Additional Oncogenic Alterations in RAS-Driven Differentiated Thyroid Cancers Associate with Worse Clinicopathologic Outcomes

This article is featured in Highlights of This Issue, p. 2559

Thyroid cancer is the most common endocrine malignancy, and its incidence has been increasing over the last decades. Although prognosis remains excellent for the majority, there is a small percentage of patients that present with aggressive disease with increased risk of recurrence, and even increased mortality.

Our understanding of the genomic basis of thyroid cancer has expanded over the last decade, especially following the landmark genomic study by The Cancer Genome Atlas (1). The RAS oncogenes (NRAS>HRAS>KRAS) have a well-established role in human thyroid tumorigenesis (2), and are the second most prevalent alterations in thyroid tumors, following BRAFV600E. Mutations in RAS genes are initiating events that drive thyroid cancer transformation via constitutive activation of the MAPK pathway, but their oncogenic potential is lower than BRAFV600E alterations (3). Consequently, RAS mutations are detected amongst the whole spectrum of follicular-derived thyroid neoplasms, from follicular adenomas (20%–25%) to thyroid cancers [30%–45% of follicular thyroid cancer (FTC) and follicular variant of papillary thyroid cancer (FVPTC)] to more aggressive thyroid malignancies [20–40% poorly differentiated thyroid cancer (PDTC), and 10–20% anaplastic thyroid cancer (ATC)]. RAS mutation analysis has been investigated extensively for diagnostic and prognostic molecular use (4). However, given the heterogeneity described above, their potential utility as biomarkers has so far been limited (5).

The genomic and transcriptomic features of more advanced types of thyroid cancer, such as ATC and PDTC, suggest that the gradual accumulation of key oncogenic alterations leads to transformation from indolent lesions to more aggressive cancers. A prime example is TERT promoter mutations which have been shown to confer worse clinicopathologic outcomes when discovered in addition to BRAFV600E and RAS-mutated papillary thyroid cancers (PTC; ref. 6). Recently, our group has shown that mutations in the PI3K pathway are independently associated with disease-specific mortality in BRAFV600E-mutated PTCs (7). To our knowledge, no studies have comprehensively examined the prognostic consequences of an extensive list of additional oncogenic mutations in RAS-mutated thyroid cancers.

Understanding the spectrum of RAS-mutated thyroid cancers would help clinicians individualize management, both regarding treatment as well as for better prognostication and individualized follow-up. In this study, we combined data from targeted next-generation sequencing (NGS) with a complete clinicopathologic analysis to identify tumors with more aggressive features and higher risk of death. We propose that a comprehensive genomic evaluation at initial diagnosis can serve as a clinical tool to define risk and inform management decisions in the clinically heterogeneous group of RAS-mutated thyroid tumors.

We performed a clinical–genomic analysis of consecutive thyroid cancer patients who presented and received treatment between 2014 and 2021 at Brigham and Women's Hospital and/or Dana Farber Cancer Institute, and in whom OncoPanel sequencing was performed on their tumor sample. All clinical and histologic data were obtained at time of diagnosis, specifically age, sex, tumor characteristics (tumor–node–metastasis staging, extrathyroidal extension, tumor size), and all available data on each patient's disease course (type of treatment, recurrence/persistence, mortality). The American Thyroid Association (ATA) risk for disease recurrence, the American Joint Committee on Cancer (AJCC) stage of disease at diagnosis, and the ATA response to therapy were defined in accordance with the 2016 ATA guidelines on thyroid cancer and the 8th AJCC edition staging system (8, 9). Encapsulated FVPTCs that were diagnosed before 2016 were rereviewed by our Pathologists, and reclassified when appropriate as noninvasive follicular thyroid neoplasm with papillary-like nuclear features (NIFTP). NIFTPs were included in the analysis and grouped with thyroid cancers, as in the latest WHO classification are not considered benign and in clinical practice are managed like low-risk thyroid cancers (10). Patients with structural evidence of disease after thyroidectomy [and radioactive iodine (RAI) ablation (when RAI was administered)] within 1 year were defined as having persistent disease. Recurrence was defined as new structural findings after thyroidectomy [and RAI ablation (when RAI was administered)] at later follow-up. Use of systemic therapy included tyrosine kinase inhibitors, chemotherapy, and/or immunotherapy for advanced disease, while progression of disease was defined as the presence of new structural findings on follow-up, or a > 20% increase in the disease burden necessitating treatment intervention such as surgery, external beam radiotherapy (XRT) or systemic therapy.

OncoPanel is a custom-targeted NGS assay designed to detect genetic alterations across the entire coding sequence of a panel of genes known to be implicated in cancer biology, including single nucleotide changes, insertions, deletions, copy-number alterations (CNA), and structural variants. In its latest version, OncoPanel targets 447 genes for mutation and CNA analysis, plus 60 selected genes targeted for rearrangement detection. To identify impactful alterations, we reviewed the relevant literature as well as cancer mutation databases (11) to define a panel of key additional genomic alterations. Full process of selection has been described previously (7). Thus, we only report on genetic alterations likely to be relevant in thyroid cancer progression.

Patients were grouped on the basis of their somatic NGS profile to a solitary RAS and a RAS+additional oncogenic alterations cohort. One patient with PTC had a NRAS and a BRAFV600E mutation, and this patient was excluded from further analysis because of uncertainty regarding the driver mutation of that tumor.

The study was conducted in accordance with the Declaration of Helsinki. Approval from the Brigham and Women's Hospital institutional review board was obtained (IRB#1999P02899 and 2000P000167). Informed written consent was obtained from each subject.

Continuous variables were described as the mean and SD, while categorical variables as frequencies and percentages (%). Continuous variables between two groups were compared with Student t test, the Mann–Whitney or one-way ANOVA. Categorical variables were compared using the Pearson χ² test. For survival analysis, a Kaplan–Meier curve was built, and the log-rank test was used to assess statistical significance. Cox regression analysis was used to identify the factors independently associated with disease specific mortality. Mutation plots were generated using the OncoPrinter tool, available at the cBioPortal (RRID:SCR_014555; refs. 12, 13).

Data generated during this study are available through the corresponding author upon reasonable request. The list with all genetic alterations detected in all patients included in this study is available in Supplementary Table S1.

Patient and tumor characteristics are shown in Table 1. The cohort included 78 patients with RAS-mutant thyroid cancer (Fig. 1); 65% PTC (predominantly FVPTC), while 13% were FTC, 1% Hurthle cell carcinoma or oncocytic carcinoma (HCC), 4% PDTC, 8% ATC, and 9% were reclassified as NIFTP. Mean age was 52.3 ± 15.3 years, and 63% (n = 49) were female.

Genomic sequencing of all RAS-mutant thyroid cancers detected a median ± interquartile range of 4±4 variants per tumor (range 1–15). Importantly, 22% (17/78) of patients harbored additional oncogenic mutations. The Oncoprint and curated list of mutations for these tumors are presented in Fig. 2A; Table 2 (full list of detected variants is shown in Supplementary Table S1). These selected additional mutations were clonal events, as shown by the fact that their minor allele frequencies (MAF) were similar to those for the RAS driver mutation (median MAFs: 35% vs. 32%, respectively, Mann–Whitney P = 0.9083 ns; Fig. 2B).

All patients (100%, 6/6) with RAS-mutated ATC were found to have an additional oncogenic mutation, most commonly secondary TP53 mutations identified in 83%. As expected, patients with ATC had a poor prognosis, with a mortality rate of 83% (5/6). The single alive patient at the time records were reviewed for this study was the only patient that did not have a TP53 mutation (alive at 96 months post-diagnosis). The median survival time from diagnosis was 18 months. Five of 6 patients with ATC had Stage IVc disease at diagnosis with distant metastases present primarily in the lungs. These patients’ and tumors’ characteristics are presented in Supplementary Table S2.

Our cohort also included 3 patients with PDTC, two of whom had a solitary RAS mutation (i.e., no other mutations in our highly curated set of oncogenic alterations) while one harbored an NRAS plus a CDKN2A frameshift mutation. The patient with NRAS+CDKN2A mutations died of thyroid cancer 96 months post-diagnosis, while the 2 patients with solitary RAS mutation were alive at 156 and 25 months post-diagnosis, respectively. One of those 2 patients had a more aggressive clinical course with several distant bone metastases, while the second patient demonstrated excellent response to therapy with no detectable disease at 25 months post-surgery, likely reflecting timely intervention early in the disease course.

Separate from those with ATC or PDTC, 69 patients with differentiated thyroid cancer (DTC) were analyzed. Fifty-nine DTCs harbored a solitary RAS mutation, while 10 harbored RAS and additional oncogenic mutations. The specific mutations can be found in Fig. 2A; Table 2.

When comparing patients with additional mutations in genes known to be involved in thyroid cancer progression (RAS+additional mutations group, N = 10) versus patients harboring solitary RAS mutations (N = 59), the two groups had similar age and sex, with a higher percentage of FTCs in the former group (50% vs. 8.5%; P = 0.002). The risk profiles differed significantly between the two groups (detailed in Table 3), with the RAS+additional mutations group having larger primary tumors (4.7±2.2 cm vs. 2.5±1.8 cm; P = 0.002), higher AJCC stage (Stage III or Stage IV) at presentation [67% (N = 6) vs. 3% (N = 2); P < 0.001] and a higher percentage of tumors at high risk for disease recurrence per ATA staging [77% (N = 7) vs. 12% (N = 7); P < 0.001; Fig. 3A]. Sixty percent (n = 6) of patients in the RAS+additional mutations group had distant metastases at presentation versus 8% (n = 5) in the RAS-only group (P = 0.002).

Importantly, the patients in the RAS+additional mutations group had worse disease outcomes. The rates of recurrent or persistent disease were significantly higher when compared with the solitary RAS group (87.5% vs. 11%; P < 0.001; Fig. 3B). The more aggressive nature of the tumors with additional mutations was also evident by the rates of patients requiring systemic therapy [56% (n = 5) vs. 4% (n = 2); P < 0.001] and XRT [56% (n = 5) vs. 4% (n = 2); P < 0.001]. Mortality in the RAS+additional mutations group was 10-fold higher (20% vs. 1.8% over the course of follow up). Five-year disease specific survival was 98% for solitary RAS-mutated DTCs group, though reduced to 70% among the RAS plus additional mutations cohort (P = 0.007; Fig. 3C). No patient with ATA low risk for recurrence tumor that had a solitary RAS mutation developed any recurrence, distant metastasis or died from thyroid cancer.

Notably, on a Cox regression analysis, harboring additional mutations was associated with disease-specific mortality [HR, 12.27 (1.11–135.43); P = 0.041]. When we included in the model advanced AJCC stage (Stage ≥ 3) and sex, the additional mutations remained independently associated with disease specific mortality [HR, 28.03 (1.54–509.33); P = 0.024]. The Kaplan–Meier curve comparing the two genomic profiles after being adjusted for advanced AJCC stage is presented in Fig. 3D.

Finally, we examined whether copy-number losses of chromosome 22q arm, which have been reported in RAS-mutant FVPTC and PDTC (1, 14), associated with more aggressive features in our DTC cohort. As shown in Supplementary Table S3, no significant differences were found in the clinicopathologic characteristics of DTCs which were diploid for chromosome 22q (N = 44) versus those harboring heterozygous losses (N = 25).

Genomic sequencing of thyroid cancer in routine clinical practice is becoming increasingly available, particularly for advanced disease and as a prerequisite for certain targeted therapies (15). The interpretation of findings and translational use of such data, however, are not straightforward. In the current manuscript, we present data from a comprehensive clinicopathologic and genomic study of 78 patients with RAS-mutated thyroid cancers. The ATCs of our cohort had a poor prognosis, and all of them harbored additional oncogenic mutations. More importantly, we showed that even amongst well-differentiated RAS-mutant thyroid cancers additional oncogenic mutations predict a more aggressive clinicopathologic behavior, more advanced histologic features, worse prognosis with significantly higher rates of recurrent/persistent disease, and higher disease-specific mortality. In a Cox regression analysis, we showed that the genomic status was associated with mortality after correcting for AJCC stage and sex. Our data highlight the value of genomic characterization amongst the very heterogeneous RAS-mutant DTC cohort as it can assist in determining the appropriate cancer risk and management.

Oncogenic mutations at RAS result in constitutive, aberrant activation of the downstream MAPK and PI3K/AKT signaling pathways, a critical event in thyroid tumorigenesis. Yet, in contrast with BRAFV600E, RAS mutants respond to feedback inhibitory signals that temper the extent of MAPK activation (16, 17). As a result, RAS mutations appear throughout the entire spectrum of follicular-derived neoplasms (18). It is postulated that the RAS-mutant follicular adenoma is the precursor lesion for the development of RAS-mutant FTC and FVPTC (19). Although certain studies have described that RAS may confer a more aggressive phenotype (20), more recent evidence do not describe RAS mutations as predictors of disease specific-mortality (21). Therefore, the identification of RAS mutations has not been proven to reliably define diagnosis and prognosis (5), and we need additional tools to identify more aggressive tumors earlier in the disease course. In the current study, we demonstrate that the coexistence of RAS with other select genetic alterations can serve as a useful biomarker of aggressive behavior.

Our study provides an in-depth genomic analysis with a full clinicopathologic panel of characteristics in a large cohort of RAS-mutated thyroid cancers. Our results are congruent with others that have shown RAS and other key oncogenic mutations to be negative predictors in ATC (22, 23). In RAS-mutant DTC, our knowledge of genomic synergy has been limited to the negative prognostic significance conferred by the co-occurrence of RAS with TERT promoter mutations in DTC (6, 24). However, our data are novel in expanding that notion to additional oncogenic mutations in RAS-driven DTC. We used our Oncopanel NGS to perform a comprehensive characterization of hundreds of cancer genes followed by a stringent curation of mutations relevant for thyroid cancer progression and biology. Although our cohort was representative of the entire spectrum of RAS-mutant thyroid cancer, it was enriched with the intermediate-to-high-risk thyroid cancers that often pose the greatest clinical management challenges. This also allowed us to record more genetic events in these otherwise genomically quiet tumors. An important consideration would be NIFTPs, that were described in 2016, and shown to mostly represent indolent tumors (25, 26). While it would be interesting in future studies with larger cohorts to examine those specifically, we did not exclude those tumors as there is still diagnostic uncertainty with the current WHO classification not classifying them as benign, and largely in clinical practice those are treated in the same manner as nonaggressive FTCs and FVPTCs (26).

Currently, clinical management decisions are primarily made on the basis of histology and the ATA risk of disease recurrence, and secondarily based on the response to therapy (8). However, there is uncertainty regarding therapy and follow up in tumors falling in the intermediate risk for disease recurrence. Genomic characterization of these tumors can assist in better risk stratification, as intermediate-risk tumors with RAS+additional oncogenic mutations might better be classified as high-risk, acknowledging that more traditional pathologic evaluation now converges with molecular characterization for individualized patient care. Similarly, in the current study we demonstrate that no patient with ATA low-risk for recurrence tumor that had a solitary RAS mutation developed any recurrence, distant metastasis or died from thyroid cancer. Therefore, results of the current study can also assist in reducing follow up burden for patients and clinicians in this subpopulation.

The concept of additional oncogenic mutations has also been explored in thyroid cancer preclinical models. In transgenic mouse models, Hras (under the control of bovine thyroglobulin promoter) and Kras (under high TSH stimulation) mutations had low oncogenic potential, but they showed they could act as a predisposing factor (27, 28). Supporting this hypothesis, in vivo engineering of specific oncogenic mutations has been proven to promote RAS-driven thyroid lesions towards aggressive tumors (29–32), with the prime example being the combination of oncogenic Ras and p53 loss that led to the development of PDTCs and ATCs (33, 34).

Comprehensive genomic profiling of tumors does not only result in improved prognostication, but also promotes individualized oncologic management. In our cohort, 81% of patients with RAS+additional oncogenic mutations received some form of systemic therapy, with mostly poor outcomes as described above. Thus, more targeted approaches are necessary. Although direct targeting of RAS proteins used to be considered impossible because of the lack of drug-binding pockets on the surface of RAS proteins, more recently several strategies have been employed including targeting upstream and downstream proteins, RAS directly, as well as RNA interference (35). While KRAS inhibitors have recently been developed, studied in a clinical trial setting and approved by the FDA, the potential for long-term benefit of KRAS inhibitors monotherapy is modest with the development of resistance, and therefore the identification of additional oncogenic mutations is crucial in the choice of combination therapies (36). Targeting the PI3K/AKT/mTOR pathway has shown promise in preclinical models but clinical application mostly failed due to toxicities (37–39). The discovery of additional actionable mutations can nevertheless expand the arsenal of available options in advanced thyroid malignancies.

We acknowledge limitations to our study. Our cohort size remains relatively modest especially as we attempted to define subgroups with additional oncogenic mutations, and thus this analysis could not be performed. However, this is one of the largest cohorts with full genomic and clinicopathologic characterization published, and the prevalence of those tumors are proportional to what we would expect when comparing to our previous work on BRAFV600E-mutant PTC (7). Future multicenter studies can be very useful in studying the specific risk of different additional oncogenic mutations. Although we recorded high quality follow-up data, outcome information was not available in 10% of subjects. These two factors limit the statistical power of our findings and thus it would be very important to confirm them in larger patient cohorts. Moreover, the OncoPanel platform does not cover the EIF1AX gene. Because EIF1AX mutations have been shown to co-occur with RAS mutations and to induce dedifferentiation in a subset of RAS-mutant thyroid cancers (14, 30, 40, 41), it is reasonable to think that some of the RAS-mutant tumors in our cohort also harbor EIF1AX mutations.

In conclusion, these data provide the most extensive clinicopathologic and genomic characterization of RAS-mutant thyroid cancers. We have demonstrated that additional oncogenic mutations are not only associated with a more aggressive histologic phenotype, but also with increased recurrence and mortality rates in RAS-mutated DTC. Given the diagnostic uncertainty of solitary RAS mutations, we believe that a comprehensive genomic characterization can improve prognostication in patients with thyroid cancer. This, in turn, can help guide management and follow-up clinical decisions. We certainly acknowledge the cost and availability issues in clinical practice, but that would be another step towards a personalized care-delivery model for patients with thyroid cancer.

T. Pappa reports grants from Harvard Catalyst during the conduct of the study. K. Sehgal reports grants and personal fees from Merck and personal fees from Exelexis Inc., Equinox Group Inc., Scholar Rock, and MedScape outside the submitted work. E.K. Alexander reports personal fees from Veracyte Inc. outside the submitted work. No disclosures were reported by the other authors.

A. Bikas: Conceptualization, data curation, formal analysis, investigation, visualization, methodology, writing–original draft, writing–review and editing. S. Ahmadi: Data curation, formal analysis, investigation, writing–review and editing. T. Pappa: Data curation, formal analysis, investigation, writing–review and editing. E. Marqusee: Formal analysis, writing–review and editing. K. Wong: Formal analysis, writing–review and editing. M.A. Nehs: Resources, writing–review and editing. N.L. Cho: Resources, writing–review and editing. J. Haase: Investigation, writing–review and editing. G.M. Doherty: Resources, writing–review and editing. K. Sehgal: Resources, project administration, writing–review and editing. J.A. Barletta: Resources, validation, methodology, project administration, writing–review and editing. E.K. Alexander: Conceptualization, resources, supervision, project administration, writing–review and editing. I. Landa: Conceptualization, resources, formal analysis, supervision, visualization, methodology, writing–original draft, writing–review and editing.

No specific outside funding was received for this work. T. Pappa is partially supported by the KL2/CMeRIT award - Harvard Catalyst/The Harvard Clinical and Translational Science Center (National Center for Advancing Translational Sciences, NIH Award UL1 TR002541). I. Landa is partially supported by the NCI Career Transition Award, grant number 1K22CA230381.

This work was presented as an oral presentation in the 91st Annual Meeting of the American Thyroid Association in 10/2022 (Montreal, Canada).

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.