Mechanistic Insights of Thyroid Cancer Progression

Abstract

Differentiated thyroid cancers (DTCs) are primarily initiated by mutations that activate the MAPK signaling cascade, typically at BRAF or RAS oncoproteins. DTCs can evolve to more aggressive forms, specifically, poorly differentiated (PDTC) and anaplastic thyroid cancers (ATC), by acquiring additional genetic alterations which deregulate key pathways. In this review, we focused on bona fide mutations involved in thyroid cancer progression for which consistent mechanistic data exist. Here we summarized the relevant literature, spanning approximately 2 decades, highlighting genetic alterations that are unquestionably enriched in PDTC/ATC. We describe the relevant functional data obtained in multiple in vitro and in vivo thyroid cancer models employed to study genetic alterations in the following genes and functional groups: TP53, effectors of the PI3K/AKT pathway, TERT promoter, members of the SWI/SNF chromatin remodeling complex, NF2, and EIF1AX. In addition, we briefly discuss other genetic alterations that are selected in aggressive thyroid tumors but for which mechanistic data is still either limited or nonexistent. Overall, we argue for the importance conveyed by preclinical studies for the clinical translation of genomic knowledge of thyroid cancers.

In the last decade, the implementation of next-generation sequencing (NGS) technologies to characterize the genome of patients’ tumors provided an exponential increase in our knowledge of the genetic mutations that drive cancer progression. Compared with other cancers, thyroid tumors are genetically simple entities (1, 2) whose initiation and evolution are driven by a handful of genomic events, facilitating the study of the biological consequences of specific genetic alterations. Yet, by definition, the mechanistic characterization of oncogenic mutations requires multi-year studies and the implementation of suitable in vitro and in vivo models. In this review, we provide a description of a curated list of genomic events involved in thyroid cancer progression for which solid mechanistic insights exist (Fig. 1). We believe that the availability of functional data in preclinical models plays an essential role toward the individualization of the clinical management of thyroid cancer patients.

Figure 1.

Timeline highlighting key contributions in our knowledge of the genetic events that determine thyroid cancer progression, including reports of these mutations in patient cohorts (DNA strand cartoon), in vitro studies (culture dish) and animal models (mouse). Brief summaries of selected publications are provided. References to published papers are shown in brackets. See main text for a detailed description of these works, as well as for a more comprehensive list of studies. Abbreviations: ATC, anaplastic thyroid cancer; NGS, next-generation sequencing; PDTC, poorly differentiated thyroid cancer.

Follicular Cell–Derived Thyroid Cancers

Thyroid cancers are the most frequent endocrine malignancy (3, 4). Tumors arising from thyroid follicular cells include well-differentiated and poorly-/undifferentiated types. According to the 2022 World Health Organization (WHO) classification of thyroid tumors, the differentiated thyroid cancers (DTC) encompass the papillary (PTC), follicular (FTC) and oncocytic (OC; previously known as Hürthle) thyroid cancer subtypes (5). PTC is by far the most prevalent thyroid cancer type, accounting for ∼85% of cases and including several subtypes, followed at a great distance by FTC (∼5%), which has been historically associated with iodine-deficient regions (6, 7). Both PTC and FTC typically have good prognosis and are highly treatable cancers, with some exceptions. Conversely, poorly differentiated (PDTC; < 5% of cases), and particularly anaplastic thyroid cancers (ATC; ∼2%) are rare but much more aggressive tumors, ranging from intermediate (in PDTC patients) to dismal prognosis (ATC). The 2022 WHO classification defined differentiated high-grade thyroid carcinoma (DHGTC), which retains well-differentiated histopathological features but has a prognosis similar to PDTC (5). Tumors occurring from parafollicular or C cells, which represent a different cell lineage within the thyroid gland, are termed medullary thyroid cancers (MTC; < 5% of cases) and will not be discussed in this review.

Genetic Drivers of Differentiated Thyroid Cancer Initiation

The Cancer Genome Atlas (TCGA) characterized 500 adult PTCs, defining their genomic and transcriptomic hallmarks (8). About 90% of PTCs harbor clonal nonoverlapping alterations in genes encoding effectors of the mitogen-activated protein kinase (MAPK) pathway, a major hub for cell proliferation. Constitutive stimulation of MAPK signaling in thyroid cancer is achieved via activating mutations in BRAF, predominantly leading to the p.V600E substitution, which occur in ∼60% of PTCs. Oncogenic mutations in NRAS, HRAS, and KRAS are found in 13% of tumors, whereas genetic rearrangements involving the receptor tyrosine kinases RET, NTRK1/3, and ALK occur in 16% of PTCs (8). FTCs, on their part, are predominantly driven by mutations in RAS genes, although a subset of tumors are initiated by PAX8-PPARG rearrangements or TSHR (thyroid-stimulating hormone receptor) mutations (9-11). The fact that all these genomic events are mutually exclusive and that most of them target the same pathway strongly supports the notion of a single alteration as the initiating event in thyroid tumors.

The degree of MAPK stimulation varies depending on the genetic driver; for instance, BRAF^V600E has a stronger activating potential compared to RAS mutations, and this is inversely correlated with the expression of markers of thyroid differentiation, including of the sodium-iodide symporter (NIS), which ultimately determines the ability of thyroid cancers to respond to radioactive iodine (RAI) therapy (6, 8). Overall, the MAPK pathway is the central process in the pathogenesis of DTCs, which are increasingly referred to as “BRAF^V600E-like” and “RAS-like” tumors, as their driver event determines an important part of their properties via defined transcriptomic features. Genomic and transcriptomic heterogeneity within those categories still exists, impacting their clinical outcomes, particularly as these tumors progress to more advanced forms, such as DHGTC/PDTC and ATC.

Genomic Determinants of Thyroid Cancer Progression

The observation that the majority of PDTC and ATC harbor mutations in BRAF or RAS oncogenes supports a model of tumorigenesis whereby most of these tumors arise from PTC/FTC via the accumulation of key genetic alterations. Whole exome sequencing of multiple regions of thyroid tumors displaying PTC and ATC components, as well as recent single-cell RNA sequencing efforts, have defined common ancestors with truncal mutations, although an early PTC-ATC divergence seems to be the rule (12-14). Building upon the idea of this continuum in thyroid dedifferentiation, NGS studies showed that PDTC/ATC display an extended set of alterations, beyond their BRAF and RAS drivers, compared with PTC/FTC (15-17). A subset of those alterations, shown in Fig. 2, have been evaluated in experimental systems, most elegantly in genetically modified mouse models, for their ability to induce thyroid cancer progression (for a detailed review of in vivo models, see references (29-32)). In the following section, we elaborate on those genetic alterations found in advanced thyroid cancers for which consistent mechanistic data exist (summarized in Fig. 3).

Figure 2.

Frequency of mutations in the specified genes, pathways, or functional groups across the indicated thyroid cancer subtypes. Data was obtained from studies, published from 2014 to 2022, using next-generation sequencing platforms in nonoverlapping thyroid cancer cohorts of at least 15 patients per subtype (references (8, 10, 11, 15-28)). Each dot represents an independent patient cohort. A, TP53 mutations. B, Mutations in effectors of the PI3K/AKT pathway (AKT1, PIK3CA, and PTEN genes). C, TERT promoter mutations. D, Mutations in members of the SWI/SNF chromatin remodeling complex (ARID1A, ARID1B, ARID2, SMARCB1, and PBRM1 genes). E, NF2 mutations. F, (left) EIF1AX mutations; (right) Co-occurrence of EIF1AX + RAS mutations, expressed as a share of tumors with mutations in either of these genes within each subtype also harboring mutations in the other locus (data from selected references (8, 15, 16, 19)). Abbreviations: ATC, anaplastic thyroid cancer; PDTC, poorly differentiated thyroid cancer; PTC, papillary thyroid cancer.

Figure 3.

Schematic representation of the mechanistic consequences of the genetic alterations in the MAPK pathway and in the main processes involved in thyroid cancer progression. Stars indicate nodes where mutations in these pathways occur in thyroid tumors.

Mechanistic Insights of Mutations Inducing Thyroid Cancer Progression

TP53 Mutations: Gatekeeper of Cell Division and of ATC Transformation

Loss-of-function mutations in the TP53 gene are virtually absent in most DTCs, become more frequent in metastatic PTC/FTC and PDTC, and constitute the most prevalent genetic alterations in ATC (Fig. 2A). TP53 mutations were the first alterations associated with undifferentiated thyroid cancers (33, 34) and are now considered a hallmark of PTC-to-ATC evolution, as exemplified by tumors with both PTC and ATC foci, the latter harboring TP53 losses (12). TP53 disruption constitutes an unequivocal marker of poor prognosis (11, 18, 35).

The TP53 gene encodes p53, a well-known tumor suppressor whose functions comprise responding to cellular stresses to induce cell cycle arrest, DNA repair, and apoptosis, thus preventing abnormal cells from further propagation (36, 37). Loss of p53 allows cancer cells to circumvent these restrictions, which explains why multiple tumors select for genetic inactivation of this locus (38). The role of p53 disruption in thyroid cancers was first evaluated in in vitro systems. Early efforts assessed the role of p53 loss in thyroid differentiation, showing that TP53 mutations induced loss of thyroid transcription factor PAX8 (paired box 8) in a rat thyroid cell line. Similarly, the re-introduction of wild-type p53 in human thyroid cancer cells with an endogenous TP53 mutation induced re-expression of thyroid lineage markers PAX8 and TPO (thyroid peroxidase) (39, 40). Additional experiments in irradiated thyroid cancer cells showed that wild-type p53, but not mutant p53, facilitated DNA repair activities (41).

The first in vivo evidence of p53 loss inducing ATC transformation was reported in the context of thyroid-specific activation of ret/PTC1 fusion (42). In this study, constitutive genetic inactivation of Trp53 (mouse ortholog of human TP53 gene) in homozygosis generated multiple tumors (eg, lymphomas and sarcomas), and also induced anaplasia and metastasis in ret/PTC1-driven thyroid tumors from older animals. McFadden and colleagues employed inducible thyroid-specific mouse models in which oncogenic alterations in Braf and Trp53 were evaluated, thus mimicking the most frequent genetic combination in human ATCs. In this work, both homozygous deletion of Trp53 and hemizygous expression of oncogenic p.R270H mutation induced Braf^V600E-driven PTCs to progress to ATCs (43). Murine ATC recapitulated the clinical features and molecular hallmarks of their human counterparts: explosive growth, metastatic dissemination, epithelial-to-mesenchymal transition (EMT), and thyroid de-differentiation. Expression profiling of mouse Braf + Trp53 showed that these tumors activated embryonic transcriptional programs and PI3K/mTOR signaling without genetic defects in effectors of this pathway (see next section “Activation of PI3K/AKT/mTOR Pathway: Partnering With MAPK in Thyroid Cancer Progression”), and that combined Raf plus Mek inhibition improved therapeutic response of ATCs (43). Zou and colleagues generated another model and showed that p53 loss removed a critical barrier in oncogene-induced senescence to promote PTC-to-ATC progression and PI3K activation in Braf^V600E-initiated thyroid tumors (44). Knauf, Luckett et al, on their part, employed an alternative model to target Braf^V600E and Trp53 loss to the thyroid, also generating mouse ATCs that recapitulated human disease, that is, rapid growth, spindle cell histology, invasiveness, and metastatic potential (45). Compared to Braf alone, murine Braf + Trp53 tumors displayed an increase in MAPK pathway transcriptomic output, which portends potential clinical ramifications. These tumors also phenocopied their human counterparts, displaying a heavy infiltration of tumor-associated macrophages (45-47). Blockade of oncogenic Raf caused temporary tumor regression, but ATCs recurred in a Braf-independent fashion, that is, activating MAPK via recurrent amplifications targeting the Met gene, thus suggesting that p53 loss can cooperate with a variety of genetic drivers toward ATC transformation. In fact, another study from the same group showed that concomitant thyroid-specific Hras^G12V and p53 loss can generate aggressive tumors, ranging from PDTC to ATC (48). Interestingly, these tumors acquired copy number alterations targeting specific chromosomes, suggesting a direct effect between homozygous p53 loss and genomic instability. Although it remains to be fully proven, the former suggests that p53 inactivation in thyroid cancer pathogenesis might follow a specific sequence of events, as recently reported in murine pancreatic adenocarcinoma models (49). Likewise, deletion of p53 in a Kras^G12D-driven thyroid cancer model generated murine PDTC/ATC and induced anti-apoptotic effects via overexpression of Bcl2 family members (50). Finally, the engineering of p53 loss in the thyroid has also been shown to induce dedifferentiation from PTC to PDTC/ATC, EMT, and invasiveness in mouse tumors driven by oncogenic STRN-ALK rearrangements (51). Overall, p53 loss in ATC, as is the case in other tumors, seems to promote cancer progression by removing critical checkpoints on proper genomic and cell cycle states, favoring genomic instability, tolerating DNA damage, and possibly allowing the accumulation of additional oncogenic mutations. Given the importance of MAPK signaling in thyroid cancer, p53 loss in this lineage likely prevents senescence in cells with massive MAPK output levels.

Activation of PI3K/AKT/mTOR Pathway: Partnering With MAPK in Thyroid Cancer Progression

Aberrant activation of the PI3K/AKT/mTOR pathway is a common event in cancer, fueling tumor's unrestricted growth abilities through enhanced protein synthesis and overall reprogramming of cancer cell metabolism (52). Activation of this pathway in thyroid cancers was traditionally associated with a follicular growth pattern, partly because RAS mutations, which are enriched in tumors with this histology, can activate the PI3K pathway distally (53-55). Targeted sequencing of key effectors of the PI3K pathway, primarily PIK3CA, AKT1, and PTEN, soon expanded the spectrum of thyroid tumors with PI3K activation and showed that these mutations were enriched in PDTC and ATC (56-59). Data from NGS studies confirmed these associations, particularly for PIK3CA and AKT1, which are clearly selected in advanced disease (16, 17, 19, 60), whereas PTEN alterations, when in isolation, also occur in a subset of FTCs (Fig. 2B) and in follicular adenomas. In PDTC, and particularly in ATC, genetic defects of PI3K pathway effectors show specific combinatorial trends with MAPK-activating drivers: oncogenic PIK3CA mutations tend to cluster in BRAF^V600E-mutant tumors, whereas PTEN losses co-occur with RAS and NF1 mutations (16).

The reasons why the PI3K pathway gets activated predominantly at different nodes in BRAF vs RAS-mutant disease remain elusive, but these specific combinations have been proven to induce thyroid cancer progression in mouse models. In this regard, a study showed that targeting Pten loss to the mouse thyroid induced thyrocyte proliferation, which resulted in goiter and follicular adenomas but was insufficient to initiate tumorigenesis (61). However, the combination of oncogenic Kras and Pten loss induced invasive FTCs (62). In this model, Kras+Pten FTCs showed enhanced MAPK signaling (increased phospho-Erk levels) and overexpression of cyclin D1 transcripts, suggesting an effect on cell cycle progression. Transcriptomic profiling of murine Kras+Pten specimens by expression arrays pointed at an extended list of deregulated processes potentially involved in the transition from follicular adenoma to FTC (63). An alternative model, combining Hras^G12V and homozygous Pten loss generated metastatic FTCs which displayed an immunosuppressive tumor microenvironment (64). Similarly, mouse models in which Pten loss was engineered in Braf^V600E backgrounds induced enhanced proliferation and progression to PDTC/ATC, although this allelic combination tends to be mutually exclusive in human thyroid tumors (65-68). Of note, genetic inactivation of Pten and Trp53, which occurs in a subset of human ATCs, has been proven to induce murine ATCs, which in turn displayed deregulation of cell cycle and were highly glycolytic (69). Charles and colleagues, on their part, engineered a kinase domain mutation at murine Pik3ca^H1047R in an inducible Braf^V600E mouse model, demonstrating the cooperative effects of these 2 mutations in ATC pathogenesis (65). This allelic combination led to highly proliferative tumors, EMT, and a marked decrease in mouse survival.

Overall, PI3K pathway activation, in a constitutive MAPK background, as in the described mouse models (62, 65), their in vitro derivatives (70-72), as well as in human thyroid cancer cell lines (73-77), showed enhanced MAPK signaling and/or sensitized cells to combined pharmacological blockade of MAPK and PI3K signaling. Although the preclinical rationale for combining MAPK and PI3K inhibitors is strong, so far it has not translated into clinical practice in cancer patients harboring these mutations, primarily due to significant toxicities (78). Nevertheless, activation of the PI3K/AKT pathway, via mutations in RAS, PIK3CA, AKT1, or PTEN, as well as nonmutational processes (eg, b-catenin transcriptional activity or microRNA 146b overexpression (79, 80)), constitutes an essential process in thyroid tumorigenesis and progression.

TERT Promoter Mutations: Telomerase Re-expression, Telomere Maintenance, and Beyond

Mutations in the proximal promoter of the TERT (telomerase reverse transcriptase) gene are one of the most prevalent alterations in thyroid cancers. TERT promoter mutations were initially discovered in melanomas (81, 82) and then identified in gliomas and hepatocellular, urothelial, and thyroid carcinomas, among others (83-86). Pan-cancer whole genome sequencing confirmed the TERT promoter as the most frequently mutated noncoding locus in cancer (87). TERT promoter mutations in thyroid tumors are bona fide biomarkers of disease progression: they are relatively infrequent in PTCs and FTCs, they become more frequent in metastatic PTCs/FTCs, and PDTCs, and they are very prevalent in the highly aggressive ATCs (Fig. 2C). TERT promoter mutations tend to co-occur with either BRAF or RAS mutations across thyroid cancers, strongly suggesting that tumors select for these genetic combinations to progress. Multiple studies provided evidence of the value that the assessment of TERT mutations, either alone or, most efficiently, in combination with BRAF/RAS, can provide for risk stratification and prognostication of thyroid cancer patients (88-93).

TERT promoter mutations occur in 2 hotspots at c.-124C>T and c.-146C>T and are mutually exclusive, suggesting a common effect. Tumors harboring TERT promoter mutations reactivate TERT transcription, which is otherwise repressed in adult normal cells. Recruitment of specific transcription factors, and possibly other epigenetic regulators, occurs upon acquisition of TERT promoter mutations and facilitates TERT re-expression (86, 94-99). The re-expression of telomerase, either via the acquisition of these mutations or by other mechanisms, is considered a hallmark of cancer and prompts downstream effects (100, 101). TERT encodes the catalytic subunit of the telomerase complex, whose function is to maintain critically short chromosomal ends (“telomeres”), thus enhancing replicative immortality. Multiple other extra-telomeric effects of TERT have been described in various cell contexts (102, 103).

A recent study showed that upregulation of both TERT and TERC (the RNA template component of the telomerase complex) correlates with clinically aggressive and highly proliferative thyroid tumors (104). Tumors with TERT promoter mutations, alone or in combination with other means of TERT reactivation (eg, promoter hypermethylation or TERT gene duplication), had shorter median telomere length compared to their TERT wild-type counterparts, indirectly suggesting that telomerase reactivation allows tumor cells to tolerate critically short telomeres. In this study, shorter telomeres associated with a more relaxed chromatin state at chromosome 5p end (and other sub-telomeric regions) which in turn was linked with increased expression of genes located in those loci, including TERT. These point to a TERT main function in telomere maintenance as a mechanism likely operating in thyroid cancers. In parallel, we explored the role of Tert reactivation in a Braf^V600E-driven murine model of thyroid cancer, highlighting nontelomeric functions in Tert-mediated thyroid cancer progression. Our group engineered a Tert promoter mutation in the equivalent mouse locus and showed that telomerase reactivation induced progression from PTC to DHGTC/PDTC/ATC (105). Thyroid cancer progression tracked with increased Tert transcription, but not with critical changes in telomere length, likely due to the known fact of mice having unusually long telomeres relative to humans (106). However, RNA sequencing (RNAseq) of Tert-re-expressed, Braf^V600E-driven mouse tumors showed an overactivation of well-known pathways in thyroid pathogenesis such as the MAPK and PI3K/AKT/mTOR signaling (vs Braf^V600E alone) (105). RNAseq also unveiled a potential role for cytokine and chemokine signaling, primarily via the NF-kB pathway (a key hub of tumor-elicited inflammation (107)), in mouse tumors with telomerase reactivation. Of note, a re-analysis of TCGA RNAseq data also pointed to chemokine and cytokine signaling overactivation in human PTCs with BRAF^V600E mutation and TERT expression (97), whereas the TERT-mediated regulation of NF-kB transcriptional programs had been reported in other contexts (108, 109). These observations in murine models remain to be fully validated in human thyroid cancers with TERT promoter mutations. Overall, it is likely that both telomeric and nontelomeric effects mediated by TERT re-expression operate in patients’ tumors, ultimately mediating thyroid cancer progression.

SWI/SNF: Chromatin Reconfiguration and Refractoriness to Radioiodine Treatment

The acquisition of mutations in genes encoding epigenetic modifiers, most notably subunits of the SWI/SNF chromatin remodeling complexes, is another prominent feature in the progression of DTCs to PDTCs and ATCs (16). Mutations of various SWI/SNF complex members occur in 20% of all human cancers (110), pointing at epigenetic deregulation as a key process in tumor progression. Up to 40% of ATCs display SWI/SNF mutations, most of those truncating events, and this might be an underestimation since copy number losses of these loci are not always systematically considered (Fig. 2D). The molecular function of SWI/SNF complexes consists of mobilizing nucleosomes and remodeling chromatin through ATP hydrolysis. SWI/SNF activity promotes terminal differentiation in a variety of contexts (111), whereas its disruption induces stem cell–like characteristics and tissue regeneration (112).

The first functional evidence of the importance of SWI/SNF in thyroid pathogenesis came from an in vivo transposon mutagenesis screen (known as the Sleeping Beauty system), which showed that disruptions of chromatin modifiers, including Swi/Snf subunits, significantly cooperate with oncogenic Hras in progression to PDTC (113). More recently, Saqcena, Leandro-García, et al defined the consequences of loss of different Swi/Snf complex members in a Braf^V600E murine thyroid cancer model (114). Whereas mice with endogenous Braf^V600E in isolation develop PTCs (115), animals with concomitant homozygous loss of either Arid1a, Arid2, or Smarcb1 (all key members of the Swi/Snf complex) developed PDTC or ATCs with short latency and decreased survival, and in the case of Braf+Smarcb1 mice, displayed distant metastases (114). Swi/Snf disruption caused distinct chromatin and transcriptional changes, resulting in thyroid dedifferentiation and the acquisition of a chromatin state that activates specific transcriptional programs. Furthermore, the loss of Swi/Snf in Braf^V600E-driven thyroid cancer impaired the ability of MAPK pathway inhibitors to block the pathway's transcriptional output, thus affecting their ability to restore thyroid differentiated function and response to RAI therapy. The former constitutes a stark contrast with previous studies in Braf^V600E models, in which potent MAPK signaling inhibition restored thyroid differentiation gene expression, increased RAI uptake, and improved responses to RAI therapy (116-118).

In conclusion, SWI/SNF disruption appears to lock thyroid cells into a dedifferentiated state that is no longer reversible by blocking the MAPK pathway. This suggests that SWI/SNF mutations may be markers of resistance to redifferentiation strategies, as was recently proposed in a pilot redifferentiation clinical trial for patients with BRAF^V600E RAI-refractory thyroid cancer (119), as well as in a study delineating the genomic characteristics of RAI responses in metastatic thyroid cancers (120).

NF2 Disruption: YAP-mediated Inactivation of the Hippo Pathway

The Hippo pathway is an evolutionarily conserved kinase cascade that suppresses tissue overgrowth by phosphorylating YAP (Yes1 associated transcriptional regulator), sequestering it in the cytoplasm and impairing its ability to promote TEAD-dependent transcription of genes involved in proliferation and survival (121). Mutations in Hippo canonical effectors are uncommon in cancer, with the exception of NF2, which is mutated in a subset of ATCs (15, 16) but almost absent in DTCs (Fig. 2E). Most NF2 mutations lead to early truncation of the encoded protein, known as Merlin, in line with its tumor suppressor properties (122). In addition, arm-level deletions of chromosome 22q, spanning NF2 and other tumor suppressor genes, occur in a subset of RAS-mutant PTCs and PDTCs, pointing at an additional mechanism by which a larger share of thyroid tumors lose Merlin activity (8, 16, 123).

García-Rendueles et al explored the effects of targeting thyroid-specific Nf2 loss in Hras^G12V mice, showing that the combination of these oncogenic alterations induced PDTC phenotypes (124). These tumors showed high YAP nuclear expression. Nf2/Merlin loss increased transcription of YAP/TEAD-regulated genes, including the 3 Ras isoforms (both mutant and wild-type alleles), resulting in an increased MAPK signaling output and in a dependency on this pathway for viability, which in turn increased sensitivity to MEK inhibition (121). The mechanisms underlying YAP activation in thyroid cancer are, however, not completely understood, since some thyroid cancers show low-to-null Merlin expression without NF2 mutations. A recent study from the same group tested a large panel of human thyroid cancer cell lines and showed that YAP nuclear localization is highly prevalent in thyroid cancer, it co-occurs with BRAF and RAS mutations, but it is only associated with upstream canonical Hippo pathway gene alterations in a minority of cases (125). The authors created a mouse model with inducible constitutively nuclear YAP^S127A in thyroid cells, either alone or in combination with endogenous expression of either Hras^G12V or Braf^V600E. They showed that constitutively nuclear YAP is sufficient to cause thyroid tumor formation and collaborates with Hras^G12V or Braf^V600E to confer a higher tumor grade, promote distant metastasis, and increase mortality. In contrast to the increased sensitivity to MEK inhibition observed in the Hras^G12V/Nf2 loss mice, they observed that constitutive nuclear YAP localization confers resistance to RAF inhibitor vemurafenib in human and murine BRAF^V600E thyroid cancer cell lines, owing to attenuation of vemurafenib's negative feedback on the NRG1-HER3/HER2 pathway, with subsequent ERK reactivation (125, 126). These findings point to YAP as a key player in the resistance to RAF kinase inhibition in BRAF mutant thyroid cancer. On top of these NF2 mutation-independent, YAP-mediated roles of the Hippo pathway in thyroid cancer pathogenesis, other studies in thyroid cancer cell lines have further suggested typical tumor suppressor gene roles for Merlin, including its ability to suppress cell proliferation and colony formation (127). The fact that NF2 is a target of miR-146b derivatives, a microRNA commonly overexpressed in thyroid cancers, further expands the mechanisms by which thyroid cancer cells downregulate NF2 (128). Overall, NF2 loss, YAP nuclear localization and/or activation of the Hippo pathway, play a role in the progression of both oncogenic BRAF- and RAS-driven thyroid cancers.

EIF1AX: The Role of Aberrant Initiation of Protein Translation

EIF1AX mutations were first identified in uveal melanomas (129). In thyroid specimens, they are found across the spectrum of the disease, particularly in those maintaining the follicular architecture, such as follicular adenomas and FTCs (11, 130). EIF1AX mutations also occur in around 1% of PTCs (8). Crucially, in all these neoplasms, they are often seen in isolation and tend to be mutually exclusive with other drivers. In contrast, EIF1AX mutations are not only more frequent in PDTCs and ATCs (15, 16) (Fig. 2F, left panel), but they show a strong association with RAS mutations in advanced thyroid cancers (Fig. 2F, right panel), suggesting that these events cooperate in tumor progression. EIF1AX mutations seen in other tumor lineages are missense changes targeting the N-terminal domain of the protein (129, 131, 132). In contrast, most EIF1AX mutations in thyroid tumors target a hotspot in a splice site in the C-terminal region (EIF1AX^A113splice) (8, 15, 16). EIF1AX is an essential component of the so-called protein translation preinitiation complex (PIC), and it plays a crucial role assembling the multiple elements (messenger and transfer RNAs, ribosomal subunits and initiation factors) to initiate protein translation (133).

Krishnamoorthy and colleagues studied the role of mutant EIF1AX in thyroid cancer progression in several in vitro and in vivo models (134). They discovered that EIF1AX^A113splice, which is confined to thyroid cancer, abolishes the acceptor site of exon 6, resulting in 2 alternatively spliced transcripts, one of which resulting in the exclusion of 12 amino acids through the use of a cryptic site in exon 6. They demonstrated that the cryptic spliced transcript was responsible for the oncogenic transformation in vitro. These authors went ahead and engineered mice with thyroid-specific inducible expression of Eif1ax cryptic splice transcript and found that these animals primarily developed thyroid hyperplasia, as well as low-penetrant PTCs (134). However, when expressed in a Hras^G12V context, a subset of animals developed lesions that progressed to PDTC, consistent with the histologic characteristics of human advanced thyroid tumors harboring RAS + EIF1AX mutations.

Subsequently, using isogenic human thyroid cancer cell lines, Krishnamoorthy et al showed that EIF1AX^A113splice mutants displayed a higher affinity to components of the translation PIC (preinitiation complex), leading to higher complex stability and increased global protein synthesis (134). This effect was mediated by inducing translation of ATF4, a sensor of cellular stress, such as amino acid deficiency. In xenograft experiments they showed that EIF1AX^A113splice cooperated with mutant RAS to stabilize c-MYC, which in turn sensitized cells to MEK or BRD4 inhibition. In addition, RAS and EIF1AX cooperated to activate mTOR by an increased influx of amino acids (in a PI3K-independent manner), conferring sensitivity to mTOR inhibition. Pharmacological blockade of mTOR with either MEK or MYC inhibition induced enhanced tumor shrinkage in RAS + EIF1AX^A113splice xenografts (134). Overall, these findings provide a strong rationale for the use of combined mTOR and MEK inhibitors in advanced tumors harboring RAS + EIF1AX mutations, although this has not yet been explored in patients.

Other Mutations in Thyroid Cancer Progression

In addition to the mutations described above, for which insights into their functional consequences are available (Fig. 3), the following genetic alterations clearly track with thyroid cancer progression, but so far only limited or no mechanistic data in experimental thyroid models exist:

CDKN2A/p16 and RB1

Loss-of-function mutations and deletions of CDKN2A (cyclin dependent kinase inhibitor 2A) gene are enriched in a subset of aggressive DTC and in ATC (17, 20, 135). CDKN2A encodes the bona fide tumor suppressor p16, a protein that induces cell cycle arrest via CDK4/6 inhibition, enhances p53 actions and promotes apoptosis. Although in vivo models of p16 disruption in thyroid cancer are lacking, in vitro experiments showed that CDKN2A loss is a mechanism of resistance to RAF inhibitor but sensitizes thyroid cells to BCL2/MCL1 or CDK4/6 inhibitors (136-138). In addition, truncating mutations of RB1, the gene encoding the retinoblastoma tumor suppressor protein, occur in 5% to 10% of ATCs, and likely define a separate cluster of this tumor type (16, 17). Loss of RB1 staining has been associated with metastatic PTC (139). RB1 is a negative regulator of cell cycle, and its loss likely has comparable effects for thyroid cancer cells than those affecting p53 or p16 loci. In addition, a study in rat thyroid cells showed that RB1 is a transcriptional co-activator of thyroid transcription factor PAX8 (140).

Epigenetic deregulation

Beyond SWI/SNF members, NGS of PDTC/ATC showed that mutations in other families of epigenetic regulators are selected in advanced thyroid tumors (16). These include histone methyltransferases (KMT2A, KMT2C, KMT2D, and SETD2 genes) and histone acetyltransferases (CREBBP and EP300). Based on similar loss-of-function alterations of these genes in experimental models from other tumor lineages (141, 142), thyroid cancers with these defects are expected to reconfigure the histone marks of specific genomic regions, deregulating discrete transcriptional programs. However, no functional data in thyroid tumor models exist.

Sensors of DNA damage

Truncating mutations in ATM (ATM serine/threonine kinase) and in members of the DNA mismatch repair (MMR) pathway (eg, MLH1, MSH2, MSH6 genes) predominantly occur in ATCs (15, 16). Loss of DNA repair abilities likely allows ATCs for greater genome instability and the accumulation of genomic defects, as it has been reported in other cancers (143, 144), but the specific roles of these alterations in thyroid cancer progression remain the be functionally dissected. Of note, a comparative study in thyroid cancer cell lines showed that FTC133 and T243, which have truncating alterations in MMR genes, harbored a higher tumor mutation burden than those with intact MMR function, and T243 cells displayed microsatellite instability by STR profiling (145). Anecdotal evidence of ATCs with MMR defects ranking high in mutational burden in several patient cohorts has also been reported (15, 16, 146).

Deregulation of mRNA splicing

Mutations in RBM10 are enriched in metastatic PTCs (17, 21, 147). De novo mutations in RBM genes have been described in vemurafenib-resistant thyroid cancer cells (137). A recent conference communication showed that loss of RBM10, which encodes a regulator of mRNA splicing, leads to alternative splicing events that promote extracellular matrix reconfiguration and enhanced metastatic abilities (148).

Closing Remarks

The generalization of NGS studies represented a big push in our understanding of the genomic defects determining thyroid cancer progression. From the discovery standpoint, there is still a small proportion of advanced thyroid cancers for which drivers and/or determinants of progression remain to be identified—the so-called dark matter. In this regard, a more detailed characterization of recurrent copy number events might help pinpoint additional genomic alterations. Nevertheless, multiple studies using in vivo models have demonstrated the oncogenic capacity of the gene mutations identified in advanced thyroid cancer patients. These preclinical studies have also unveiled dysregulated processes and potential therapeutic vulnerabilities for some of the oncoproteins involved in thyroid cancer progression. Clinical translation of this knowledge lags behind, partly due to the lack of specific inhibitors for some of these targets, and in part because of the obvious safeguards needed to design clinical trials. Recent Food and Drug Administration approvals of targeted therapies against thyroid cancer drivers such as BRAF^V600E and RET (149, 150), which were discovered decades ago and have been extensively studied in preclinical models, provide a realistic example of the discovery-to-application timeline. The continuous improvement of technologies for comprehensive tumor profiling and for generation of experimental models will surely accelerate the closing of this gap.

Abbreviations

ATC

anaplastic thyroid cancer

DHGTC

differentiated high-grade thyroid carcinoma

DTC

differentiated thyroid carcinoma

EMT

epithelial-to-mesenchymal transition

FTC

follicular thyroid cancer

MAPK

mitogen-activated protein kinase

MMR

mismatch repair

NGS

next-generation sequencing

PDTC

poorly differentiated thyroid cancer

PTC

papillary thyroid carcinoma

Funding

L.J.L.G. is a La Caixa Junior Leader fellow (LCF/BQ/PI20/11760011). I.L. is supported by the National Cancer Institute (NCI) Career Transition Award, grant number 1K22CA230381.

Disclosures

The authors declare that no conflicts of interest exist.

Data Availability Statement

Data sharing is not applicable to this article as no datasets were generated or analyzed during the current study.