Targeted Expression of BRAF^V600E in Thyroid Cells of Transgenic Mice Results in Papillary Thyroid Cancers that Undergo Dedifferentiation

The BRAF^T1799A mutation is the most common genetic alteration in papillary thyroid carcinomas (PTC). It is also found in a subset of papillary microcarcinomas, consistent with a role in tumor initiation. PTCs with BRAF^T1799A are often invasive and present at a more advanced stage. BRAF^T1799A is found with high prevalence in tall-cell variant PTCs and in poorly differentiated and undifferentiated carcinomas arising from PTCs. To explore the role of BRAF^V600E in thyroid cancer pathogenesis, we targeted its expression to thyroid cells of transgenic FVB/N mice with a bovine thyroglobulin promoter. Two Tg-BRAF^V600E lines (Tg-BRAF2 and Tg-BRAF3) were propagated for detailed analysis. Tg-BRAF2 and Tg-BRAF3 mice had increased thyroid-stimulating hormone levels (>7- and ∼2-fold, respectively). This likely resulted from decreased expression of thyroid peroxidase, sodium iodine symporter, and thyroglobulin. All lines seemed to successfully compensate for thyroid dysfunction, as serum thyroxine/triiodothyronine and somatic growth were normal. Thyroid glands of transgenic mice were markedly enlarged by 5 weeks of age. In Tg-BRAF2 mice, PTCs were present at 12 and 22 weeks in 14 of 15 and 13 of 14 animals, respectively, with 83% exhibiting tall-cell features, 83% areas of invasion, and 48% foci of poorly differentiated carcinoma. Tg-BRAF3 mice also developed PTCs, albeit with lower prevalence (3 of 12 and 4 of 9 at 12 and 22 weeks, respectively). Tg-BRAF2 mice had a 30% decrease in survival at 5 months. In summary, thyroid-specific expression of BRAF^V600E induces goiter and invasive PTC, which transitions to poorly differentiated carcinomas. This closely recapitulates the phenotype of BRAF-positive PTCs in humans and supports a key role for this oncogene in its pathogenesis.

Papillary thyroid carcinoma (PTC) is the most common type of differentiated thyroid carcinoma (1). Many of the genetic events involved in the initiation of PTC have been identified (2). Notable among them are the RET/PTC oncogenes, which until recently were the most characteristic oncogenic event in this tumor type, particularly in pediatric cancers and in patients with prior exposure to ionizing radiation (3). Recently, an activating mutation of BRAF was found in 36% of PTC, making it the most common oncogene thus far identified in sporadic forms of the disease (4). The mutation was exclusively a thymine-to-adenine transversion at position 1799, previously designated as 1796, leading to a valine-to-glutamate substitution at residue 600 (V600E), formerly designated as V599 (5). The BRAF mutations are unique to PTC and not found in other types of well-differentiated thyroid follicular neoplasm. The initial observations have now been confirmed by other reports finding a prevalence of this mutation in 29% to 83% of PTCs (4, 6–13). Overall, of the 916 PTC reported to date, 42% were positive for the BRAF^T1799A mutation.

There are three isoforms of the serine-threonine kinase RAF in mammalian cells: ARAF, BRAF, and CRAF or RAF1. CRAF is expressed ubiquitously, whereas BRAF is expressed at higher levels in hemopoietic cells, neurons, and testes (14). BRAF is also the predominant isoform in thyroid follicular cells.⁶

Although all RAF isoforms activate mitogen-activated protein (MAP)/extracellular signal–regulated kinase (ERK) kinase (MEK), they are differentially activated by oncogenic RAS. In addition, BRAF has higher affinity for MEK1 and MEK2 and is more efficient in phosphorylating MEKs than other RAF isoforms (15). The initial discovery of BRAF mutation indicated a high prevalence of this event in malignant melanomas (16) and in a smaller subset of colorectal and ovarian cancers (16). Recent resolution of the crystal structure of the wild type and BRAF^V600E kinase domains helps understand the mechanisms of mutational activation of the protein (17). BRAF exhibits the characteristic bilobar structure of protein kinases. In its inactive conformation, BRAF residues G597-V601 in the activation loop form hydrophobic interactions with residues G465-V472 in the ATP binding site (P loop), resulting in a structure that is not aligned for binding to ATP or substrate. Oncogenic mutations in the activation loop or the P loop disrupt their interaction and destabilize the inactive conformation. Most, but not all of known oncogenic BRAF substitutions allow the formation of new interactions that fold the kinase into a catalytically competent structure (reviewed in refs. 18, 19). Paradoxically, some of the oncogenic BRAF mutants impair in vitro kinase activity (17). Despite this, these low-activity kinase BRAF mutants induce ERK phosphorylation, which is due to activation of CRAF, presumably by heterodimerization (17).

Several groups have examined PTCs for the presence of RET/PTC, BRAF, and RAS mutations, all of which can activate the MAP kinase (MAPK) signaling pathway (4, 6, 20). Altogether, 177 PTCs were studied, and one of these alterations was present in about 70% of tumors. However, there was no single PTC with a mutation involving more than one of these genes. The lack of concordance provides compelling genetic evidence for a requirement of mutation of MAPK signaling components for transformation to PTC. This is consistent with evidence that RET/PTC-induced dedifferentiation (21) and thyroid-stimulating hormone (TSH)–independent growth (22) are dependent on activation of the MAPK pathway in thyroid cell lines.

It is now clear that PTCs with BRAF mutations have distinct phenotypic and biological properties. Although most have a classic papillary architecture, almost all tall-cell variant PTCs are positive for BRAF^T1799A (8, 20). PTCs with a BRAF^T1799A mutation present more commonly at an advanced stage of the disease (8, 10) and with extrathyroidal extension (8). Undifferentiated (anaplastic) and poorly differentiated carcinomas arising from preexisting PTCs have a significant prevalence of BRAF mutations, whereas those arising from preexisting follicular carcinoma do not (8, 10, 23). In addition, a subset of microscopic PTC, thought to be early precursors of these cancers, harbors BRAF mutations, indicating that this oncogene may be activated during tumor initiation (8). These data suggest that BRAF mutations may be a tumor-initiating event in PTC and associated with tumor dedifferentiation and more aggressive behavior. To test this in an animal model, BRAF^V600E expression was targeted to thyroid follicular cells of transgenic FVB/N mice with a bovine thyroglobulin promoter. Here we provide the detailed characterization of two lines with thyroid-specific expression of BRAF^V600E.

Creation of Tg-BRAF mice. The human BRAF^T1799A cDNA with an NH₂-terminal myc tag (gift from Richard Marais, Institute of Cancer Research, United Kingdom) was cloned into the BamHI site of pSG5 downstream of the β-globin intron. The ClaI/SalI fragment from pGS5/myc-BRAF^T1799A containing the β-globin intron and myc-BRAF^T1799A, was cloned downstream of the bovine thyroglobulin promoter into the ClaI/SalI site of pSKbTg (Fig. 1A; ref. 24). The SpeI/SalI fragment was injected into fertilized FVB/N mouse eggs, which were implanted into pseudopregnant female mice. The pups were confirmed by PCR and Southern blotting to have integrated the transgene. Lines were then generated from founder animals by crossing them with wild-type FVB/N mice.

Real-time reverse transcription-PCR. Thyroid lobes were surgically removed, weighed, and immediately placed in liquid N₂. RNA was isolated using Tri-reagent (Molecular Research Center, Inc., Cincinnati, OH) and 0.6 μg was reverse transcribed with SuperScript III (Invitrogen, Carlsbad, CA) in the presence of random hexamers to generate cDNA. Quantitative reverse transcription-PCR (RT-PCR) was done using QuantiTect SYBR Green PCR Kit (Qiagen, Valencia, CA) and primer pairs for β-actin (5′-ctgaaccctaaggccaaccgtg-3′ and 5′-ggcatacagggacagcacagcc-3′); myc-tagged BRAF (5′-caagcttatcgatcctgagaact-3′ and 5′-ggctccatgtccccgttgaa-3′); thyroglobulin (5′-tgtcccaccaagtgtgaaaa-3′ and 5′-ccaaggaaagcttgttcagc-3′); sodium iodide symporter, NIS (5′-gctcagtctcgctcaaaacc-3′ and 5′-cgtgtgacaggccacataac-3′); thyroid peroxidase, TPO (5′-tgacttccaggagcacacag-3′ and 5′-gcaagttcagtgatgccaga-3′); and the TSH receptor, TSHR (5′-ctctcttacccgagccactg-3′ and 5′-ttgtcacccggatcttcttc-3′). The cycle threshold values for β-actin and the target gene were determined using Lightcycler (Cepheid, Sunnyvale, CA) and used to calculate the normalized relative expression using the QGENE program (25).

Western blots. Thyroid tissue was placed in buffer B [20 mmol/L Tris-HCl (pH 7.5), 1 mmol/L EDTA, 1 mmol/L EGTA, 1.0% Triton X-100, 50 mmol/L NaF, 1 mmol/L Na orthovanadate, and protease inhibitor cocktail; Sigma, St. Louis, MO] and homogenized using a polytron. Protein lysates were centrifuged, supernatant collected, and protein concentration determined using micro-bicinchoninic acid (Pierce, Rockford, IL). Protein lysates were subjected to SDS PAGE, transferred to nitrocellulose membranes and probed with antibodies (Santa Cruz Biotechnology, Santa Cruz, CA) to phospho-ERK1/2, total ERK, or BRAF. Membranes were hybridized with species-specific horseradish peroxidase–conjugated antibodies (Santa Cruz Biotechnology), and bands visualized with Enhanced Chemiluminescence as directed by manufacturer (Amersham, Piscataway, NJ).

RIAs. Blood from mice was collected immediately after euthanasia with CO₂. Blood was incubated on ice for 1 hour, centrifuged at 4°C for 15 minutes, serum removed, and stored at −70°C until assayed. Serum TSHs were determined as previously described (26). The serum levels of total thyroxine and triiodothyronine were measured by solid-phase RIAs (Diagnostic Products, Los Angeles, CA) and the free thyroxine was estimated from the total thyroxine and the resin thyroxine uptake ratio and expressed as the free thyroxine index (FT4I; ref. 27).

Histology and immunohistochemistry. Thyroid tissues were fixed in 10% neutral-buffered formalin and embedded in paraffin. Five-micrometer-thick sections were prepared and stained with H&E. Immunostaining was done on paraffin sections using avidine-streptavidin immunoperoxidase method with rabbit antimouse thyroglobulin immunoglobulin G (a gift from Paul Kim, University of Cincinnati) at a dilution of 1:100 on automated ES system (Ventana Medical Systems, Inc.,Tuscon, AZ) without additional antigen retrieval.

Statistical analysis. Difference between male and female transgenic animals in the frequency of histologic features was assessed by Fisher's exact test. Difference in survival was assessed by log-rank test. For all others, a two-tailed t test was used.

Of seven original Tg-BRAF^V600E founders showing integration of the intact transgene as determined by Southern blotting, four had transgenic litters with low or undetectable expression of myc-tagged BRAF in thyroid tissue. One founder was runted and died at 6 weeks of age without progeny, despite treatment with thyroid hormone given on the presumptive diagnosis of hypothyroidism. Examination of this animal revealed a large goiter, papillary thyroid cancer and elevated TSH levels. The remaining two lines, Tg-BRAF2 and Tg-BRAF3, with thyroid-specific expression of BRAF^V600E were propagated and characterized in detail. Quantitative RT-PCR with primers spanning the β-globin intron (Fig. 1A) and thus specific to the BRAF^V600E transcript, showed the presence of BRAF^V600E mRNA in thyroid tissues of both Tg-BRAF2 and Tg-BRAF3 mice, with significantly higher levels of expression in Tg-BRAF2 animals (Fig. 1B). Western blots of protein extracts from thyroid tissues of Tg-BRAF2 mice confirmed the increase in BRAF levels, with males having higher expression than females (Fig. 1C). The robust increase in phospho-ERK (pERK) levels (males > females) confirms expression of a functional protein. The weaker increase in pERK levels in Tg-BRAF3 mice was consistent with lower expression of BRAF^V600E in this line (Fig. 1C).

Because mice with thyroid-specific expression of RET/PTC1 (28, 29), an upstream activator of BRAF,⁷

were found to have stunted growth due to hypothyroidism, we monitored growth in animals with thyroid-specific expression of BRAF^V600E. For this, weight of transgenic mice and wild-type littermates was followed for 12 weeks. Tg-BRAF3 mice had growth rates indistinguishable from controls (Fig. 2A). Tg-BRAF2 mice initially grew marginally slower than nontransgenic littermates. The difference in body weight was more pronounced in males than in females and proved to be transient, as it was not observed past 8 weeks (Fig. 2A). To determine the effects of BRAF^V600E expression on thyroid function, serum levels of TSH, total triiodothyronine, total thyroxine, and free thyroxine levels were determined. At 5 weeks, TSH levels in male Tg-BRAF2 mice were on average 80-fold greater than those of age/sex-matched wild-type littermates. TSH levels markedly declined, although they still remained significantly elevated at 8 and 12 weeks (Table 1). In female Tg-BRAF2 mice, serum TSH levels were 5- to 8-fold above those of age/sex-matched wild-type littermates and did not vary significantly with time. Tg-BRAF3 mice had a 2-fold increase in serum TSH with no significant difference between males and females. Despite the elevated TSH levels, serum total thyroxine and free thyroxine in Tg-BRAF2 and Tg-BRAF3 mice were not significantly lower than sex- and age-matched wild-type littermates at any time point (Table 1). Serum total triiodothyronine levels trended lower at all time points in both male and female Tg-BRAF2 mice but were only statistically significant in 8-week-old male animals (Table 1). Triiodothyronine levels in Tg-BRAF3 mice were not statistically different from sex- and age-matched wild-type littermates.

Oncogenic activation of the MAPK pathway results in decreased expression of thyroglobulin, NIS, and TPO in clonal well-differentiated thyroid cell lines (21, 30, 31). In addition, the decrease in thyroid function in Tg-RET/PTC mice has been attributed, at least in part, to reduced expression of NIS (28). Consistent with these observations, in 5-week-old male Tg-BRAF2 mice TPO, thyroglobulin, and NIS mRNA levels were 45%, 75%, and 95% lower than those found in wild-type littermates, respectively (Fig. 2B). Similar findings were obtained in 8-week male and 5-week female Tg-BRAF2 mice (data not shown). The decrease in mRNA levels of these genes is not due to changes in expression of the TSH receptor, as TSH receptor mRNA levels were not lower than wild-type littermates in 5-week-old male Tg-BRAF2 mice, which have the highest serum TSH levels (data not shown). In 5-week-old Tg-BRAF3 mice, expression of these gene products did not vary significantly from wild-type littermates, although there was a trend towards a decrease in NIS and TPO (Fig. 2B). Taken together, these data indicate that Tg-BRAF mice were euthyroid, having compensated for the BRAF-induced thyroid dysfunction through increased TSH levels, and as shown below, goiter development.

Thyroid glands of Tg-BRAF2 mice, and to a lesser extent Tg-BRAF3 mice, were markedly larger than age- and sex-matched wild-type littermates (Fig. 3). Histologic examination of thyroid glands from 12- and 22-week-old Tg-BRAF2 mice revealed multifocal tumors typically involving both lobes of the gland and having mixed papillary and follicular growth pattern (Fig. 4A). The tumor cells showed nuclear features characteristic of human papillary carcinomas, including nuclear enlargement, overlapping and crowding, irregular nuclear contours, and occasional chromatin clearing and nuclear grooves (Fig. 4B and C). In addition, almost all tumors in Tg-BRAF2 mice had focal areas showing well-defined tall-cell features (Fig. 4D; Table 2). Approximately 50% of the 12- and 22-week-old animals revealed focal areas of dedifferentiation composed of solid sheets of spindle cells lacking characteristic nuclear features of papillary carcinoma and showing no evidence of follicular architecture or colloid formation (Fig. 4E and F). Mitotic figures were frequently seen in these areas. Microscopic appearance of these areas was comparable with those of human poorly differentiated carcinoma, whereas no severe nuclear atypia or tumor necrosis, characteristic of human anaplastic (undifferentiated) carcinoma, was observed. Areas of dedifferentiation were frequently surrounded by well-differentiated tumor with prominent tall-cell appearance. Thyroglobulin staining of thyroid section showed reduced staining in the poorly differentiated areas (data not shown); however, the reduction in thyroglobulin levels did not correlate with the degree of morphologic dedifferentiation. The tumors in Tg-BRAF2 mice showed aggressive malignant features as they frequently invaded blood vessels and thyroid gland capsule with extrathyroidal extension into adjacent adipose tissue and skeletal muscle (Fig. 4G-I; Table 2). The frequency of dedifferentiation and vascular invasion were higher in male than in female animals (P = 0.026 and 0.0052, respectively). In Tg-BRAF3 mice, 3 of 12 and 4 of 9 animals at 12 and 22 weeks, respectively, had small tumor foci with follicular architecture and cytologic features of PTC (Table 2). One of the PTC foci presented with focal tall-cell features. None of the tumors showed extrathyroidal extension, vascular invasion, or progression to poorly differentiated carcinoma.

The aggressive cancer in Tg-BRAF2 mice was associated with a 30% decrease in survival at 22 weeks. No significant change in survival was found for Tg-BRAF3 mice (Fig. 5). To determine whether the decreased survival of the Tg-BRAF2 mice was due to distant metastasis, we examined lungs and hearts, two organs found to be a common site of thyroid cancer metastasis in another animal model for thyroid cancer (32). However, no evidence of metastasis to these tissues was found. The lungs from the deceased animals showed microscopic features of acute asphyxiation, suggesting that their death may have been caused by upper airway compression.

The BRAF^T1799A mutation is the most common oncogenic event thus far identified in PTC (4, 6–13). This mutation can occur early in tumor development, as it is present in papillary microcarcinomas, which are thought to represent an early stage of PTC (8). The lack of overlap between RET/PTC, RAS, and BRAF mutations in PTC (4, 6, 20) indicates that a single oncogenic hit along the MAPK pathway provides sufficient activation of this pathway that, when combined with other changes, results in transformation. Despite this, the nature of the oncogenic event has clear effects on the phenotypic features and biological behavior of the resulting cancer. For example, PTC with RAS mutations are almost exclusively of the follicular variant (33), whereas those with RET/PTC3 rearrangements are associated with the solid variant of the disease (34). PTC with BRAF^T1799A are associated with a classic histotype and with tall-cell variant, which has been associated with more aggressive tumor behavior (8, 20). Furthermore, RET/PTC rearrangements are rarely if ever found in poorly differentiated or anaplastic thyroid carcinomas. By contrast, BRAF^T1799A is commonly found in poorly differentiated and undifferentiated carcinomas arising from preexisting PTC (8), suggesting that this oncogenic event promotes tumor dedifferentiation and confers patients with a significantly worse prognosis.

Some of the histologic features associated with a particular oncogenic hit in human PTC have also been found in their corresponding mouse models. Thus, mice with thyroid-specific expression of RET/PTC3 develop solid variant PTC (35), and those with expression of RET/PTC1 develop tumors with a more classic PTC architecture (36). Consistent with the human data, PTC that develop in either Tg-RET/PTC1 or Tg-RET/PTC3 mice do not progress to poorly differentiated carcinomas (35, 36) unless crossed with p53^−/− mice (37, 38). We show here that mice with thyroid-specific expression of BRAF^V600E develop PTCs that closely recapitulate the properties BRAF-positive human PTC (i.e., classic PTC architecture, tall-cell features, and high potential for invasiveness). Moreover, in the majority of the mice from the Tg-BRAF2 line there is progression to poorly differentiated carcinomas. These results indicate that BRAF^V600E can serve as a tumor initiator and promote progression to poorly differentiated carcinomas. By contrast to human poorly differentiated or anaplastic cancers that typically exhibit almost complete loss of thyroglobulin-immunoreactive cells (39), the poorly differentiated foci present in the Tg-BRAF2 mice still expressed thyroglobulin, albeit at lower levels than surrounding well-differentiated areas (data not shown). However, this difference may simply represent a limitation of this particular mouse model. As transgene expression is driven by the thyroglobulin gene promoter and dedifferentiation results in reduction in thyroglobulin promoter activity, this will be associated with a corresponding decrease in BRAF^V600E expression. It is tempting to speculate that persistence of thyroglobulin expression indicates a continued requirement for BRAF^V600E even in the poorly differentiated cancers.

Lower body weight in the younger Tg-BRAF2 mice suggests that the animals may have been hypothyroid at some point during development. However, despite markedly elevated TSH levels, serum thyroxine and free thyroxine levels were normal in both transgenic mouse lines at all times. There was a subtle decrease in triiodothyronine levels at one of the time points, but this was not sustained. The most likely explanation for the elevated TSH is that the Tg-BRAF mice were markedly hypothyroid in late gestation and in the early neonatal period. This was likely compensated by gradual thyroid enlargement, which by 5 weeks of age was at least 6-fold greater than nontransgenic littermates. As we did not measure free thyroid hormone levels at these early time points, the period of hypothyroidism was not detected. Thus, the data are consistent with gradual development of a compensated hypothyroid state, arising because of BRAF-induced impairment of TSH action in thyroid follicular cells.⁷ Higher TSH levels are required to reach a new steady state, which is arrived at in part through development of goiter. Perhaps, successful compensation was possible in Tg-BRAF animals because TSHR gene expression was unaffected hence allowing for adequate TSH responsiveness. This is in contrast to what occurs in mice with haploinsufficiency of TTF1 (also called TITF1, T/EBP, and NKX2.1), a transcription factor containing a homeobox domain that is required for normal thyroid development (40). These animals have decreased TSHR, and also have higher TSH levels, which is not sufficient to make these animals euthyroid. It is tempting to speculate that this is because these animals do not have enlarged thyroid glands, which would be required to return them to euthyroidism.

The transgenic line with higher penetrance of PTC also had higher TSH levels, and it is likely that TSH was an important contributing factor to the development of these cancers. This is because of the following: (a) Expression of the transgene is driven by the thyroglobulin promoter. Thus, the higher the TSH, the higher the expression of the mutant BRAF. (b) The growth promoting effects of TSH. TSH cooperates with growth factors or oncoproteins that activate the MAPK pathway to promote thyroid cell growth (41). This is of clinical importance, because even physiologic levels of TSH are believed to promote PTC progression and recurrence (42, 43). Indeed, TSH increments within the physiologic range are an independent variable that increases relative risk of PTC in humans (44). Hence, the Tg-BRAF model of PTC results from an accentuation of growth promoting signals that are normally involved in development of these cancers in humans. The higher intensity of these signals likely accounts for the high penetrance and the short latency.

In summary, thyroid-specific expression of BRAF^V600E induces goiter and invasive PTC with tall-cell features, which later transitions to poorly differentiated carcinomas. This closely recapitulates the phenotype of BRAF^V600E-positive PTCs in humans and supports a key role for this oncogene in its pathogenesis. Indeed, the similarity between the histopathologic features of the Tg-BRAF^V600E PTC and their human counterparts is striking. It is likely that this animal model will serve as an important tool for further understanding of molecular events associated with dedifferentiation of PTC. This is of clinical significance because poorly differentiated and anaplastic thyroid carcinomas are uniformly associated with high mortality and there are no reliable clinical or pathologic indicators that predict predisposition to undergo progression to poorly differentiated forms of the disease. In addition, this model should be useful for testing potential therapeutic strategies for treatment of PTC as well as those aimed to prevent tumor dedifferentiation.

Grant support: NIH grants CA50706 (J.A. Fagin) and DK15070 (S. Refetoff), American Cancer Society grant RSG-03-027-01-CCE (Y.E. Nikiforov), American Cancer Society Ohio Chapter grant 06001011 (J.A. Knauf), Nakayama Foundation for Human Science (N. Mitsutake), and SUMITOMO Life Social Welfare Services Foundation (N. Mitsutake).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

We thank Richard Marais and Gilbert Vassart for providing the BRAF^V600E cDNA and bovine thyroglobulin promoter constructs, respectively.