RET Gene Fusions in Malignancies of the Thyroid and Other Tissues.

Following the identification of the BCR-ABL1 (Breakpoint Cluster Region-ABelson murine Leukemia) fusion in chronic myelogenous leukemia, gene fusions generating chimeric oncoproteins have been recognized as common genomic structural variations in human malignancies. This is, in particular, a frequent mechanism in the oncogenic conversion of protein kinases. Gene fusion was the first mechanism identified for the oncogenic activation of the receptor tyrosine kinase RET (REarranged during Transfection), initially discovered in papillary thyroid carcinoma (PTC). More recently, the advent of highly sensitive massive parallel (next generation sequencing, NGS) sequencing of tumor DNA or cell-free (cfDNA) circulating tumor DNA, allowed for the detection of RET fusions in many other solid and hematopoietic malignancies. This review summarizes the role of RET fusions in the pathogenesis of human cancer.

Structurally, the RET protein is composed by an extracellular (EC), a transmembrane (TM), and an intracellular (IC) portion ( Figure 1). RET-EC contains 4 cadherin-like (CLD) and one cysteine-rich (CRD) domains, that are involved in binding to the bipartite ligand [8]. RET-IC contains the tyrosine kinase domain (TKD) that is split into two subdomains [7,10]. This is followed by a C-terminal tail that is subject to alternative splicing generating different isoforms, the most abundant being RET9 and RET51 (depending whether they contain 9 or 51 residues starting from glycine 1063 in exon 19) [3].

Functional Consequence of RET Gene Fusions
RTK fusions in cancer may either result in the juxtaposition of a N-terminal partner to the Cterminal portion of the RTK, including its catalytic domain (so called 3' kinase fusion), or, vice versa, of the N-terminal portion of the RTK, with its catalytic domain, to the C-terminal of a fusion partner (5' kinase fusion) [77]. In both cases, the retention of an intact kinase domain in the fusion product is essential to distinguish cancer-driving RTK fusions from random chimeric products secondary to genetic instability [77]. 3' kinase fusions are the most common; however, examples of 5' kinase fusions have also been described, such as FGFR2 and FGFR3 (fibroblast growth factor receptors) fusions in cholangiocarcinoma and other malignancies [78] as well as the EGFR (epidermal growth factor receptor)-RAD51 (Radiation-sensitive 51) and -PURB (Purin Rich Beta) fusions in lung cancer [79].  Upon activation, several tyrosine residues of RET-IC undergo phosphorylation and mediate intracellular signal transduction. Thus, tyrosines Y900, Y905, Y981, Y1015, Y1062, and Y1096 (this one is specific for RET51) have been involved in functional RET signaling. Phosphorylated tyrosine 1062 (Y1062), in particular, recruits a multitude of adaptors such as SHC1/3, FRS2, IRS1/2, and DOK1/4/5 that, in turn, mediate the activation of RAS (Rat Sarcoma)-MAPK (Mitogen-Activated Protein Kinases) and PI3K (Phosphatidylinositol-3 Kinase)-AKT (Protein Kinase B) pathways [3][4][5][6][7].

RET Oncogenic Conversion in Human Neoplasms
Different RET molecular lesions have been described in tumors at either germline or somatic levels. These include gene amplification, fusion, as well as single base substitutions/small insertions/deletions either in sequences encoding RET-EC or -IC. Germline or somatic single base substitutions/small insertions/deletions in RET are characteristic of sporadic or familial (MEN2-multiple endocrine neoplasia type 2 associated) medullary thyroid carcinoma (MTC), respectively. Instead, RET fusions, occurring at the somatic level, are typical of papillary thyroid carcinoma, lung adenocarcinoma, and few other cancers. This notion has made RET an attractive molecular target for small molecule tyrosine kinase inhibitors (TKI) [11][12][13][14]. In this frame, novel selective RET TKIs have featured promising results in clinical investigation [14,15].
This review addresses, in particular, the role of RET gene fusions in cancer. Table 1 lists RET fusions so far described and Figure 1 depicts the protein structure of RET and its fusion partners.   1 Some RET fusions (EPHA5-RET, KHDRBS1-RET, PICALM-RET, SQSTM1-RET) are listed here but not reported in Figure 1 because of the lack of information about the involved exon. Moreover, additional RET fusion partners have been recently listed but without molecular details: CLIP1 and PRKG1 (15) and EML4 and PARD3 (14). 2

Functional Consequence of RET Gene Fusions
RTK fusions in cancer may either result in the juxtaposition of a N-terminal partner to the C-terminal portion of the RTK, including its catalytic domain (so called 3' kinase fusion), or, vice versa, of the N-terminal portion of the RTK, with its catalytic domain, to the C-terminal of a fusion partner (5' kinase fusion) [77]. In both cases, the retention of an intact kinase domain in the fusion product is essential to distinguish cancer-driving RTK fusions from random chimeric products secondary to genetic instability [77]. 3' kinase fusions are the most common; however, examples of 5' kinase fusions have also been described, such as FGFR2 and FGFR3 (fibroblast growth factor receptors) fusions in cholangiocarcinoma and other malignancies [78] as well as the EGFR (epidermal growth factor receptor)-RAD51 (Radiation-sensitive 51) and -PURB (Purin Rich Beta) fusions in lung cancer [79]. One argument supporting the selection of 3' kinase fusions in cancer may be that this type of fusion is able to move the kinase domain under the control of the transcriptional promoter of the fusion partners, thus fostering aberrant kinase overexpression (see also below) [77]. The cancer-associated RET fusions described so far are of the 3' kinase fusion type, involving a 5'-terminal partner coding sequence fused to the 3'-terminal RET kinase domain coding sequence. Breakpoints in RET and its fusion partners typically occur in intronic regions and are able to preserve, upon mRNA splicing, an open reading frame. Most commonly, secondary to a breakpoint in RET intron 11, the RET coding sequence from exon 12 to the STOP codon is included in the fusion. However, in rare instances, the fusion product starts from RET alternative exons, such as exon 3, 7, 9 (EC portion), 10 (TM: transmembrane segment), or 11 (IC portion).
RET gene fusions lead to kinase activation, in turn augmenting signal transduction along classical pathways such as the MAPK and the PI3K/AKT ones [4,5,80]. This causes gain of RET oncogenic activity, as shown for the most frequent fusions (CCDC6-RET, NCOA4-RET, KIF5B-RET) using cell-based assays as well as constitutive or conditional transgenic mouse models [28,29,50,51,[81][82][83][84]. Various mechanisms may contribute to fusion-mediated RET activation, including: (i) as mentioned above, the increased kinase expression, due to the replacement of the 5'-upstream RET promoter, that is normally active mainly in neuronal cell lineages, with that of the fusion partners [85], and (ii) the dimerization/oligomerization of the RET kinase domain mediated by protein-protein interaction motifs, typically coiled-coil domains, present in the fusion partners, that leads to ligand-independent kinase activation [65,86]. In addition, loss of autoinhibitory N-terminal portions and the altered conformation of the rearranged kinase may also contribute to RET kinase activation [87].
Besides the aberrant expression and the gain of intrinsic RET kinase activity, it is possible that other mechanisms contribute to the oncogenic role of RET fusions. First, the altered intracellular localization, due to the loss of the RET signal peptide and TM, may participate to RET transforming ability, for example by facilitating rearranged RET coupling to intracellular signal transducers. In addition, the RET fusion partner may act as a protein-protein interaction platform able to modify RET kinase-substrate(s) binding. Accordingly, in the KIF5B (Kinesin-1 heavy chain)-RET fusion, commonly found in lung carcinoma (see below), the N-terminal KIF5B backbone, containing the kinesin motor domain, was demonstrated to activate, through RAB GTPase vesicles, multiple RTKs, including EGFR and FGFR, and their downstream signaling [88]. Furthermore, the interaction of the chimeric RET kinase with another RTK may potentiate signaling and reduce efficacy of RET-targeted TKIs. In this frame, it has been shown that the physical interaction between CCDC6-RET and the EGFR facilitates signaling and reduces efficacy of RET kinase inhibition in a lung cancer cell line [89]. Finally, the concurrent altered function of the RET fusion partner may also contribute to neoplastic transformation. Accordingly, several RET fusion partners have been shown to exert relevant homeostatic functions (Table 1). PRKAR1A, one of the RET fusion partners, is the tumor suppressor protein involved in inheritance of Carney complex, a tumor-prone syndrome [90]. Similarly, another RET fusion partner, NCOA4, is a multifunctional protein that is involved in the regulation of DNA synthesis by inhibiting replication origin activation. Alteration of such a function in tumors harboring the NCOA4-RET fusion may contribute to its oncogenic activity [91]. Finally, CCDC6 exerts proapoptotic and DNA damage response activities and perturbation of these functions may participate to the oncogenic role of the corresponding RET fusion [92].

Genomic Mechanism of RET Gene Fusions
Different inter-chromosomal (translocations) or intra-chromosomal (inversions, tandem duplications, and interstitial deletions) structural variations can result in gene fusion events [87,93]. As discussed below, the most common RET fusions, in papillary thyroid carcinoma (PTC) and in lung adenocarcinoma (LADC) are CCDC6-RET and NCOA4-RET (primarily in PTC) and KIF5B-RET (primarily in LADC). CCDC6 and NCOA4 genes both map on the long arm of chromosome 10, at 10q21.2 or 10q11.22, respectively (Table 1). RET (that maps at 10q11.21) is transcribed in an opposite orientation with respect to CCDC6 and NCOA4. Thus, a 10q paracentric (not including the centromere, e.g., with both breakpoints in the same chromosome arm) inversion is the plausible genomic mechanism underlying CCDC6-RET and NCOA4-RET fusions [94]. Instead, KIF5B maps at 10p11.22; therefore, a pericentric (including the centromere, e.g., with a breakpoint in each chromosome arm) inversion of chromosome 10 is the mechanism of KIF5B-RET fusion (Table 1) [29,50,51,95]. In principle, inversions may be reciprocal (both fusion products) or nonreciprocal (only one fusion product). While in PTC, the majority of the inversions were reciprocal, in LADC, nonreciprocal inversions have also been documented [95].
RET gene fusions are thought to occur secondary to DNA double-strand breaks (DSB) followed by illegitimate repair of the two unrelated genes [13,18,95,96]. DNA repair mechanisms that can be responsible include NHEJ (nonhomologous end joining), which requires no homology or just very short homology sequences and inserts a few nucleotides at the fusion point, and BIR (break-induced replication), that instead requires long homology DNA stretches [95,96]. Sequencing of RET breakpoints in LADC suggested that both mechanisms can be involved, while, at least in radiation-associated PTC, NHEJ was the prevalent one [95,97].
In turn, DSBs can be induced by endogenous sources of DNA damage, such as reactive oxygen species (ROS) and replication stress, or exogenous sources, such as ionizing radiation (IR), see also below [98,99]. Chromosomal fragile sites are DSB-prone genome regions, in particular, under conditions that reduce DNA replication or upon the exposure to several chemical agents. It is worth noting that RET and NCOA4 are located within the fragile site FRA10G, and CCDC6 is located within the fragile site FRA10C, a fact that may facilitate their rearrangement [100][101][102][103]. Moreover, breakpoints in RET intron 11 map close to DNA topoisomerase 1/2 predicted cleavage sites, suggesting that DNA topoisomerases may play a role in RET fusion events [102]. Accordingly, during replication and transcription, DNA is unwound by DNA helicase, resulting in torsion that can be removed by topoisomerases through transient introduction of DNA breakage, thus potentially favoring gene rearrangements in cancer cells [102]. Finally, tissue-specific organization of chromatin territories may affect the frequency as well as the specificity of RET fusion with specific gene partners. In this frame, the overlap in thyrocyte interphase chromatin of CCDC6, NCOA4, and RET loci is important to favor their fusion [100].
Chromoplexy occurs when several DSBs are generated at the same time in different chromosomes. This is followed by chains of rearrangements, resulting in the simultaneous shuffling of several chromosomal fragments. Chromoplexy is typically found in prostate carcinoma and lymphoid malignancies [93]. However, in the recent pan-cancer analysis of whole genomes, it was found that 4 of the 13 fusion genes identified in thyroid carcinoma, including 2 RET fusions, were caused by chromoplexy [93].
PTC is typically associated with lesions in genes causing unscheduled activation of the MAPK (mitogen-activated protein kinase) cascade. Indeed, besides RET and other RTKs, common PTC-driver events are represented by gain-of-function mutations of RAS small GTPases, and, most commonly, BRAF kinase. The mutually exclusive nature of these lesions supports the notion they act along a common signaling mechanism [104]. In the TCGA study, RET-, similarly to BRAF-positive samples, were "classic" papillary cases, whereas RAS-positive samples typically featured the follicular variant phenotype [24].
RET fusions occur in~10%-20% of sporadic PTCs [15,18]. In the TCGA study, enrolling almost 500 PTC samples,~6.8% of cases harbored a RET fusion [24]. RET fusions are more common in radiation-associated than in "sporadic" cases. Several evidences support the possibility that both internal (to 131 I and other isotopes) and external irradiation may favor these genomic events [105]. In 1986, large amounts of 131 I and other isotopes, released by the Chernobyl nuclear power plant, contaminated regions of Belarus, Russia, and Ukraine, thus resulting in a sharp PTC incidence increase in children and adolescents. In one study,~58% of post-Chernobyl PTCs in patients who were <10 years old at the time of the accident harbored a RET fusion [106]. Prevalence of RET fusions tended to decline after a longer interval from the nuclear accident [16,107]. Moreover, as many as 50% of PTCs in atomic bomb survivors, who were exposed to high radiation doses (>0.5 Gy), harbored RET fusions [108]. Accordingly, RET fusions occur dose-dependently upon irradiation with or X-ray thyrocytes in culture [24,98,109].
RET fusions have been reported more commonly (and the BRAFV600E mutation more rarely) in pediatric than in adult thyroid cancer patients [16,110,111]. In young patients, the NCOA4-RET fusion occurred more commonly in radiation-associated, particularly in early cases, while CCDC6-RET was more common in "sporadic" cases [112].
As mentioned above, the most typical oncogenic drivers of MTC, a thyroid carcinoma arising from C-cells, are RET point mutations, both in sporadic and familial cases [4][5][6][7][16][17][18]. It is worth noting that RET is normally expressed in C-cells and therefore, its oncogenic conversion does not need the acquisition of a novel transcriptional promoter caused by a gene fusion. However, one MTC case harboring a RET fusion, MYH13-RET, has been recently reported [55].
In terms of the clinical relevance of RET fusions in thyroid cancer, besides the possible therapeutic use of RET TKIs, it should be noted that fine-needle aspiration cytology is commonly used for pre-operative assessment of thyroid nodules, though it is unable to reach a definitive diagnosis in up to 25% of the cases. In these cases, molecular diagnostic methods have been developed to help diagnosis by determining the presence of different oncogenic mutations, including RET fusions [113].
In LADC, RET fusions are reported to be associated with young age (<60 years), female gender, Asian ethnicity, and minimal tobacco exposure [17]. RET fusions were also associated with poor differentiation, solid subtype, presence of signet ring cells, small tumor size, early lymph node metastases, low levels of PD-L1 expression, and poor response to immunotherapy [13,15,17]. In vitro irradiation with -rays generated KIF5B-RET fusion in lung cells, thus pointing, similar to thyroid cancer, to radiation as a possible risk factor for RET fusion in lung cancer [117].
RET is a transcriptional target of estrogen receptor alpha (ESR1) and several studies have reported its overexpression, particularly in ER+ breast cancer as well as its association with endocrine resistance [124]. In a recent study based on targeted genomic profiling of 9693 cases, RET genomic alterations, including 16 rearrangements, were observed in 1.2% of breast cancers. These rearrangements featured the classical CCDC6-RET and NCOA4-RET fusions, the new uncharacterized RASGEF1A-RET and ZNF485-RET fusions, and rearrangements resulting in tandem duplications that involve exons 12-19 of RET [32].
RET fusions have been identified in <1% of colorectal carcinomas (CRC) [33,34,66]. The most common fusions in CRC are CCDC6-RET and NCOA4-RET. More rare fusions involving TNIP1-, SNRNP70-, GEMIN5-and RRBP1 as N-terminal partners were also described (Table 1) [34,35]. In CRC, RET fusions were more frequent in old patients, right-sided, RAS/ BRAF wild-type, and MSI (microsatellite instable)-high tumors and were associated with negative prognosis. Accordingly, as many as 26% of patients with right-sided RAS and BRAF wild-type tumors harbored a RET rearrangement. This fraction significantly increased when only MSI-high CRCs were considered [34].
Salivary gland carcinomas are rare and heterogeneous tumors with several different subtypes. Secretory carcinoma, originally described as mammary analogue secretory carcinoma (MASC), is a salivary gland low-grade carcinoma typically associated with ETV6-NTRK3 fusion [42]. Recently, some secretory carcinomas, affecting the parotid or the submandibular gland, and negative for ETV6-NTRK3, were found to harbor ETV6-or NCOA4-RET fusions [42]. Recurrent rearrangements involving the RET gene were also identified in a subset of intraductal carcinoma (IC), another salivary duct carcinoma that shares some morphologic and immunophenotypical features with secretory carcinoma. Specifically, more than 40% of the IC cases harbored RET fusions, including NCOA4-RET, TRIM27-RET, and KIAA1217-RET [59,60,125].

Conclusions
Though initially believed to be PTC-specific, RET gene fusions have more recently been identified in several cancer types, including lung, colon, and breast carcinoma. In terms of clinical translation, this has significantly raised the interest for RET as far as the possibility of developing molecular diagnostic methods as well as targeted treatments with RET TKIs. On the other hand, from a mechanistic point-of-view, RET fusion was revealed to be an interesting model to study the mechanism of gene rearrangement and the role of radiation in cancer. Funding: This study was supported by the Associazione Italiana per la Ricerca sul Cancro (AIRC) IG 20793, by the POR Campania FESR 2014-2020 "SATIN" grant and by the NIH 1R01CA197178-01A1R grant.

Conflicts of Interest:
The authors declare no conflict of interest.