Characterization of Driver Mutations in Anaplastic Thyroid Carcinoma Identifies RAS and PIK3CA Mutations as Negative Survival Predictors

Anaplastic thyroid carcinoma (ATC) is rare but highly aggressive. We investigated the association of selected driver mutations, including BRAF, RAS, PIK3CA, TERT promoter, TP53, POLE, and mismatch repair deficiency (MMR-D) with the clinicopathological features of ATC to identify prognostic and predictive biomarkers. Thirty-nine retrospective cases from pathology archives were enrolled for clinicopathology analysis and immunohistochemistry, and 27 cases had sufficient specimens for further molecular testing using targeted next-generation sequencing and mass spectrometry. BRAFV600E and RAS mutations were identified in 25.9% and 40.7% of ATC, respectively. BRAFV600E mutation was significantly associated with coexisting papillary thyroid carcinoma (p = 0.009) and RAS mutations with female gender (p = 0.012). In univariant analysis, the non-BRAF/RAS tumors were significantly associated with the presence of a sarcomatoid pattern (p = 0.045). PIK3CA, TERT promoter, and TP53 mutations were identified in 14.8%, 81.5%, and 70.4% of cases, respectively. No MMR-D or POLE mutations were detected. In survival analyses, RAS and PIK3CA mutations were significantly associated with inferior outcomes (p = 0.03 and p = 0.006, respectively). In conclusion, driver mutations in ATC are associated with distinct clinicopathological features. RAS and PIK3CA mutations were negative predictors for patient survival. Emerging therapeutic agents targeting BRAF, RAS, and PI3 kinase may benefit a substantial proportion of ATC patients.


Introduction
Thyroid carcinoma is the most common malignancy of the endocrine system [1]. Anaplastic thyroid carcinoma (ATC) is an aggressive tumor comprising 1%~2% of thyroid carcinoma [2]. Most patients present with advanced disease at the initial clinical encounter, and up to 90% of them die within a year after diagnosis [3,4]. ATC is managed by a multi-modality treatment approach combining surgical resection and chemoradiation. However, the effectiveness of the current treatment approach is limited [2]. As such, identification of actionable treatment targets, as well as biological predictors for clinical behavior in ATCs, is of value to improve clinical outcomes.

Molecular Profiles of ATC
Twenty-seven ATC cases with sufficient formalin-fixed paraffin-embedded (FFPE) materials for DNA extraction were subjected to targeted next-generation sequencing (NGS) and mass spectrometry, and the results are summarized (Figure 2, Supplementary Data 1 and Data 2). BRAF V600E (c.T1799A) was identified in 25.9% of the cases. All ATC with BRAF V600E mutation detected by NGS were positive for the BRAF VE1 immunohistochemistry. RAS-activating mutations at codon 12, 13, and 61 were analyzed. We found overall RAS mutations in 40.7% of the cases, and among the 11 cases, 8 were found in N-RAS (Q61R in 6, Q61K in 1, and G12D in 1) and 3 in K-RAS (Q61R in 2 and G12A in 1). H-RAS mutation was not identified. To investigate the potential association between driver mutations and clinicopathological features, the 27 cases were stratified into three molecular groups: BRAF-mutated, RAS-mutated, and non-BRAF/RAS group (Table 3). We found that RAS mutation was associated with female gender (90.9% versus 44.4% and 28.6% in non-BRAF/RAS and BRAF groups respectively, p = 0.012). The non-BRAF/RAS group was associated with the presence of a sarcomatoid component morphologically (77.8% versus 45.5% in RAS and 14.3% in BRAF groups, respectively, p = 0.045). Association between RAS mutations and female gender remains significant (odds ratio 20.31, p = 0.022, supplementary Table S1) in multivariate analysis, while the non-BRAF/RAS group becomes marginally associated with the presence of a sarcomatoid pattern (odds ratio 23.17, p = 0.089, Supplementary Table S2). TERT promoter C288T mutation was present in 81.5% of cases, while C250T was not detected in any of the cases. TP53 missense mutation was found in 70.4% of ATC. PIK3CA mutation was present in 14.8% of ATC, including one E542K, one E545K, and two H1047R mutations. Mismatch repair deficiency and POLE exonuclease domain mutation potentially led to a hypermutation phenotype in tumors. We found that mismatch repair (MMR) protein expression was retained in all of the cases ( Table 2). We did not identify mutations in the exonuclease domain of POLE in any of the ATC.

Molecular Profiles of ATC
Twenty-seven ATC cases with sufficient formalin-fixed paraffin-embedded (FFPE) materials for DNA extraction were subjected to targeted next-generation sequencing (NGS) and mass spectrometry, and the results are summarized (Figure 2, Supplementary Data 1 and Data 2). BRAF V600E (c.T1799A) was identified in 25.9% of the cases. All ATC with BRAF V600E mutation detected by NGS were positive for the BRAF VE1 immunohistochemistry. RAS-activating mutations at codon 12, 13, and 61 were analyzed. We found overall RAS mutations in 40.7% of the cases, and among the 11 cases, 8 were found in N-RAS (Q61R in 6, Q61K in 1, and G12D in 1) and 3 in K-RAS (Q61R in 2 and G12A in 1). H-RAS mutation was not identified. To investigate the potential association between driver mutations and clinicopathological features, the 27 cases were stratified into three molecular groups: BRAF-mutated, RAS-mutated, and non-BRAF/RAS group (Table 3). We found that RAS mutation was associated with female gender (90.9% versus 44.4% and 28.6% in non-BRAF/RAS and BRAF groups respectively, p = 0.012). The non-BRAF/RAS group was associated with the presence of a sarcomatoid component morphologically (77.8% versus 45.5% in RAS and 14.3% in BRAF groups, respectively, p = 0.045). Association between RAS mutations and female gender remains significant (odds ratio 20.31, p = 0.022, supplementary Table S1) in multivariate analysis, while the non-BRAF/RAS group becomes marginally associated with the presence of a sarcomatoid pattern (odds ratio 23.17, p = 0.089, Supplementary Table S2). TERT promoter C288T mutation was present in 81.5% of cases, while C250T was not detected in any of the cases. TP53 missense mutation was found in 70.4% of ATC. PIK3CA mutation was present in 14.8% of ATC, including one E542K, one E545K, and two H1047R mutations. Mismatch repair deficiency and POLE exonuclease domain mutation potentially led to a hypermutation phenotype in tumors. We found that mismatch repair (MMR) protein expression was retained in all of the cases ( Table 2). We did not identify mutations in the exonuclease domain of POLE in any of the ATC.

Survival Analysis of ATC Cohort
Follow-up data of all cases was obtained, and the median survival interval was 54 days (1-1079). We constructed Kaplan-Meier survival curves with stratification on molecular alternations including BRAF V600E , RAS, TERT promoter, TP53, and PIK3CA mutations (Figure 3). RAS and PIK3CA mutations were significantly associated with an inferior survival (median 94 versus 41 days, p = 0.03 and median 94 versus 29 days, p = 0.006 respectively). None of the other mutations correlate with patient survival.

Survival Analysis of ATC Cohort
Follow-up data of all cases was obtained, and the median survival interval was 54 days (1-1079). We constructed Kaplan-Meier survival curves with stratification on molecular alternations including BRAF V600E , RAS, TERT promoter, TP53, and PIK3CA mutations (Figure 3). RAS and PIK3CA mutations were significantly associated with an inferior survival (median 94 versus 41 days, p = 0.03 and median 94 versus 29 days, p = 0.006 respectively). None of the other mutations correlate with patient survival. , BRAF V600E mutation (c), TERT promoter mutation (d), and TP53 non-synonymous mutations (e). Survival differences were analyzed by log-rank test with significance at p value < 0.05 (bold type). WT, wild-type; pTERT, TERT promoter.

Discussion
In the current study, we analyzed a cohort of ATC and identified RAS mutations in 40.7% and BRAF V600E mutation in 25.9% of ATC. The two mutations were mutually exclusive and detected in up to 70% of ATC. ATC stratified by BRAF V600E and RAS mutations showed distinct clinicopathological features. BRAF V600E mutation was significantly associated with coexisting PTC. RAS-mutated ATC was positively associated with female gender and an inferior survival outcome. Non-RAS/BRAF tumors tended to show a sarcomatoid morphological pattern. PIK3CA mutations were found in 14.8% of ATC and were also identified as a negative prognostic indicator. TERT promoter and TP53 mutations were frequent mutations in ATC, identified in 81.5% and 70.4% of the tumors respectively. MMR deficiency and POLE exonuclease domain mutation were not detected in our cohort. Except for RAS and PIK3CA mutations, none of the other molecular changes investigated in this study showed prognostic significance.
Genetic alternation rates of selected genes in ATC reported in the literature are summarized in Table 4. RAS mutation rate in ATC was reported ranging from 18% to 44.4%. In our cohort, N-RAS Survival differences were analyzed by log-rank test with significance at p value < 0.05 (bold type). WT, wild-type; pTERT, TERT promoter.

Discussion
In the current study, we analyzed a cohort of ATC and identified RAS mutations in 40.7% and BRAF V600E mutation in 25.9% of ATC. The two mutations were mutually exclusive and detected in up to 70% of ATC. ATC stratified by BRAF V600E and RAS mutations showed distinct clinicopathological features. BRAF V600E mutation was significantly associated with coexisting PTC. RAS-mutated ATC was positively associated with female gender and an inferior survival outcome. Non-RAS/BRAF tumors tended to show a sarcomatoid morphological pattern. PIK3CA mutations were found in 14.8% of ATC and were also identified as a negative prognostic indicator. TERT promoter and TP53 mutations were frequent mutations in ATC, identified in 81.5% and 70.4% of the tumors respectively. MMR deficiency and POLE exonuclease domain mutation were not detected in our cohort. Except for RAS and PIK3CA mutations, none of the other molecular changes investigated in this study showed prognostic significance.
Genetic alternation rates of selected genes in ATC reported in the literature are summarized in Table 4. RAS mutation rate in ATC was reported ranging from 18% to 44.4%. In our cohort, N-RAS mutation was the most frequent (eight cases, 29.6%) among the three RAS subtypes, concordant with findings in previous reports [9,13,15,16]. K-RAS mutation was present in three cases (11.1%), while H-RAS mutation was not detected. Mutations at codon 61 were the most common, followed by codon 12. RAS-mutated ATC emerged with distinct clinicopathological features. RAS mutations were strongly associated with female gender in both univariate and multivariate analyses (p = 0.012 and 0.022) and showed significantly poor survival outcomes (p = 0.03). It has been reported that the co-occurrence of RAS and TERT promoter mutations was associated with an inferior survival outcome [8,18]. Our study was the first one to demonstrate that RAS mutations alone predicted a significantly poor outcome. On the other hand, BRAF V600E mutation rate in ATC was reported ranging from 12.8% to 56% (Table 4). In one Korean report where PTC is prevalent, the BRAF V600E mutation in ATC was up to 91% [14]. In our cohort, BRAF V600E mutation was also strongly associated with a coexisting PTC component (p = 0.009). In the 10 ATC with coexisting PTC, up to 60% harbored BRAF V600E mutation. Therapies targeting BRAF V600E are being utilized for the treatment of ATC. In BRAF V600E -mutated ATC, combination treatment with Dabrafenib and Trametinib has been proven robust as an adjuvant therapy to improve clinical outcomes and a neoadjuvant treatment to facilitate complete surgical resection for local control [24,25]. Direct inhibitors of RAS have not yet achieved clinical success. Recently, small molecules targeting KRAS G12C have entered phase 1 and 2 clinical trials for the treatment of colorectal carcinoma [26]. Targeting RAS may become a potential strategy in the management of advanced malignancy. Among the 27 cases subjected to NGS studies, nine cases were absent of either RAS or BRAF V600E mutations. Whether these tumors harbor other rare targetable driver fusion events found in WDTC, such as RET-PTC, PAX8-PPARG, NTRK, and ALK rearrangement, requires further analysis [27].
Mutation rates of PIK3CA, TERT promoter, and TP53 are higher in ATC as compared to WDTC [19]. These gene mutations often co-occur with BRAF V600E and RAS mutations and are thought to be acquired during thyroid cancer de-differentiation [19]. In the 18 cases with the driver BRAF V600E or RAS mutations, only one was absent of all PIK3CA, TERT promoter, and TP53 mutations. PIK3CA mutation rates were reported ranging from 0% to 44% in the literature (Table 4). We identified PIK3CA mutations in 14.8% of our cohort, with two cases co-occurring with BRAF V600E mutation and one with RAS mutation. Moreover, we identified PIK3CA mutation significantly associated with an inferior survival outcome (p = 0.006). Duan et al. also reported PIK3CA as a negative predictor of patient survival in ATC, and co-occurrence of PIK3CA and TERT promoter mutations had a higher risk of mortality than patients with only one of either mutation [10]. PI3 kinase pathway and MAP kinase pathway are divergent pathways downstream of RAS signaling. Pre-clinical studies had identified combined treatment with MEK and PI3 kinase inhibitors synergized to target BRAF V600E and PIK3CA double-mutant ATC cell lines and genetically engineered mice [28]. PI3 kinase inhibitors have also been evaluated in clinical trials for breast, colorectal, and lung cancers [29]. Whether targeting PI3 kinase serves as an effective treatment strategy for ATC remains to be elucidated. Mutations of TERT promoter and TP53 are the genetic hallmark of ATC [2,12,20]. In ATC, the rate of TERT promoter and TP53-inactivating mutations ranged from 32% to 75% and 25% to 65.6%, respectively (Table 4). Moreover, TERT promoter mutation had been recurrently reported to associate with shorter survival [8,10,18]. Xu et al. and Duan et al. both showed that TERT promoter mutation alone served as a negative prognostic indicator [8,10]. Xu et al. further found that patient outcome was particularly grave in tumors with co-occurrence of TERT promoter mutation with RAS or BRAF V600E mutation [8]. Yet no prognostic effect of TERT promoter mutation could be observed in our analysis. This could be attributed to a high mutation rate of TERT promoter in our cohort.
Previous studies reported an increase in somatic mutation numbers in ATC [15,18]. In colorectal and endometrial carcinoma, it was well demonstrated that MMR-D or POLE exonuclease domain mutation could lead to a marked increase of somatic mutations in tumors and predict response to anti-PD-L1 treatment [30][31][32]. Prior studies demonstrated MMR-D in up to 13.6% of ATC [8,15,22,33], and POLE missense mutation in 4% and 9% ( Table 4). The mutation sites of POLE were not specified in previous studies. In our cohort, all ATC had preserved MMR protein expression, and we did not detect any non-synonymous mutations specifically in the POLE exonuclease domain. Our findings along with previous reports suggested that MMR-D and POLE exonuclease domain mutations were rare in ATC and had limited implications regarding anti-PD-L1 therapy.
In this study, we identified distinct clinicopathological features in ATC stratified by BRAF V600E and RAS mutations. BRAF V600E mutation was significantly associated with coexisting PTC, while RAS-mutated tumors were strongly associated with female gender. Of note, RAS and PIK3CA mutations were significantly associated with worse prognostic outcomes in ATC. Genetic hallmarks including TERT promoter and TP53 mutations were prevalent, while MMR-D or POLE exonuclease domain mutations were not identified in our ATC cohort. The molecular profile of these selected driver mutations has clinical implications regarding the emerging targeted therapies in this particularly aggressive cancer.

Case Selection
This study was carried out in accordance to the Declaration of Helsinki and approved by the Institutional Review Board of Taipei Veterans General hospital, which granted exemption of informed consent for tissue procurement through the Biobank of Taipei Veterans General Hospital after an unlinked anonymous process (IRB no.: 2018-06-008BC, June 7th, 2019; Biobank no.: 10707). A total of 39 ATC cases from four participating hospitals in the period between 1997 and 2018 were enrolled. All pathology slides were reviewed, and the diagnosis of ATC in each case was confirmed according to the 2017 WHO Classification of Tumors of Endocrine Organs [35]. Histologic patterns, including epithelial, giant cell, and sarcomatoid, as well as the presence of coexisting differentiated carcinoma component, were recorded. Clinical information including age, gender, stage at presentation, specimen obtaining procedure, treatment modality, and follow-up information was obtained from the biobank database.

Construction of Tissue Microarray and Immunohistochemistry
For each case, representative formalin-fixed paraffin-embedded (FFPE) ATC tissue was selected, and two 2-mm cores were obtained for tissue microarray construction. Four-µm tissue microarray sections were used for H&E and immunohistochemistry. Immunohistochemical stains for PAX8 incubation with primary antibody followed by secondary antibody, and chromogen visualization. For PAX8 and TTF-1, positive staining was defined as the presence of positive nuclear staining in more than 5% of ATC tumor cells. For BRAF VE1, positive staining was defined as diffuse cytoplasmic staining of moderate to strong intensity. Loss of MMR protein expression was defined as complete absence of nuclear staining in the entire tumor area with lymphocytes and vascular endothelial cells serving as positive internal control.

DNA Extraction
Cases with sufficient FFPE tissue were subjected to molecular testing. Representative FFPE blocks were used for DNA extraction. Four to six 10 µm-sections on glass slides were used for each case based on tissue size. DNA extraction was performed using QIAamp DNA FFPE tissue kit (Qiagen, Germantown, MD, USA) based on the manufacturer's instruction. In brief, the tumor region on section was applied with lysis buffer, scraped off the glass slide, and collected into a microcentrifuge tube in a total volume of 180 µL. The sample was then incubated with 20 µL proteinase K at 56 • C for 1 h and then mixed with 200 µL buffer AL and 200 µL ethanol. The mixture was then transferred to the MinElute column, centrifuged to remove the buffer, and passed sequentially by buffer AW1 and AW2. Finally, 60 µL of buffer ATE was added to the column and the eluate containing extracted DNA was collected.

Targeted Next-Generation Sequencing (NGS)
We performed targeted NGS panel to identify mutation at BRAF exon 15 (NM_004333.4), RAS codon 12, 13, and 61 (H-RAS NM_005343.2, N-RAS NM_002524.4, and K-RAS NM_033360.2) [36], POLE exonuclease domain encompassing codons 268 to 471 (NM_006231.2) [37], TERT promoter (NC_000005.9) [12], and TP53 whole coding region (NM_000546.5). The amount of 250 ng tumor genomic DNA of each individual was used for library construction of the targeted gene regions using QIAseq target DNA system (Qiagen). Briefly, DNA was enzymatically fragmented and end-repaired in a 25 µL reaction containing 2.5 µL 10× fragmentation buffer and 5 µL fragmentation enzyme mix. The reaction was carried out at 4 • C for 1 min, 32 • C for 24 min, and 65 • C for 30 min. Immediately after the reaction, 10 µL 5× ligation buffer, 5 µL DNA ligase, 2.8 µL 25 µM barcoded adapters, and water were added to reach a total volume at 50 µL. The reaction was incubated at 20 • C for 15 min. To ensure complete removal of free barcoded adapters, each reaction was purified for two rounds using beads. To amplify the barcoded DNA fragments, the purified DNA was then mixed with 10 nM each of target primer (a total of 72 target specific primers), 400 nM Forward primer, 1× PCR buffer, and 0.8 µL HotStarTaq DNA polymerase in a 20 µL reaction. The PCR was carried with the following conditions: 95 • C for 13 min, 98 • C for 2 min; six cycles of 98 • C for 15 s and 65 • C for 15 min; 72 • C for 5 min. The reaction was then incubated with beads to remove remaining excess primers. The DNA product was subjected to another round of amplification by mixing with 400 nM Universal primer, 400 nM Index primer, 1X PCR buffer, and 1 µL HotStarTaq DNA polymerase in a volume of 20 µL, with the following PCR conditions: 95 • C for 13 min, 98 • C for 2 min, 20 cycles of 98 • C for 15 s and 60 • C for 2 min; 72 • C for 5 min. The DNA library was purified by incubation with beads, and the final pool subjected to targeted sequencing using Illumina MiniSeq sequencer (Illumina, San Diego, CA, USA). The raw outputs of each case were>16 Mb, and the average depth of target regions was more than 500 folds. The sequence of each read was trimmed based on the quality score (Q30), and reads with length less than 45 bp were discarded in subsequent analysis. Reads were aligned to the human hg19 reference genome using BWA-MEM [38]. GATK Unified Genotyper was used for variant calling [39]. Variant Effect Predictor was used to annotating the variants [40]. Variants with an allelic frequency ≤5% were excluded. Only non-synonymous mutations reported in the Sanger Institute Catalogue were taken into account.

Mass Spectrometry for PI3KCA Mutations
Analysis of hot spot somatic mutations of PI3KCA in exon 9 and exon 20 (NM_006218.4) [41] was performed using Agena MassARRAY platform with iPLEX reagent chemistry (Agena, San Diego, CA, USA). The genotypes of the codons assayed are listed in Supplementary Table S3. The specific PCR primer and extension primer sequences were designed with Assay Designer software package (v.4.0). Multiplex PCR reaction was carried by mixing 10 ng genomic DNA sample, 1 U Taq polymerase, 500 nmol each of PCR primers, and 2.5 mM dNTP in 5 µL reaction volume and amplified in a thermocycler for 45 cycles. Unincorporated dNTPs were deactivated using 0.3 U of shrimp alkaline phosphatase. Single base extension reaction was performed by adding a mix of specific extension primers, dideoxynucleotide terminators, and extension enzyme, followed by thermocycling according to manufacturer's instruction. The specific extension primers anneal directly to the mutation sites, and were extended and terminated by the dideoxynucleotide terminators complementary to the target somatic mutational site. The reaction was purified by incubation with cation exchange resin, and 7 µL of the purified product was loaded onto a matrix pad of a SpectroCHIP where the analyte crystalizes with a pre-spotted MALDI matrix. SpectroCHIPs were loaded in MassARRAY Analyzer 4 where ionization and signal detection were carried, and base calling reported by clustering analysis with TYPER 4.0 software.

Statistical Analysis
Categorical data were analyzed using Fisher exact test and logistic regression for univariate and multivariate analyses. Survival curves were built using the Kaplan-Meier method, and log-rank test was applied to test for differences between the variables. The p-values were two-sided, and a value <0.05 was considered significant. All analyses were performed using R software (version 3.5.1, R Foundation for Statistical Computing, Vienna, Austria).

Conclusions
This study illustrates clinicopathological features in ATC stratified by BRAF V600E and RAS mutations. Importantly, RAS and PIK3CA mutations were found to be negative prognostic indicators in ATC. In concordance with previous reports, we demonstrated that genetic hallmarks in ATC including TERT promoter and TP53 mutations were prevalent in ATC. MMR-D or POLE exonuclease domain mutations were not identified in our study cohort. Acknowledging molecular profiles of ATC may have clinical implications regarding emerging targeted therapies, including BRAF, RAS, and PI3 kinase inhibitors.
Supplementary Materials: The following are available online at http://www.mdpi.com/2072-6694/12/7/1973/s1, Supplementary data 1: Mutational profile of anaplastic thyroid carcinoma cases detected by next-generation sequencing, Supplemental data 2: PIK3CA mutations in anaplastic thyroid carcinoma cases detected by mass spectrometry, Table S1: Clinicopathological features of RAS-mutated and RAS wild-type ATC cases, Table S2: Clinicopathological features of ATC cases wild type for RAS and BRAFV600E and cases with either one mutation, Table S3: List of PIK3CA mass spectrometry assay panel.