PDX models directly use patient-derived tissues, recreate human cancer gene expression, tumor heterogeneity, and microenvironment, and are expected to inform drug response better than standard xenograft models [19]. These models are successfully established in many cancers but are still limited in thyroid cancer research. Maniakas et al. [19] generated PDX model using six ATC tumors: the tumor implanted in the flank of immunodeficient athymic mice and expanded for at least four generations. Mutation, histopathological characterization of PDX models showed good fidelity with the original tumor. The take rate was about 30% [19]. This group of researchers reported the result of high-throughput screening multiple drugs using these models and matching cell lines and suggested the feasibility of using this systematic approach for preclinical in vivo drug testing or personalized therapeutics development [29].

The syngeneic model is established by implanting mouse cancer cells or tissues into immunocompetent mice of the same genetic background. Like the xenograft model, the syngeneic model can be ectopic (subcutaneous), orthotopic (thyroid bed), or metastatic (tail vein or intracardiac) [21]. Vanden Borre et al. [30] reported the development of syngeneic ATC orthotopic mouse model using cell lines derived from tumors arising in GEMM with thyroid-specific expression of mutant BRAF and deletion of either TP53 or phosphatase and tensin homolog (PTEN), which were implanted into immunocompetent syngeneic B6129SF1/J mice. Caperton et al. [31] reported the development of syngeneic follicular thyroid cancer (FTC) ectopic mouse model using tumor cells from GEMM with mutant HRas and PTEN inactivation. These models showed very rapid tumor formation with significant immune cell infiltration in the tumor [30-32]. The syngeneic model has a significant advantage over the xenograft model in that researchers can study tumor growth with an intact immune system and investigate the immunemodulatory effect of therapies. The short latency of tumor development also makes this an attractive model [21,30]. The major drawback of this model is that the immunoediting process, so critical in human tumor development, may not proceed due to very rapid tumor formation, which introduces a model bias in the response of immunotherapeutic agents [33]. Because of very rapid tumor growth in most syngeneic mouse models, distant metastasis cannot be easily studied [30].

The GEMM is established by genetic engineering tools and the Cre-loxP system is the most widely used [34]. The expression of oncogenes and/or loss of tumor suppressor genes can be spatially edited by controlling the expression of Cre recombinase using cell-specific regulatory elements such as thyroglobulin (Tg) or thyroid peroxidase (TPO), as well as temporally controlled by further engineering of Cre and exogenous inducers like tamoxifen (Table 4) [35].

As BRAF V600E mutations account for the majority in PTC and ATC tumors, many GEMMs contain BRAF V600E [2]. The first BRAF GEMM used the Tg promoter and constitutive BRAF V600E cDNA overexpression resulting PTC with high penetrance. However, the BRAF transgene led to dedifferentiation, which exhibited a loss of Tg expression, causing a negative feedback loop on the Tg-BRAF transgene [36]. To overcome this, a subsequent model used tamoxifen-inducible Cre (CreER^T2) under the control of the Tg promoter and Cre-activating BRAF V600E (Braf^CA) allele. PTCs were seen 6 months after tamoxifen injection and progressed to advanced DTC at 12 months. However, distant metastases were not detected [37]. Chakravarty et al. [38] also reported that conditional activation of BRAF V600E with doxycycline in the thyroid under the control of the Tg promoter resulted in mitogen-activated protein kinase kinase (MEK) and extracellular signal-regulated kinase (ERK) activation and PTC formation after 1 week of drug exposure. The treatment with a MEK inhibitor in this model restored tumor radioiodine uptake [38], leading to many human studies exploring BRAF and MEK inhibition to improve radioiodine uptake in humans with DTC [39,40]. All these BRAF tumor models showed decreased thyroid function and elevated thyroid stimulating hormone (TSH) due to dedifferentiation [36-38,41]. Many investigators treat these mouse models with levothyroxine to mitigate the effect of high TSH on the model and better approximate the human condition. McFadden et al. [42] reported that BRAF V600E initiates PTC in the adult mouse thyroid and additional TP53 loss enables the progression to ATC. PTC tumors developed in mice with TPO-CreER^T2/Braf^CA/+ at 12 weeks postinduction. Crossing this strain with the mice with homozygous deletion of p53 (Trp53^f/f) further accelerated to overt ATC. This model is referred to as TBP. An explosive tumor growth was seen approximately 100 days after tamoxifen treatment. They also showed that the selective BRAF inhibitor was not effective in controlling these ATC tumors and the combination of MEK/BRAF inhibitor was more effective [42]. Additional activation of phosphatidylinositol-4,5-bisphosphate 3-kinase catalytic subunit alpha (PI3KCA) (Tg-CreER^T2/Braf^CA/PIK3CA^H1047R) or knock-out of PTEN (Tg-CreER^T2/Braf^CA/PTEN^f/f) also resulted in the development of ATCs [43].

RET/PTC or neurotrophic tyrosine receptor kinase (NTRK) rearrangement are also drivers of PTC [44,45]. The RET/PTC1 [46,47] and RET-PTC3 [48] transgenic mouse models using Tg promoter were reported, and both resulted in the development of PTC. However, these RET/PTC rearrangement in mice are germline, and PTC development and tumor progression may be more advanced in these mouse models compared with sporadic genetic alterations in humans. Buckwalter et al. [49] reported that mutation of major tyrosine signaling sites in RET/PTC1 transgenic mice didn’t block the development of the tumor. This suggests that other signaling pathways, or the microenvironment may play a major role in tumor formation [49]. The fusion of translocated promoter region (TPR) with NTKR1 (TRK-T1) developed follicular hyperplasia or PTC in about 50% of the mice [50].

RAS activation is a well-known driver in FTC. An early mouse model containing a mutant K-Ras^G12V gene under the control of rat Tg promoter showed no thyroid cancer development, and only one of these mice developed FTC following treatment with goitrogens (aminotriazol and sodium perchlorate in the drinking water) for 6 months [51]. This result suggests that RAS mutation alone or this degree of RAS signaling is insufficient to induce FTC and other alterations are required. One study using transgenic mice harboring the human N-Ras^Q61K under the control of the bovine Tg promoter reported that approximately 40% of the mice developed invasive FTC, in some with a mixed papillary morphology and about 25% showed de-differentiation with distant metastases [52]. These data suggest that secondary mutations may occur, or multiple copies of the transgene could be incorporated and lead to the abnormally high transcript expression [24]. These older transgenic mouse models of FTC have a limitation in that the cancer phenotype resulting from overexpression of a mutated RAS gene is different from the endogenous mutant RAS expressed at the physiological level. Miller et al. [53] crossed a mouse strain with TPO-Cre and an oncogenic K-Ras^G12D which is conditionally expressed through Cre-mediated deletion of a floxed STOP cassette preventing K-Ras^G12D expression (lox-stop-lox [LSL]-K-Ras), with a mouse strain with PTEN loss (PTEN^f/f). All of the double-mutant mice (TPO-Cre/LSL-K-Ras^G12D/PTEN^f/f) rapidly developed FTC and none survived over age 4 months [53]. Thyroid-specific double protein kinase cAMP-dependent type I regulatory subunit alpha (PRKAR1A)-PTEN knock-out mice also developed FTC [54]. Serum TSH levels were suppressed and thyroxine was elevated in these double-mutant mice, indicating development of thyrotoxicosis from functional FTC [53,54]. Thyroid-specific expression of H-Ras^G12V and PTEN inactivation (TPO-Cre/H-Ras^G12V/PTEN^f/f) also leads to the development of high grade FTCs or poorly differentiated thyroid cancer (PDTC) showing gross extrathyroidal extension and/or lymphovascular invasion, and lung metastasis [55].

A dominant negative mutation (PV) into the thyroid hormone nuclear receptor β (TRβ) generates another FTC mouse model. TRβ^PV/PV mice reproduce human thyroid hormone resistance syndrome. T3 is unbale to bind mutant TRβ, and follicular cells become hyperplastic with elevated TSH levels. Invasive FTC was observed at 4 to 5 months of age and distant metastases to lung or heart was seen over 5 months of age. TRβ^PV/− mice from crossing TRβ^PV/PV mice with TRβ^−/− mice also showed spontaneous FTC with lung metastasis [56]. The phosphoinositide 3-kinase (PI3K)-AKT signaling was activated in this model and PI3K inhibitor treatment inhibited FTC progression in these mice [57].

The GEMM model best interrogates the tumor microenvironment and the interaction between the tumor and immune system because the tumor is developed more slowly over time, not grafted [25,26,33]. Furthermore, this model helps to dissect the effect of each genetic alteration on tumor initiation and progression [2]. This model has limitations including long latency to tumor formation, complex breeding schemes, and high costs [25].