and widely invasive (wiFTC) subtypes, according to the 2017 World Health Organization guidelines. The genetic signatures of these subtypes may be crucial for diagnosis, prognosis, and treatment but have not been described. Objective Identify and describe the genetic underpinnings of subtypes of FTC. Methods Thirty-nine tumors, comprising 12 miFTCs, 17 eaFTCs, and 10 wiFTCs, were whole-exome sequenced and analyzed. Somatic mutations, constitutional sequence variants, somatic copy number alterations, and mutational signatures were described. Clinicopathologic parameters and mutational profiles were assessed for associations with patient outcomes. Results Total mutation burden was consistent across FTC subtypes, with a median of 10 (range 1 to 44) nonsynonymous somatic mutations per tumor. Overall, 20.5% of specimens had a mutation in the RAS subfamily (HRAS, KRAS, or NRAS), with no notable difference between subtypes. Mutations in TSHR, DICER1, EIF1AX, KDM5C, NF1, PTEN, and TP53 were also noted to be recurrent across the cohort. Clonality analysis demonstrated more subclones in wiFTC. Survival analysis demonstrated worse disease-specific survival in the eaFTC and wiFTC cohorts, with no recurrences or deaths for patients with miFTC. Mutation burden was associated with worse prognosis, independent of histopathological classification. Conclusions Though the number and variety of somatic variants are similar in the different histopathological subtypes of FTC in our study, mutational burden was an independent predictor of mortality and recurrence. Follicular thyroid carcinoma (FTC) is a well-differentiated endocrine malignancy that accounts for 10% of all thyroid cancers (1–3). FTC is ∼2.5 times more common in women, similar to papillary thyroid carcinoma (PTC) (2, 4). The 5-year survival of FTC is ∼88%, but drops to 78% at 10 years (5). FTCs have traditionally been classified as minimally invasive (miFTC) or widely invasive FTC (wiFTC) based on the presence of vascular and/or capsular infiltration (6). In 2017, the World Health Organization (WHO) unveiled new guidelines that include an intermediate histopathologic type: encapsulated angioinvasive FTC (eaFTC) (7). Although determination of invasive status is critical for prognostication, even the new classification relies on subjective pathological evaluation of the degree of invasion. The earlier system, despite its limitations, has been demonstrated to predict considerable differences in tumor recurrence, metastatic potential, and mortality, though the 2017 criteria have not yet been similarly validated (8–11). Although the genomic landscape and driver events in PTC have been well described (12), the molecular etiology of FTC is less well known, particularly wiFTC and eaFTC. Recurrent somatic mutations in the RAS family (NRAS, KRAS, and HRAS) have been reported in FTC, usually in the 61st codon (13–15). Most common is NRAS, mutated in 15% to 40% of FTCs (16, 17). RAS family mutations hold clinical significance, as they have been shown to increase metastatic potential and disease-specific mortality (18). The PAX8-PPARγ fusion gene is identified in about one-third of FTCs, with estimates ranging from 12% to 56% (19, 20). Although PAX8-PPARγ likely contributes to follicular tumorigenesis, it does not appear to impact prognosis (20). The increased availability and accuracy of next-generation sequencing technology has allowed recent advances in understanding the mutational landscape of FTC and the major differences between FTC and follicular thyroid adenoma (FTA) (13–15). However, nearly all prior sequencing studies in FTC have been performed on miFTC, with limited applicability to the rarer but much deadlier wiFTC. Moreover, there have been few data characterizing the molecular underpinnings of invasion or the genetic distinctions between the different categories of FTC, particularly since the recent introduction of eaFTC in the WHO 2017 guidelines. The lack of genetic markers associated with each category limits the use of the histopathological classification of follicular tumors to surgical specimens rather than fine-needle aspiration (FNA) biopsies. Our study uses next-generation sequencing techniques and bioinformatics tools to investigate the genomic landscape of FTC, with particular attention to the recently described WHO 2017 histopathological categories of invasiveness. We investigate whether the divergent behavior of these tumors in clinical practice is associated with distinct molecular profiles. Materials and Methods Patient cohort and sample acquisition The 39 patients recruited for this study received surgical treatment at Yale New Haven Hospital (n = 24; FTC600 series) or Karolinska University Hospital (n = 15; FTC1 series) between 2002 and 2013 (Supplemental Table 1). All samples were independently reviewed by a minimum of two experienced endocrine pathologists for histopathological confirmation, and poorly differentiated thyroid cancer was ruled out by the Turin criteria, in accordance with the 2017 recommendations from the WHO. Matched normal samples for each tumor were obtained from adjacent histologically normal thyroid or blood leukocyte DNA. The diagnosis and degree of invasion were confirmed according to the 2017 guidelines established by the WHO (7). Although the 2017 WHO guidelines distinguish Hürthle cell (oxyphilic) carcinomas from FTCs, both Hürthle cell and conventional FTCs were included in this study, as both exhibit invasive behavior. Informed consent was obtained from all patients involved in this study. The acquisition and use of protected health information and tissue specimens were performed as specified by the Health Insurance Portability and Accountability Act (Yale) or Swedish Act on Biobanks (Karolinska). The study was approved by the Yale University and Karolinska Institutet Institutional Review Boards. Whole-exome sequencing Genomic DNA was extracted from formalin-fixed paraffin-embedded (FFPE) or fresh-frozen tissue. Three 1-mm–thick tissue cores were obtained per block for FFPE samples, with paraffin enzymatically removed and genomic DNA prepared using a proprietary in-house method at the Yale Center for Genome Analysis. Before and after coring, FFPE blocks were sectioned, stained, and analyzed via light microscopy to ensure tissue identity. For fresh-frozen samples, after tissue disruption via sonication, genomic DNA was isolated using the DNeasy Blood & Tissue Kit (Qiagen) in accordance with the manufacturer’s instructions. The quality and quantity of isolated DNA were then measured via spectrophotometry (NanoDrop Technologies). Tumor and normal genomic DNA pairs were then subjected to exome capture and sequencing as previously described (21). Adaptors of known sequence were ligated to genomic DNA fragments that were amplified by ligation-mediated PCR. Specimens were then subjected to capture using the NimbleGen 2.1M human exome array. Exome-specific DNA was eluted and then underwent 74-base paired-end sequencing on the HiSeq 2000 instrument per the specifications of the manufacturer (Illumina). Removal of PCR duplicates was performed with Picard Tools (http://broadinstitute.github.io/picard). Though 42 tumor-normal pairs originally generated libraries for whole-exome sequencing (WES), 39 of these passed quality metrics and were included in subsequent analyses. The Burrow-Wheeler Aligner-MEM program was used to map reads to the human reference genome GRCh37/hg19 (22). Somatic and germline single nucleotide variants and short insertions or deletions were called by MuTect2 using the Bayesian classifier followed by in-house filtering scripts to increase variant calling specificity (23). Likely false-positives were excluded using the D-ToxoG filter, and variant allele frequencies <0.10 were manually censored for variant calling (24, 25). Tumor purity was calculated with a nested approach, taking into account B-allele frequency in loss-of-heterozygosity regions, followed by tumor driver allele frequency and average overall variant allele frequency (26). Known variants in annotated databases [1000 Genomes (27, 28); and NHLBI ESP6500 (Exome Variant Server, NHLBI GO Exome Sequencing Project, Seattle, WA; http://evs.gs.washington.edu/EVS/) and 2577 noncancer exomes sequenced at Yale] were excluded, and novel exonic variants were evaluated for impact on transcriptional and/or translational processing. Copy-number variation detection EXCAVATOR was used to examine somatic copy number variation (CNV) in tumors (29). Briefly, EXCAVATOR normalizes the nonuniform WES read depth taking guanine/cytosine content into account and thereby calculates the ratios of normalized read depth between tumor and normal for the exome capture intervals and performs segmentation of copy number (CN) data using a novel heterogeneous hidden Markov model algorithm. Somatic CN gain and loss were defined relative to the matched normal samples. The significance of individual CNV peaks was evaluated using GISTIC (30). Fusion gene detection in FTC Fresh-frozen tumor samples (n = 13) were assayed for the presence of three chromosomal translocations common to thyroid carcinoma: PAX8-PPARγ, RET-PTC1, or RET-PTC3. The remaining samples from the WES cohort were excluded due to low RNA yield or quality or were in FFPE and not amenable to reliable RNA gene fusion detection. After isolating RNA using an RNeasy Plus Mini Kit (Qiagen), spectrophotometry was used to determine the concentration and purity of RNA (NanoDrop Technologies). We then assayed the samples for the defined chromosomal translocations with the Thyroid Cancer Fusion Gene Detection Kit (THRNA-RT32; EntroGen, Inc.). In this protocol, cDNA synthesis and gene amplification are performed simultaneously. The products are exposed to dual-labeled hydrolysis probes designed to detect the amplification products. Fusion gene amplification was detected via a FAM-BHQ–labeled probe in a CFX96 Real-Time System thermocycler (Bio-Rad). Additionally, exome sequencing data were investigated for evidence of somatic gene fusions using the DELLY2 and Manta packages and filtered for recurrent or previously described fusion events (31, 32). Mutational signatures, clonality, and pathway analyses Mutational signatures, as described by Alexandrov et al. (33), were also analyzed. Signatures of the individual samples and across each type of tumor were constructed through the compilation and quantification of somatic variant patterns using the deconstructSigs algorithm (34). SciClone was used to detect clonal populations of tumor cells in each sample (35). In brief, this algorithm clusters mutations by variant allele frequency to predict subclonal populations within the tumor, focusing on CN neutral regions. For clonality analysis, mutations at a more generous variant allele frequency cutoff of 0.05 were included. Additionally, using the network and pathway analysis platforms from STRING (https://string-db.org), GeNets (https://apps.broadinstitute.org/genets#), and Kyoto Encyclopedia of Genes and Genomes (KEGG; http://www.genome.jp/kegg/mapper.html), all called somatic alterations were evaluated to determine if genetic events were enriched in specific biologic pathways. Networks were mapped with any disconnected nodes excluded for visual clarity. The STRING database maps connections between proteins based on experimental data as well as text mining of major biomedical databases (36). GeNets analysis uses a meta-network of several other protein interaction tools and databases to map connections between genes of interest, whereas KEGG lists all pathways enriched in each set of genes, ranked by number of genes altered per pathway (37). Statistical analysis Clinicopathologic parameters including patient age, sex, tumor size (largest diameter), and American Joint Committee on Cancer (AJCC) stage (7th and 8th editions) (38) and genetic parameters including total count of somatic variants, clonality, and number of cancer-associated driver events were assessed for association with invasive status. All variables were assumed to be nonparametrically distributed due to small sample size. Continuous variables were analyzed using the Kruskal-Wallis test and/or two-tailed Mann-Whitney test, as appropriate, with Dunn correction for multiple testing. Fisher exact test with Freeman-Halton extension was used for categorical variables. Disease-specific survival (i.e., time to death due to FTC) was illustrated with Kaplan-Meier curves, and Cox proportional hazards regression was used for univariate and multivariate survival analysis. All statistical analyses were performed using Prism 7 (GraphPad Software), except for three-category Fisher exact tests, which were performed using VassarStats (vassarstats.net) and Cox proportional hazards analyses, which were performed in R (39). Results Patient and tumor characteristics Twenty-six of the 39 patients included in the exome analysis were female (67%), and 13 were male (33%). The mean age of the population was 55.2 years old (median 54; range 14 to 86 years). The overall median tumor diameter was 3.6 cm, and the majority of patients had stage I or II disease (92%; 36 out of 39) based on AJCC 8th edition. Following surgery, patients were clinically monitored for disease progression or recurrence for a median duration of 5.8 years. The overall recurrence rate and disease-specific mortality rate were both 15% (6 out of 39), as all patients in our cohort who had disease recurrence ultimately died of FTC; these figures are concordant with previously reported trends (5). All patients had surgery, and most had radioactive iodine treatment, whereas only one had cytotoxic or targeted chemotherapy (Supplemental Table 1). On average, miFTCs were smaller and found in younger patients, though these differences were not statistically significant. Relative to miFTC and eaFTC, wiFTC had a higher AJCC stage at time of surgery (P = 0.013) and were more likely to recur or be the cause of death [P = 0.0017; Table 1; Fig. 1(a)]. Table 1. Clinicopathologic Features of Patients With miFTC, eaFTC, and wiFTC Variable All (N = 39) miFTC (n = 12) eaFTC (n = 17) wiFTC (n = 10) P Value a Age at diagnosis, y, mean ± SD 55.2 ± 16.1 48.1 ± 15.6 58.0 ± 15.9 58.9 ± 15.7 0.24 Female sex, n/total n (%) 26/39 (67) 8/12 (67) 12/17 (71) 6/10 (60) 0.91 Tumor diameter, cm, mean ± SD 3.6 ± 1.5 3.0 ± 0.9 3.9 ± 2.0 3.8 ± 0.9 0.29 AJCC stage greater than II, n/total n (%) 3/39 (8) 0/12 (0) 0/17 (0) 3/10 (30) 0.013 (Pmi-ea > 0.99; Pmi-wi = 0.078; Pea-wi = 0.041) Disease recurrence, n/total n (%) 6/39 (15) 0/12 (0) 1/17 (6) 5/9 (56) 0.0017 (Pmi-ea > 0.99; Pmi-wi = 0.0062; Pea-wi = 0.0097) Death due to FTC, n/total n (%) 6/39 (15) 0/12 (0) 1/17 (6) 5/9 (56) 0.0017 (Pmi-ea > 0.99; Pmi-wi = 0.0062; Pea-wi = 0.0097) Disease-specific survival, 5-year, n/total n (%) 22/24 (92) 7/7 (100) 9/9 (100) 7/9 (78) 0.31 Disease-specific survival, 10-year, n/total n (%) 8/14 (57) 4/4 (100) 3/4 (75) 1/6 (17) 0.052 Variable All (N = 39) miFTC (n = 12) eaFTC (n = 17) wiFTC (n = 10) P Value a Age at diagnosis, y, mean ± SD 55.2 ± 16.1 48.1 ± 15.6 58.0 ± 15.9 58.9 ± 15.7 0.24 Female sex, n/total n (%) 26/39 (67) 8/12 (67) 12/17 (71) 6/10 (60) 0.91 Tumor diameter, cm, mean ± SD 3.6 ± 1.5 3.0 ± 0.9 3.9 ± 2.0 3.8 ± 0.9 0.29 AJCC stage greater than II, n/total n (%) 3/39 (8) 0/12 (0) 0/17 (0) 3/10 (30) 0.013 (Pmi-ea > 0.99; Pmi-wi = 0.078; Pea-wi = 0.041) Disease recurrence, n/total n (%) 6/39 (15) 0/12 (0) 1/17 (6) 5/9 (56) 0.0017 (Pmi-ea > 0.99; Pmi-wi = 0.0062; Pea-wi = 0.0097) Death due to FTC, n/total n (%) 6/39 (15) 0/12 (0) 1/17 (6) 5/9 (56) 0.0017 (Pmi-ea > 0.99; Pmi-wi = 0.0062; Pea-wi = 0.0097) Disease-specific survival, 5-year, n/total n (%) 22/24 (92) 7/7 (100) 9/9 (100) 7/9 (78) 0.31 Disease-specific survival, 10-year, n/total n (%) 8/14 (57) 4/4 (100) 3/4 (75) 1/6 (17) 0.052 P values <0.05 shown in bold. a Kruskal-Wallis followed by corrected Dunn test for continuous variables and Fisher exact test (with Freeman-Halton extension) for categorical variables. P values in parentheses are pairwise comparisons between miFTC, eaFTC, and wiFTC when significance was found in the overall three-way analysis. View Large Table 1. Clinicopathologic Features of Patients With miFTC, eaFTC, and wiFTC Variable All (N = 39) miFTC (n = 12) eaFTC (n = 17) wiFTC (n = 10) P Value a Age at diagnosis, y, mean ± SD 55.2 ± 16.1 48.1 ± 15.6 58.0 ± 15.9 58.9 ± 15.7 0.24 Female sex, n/total n (%) 26/39 (67) 8/12 (67) 12/17 (71) 6/10 (60) 0.91 Tumor diameter, cm, mean ± SD 3.6 ± 1.5 3.0 ± 0.9 3.9 ± 2.0 3.8 ± 0.9 0.29 AJCC stage greater than II, n/total n (%) 3/39 (8) 0/12 (0) 0/17 (0) 3/10 (30) 0.013 (Pmi-ea > 0.99; Pmi-wi = 0.078; Pea-wi = 0.041) Disease recurrence, n/total n (%) 6/39 (15) 0/12 (0) 1/17 (6) 5/9 (56) 0.0017 (Pmi-ea > 0.99; Pmi-wi = 0.0062; Pea-wi = 0.0097) Death due to FTC, n/total n (%) 6/39 (15) 0/12 (0) 1/17 (6) 5/9 (56) 0.0017 (Pmi-ea > 0.99; Pmi-wi = 0.0062; Pea-wi = 0.0097) Disease-specific survival, 5-year, n/total n (%) 22/24 (92) 7/7 (100) 9/9 (100) 7/9 (78) 0.31 Disease-specific survival, 10-year, n/total n (%) 8/14 (57) 4/4 (100) 3/4 (75) 1/6 (17) 0.052 Variable All (N = 39) miFTC (n = 12) eaFTC (n = 17) wiFTC (n = 10) P Value a Age at diagnosis, y, mean ± SD 55.2 ± 16.1 48.1 ± 15.6 58.0 ± 15.9 58.9 ± 15.7 0.24 Female sex, n/total n (%) 26/39 (67) 8/12 (67) 12/17 (71) 6/10 (60) 0.91 Tumor diameter, cm, mean ± SD 3.6 ± 1.5 3.0 ± 0.9 3.9 ± 2.0 3.8 ± 0.9 0.29 AJCC stage greater than II, n/total n (%) 3/39 (8) 0/12 (0) 0/17 (0) 3/10 (30) 0.013 (Pmi-ea > 0.99; Pmi-wi = 0.078; Pea-wi = 0.041) Disease recurrence, n/total n (%) 6/39 (15) 0/12 (0) 1/17 (6) 5/9 (56) 0.0017 (Pmi-ea > 0.99; Pmi-wi = 0.0062; Pea-wi = 0.0097) Death due to FTC, n/total n (%) 6/39 (15) 0/12 (0) 1/17 (6) 5/9 (56) 0.0017 (Pmi-ea > 0.99; Pmi-wi = 0.0062; Pea-wi = 0.0097) Disease-specific survival, 5-year, n/total n (%) 22/24 (92) 7/7 (100) 9/9 (100) 7/9 (78) 0.31 Disease-specific survival, 10-year, n/total n (%) 8/14 (57) 4/4 (100) 3/4 (75) 1/6 (17) 0.052 P values <0.05 shown in bold. a Kruskal-Wallis followed by corrected Dunn test for continuous variables and Fisher exact test (with Freeman-Halton extension) for categorical variables. P values in parentheses are pairwise comparisons between miFTC, eaFTC, and wiFTC when significance was found in the overall three-way analysis. View Large Figure 1. View largeDownload slide Clinicopathological and mutational characteristics of FTC. (a) Clinical and pathological parameters of FTCs, grouped by histologic invasion subtype. Disease-specific survival is shown; cases lost to follow-up or who died for another reason were censored. (b) Somatic mutation profiles of recurrent and driver mutations as generated by WES. The top row contains the total number of nonsynonymous variants per tumor, including single nucleotide variants and short insertion/deletion events. Each subsequent row contains the mutations in a specific gene. Bars to the right illustrate prevalence in our cohort. Driver genes were selected based on Vogelstein et al. (41) and supplemented with purported FTC drivers from the three most recent sequencing studies in FTC: BRAF, BRIP1, CNOT1, DICER1, EIF1AX, EZH1, HRAS, IDH1, IGF2BP3, KDM5C, KMT2C, KRAS–MAP4K3, NF1, NRAS, PTEN, SOS1, SPOP, STAG2, TCF12, TP53, and TSHR (13–15). Recurrent mutations were noted in at least two samples, whereas private mutations were identified in only one sample in our cohort. (c) Recurrent arm-level CN events. Cases unable to have CNV analysis performed due to low DNA quality are marked with an “X.” F, female; M, male. Figure 1. View largeDownload slide Clinicopathological and mutational characteristics of FTC. (a) Clinical and pathological parameters of FTCs, grouped by histologic invasion subtype. Disease-specific survival is shown; cases lost to follow-up or who died for another reason were censored. (b) Somatic mutation profiles of recurrent and driver mutations as generated by WES. The top row contains the total number of nonsynonymous variants per tumor, including single nucleotide variants and short insertion/deletion events. Each subsequent row contains the mutations in a specific gene. Bars to the right illustrate prevalence in our cohort. Driver genes were selected based on Vogelstein et al. (41) and supplemented with purported FTC drivers from the three most recent sequencing studies in FTC: BRAF, BRIP1, CNOT1, DICER1, EIF1AX, EZH1, HRAS, IDH1, IGF2BP3, KDM5C, KMT2C, KRAS–MAP4K3, NF1, NRAS, PTEN, SOS1, SPOP, STAG2, TCF12, TP53, and TSHR (13–15). Recurrent mutations were noted in at least two samples, whereas private mutations were identified in only one sample in our cohort. (c) Recurrent arm-level CN events. Cases unable to have CNV analysis performed due to low DNA quality are marked with an “X.” F, female; M, male. WES Thirty-nine FTC samples and their respective matched normal samples passed quality metrics after WES. Mean depth of coverage for tumor and normal samples was 258.6 and 115.4, respectively. The proportion of bases with a minimum of 20 times coverage was 96.9% and 93.1% for tumor and normal samples, respectively, with associated error rates of 0.3% and 0.3%. Mean tumor purity was calculated at 70% and not associated with the number of called somatic mutations (Supplemental Fig. 1). Overall, there was no considerable difference in total or cancer-specific somatic mutational burden across histologic subtypes, though there was a trend toward more mutations in eaFTC and wiFTC (Table 2). Mutations in RAS family genes were noted in 20.5% of samples (8 out of 39), all in codon 61. NRAS mutations were found in 5 out of 39 samples, whereas KRAS and HRAS variants were found in 2 and 1 specimens, respectively, with no difference across histologic categories [Fig. 1(b)]. TSHR mutations were identified in four tumors. Among known cancer- or FTC-specific driver genes, DICER1, EIF1AX, KDM5C, NF1, PRDM1, PTEN, and TP53 were recurrently mutated in two samples each. Although these were all too rare to be distributed in a statistically significant manner, TP53 mutations were only noted in wiFTC (P = 0.06). There was a trend toward more driver gene mutations in the wiFTC cohort. An additional set of recurrently mutated genes in our cohort have not been previously described as drivers, and though two of these (CAMTA1 and SFPQ) are in the Catalogue of Somatic Mutations in Cancer (COSMIC) cancer gene census, the specific mutations identified in our study are not listed in COSMIC (http://cancer.sanger.ac.uk/cosmic). Table 2. Genetic Features of miFTC, eaFTC, and wiFTC Variable All (n = 39) miFTC (n = 12) eaFTC (n = 17) wiFTC (n = 10) P Value a Nonsynonymous somatic variants, mean ± SD 12.5 ± 8.9 8.2 ± 5.3 15.4 ± 10.7 12.8 ± 7.6 0.08 Transition/transversion ratio per sample, mean ± SD 1.3 ± 1.2 1.4 ± 2.0 1.1 ± 0.7 1.5 ± 1.0 0.42 RAS family mutation, n/total n (%) 8/39 (21) 1/12 (8) 5/17 (29) 2/10 (20) 0.37 Cancer driver gene nonsynonymous mutations, mean ± SD 0.8 ± 1.0 0.5 ± 0.5 0.8 ± 0.8 1.2 ± 1.5 0.62 Number of clones, mean ± SD 1.9 ± 0.7 1.8 ± 0.6 1.8 ± 0.6 2.4 ± 0.7 <0.05 (Pmi-ea > 0.99; Pmi-wi = 0.15; Pea-wi = 0.05) Variable All (n = 39) miFTC (n = 12) eaFTC (n = 17) wiFTC (n = 10) P Value a Nonsynonymous somatic variants, mean ± SD 12.5 ± 8.9 8.2 ± 5.3 15.4 ± 10.7 12.8 ± 7.6 0.08 Transition/transversion ratio per sample, mean ± SD 1.3 ± 1.2 1.4 ± 2.0 1.1 ± 0.7 1.5 ± 1.0 0.42 RAS family mutation, n/total n (%) 8/39 (21) 1/12 (8) 5/17 (29) 2/10 (20) 0.37 Cancer driver gene nonsynonymous mutations, mean ± SD 0.8 ± 1.0 0.5 ± 0.5 0.8 ± 0.8 1.2 ± 1.5 0.62 Number of clones, mean ± SD 1.9 ± 0.7 1.8 ± 0.6 1.8 ± 0.6 2.4 ± 0.7 <0.05 (Pmi-ea > 0.99; Pmi-wi = 0.15; Pea-wi = 0.05) P values <0.05 shown in bold. a Kruskal-Wallis test followed by corrected Dunn test for continuous variables or Fisher exact test (with Freeman-Halton extension) for categorical variables. P values in parentheses are pairwise comparisons between miFTC, eaFTC, and wiFTC when significance was found in the overall three-way analysis. View Large Table 2. Genetic Features of miFTC, eaFTC, and wiFTC Variable All (n = 39) miFTC (n = 12) eaFTC (n = 17) wiFTC (n = 10) P Value a Nonsynonymous somatic variants, mean ± SD 12.5 ± 8.9 8.2 ± 5.3 15.4 ± 10.7 12.8 ± 7.6 0.08 Transition/transversion ratio per sample, mean ± SD 1.3 ± 1.2 1.4 ± 2.0 1.1 ± 0.7 1.5 ± 1.0 0.42 RAS family mutation, n/total n (%) 8/39 (21) 1/12 (8) 5/17 (29) 2/10 (20) 0.37 Cancer driver gene nonsynonymous mutations, mean ± SD 0.8 ± 1.0 0.5 ± 0.5 0.8 ± 0.8 1.2 ± 1.5 0.62 Number of clones, mean ± SD 1.9 ± 0.7 1.8 ± 0.6 1.8 ± 0.6 2.4 ± 0.7 <0.05 (Pmi-ea > 0.99; Pmi-wi = 0.15; Pea-wi = 0.05) Variable All (n = 39) miFTC (n = 12) eaFTC (n = 17) wiFTC (n = 10) P Value a Nonsynonymous somatic variants, mean ± SD 12.5 ± 8.9 8.2 ± 5.3 15.4 ± 10.7 12.8 ± 7.6 0.08 Transition/transversion ratio per sample, mean ± SD 1.3 ± 1.2 1.4 ± 2.0 1.1 ± 0.7 1.5 ± 1.0 0.42 RAS family mutation, n/total n (%) 8/39 (21) 1/12 (8) 5/17 (29) 2/10 (20) 0.37 Cancer driver gene nonsynonymous mutations, mean ± SD 0.8 ± 1.0 0.5 ± 0.5 0.8 ± 0.8 1.2 ± 1.5 0.62 Number of clones, mean ± SD 1.9 ± 0.7 1.8 ± 0.6 1.8 ± 0.6 2.4 ± 0.7 <0.05 (Pmi-ea > 0.99; Pmi-wi = 0.15; Pea-wi = 0.05) P values <0.05 shown in bold. a Kruskal-Wallis test followed by corrected Dunn test for continuous variables or Fisher exact test (with Freeman-Halton extension) for categorical variables. P values in parentheses are pairwise comparisons between miFTC, eaFTC, and wiFTC when significance was found in the overall three-way analysis. View Large The range of somatic mutational burden across the cohort was 1 to 44 nonsynonymous variants per tumor (mean 12.5; median 10), with no marked difference in mutational signature between subtypes (Fig. 2; Supplemental Table 2). Although 55 constitutional variants were noted in potential cancer-associated genes, none of these was in BROCA-curated thyroid cancer susceptibility genes (tests.labmed.washington.edu/BROCA) (Supplemental Table 3). There was a higher number of subclones per tumor in the wiFTC category (P = 0.05), though the number of somatic variants in our cohort is low, limiting the reliability of the clonality analysis (Supplemental Fig. 2). Figure 2. View largeDownload slide Somatic mutational burden and signature of FTC. (a) Burden of somatic variants per tumor, displayed according to histological subtype. (b) Base-change spectra of individual tumors. (c) Mutational signatures [as originally described in Alexandrov et al. (33) and curated now by COSMIC] in each subtype of FTC. Signatures were generated for each tumor individually using deconstructSigs and summed to generate the distribution shown for each subtype. “Unknown” represents base changes that could not be mapped to a known mutational signature. Figure 2. View largeDownload slide Somatic mutational burden and signature of FTC. (a) Burden of somatic variants per tumor, displayed according to histological subtype. (b) Base-change spectra of individual tumors. (c) Mutational signatures [as originally described in Alexandrov et al. (33) and curated now by COSMIC] in each subtype of FTC. Signatures were generated for each tumor individually using deconstructSigs and summed to generate the distribution shown for each subtype. “Unknown” represents base changes that could not be mapped to a known mutational signature. CN analysis Overall, FTCs were defined in our study by a general pattern of CN gain, with differences at some loci between the different tumor subtypes. Arm-level CN events were identified across the genome [Fig. 1(c); Supplemental Fig. 3A and 3B]. Although the most common arm-level CN events were gains of 5q, 7p, and 12q, recurrent CN losses at 22q were also identified. The regions with the most substantial CN gains were 7q22.3-36.3 and 12q24.33, whereas 1p36.11 demonstrated CN loss (Supplemental Fig. 3C–3E). The 7q22.3-36.3 region contains 306 genes, including the oncogene PRKAR2B (40). Interestingly, the eaFTC cohort carried recurrent CN gains in region 5q31.3-q35.3, which were not seen in the other tumor types; this region carries the genes CSF1R and NPM1, both of which are implicated as cancer drivers in other studies (41). Survival analysis Invasive status and AJCC stage were significantly associated with disease-specific survival, whereas, surprisingly, tumor size had no statistically noteworthy effect on survival in our cohort [Fig. 3(a)–3(d)]. Higher-than-mean total nonsynonymous mutation burden was associated with worse survival, as was burden of cancer driver gene mutations (41) [Fig. 3(e) and 3(f)], but not mutational burden of COSMIC census genes or clones per tumor [Fig. 3(g) and 3(h)]. Univariate Cox proportional hazards analysis demonstrated that patient age, histologic subtype, AJCC stage, total nonsynonymous mutational burden (as a continuous variable), and number of driver events were all associated with disease-specific survival. Although the number of events in our cohort was too low to support multivariate analysis of all of these factors at once, total mutational burden was found to predict mortality in a limited multivariate analysis that controlled for stage and histologic subtype (Table 3). Figure 3. View largeDownload slide Kaplan-Meier survival curves. Disease-specific survival is plotted against time to event, out to 10 years of follow-up. Tick marks along survival curves indicate cases censored at that time point. Cases were stratified on the basis of: (a) invasion (n = 12 miFTC, 17 eaFTC, and 10 wiFTC); (b) tumor size (n = 30 tumors ≤4 cm and 9 tumors >4 cm); AJCC stage: (c) 7th edition (n = 25 stage I or II, 10 stage III, and 4 stage IV) or (d) 8th edition (n = 36 stage I or II and 3 stage IV); (e) total nonsynonymous mutational burden (n = 24 with <13 variants and 15 with ≥13 variants); (f) cancer driver gene mutations as proposed by Vogelstein et al. (41) (n = 19 with 0 drivers, 17 with 1 driver, and 3 with 2 or more drivers); (g) COSMIC census gene mutations (n = 9 with 0 genes mutated, 11 with 1 gene mutated, 10 with 2 genes mutated, and 9 with 3 or more genes mutated); and (h) subclones per tumor as determined by SciClone (35) (n = 8 with 1 clone, 25 with 2 clones, and 5 with 3 or more clones). Figure 3. View largeDownload slide Kaplan-Meier survival curves. Disease-specific survival is plotted against time to event, out to 10 years of follow-up. Tick marks along survival curves indicate cases censored at that time point. Cases were stratified on the basis of: (a) invasion (n = 12 miFTC, 17 eaFTC, and 10 wiFTC); (b) tumor size (n = 30 tumors ≤4 cm and 9 tumors >4 cm); AJCC stage: (c) 7th edition (n = 25 stage I or II, 10 stage III, and 4 stage IV) or (d) 8th edition (n = 36 stage I or II and 3 stage IV); (e) total nonsynonymous mutational burden (n = 24 with <13 variants and 15 with ≥13 variants); (f) cancer driver gene mutations as proposed by Vogelstein et al. (41) (n = 19 with 0 drivers, 17 with 1 driver, and 3 with 2 or more drivers); (g) COSMIC census gene mutations (n = 9 with 0 genes mutated, 11 with 1 gene mutated, 10 with 2 genes mutated, and 9 with 3 or more genes mutated); and (h) subclones per tumor as determined by SciClone (35) (n = 8 with 1 clone, 25 with 2 clones, and 5 with 3 or more clones). Table 3. Predictors of Disease-Specific Mortality Variable Hazard Ratio 95% CI P Value a Univariate analysis  Age 1.1 1.0–1.2 0.043  Sex, male 1.3 0.24–7.2 0.76  Size, cm 1.7 0.74–3.9 0.21  Subtypeb 8.2 1.1–58.6 0.036  AJCC 8th edition stageb 2.3 1.3–4.0 0.0081  Total nonsynonymous variants 1.1 1.0–1.2 0.013  Clones 1.8 0.67–4.7 0.28  Cancer driver gene mutationsc 2.8 1.4–5.4 0.0018  FTC driver gene mutationsd 3.3 1.2–9.3 0.018  RAS gene mutation 0.53 0.06–4.5 0.54  CN loss of arm 22q 0.97 0.09–11 0.98 Multivariate analysis  Subtypeb 174 0.66–45,546 0.07  AJCC 8th edition stageb 1.8 0.64–5.0 0.27  Total nonsynonymous variants 1.4 1.06–2.0 0.02 Variable Hazard Ratio 95% CI P Value a Univariate analysis  Age 1.1 1.0–1.2 0.043  Sex, male 1.3 0.24–7.2 0.76  Size, cm 1.7 0.74–3.9 0.21  Subtypeb 8.2 1.1–58.6 0.036  AJCC 8th edition stageb 2.3 1.3–4.0 0.0081  Total nonsynonymous variants 1.1 1.0–1.2 0.013  Clones 1.8 0.67–4.7 0.28  Cancer driver gene mutationsc 2.8 1.4–5.4 0.0018  FTC driver gene mutationsd 3.3 1.2–9.3 0.018  RAS gene mutation 0.53 0.06–4.5 0.54  CN loss of arm 22q 0.97 0.09–11 0.98 Multivariate analysis  Subtypeb 174 0.66–45,546 0.07  AJCC 8th edition stageb 1.8 0.64–5.0 0.27  Total nonsynonymous variants 1.4 1.06–2.0 0.02 P values <0.05 shown in bold. a Likelihood ratio test for univariate analysis or Wald test for multivariate analysis. b Per each increment of subtype (miFTC, eaFTC, or wiFTC) or stage (I–IV). c Cancer driver gene mutations as curated by Vogelstein et al. (41). d Drawn from recent exome sequencing studies in FTC: BRAF, BRIP1, CNOT1, DICER1, EIF1AX, EZH1, HRAS, IDH1, IGF2BP3, KDM5C, KMT2C, KRAS, MAP4K3, NF1, NRAS, PTEN, SOS1, SPOP, STAG2, TCF12, TP53, and TSHR (13–15). View Large Table 3. Predictors of Disease-Specific Mortality Variable Hazard Ratio 95% CI P Value a Univariate analysis  Age 1.1 1.0–1.2 0.043  Sex, male 1.3 0.24–7.2 0.76  Size, cm 1.7 0.74–3.9 0.21  Subtypeb 8.2 1.1–58.6 0.036  AJCC 8th edition stageb 2.3 1.3–4.0 0.0081  Total nonsynonymous variants 1.1 1.0–1.2 0.013  Clones 1.8 0.67–4.7 0.28  Cancer driver gene mutationsc 2.8 1.4–5.4 0.0018  FTC driver gene mutationsd 3.3 1.2–9.3 0.018  RAS gene mutation 0.53 0.06–4.5 0.54  CN loss of arm 22q 0.97 0.09–11 0.98 Multivariate analysis  Subtypeb 174 0.66–45,546 0.07  AJCC 8th edition stageb 1.8 0.64–5.0 0.27  Total nonsynonymous variants 1.4 1.06–2.0 0.02 Variable Hazard Ratio 95% CI P Value a Univariate analysis  Age 1.1 1.0–1.2 0.043  Sex, male 1.3 0.24–7.2 0.76  Size, cm 1.7 0.74–3.9 0.21  Subtypeb 8.2 1.1–58.6 0.036  AJCC 8th edition stageb 2.3 1.3–4.0 0.0081  Total nonsynonymous variants 1.1 1.0–1.2 0.013  Clones 1.8 0.67–4.7 0.28  Cancer driver gene mutationsc 2.8 1.4–5.4 0.0018  FTC driver gene mutationsd 3.3 1.2–9.3 0.018  RAS gene mutation 0.53 0.06–4.5 0.54  CN loss of arm 22q 0.97 0.09–11 0.98 Multivariate analysis  Subtypeb 174 0.66–45,546 0.07  AJCC 8th edition stageb 1.8 0.64–5.0 0.27  Total nonsynonymous variants 1.4 1.06–2.0 0.02 P values <0.05 shown in bold. a Likelihood ratio test for univariate analysis or Wald test for multivariate analysis. b Per each increment of subtype (miFTC, eaFTC, or wiFTC) or stage (I–IV). c Cancer driver gene mutations as curated by Vogelstein et al. (41). d Drawn from recent exome sequencing studies in FTC: BRAF, BRIP1, CNOT1, DICER1, EIF1AX, EZH1, HRAS, IDH1, IGF2BP3, KDM5C, KMT2C, KRAS, MAP4K3, NF1, NRAS, PTEN, SOS1, SPOP, STAG2, TCF12, TP53, and TSHR (13–15). View Large Fusion gene analysis Of the available samples tested directly for fusion genes, only one tumor was found to have the PAX8-PPARγ fusion gene (7.7%; 1 out of 13). This widely invasive, stage II tumor was successfully resected, and the patient had 10 years of recurrence-free survival prior to the conclusion of this study. This patient had no other known cancer drivers identified in our study. The observed rate of PAX8-PPARγ fusion in this study is considerably lower than previously reported (20), most likely due to the smaller cohort sizes in each FTC category. As expected for FTC, no RET-PTC1 or RET-PTC3 fusion events were identified in our cohort. The algorithmic approach to fusion gene identification from exome sequencing data did not reveal any previously described FTC fusions or any recurrent fusion events across our cohort. Pathway analyses KEGG pathway analysis demonstrated enrichment of mutations in the MAPK signaling pathway, as expected, and several generic malignancy-associated pathways, but these were not specific to particular categories of invasiveness (Supplemental Table 4). The wiFTC cohort did carry more mutations in several pathways known to be associated with more aggressive malignancies, including cAMP signaling, mitophagy, and longevity regulation. Similarly, protein-protein interaction analysis demonstrated networks of mutations in Ras signaling, cell cycle regulation, and RNA processing, which were not specific to any of the FTC subtypes (Supplemental Fig. 4). Interestingly, each subtype of tumor often carried mutations in different genes within similar networks, hinting at a final common pathway of FTC tumorigenesis. Discussion Follicular thyroid tumors present a unique dilemma for clinicians and patients. Minimally invasive FTCs carry an excellent prognosis with low risk of recurrence or disease-specific mortality, whereas widely invasive tumors are much more aggressive even with maximally intensive treatment (8–10). The new category of eaFTC provides additional granularity in histologic diagnosis, but its effect on prognosis remains unclear. Additionally, the histologic classification requires a surgical specimen in which the capsule can be examined thoroughly, which limits the utility of FNA as a diagnostic tool. Unfortunately, even with the advent of newer molecular markers from thyroid FNA specimens, many patients must still undergo partial or total thyroidectomy to arrive at a definitive diagnosis when faced with a follicular thyroid neoplasm. Although wiFTCs have a distinctly aggressive clinical phenotype, the molecular evolution leading from a normal thyroid follicular cell to wiFTC is incompletely understood. wiFTC is relatively rare, and most next-generation sequencing studies in FTC have analyzed few widely invasive tumors. Furthermore, it is not clear based on currently available evidence whether follicular tumors progress from FTA to wiFTC in a stepwise fashion through miFTC or eaFTC intermediates or whether these subtypes represent distinct entities arising from disparate genetic backgrounds. To determine whether genetic distinctions determine the different histologic categories of FTC as defined in the 2017 WHO guidelines, our group preformed WES analysis of 39 FTC specimens across a spectrum of miFTC, eaFTC, and wiFTC cases. Our cohort is characterized by relatively modest mutational burden [mean 12.5 nonsynonymous variants per tumor or 0.38 per megabase (Mb)], concordant with other recent studies of well-differentiated thyroid cancer (mean 0.41/Mb for PTC or 0.31/Mb in FTC); although the wiFTCs behave aggressively in many cases, they do not seem to carry the large mutational burden described in anaplastic thyroid cancers (mean 2.7/Mb in anaplastic thyroid cancer or 0.72/Mb excluding hypermutators) (12, 13, 42). Mutations within RAS genes were noted in 20.5% of our cohort overall. This frequency of RAS mutations is slightly lower than generally reported in FTC, but within the range reported across prior studies. Mutations in the TSH receptor gene TSHR were common in our cohort, found in 10.3% of samples. TSHR mutations occurred in both the miFTC and wiFTC groups. Mutations in TSHR and the RAS family were mutually exclusive and did not appear to occur at significantly different rates based on invasive status in our cohort. Cancer driver genes EIF1AX and DICER1 carried mutations in two samples each; recurrent mutations in these genes were demonstrated in the recent genomic analyses of FTC (13, 15), and recurrent EIF1AX mutations were noted in a recent WES analysis of anaplastic thyroid carcinoma, the most aggressive thyroid carcinoma (42). In this cohort, one wiFTC sample was found to have a BRAFK601E mutation, without any other FTC driver events. Although somatic BRAFV600E mutations are found in ∼60% of PTCs, they are not historically associated with follicular thyroid tumors (12). This FTC had no nuclear features of PTC, confirming the diagnosis of FTC rather than follicular variant of PTC. Indeed, recent reports have identified BRAF mutations in a small proportion of FTAs (14) and specifically one BRAFK601E mutation in FTC (13). Although only a single driver gene fusion event was identified in this study, WES-indiscernible fusion events such as fusions involving introns playing a driving role in some FTCs cannot be ruled out. Future comprehensive analyses in the field will hopefully identify additional gene fusions in FTC in addition to the well-described PAX8-PPARγ fusion gene and other less common fusion events such as THADA (19, 43). Our cohort was characterized by numerous somatic arm-level copy changes, including recurrent CN loss of 22q. Losses of 22q have also been identified in recent next-generation sequencing studies of FTCs (13, 15). The functional consequences of 22q deletion have not been fully elucidated, though one study has shown that these tumors behave similarly to RAS-family mutated tumors on a transcriptional level (15). The recurrent CN gains identified in 5q in eaFTC and 7q in miFTC and wiFTC may be tumor-driving events in some cases, though the molecular mechanisms have not yet been definitively identified. Survival analysis demonstrated considerable differences in outcome based on AJCC stage and invasion subtype. Interestingly, the AJCC 7th edition staging system seemed to perform better in our cohort than the 8th edition system, likely due to our cohort’s enrichment in more aggressive widely invasive cases, which were downstaged under the new system. The genetics of the tumors also informed the survival analysis, with total mutation burden, cancer driver burden, and FTC driver burden all associated with worse prognosis. Most patients who had recurrence or mortality due to FTC had a more advanced genetic profile as well as more invasive histopathological phenotype, raising the question of which factor is more important. Our multivariate survival analysis suggests that total mutational burden may be a strong prognostic indicator independent of histopathology. Total mutational burden is already in clinical use as a biomarker to guide immunomodulatory treatment in melanoma and other cancers (44). Together with histological classification, the genetic profile of invasive follicular thyroid tumors can be used to make informed treatment decisions for patients facing this diverse group of cancers. In this study, the genetic profile seems to predict survival as a complement to the traditional histological approach, and next-generation sequencing is fast becoming affordable and practical for clinical use, particularly in high-volume centers. Whether similar results could be obtained using less-invasive biopsy specimens remains to be seen, but the approach shows promise of obviating the need for diagnostic surgery for some patients with follicular lesions. Conclusion Overall, our data demonstrate that the 2017 WHO subtypes of invasive FTC arise from comparable genetic backgrounds, in spite of the dramatic differences in clinical outcome. The differences in tumor behavior may be due to time elapsed since starting on a shared pathway toward malignancy, as tumors have time to accumulate more mutations and become more aggressive over time, without a signature genetic event defining invasiveness. Alternatively, the difference may lie in the epigenetics or noncoding genetic profile of the tumor or a background thyroid milieu that encourages the more invasive and aggressive behavior even with similar somatic mutational landscape. Some tumors have an aggressive phenotype in spite of relatively bland mutational profiles, lending further credence to the idea of nongenetic driver events in some cases. In the absence of specific genetic markers to guide prognosis in FTC, clinicians will need to rely on an integrated approach to this disease, incorporating clinical, pathological, and genetic factors to guide treatment decisions. Abbreviations: Abbreviations: AJCC American Joint Committee on Cancer CN copy number CNV copy number variation COSMIC Catalogue of Somatic Mutations in Cancer eaFTC encapsulated angioinvasive follicular thyroid carcinoma FFPE formalin-fixed paraffin-embedded FNA fine-needle aspiration FTA follicular thyroid adenoma FTC follicular thyroid carcinoma KEGG Kyoto Encyclopedia of Genes and Genomes Mb megabase miFTC minimally invasive follicular thyroid carcinoma PTC papillary thyroid carcinoma WES whole-exome sequencing WHO World Health Organization wiFTC widely invasive follicular thyroid carcinoma Acknowledgments The authors thank the Ohse Research Foundation at the Yale School of Medicine, the Stockholm County Council, the Swedish Cancer Society, and the Swedish Society for Medical Research for support of this research. Financial Support: This study was supported by the Yale School of Medicine (to T.C.), Stockholms Läns Landsting (to C.C.J.), Svenska Sällskapet för Medicinsk Forskning (to C.C.J.), Cancerfonden (to C.C.J.), and the Damon Runyon Cancer Research Foundation (to T.C.). Disclosure Summary: The authors have nothing to disclose. References 1. Aschebrook-Kilfoy B , Ward MH , Sabra MM , Devesa SS . Thyroid cancer incidence patterns in the United States by histologic type, 1992-2006 . Thyroid . 2011 ; 21 ( 2 ): 125 – 134 . 2. Enewold L , Zhu K , Ron E , Marrogi AJ , Stojadinovic A , Peoples GE , Devesa SS . Rising thyroid cancer incidence in the United States by demographic and tumor characteristics, 1980-2005 . Cancer Epidemiol Biomarkers Prev . 2009 ; 18 ( 3 ): 784 – 791 . 3. Lim H , Devesa SS , Sosa JA , Check D , Kitahara CM . Trends in thyroid cancer incidence and mortality in the United States, 1974-2013 . JAMA . 2017 ; 317 ( 13 ): 1338 – 1348 . 4. Aschebrook-Kilfoy B , Grogan RH , Ward MH , Kaplan E , Devesa SS . Follicular thyroid cancer incidence patterns in the United States, 1980-2009 . Thyroid . 2013 ; 23 ( 8 ): 1015 – 1021 . 5. James BC , Aschebrook-Kilfoy B , Cipriani N , Kaplan EL , Angelos P , Grogan RH . The incidence and survival of rare cancers of the thyroid, parathyroid, adrenal, and pancreas . Ann Surg Oncol . 2016 ; 23 ( 2 ): 424 – 433 . 6. DeLellis RA , Lloyd RV , Heitz PU , Eng C , eds. Pathology & Genetics of Tumours of Endocrine Organs. World Health Organization Classification of Tumours . Vol 8. 3rd ed . Lyon : IARC Press ; 2004 . 7. Lloyd RV , Osamura RY , Klöppel G , Rosai J , eds. WHO Classification of Tumours of Endocrine Organs . Vol 10. Geneva : WHO Press ; 2017 . 8. D’Avanzo A , Treseler P , Ituarte PH , Wong M , Streja L , Greenspan FS , Siperstein AE , Duh QY , Clark OH . Follicular thyroid carcinoma: histology and prognosis . Cancer . 2004 ; 100 ( 6 ): 1123 – 1129 . 9. Sugino K , Ito K , Nagahama M , Kitagawa W , Shibuya H , Ohkuwa K , Yano Y , Uruno T , Akaishi J , Kameyama K , Ito K . Prognosis and prognostic factors for distant metastases and tumor mortality in follicular thyroid carcinoma . Thyroid . 2011 ; 21 ( 7 ): 751 – 757 . 10. Asari R , Koperek O , Scheuba C , Riss P , Kaserer K , Hoffmann M , Niederle B . Follicular thyroid carcinoma in an iodine-replete endemic goiter region: a prospectively collected, retrospectively analyzed clinical trial . Ann Surg . 2009 ; 249 ( 6 ): 1023 – 1031 . 11. Cipriani NA , Nagar S , Kaplan SP , White MG , Antic T , Sadow PM , Aschebrook-Kilfoy B , Angelos P , Kaplan EL , Grogan RH . Follicular thyroid carcinoma: how have histologic diagnoses changed in the last half-century and what are the prognostic implications ? Thyroid . 2015 ; 25 ( 11 ): 1209 – 1216 . 12. Cancer Genome Atlas Research Network . Integrated genomic characterization of papillary thyroid carcinoma . Cell . 2014 ; 159 ( 3 ): 676 – 690 . 13. Jung SH , Kim MS , Jung CK , Park HC , Kim SY , Liu J , Bae JS , Lee SH , Kim TM , Lee SH , Chung YJ . Mutational burdens and evolutionary ages of thyroid follicular adenoma are comparable to those of follicular carcinoma . Oncotarget . 2016 ; 7 ( 43 ): 69638 – 69648 . 14. Swierniak M , Pfeifer A , Stokowy T , Rusinek D , Chekan M , Lange D , Krajewska J , Oczko-Wojciechowska M , Czarniecka A , Jarzab M , Jarzab B , Wojtas B . Somatic mutation profiling of follicular thyroid cancer by next generation sequencing . Mol Cell Endocrinol . 2016 ; 433 : 130 – 137 . 15. Yoo SK , Lee S , Kim SJ , Jee HG , Kim BA , Cho H , Song YS , Cho SW , Won JK , Shin JY , Park J , Kim JI , Lee KE , Park YJ , Seo JS . Comprehensive analysis of the transcriptional and mutational landscape of follicular and papillary thyroid cancers . PLoS Genet . 2016 ; 12 ( 8 ): e1006239 . 16. Fukahori M , Yoshida A , Hayashi H , Yoshihara M , Matsukuma S , Sakuma Y , Koizume S , Okamoto N , Kondo T , Masuda M , Miyagi Y . The associations between RAS mutations and clinical characteristics in follicular thyroid tumors: new insights from a single center and a large patient cohort . Thyroid . 2012 ; 22 ( 7 ): 683 – 689 . 17. Vuong HG , Kondo T , Oishi N , Nakazawa T , Mochizuki K , Inoue T , Tahara I , Kasai K , Hirokawa M , Tran TM , Katoh R . Genetic alterations of differentiated thyroid carcinoma in iodine-rich and iodine-deficient countries . Cancer Med . 2016 ; 5 ( 8 ): 1883 – 1889 . 18. Garcia-Rostan G , Zhao H , Camp RL , Pollan M , Herrero A , Pardo J , Wu R , Carcangiu ML , Costa J , Tallini G . ras mutations are associated with aggressive tumor phenotypes and poor prognosis in thyroid cancer . J Clin Oncol . 2003 ; 21 ( 17 ): 3226 – 3235 . 19. Nikiforova MN , Biddinger PW , Caudill CM , Kroll TG , Nikiforov YE . PAX8-PPARgamma rearrangement in thyroid tumors: RT-PCR and immunohistochemical analyses . Am J Surg Pathol . 2002 ; 26 ( 8 ): 1016 – 1023 . 20. Boos LA , Dettmer M , Schmitt A , Rudolph T , Steinert H , Moch H , Sobrinho-Simões M , Komminoth P , Perren A . Diagnostic and prognostic implications of the PAX8-PPARγ translocation in thyroid carcinomas-a TMA-based study of 226 cases . Histopathology . 2013 ; 63 ( 2 ): 234 – 241 . 21. Choi M , Scholl UI , Ji W , Liu T , Tikhonova IR , Zumbo P , Nayir A , Bakkaloğlu A , Ozen S , Sanjad S , Nelson-Williams C , Farhi A , Mane S , Lifton RP . Genetic diagnosis by whole exome capture and massively parallel DNA sequencing . Proc Natl Acad Sci USA . 2009 ; 106 ( 45 ): 19096 – 19101 . 22. Li H , Durbin R . Fast and accurate short read alignment with Burrows-Wheeler transform . Bioinformatics . 2009 ; 25 ( 14 ): 1754 – 1760 . 23. Cibulskis K , Lawrence MS , Carter SL , Sivachenko A , Jaffe D , Sougnez C , Gabriel S , Meyerson M , Lander ES , Getz G . Sensitive detection of somatic point mutations in impure and heterogeneous cancer samples . Nat Biotechnol . 2013 ; 31 ( 3 ): 213 – 219 . 24. Costello M , Pugh TJ , Fennell TJ , Stewart C , Lichtenstein L , Meldrim JC , Fostel JL , Friedrich DC , Perrin D , Dionne D , Kim S , Gabriel SB , Lander ES , Fisher S , Getz G . Discovery and characterization of artifactual mutations in deep coverage targeted capture sequencing data due to oxidative DNA damage during sample preparation . Nucleic Acids Res . 2013 ; 41 ( 6 ): e67 . 25. Chen G , Mosier S , Gocke CD , Lin MT , Eshleman JR . Cytosine deamination is a major cause of baseline noise in next-generation sequencing . Mol Diagn Ther . 2014 ; 18 ( 5 ): 587 – 593 . 26. Zhao S , Choi M , Overton JD , Bellone S , Roque DM , Cocco E , Guzzo F , English DP , Varughese J , Gasparrini S , Bortolomai I , Buza N , Hui P , Abu-Khalaf M , Ravaggi A , Bignotti E , Bandiera E , Romani C , Todeschini P , Tassi R , Zanotti L , Carrara L , Pecorelli S , Silasi DA , Ratner E , Azodi M , Schwartz PE , Rutherford TJ , Stiegler AL , Mane S , Boggon TJ , Schlessinger J , Lifton RP , Santin AD . Landscape of somatic single-nucleotide and copy-number mutations in uterine serous carcinoma . Proc Natl Acad Sci USA . 2013 ; 110 ( 8 ): 2916 – 2921 . 27. Sudmant PH , Rausch T , Gardner EJ , Handsaker RE , Abyzov A , Huddleston J , Zhang Y , Ye K , Jun G , Fritz MH , Konkel MK , Malhotra A , Stütz AM , Shi X , Casale FP , Chen J , Hormozdiari F , Dayama G , Chen K , Malig M , Chaisson MJP , Walter K , Meiers S , Kashin S , Garrison E , Auton A , Lam HYK , Mu XJ , Alkan C , Antaki D , Bae T , Cerveira E , Chines P , Chong Z , Clarke L , Dal E , Ding L , Emery S , Fan X , Gujral M , Kahveci F , Kidd JM , Kong Y , Lameijer EW , McCarthy S , Flicek P , Gibbs RA , Marth G , Mason CE , Menelaou A , Muzny DM , Nelson BJ , Noor A , Parrish NF , Pendleton M , Quitadamo A , Raeder B , Schadt EE , Romanovitch M , Schlattl A , Sebra R , Shabalin AA , Untergasser A , Walker JA , Wang M , Yu F , Zhang C , Zhang J , Zheng-Bradley X , Zhou W , Zichner T , Sebat J , Batzer MA , McCarroll SA , Mills RE , Gerstein MB , Bashir A , Stegle O , Devine SE , Lee C , Eichler EE , Korbel JO ; 1000 Genomes Project Consortium . An integrated map of structural variation in 2,504 human genomes . Nature . 2015 ; 526 ( 7571 ): 75 – 81 . 28. Auton A , Brooks LD , Durbin RM , Garrison EP , Kang HM , Korbel JO , Marchini JL , McCarthy S , McVean GA , Abecasis GR ; 1000 Genomes Project Consortium . A global reference for human genetic variation . Nature . 2015 ; 526 ( 7571 ): 68 – 74 . 29. Magi A , Tattini L , Cifola I , D’Aurizio R , Benelli M , Mangano E , Battaglia C , Bonora E , Kurg A , Seri M , Magini P , Giusti B , Romeo G , Pippucci T , De Bellis G , Abbate R , Gensini GF . EXCAVATOR: detecting copy number variants from whole-exome sequencing data . Genome Biol . 2013 ; 14 ( 10 ): R120 . 30. Mermel CH , Schumacher SE , Hill B , Meyerson ML , Beroukhim R , Getz G . GISTIC2.0 facilitates sensitive and confident localization of the targets of focal somatic copy-number alteration in human cancers . Genome Biol . 2011 ; 12 ( 4 ): R41 . 31. Chen X , Schulz-Trieglaff O , Shaw R , Barnes B , Schlesinger F , Källberg M , Cox AJ , Kruglyak S , Saunders CT . Manta: rapid detection of structural variants and indels for germline and cancer sequencing applications . Bioinformatics . 2016 ; 32 ( 8 ): 1220 – 1222 . 32. Rausch T , Zichner T , Schlattl A , Stütz AM , Benes V , Korbel JO . DELLY: structural variant discovery by integrated paired-end and split-read analysis . Bioinformatics . 2012 ; 28 ( 18 ): i333 – i339 . 33. Alexandrov LB , Nik-Zainal S , Wedge DC , Aparicio SA , Behjati S , Biankin AV , Bignell GR , Bolli N , Borg A , Børresen-Dale AL , Boyault S , Burkhardt B , Butler AP , Caldas C , Davies HR , Desmedt C , Eils R , Eyfjörd JE , Foekens JA , Greaves M , Hosoda F , Hutter B , Ilicic T , Imbeaud S , Imielinski M , Jäger N , Jones DT , Jones D , Knappskog S , Kool M , Lakhani SR , López-Otín C , Martin S , Munshi NC , Nakamura H , Northcott PA , Pajic M , Papaemmanuil E , Paradiso A , Pearson JV , Puente XS , Raine K , Ramakrishna M , Richardson AL , Richter J , Rosenstiel P , Schlesner M , Schumacher TN , Span PN , Teague JW , Totoki Y , Tutt AN , Valdés-Mas R , van Buuren MM , van ’t Veer L , Vincent-Salomon A , Waddell N , Yates LR , Zucman-Rossi J , Futreal PA , McDermott U , Lichter P , Meyerson M , Grimmond SM , Siebert R , Campo E , Shibata T , Pfister SM , Campbell PJ , Stratton MR ; Australian Pancreatic Cancer Genome Initiative; ICGC Breast Cancer Consortium; ICGC MMML-Seq Consortium; ICGC PedBrain . Signatures of mutational processes in human cancer [published correction appears in Nature. 2013;502(7470):258]. Nature . 2013 ; 500 ( 7463 ): 415 – 421 . 34. Rosenthal R , McGranahan N , Herrero J , Taylor BS , Swanton C . DeconstructSigs: delineating mutational processes in single tumors distinguishes DNA repair deficiencies and patterns of carcinoma evolution . Genome Biol . 2016 ; 17 ( 1 ): 31 . 35. Miller CA , White BS , Dees ND , Griffith M , Welch JS , Griffith OL , Vij R , Tomasson MH , Graubert TA , Walter MJ , Ellis MJ , Schierding W , DiPersio JF , Ley TJ , Mardis ER , Wilson RK , Ding L . SciClone: inferring clonal architecture and tracking the spatial and temporal patterns of tumor evolution . PLOS Comput Biol . 2014 ; 10 ( 8 ): e1003665 . 36. Szklarczyk D , Morris JH , Cook H , Kuhn M , Wyder S , Simonovic M , Santos A , Doncheva NT , Roth A , Bork P , Jensen LJ , von Mering C . The STRING database in 2017: quality-controlled protein-protein association networks, made broadly accessible . Nucleic Acids Res . 2017 ; 45 ( D1 ): D362 – D368 . 37. Kanehisa M , Goto S , Sato Y , Furumichi M , Tanabe M . KEGG for integration and interpretation of large-scale molecular data sets . Nucleic Acids Res . 2012 ; 40 ( Database issue ): D109 – D114 . 38. Tuttle RM , Haugen B , Perrier ND . Updated American Joint Committee on Cancer/tumor-node-metastasis staging system for differentiated and anaplastic thyroid cancer (eighth edition): what changed and why ? Thyroid 2017 ; 27 ( 6 ): 751 – 756 39. R: A language and environment for statistical computing [computer program]. Vienna, Austria: R Foundation for Statistical Computing: 2017. 40. Sha J , Xue W , Dong B , Pan J , Wu X , Li D , Liu D , Huang Y . PRKAR2B plays an oncogenic role in the castration-resistant prostate cancer . Oncotarget . 2017 ; 8 ( 4 ): 6114 – 6129 . 41. Vogelstein B , Papadopoulos N , Velculescu VE , Zhou S , Diaz LA Jr , Kinzler KW . Cancer genome landscapes . Science . 2013 ; 339 ( 6127 ): 1546 – 1558 . 42. Kunstman JW , Juhlin CC , Goh G , Brown TC , Stenman A , Healy JM , Rubinstein JC , Choi M , Kiss N , Nelson-Williams C , Mane S , Rimm DL , Prasad ML , Höög A , Zedenius J , Larsson C , Korah R , Lifton RP , Carling T . Characterization of the mutational landscape of anaplastic thyroid cancer via whole-exome sequencing . Hum Mol Genet . 2015 ; 24 ( 8 ): 2318 – 2329 . 43. Panebianco F , Kelly LM , Liu P , Zhong S , Dacic S , Wang X , Singhi AD , Dhir R , Chiosea SI , Kuan SF , Bhargava R , Dabbs D , Trivedi S , Gandhi M , Diaz R , Wald AI , Carty SE , Ferris RL , Lee AV , Nikiforova MN , Nikiforov YE . THADA fusion is a mechanism of IGF2BP3 activation and IGF1R signaling in thyroid cancer . Proc Natl Acad Sci USA . 2017 ; 114 ( 9 ): 2307 – 2312 . 44. Khagi Y , Goodman AM , Daniels GA , Patel SP , Sacco AG , Randall JM , Bazhenova LA , Kurzrock R . Hypermutated circulating tumor DNA: correlation with response to checkpoint inhibitor-based immunotherapy . Clin Cancer Res . 2017 ; 23 ( 19 ): 5729 – 5736 . Copyright © 2018 Endocrine Society

Journal

Journal of Clinical Endocrinology and MetabolismOxford University Press

Published: May 2, 2018

There are no references for this article.

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off