Mouse models of sporadic thyroid cancer derived from BRAFV600E alone or in combination with PTEN haploinsufficiency under physiologic TSH levels

The BRAFV600E mutation is the most prevalent driver mutation of sporadic papillary thyroid cancers (PTC). It was previously shown that prenatal or postnatal expression of BRAFV600E under elevated TSH levels induced thyroid cancers in several genetically engineered mouse models. In contrast, we found that postnatal expression of BRAFV600E under physiologic TSH levels failed to develop thyroid cancers in conditional transgenic Tg(LNL-BrafV600E) mice injected in the thyroid with adenovirus expressing Cre under control of the thyroglobulin promoter (Ad-TgP-Cre). In this study, we first demonstrated that BrafCA/+ mice carrying a Cre-activated allele of BrafV600E exhibited higher transformation efficiency than Tg(LNL-BrafV600E) mice when crossed with TPO-Cre mice. As a result, most BrafCA/+ mice injected with Ad-TgP-Cre developed thyroid cancers in 1 year. Histologic examination showed follicular or cribriform-like structures with positive TG and PAX staining and no colloid formation. Some tumors also had papillary structure component with lower TG expression. Concomitant PTEN haploinsufficiency in injected BrafCA/+;Ptenf/+ mice induced tumors predominantly exhibiting papillary structures and occasionally undifferentiated solid patterns with normal to low PAX expression and low to absent TG expression. Typical nuclear features of human PTC and extrathyroidal invasion were observed primarily in the latter mice. The percentages of pERK-, Ki67- and TUNEL-positive cells were all higher in the latter. In conclusion, we established novel thyroid cancer mouse models in which postnatal expression of BRAFV600E alone under physiologic TSH levels induces PTC. Simultaneous PTEN haploinsufficiency tends to promote tumor growth and de-differentiation.


Introduction
Sporadic thyroid cancers usually develop via abnormal activation of the RAS-RAF-MEK-ERK signaling pathway (MAPK; which relays signals from cell membrane to nucleus), primarily as a result of point mutations in the RAS/BRAF genes or chromosomal rearrangements such as RET/PTC translocations [1]. In the BRAF gene, the T1799A transverse point mutation results in a mutant BRAF, BRAF V600E , which exhibits constitutive serine/threonine kinase activity.
The carcinogenicity of BRAF V600E in the thyroid glands was first demonstrated in vivo in Tg-Braf V600E transgenic mice expressing BRAF V600E under control of thyroid-specific thyroglobulin (Tg) promoter; these mice developed thyroid cancers very early in life [2]. However, this model had various limitations, including (i) BRAF V600E was expressed in all thyroid cells from the fetal period, suggesting that this is a model of hereditary rather than sporadic thyroid cancers; (ii) serum TSH levels were elevated by BRAF V600E -mediated suppression of thyroid function, which by itself can induce thyroid goiters and sometimes tumors; and (iii) BRAF V600E expression was controlled by the Tg promotor rather than the original Braf promoter [3]. These limitations remained unsolved in subsequent mouse models of thyroid cancer. LSL-Braf V600E ;TPO-Cre mice expressed BRAF V600E in all the thyroid cells from the fetal period, with~8-to 80-fold increases in TSH, although TSH was expressed at physiologic levels under the control of the chromosomal promoter [4]. Braf CA ;Thyro::CreER mice were generated to control expression of BRAF V600E by tamoxifen in the postnatal period, but untreated mice displayed increased thyroid volumes 1 month after birth, presumably due to aberrant nuclear localization of CreER T2 in the absence of tamoxifen [5]. In that model, Braf CA mice carried a Cre-activated allele of Braf V600E [6], similar to LSL-Braf V600E mice mentioned above [7]. Leakiness of CreER in the absence of tamoxifen has also been reported [8]. Tg-rtTA/tetO-Braf V600E mice expressed BRAF V600E in all the thyroid cells, with >100-fold increases in TSH, although expression began after birth (after administration of doxycycline) [9]. Finally, Braf CA ;TPOC-reER mice were reported to develop thyroid cancers after birth (after administration of tamoxifen), although TSH increased slightly (<10-fold) [10].
To establish an ideal mouse model of sporadic thyroid cancer, we previously generated Tg (LNL-Braf V600E ) mice. Upon injection of adenovirus expressing Cre under control of the Tg promoter (Ad-TgP-Cre) into their left thyroid lobes at age of~4 weeks, these mice expressed BRAF V600E in a fraction of the thyroid cells. As such, serum TSH remained within physiologic range, and mice did not develop thyroid cancer [3]. From these data, we concluded that postnatal expression of BRAF V600E alone in a small number of thyroid cells under normal TSH levels is insufficient for thyroid cancer development. However, this model also had a drawback; a comparison of data from the previous reports [3,4] suggested that Cre-mediated DNA recombination was less efficient in Tg(LNL-Braf V600E );TPO-Cre mice than LSL-Braf V600E ;TPO-Cre mice, as serum TSH levels increased in the latter not the former.
In the present study, therefore, we first confirmed the higher transformation efficiency of Cre-mediated DNA recombination in Braf CA ;TPO-Cre mice compared with Tg(LNL-Braf V600E );TPO-Cre mice in our laboratory and then used Braf CA mice rather than Tg(LNL-Braf V600E ) mice to re-evaluate the carcinogenesis of BRAF V600E in the context of our experimental setting with Ad-TgP-Cre. Here, we show that postnatal BRAF V600E expression alone under physiologic TSH levels is sufficient for thyroid cancer development. In addition, we also studied the effect of concomitant PTEN haploinsufficiency on BRAF V600E -induced thyroid cancers and show that the simultaneous reduction of PTEN expression tends to promote tumor growth and de-differentiation. Our results also demonstrate development of thyroid hyperplasia/adenoma in Pten Δ/+ mice (but not Pten f/+ mice) injected with Ad-TgP-Cre, suggesting that

H & E staining and immunohistochemistry
Tissues were fixed in 10% neutral-buffered formalin and then embedded in paraffin. Sections (4-μm-thick) were prepared and stained with hematoxylin eosin (H & E) or immunostained with primary antibody: rabbit polyclonal anti-surfactant protein A (ab115791, Abcam, . It should be noted here that the protein recognized by anti-PAX8 mentioned above is called "PAX" throughout the paper, because, although the immunogen for this antibody was a part of human PAX8 (212 amino acids), its specificity to PAX8 has not been confirmed. The primary antibody was followed by incubation with secondary antibody: swine anti-rabbit IgG/ HRP (P0399, DAKO, Glostrup, Denmark; dilution of 1:50) or rabbit anti-mouse IgG/HGRP (PO260, DAKO; dilution of 1:100). Color was developed with 3, 3'-diaminobenzidine substrate. Slides were analyzed using an All-in-One BZ-9000 Fluorescence Microscope (Keyence, Osaka, Japan). A total of 1,500 cells were evaluated to determine the percentage of Ki67-positive cells.

Evaluation of apoptosis
Terminal deoxynucleotidyl transferase (TdT)-mediated dUTP nick end-labeling (TUNEL) was performed with the Apop-tag™ Fluorescein Direct in situ apoptosis detection kit (Merck Millipore, Darmstadt, Germany). Slides were embedded with VECTASHIELD mounting medium containing DAPI (Vector Laboratories, Burlingame, CA) and analyzed using an Allin-One BZ-9000 Fluorescence Microscope (Keyence). A total of 1,500 cells were evaluated in each sample to determine the percentage of TUNEL-positive cells.

Serum TSH measurements
Serum TSH was measured using a specific mouse TSH RIA with mouse TSH/LH reference (AFP9090D), mouse TSH antiserum (AFP98991) and rat TSH antigen (NIDDK-rTSH-I-9) as described previously [3,14]. The normal range was defined as the mean ± 3 S.D. of control untreated mice.

Statistical analyses
All data were analyzed for significant differences using the Student's t-test. A p-value of less than 0.05 was considered statistically significant.

Results
In previous research reported by us [3] and others [4], Tg(LNL-Braf V600E )#213MM (a high expressor);TPO-Cre mice exhibited a slightly (but not significantly) enlarged thyroid with focal neoplastic lesions and normal TSH levels, whereas Braf CA/+ ;TPO-Cre mice exhibited a greatly enlarged thyroid with diffuse neoplastic lesions and elevated TSH levels (Fig 1), at ages of 12 weeks. Braf CA/+ ;TPO-Cre mice, Tg(LNL-Braf V600E )#213MM;TPO-Cre mice, and controls exhibited serum TSH levels of 43.1 ± 56.6, 0.9 ± 0.2 and 1.0 ± 0.2 ng/ml, respectively, and thyroid weights of 122.0 ± 63.6, 8.0 ± 4.3 and 6.7 ± 1.8 mg, respectively. The lower transformation efficiency in Tg(LNL-Braf V600E )#213MM as compared with Braf CA mice may explain our previous failure of tumor induction in Tg(LNL-Braf V600E )#213MM mice with intrathyroidal injection of Ad-TgP-Cre in our previous study [3]. Therefore, we used Braf CA rather than Tg(LNL-Braf V600E )#213MM mice to re-evaluate the carcinogenesis of BRAF V600E with our thyroid cancer model with Ad-TgP-Cre. We also examined the carcinogenesis of PTEN haploinsufficiency using Pten f/+ mice and Braf CA/+ ;Pten f/+ mice, as reduced PTEN expression alone and in combination with BRAF V600E reportedly plays a significant role in the carcinogenesis of various organs [15][16][17]. Ad-TgP-Cre was injected into the left thyroid lobe of 4-week-old Braf CA/+ , Braf CA/+ ;Pten f/+ and Pten f/+ mice (designated as Braf thyr-V600E , Braf thyr-V600E ;Pten thyr-Δ/+ and Pten thyr-Δ/+ mice, respectively). Because it was totally unknown whether thyroid tumors developed and if so when, we decided to observe the mice either until some symptoms appeared or for 26 and 52 weeks. Because no symptom developed, the mice were sacrificed at 2 time points, as originally scheduled. The thyroid lobe was macroscopically normal in all mice at 26 weeks (data not shown), but at 52 weeks, the left lobe was enlarged in Braf thyr-V600E mice (8/9) and Braf thyr-V600E ; Pten thyr-Δ/+ mice (9/9), but not Pten thyr-Δ/+ mice (0/7) ( Table 1, Fig 2). The left lobes weighed 24.0 ± 21.0 mg in Braf thyr-V600E ;Pten thyr-Δ/+ mice and 12.1 ± 6.5 mg in Braf thyr-V600E mice vs.~2 mg in the right lobe of these mice (p<0.01) and also each lobe of the controls. The left lobe tended to be heavier in Braf thyr-V600E ;Pten thyr-Δ/+ mice compared with Braf thyr-V600E mice, but the difference was not statistically significant (Fig 2). Microscopically, all of the thyroid glands obtained at 26 weeks were intact, but the tumors encompassed almost the entire thyroid gland, and almost no normal thyroid architecture was observed in the periphery of the thyroids in Braf thyr-V600E and Braf thyr-V600E ;Pten thyr-Δ/+ mice (Figs 3 and 4) at 52 weeks. Tumors in the majority of Braf thyr-V600E mice exhibited a follicular or cribriform-like structure consisting of atypical epithelial cells with hyperchromatic swollen nuclei and no colloid formation. They also showed a hobnail pattern (represented by Braf thyr-V600E mouse No. 1 in Fig 3), suggesting a loss of the tight cell to cell adhesion [18]. A hobnail pattern has not been reported in other PTC mouse models, with the exception of Rusinek and colleagues [19], who found this pattern in a small fraction of their transgenic Tg-2HA-Braf V600E mice, which are similar to Tg-Braf V600E mice [2]. In human PTC, this pattern of pathology is usually associated with an aggressive phenotype [20,21]. Immunohistochemical analysis demonstrated clear TG and PAX staining of tumor cells (represented by Braf thyr-V600E mouse No. 1 in Fig 3). Two tumors from Braf thyr-V600E mice also contained a component of papillary structures and expressed the similar levels of PAX but decreased levels of TG (represented by Braf thyr-V600E mouse No. 3 in Fig 3). In contrast, all of the tumors in Braf thyr-V600E ;Pten thyr-Δ/+ mice showed predominantly papillary structures with sporadic undifferentiated areas exhibiting solid growth pattern of atypical cells with a number of mitotic figures. The nuclei were hyperchromatic, varying in size, and oval to spindle-shaped. No necrosis of single cells was observed. PAX expression was normal to low, and TG expression was low to absent (represented by No. 2 and No. 6 in Fig 4). Accompanying extrathyroidal invasion was occasionally observed (Fig 5A). Typical nuclear features of human PTC, such as intranuclear cytoplasmic inclusion and nuclear groove, were frequently observed in tumors of Braf thyr-V600E ;Pten thyr-Δ/+ mice (Fig 5B and 5C).
Ad-TgP-Cre-mediated BRAF V600E expression and decreased PTEN expression were confirmed by immunohistochemistry (Fig 6). Thus, BRAF V600E was expressed in thyroid cancer but not in the normal thyroid, although the basement membrane-like region stained non-specifically stained in the normal thyroid glands. Expression of PTEN was clearly observed in the thyroids of Pten +/+ and Braf thyr-V600E mice, but barely detectable in Pten Δ/+ and Braf thyr-V600E ; Pten thyr-Δ/+ mice.
Thyroid tumors exhibiting (i) typical nuclear features of human PTC such as intranuclear cytoplasmic inclusions and nuclear grooves and/or (ii) invasion of the extrathyroidal tissues surrounding the thyroid glands were readily diagnosed as cancers. Some tumors in Braf thyr-V600E mice not exhibiting these features were also judged as cancers, because they had malignant characteristics such as structural atypia, including cribriform-like, papillary, and solid growth of atypical follicular cells with hyperchromatic swollen nuclei, which occasionally showed a hobnail pattern. Higher cell proliferation indices determined by Ki67 staining (22.5 ± 10.2 vs. 5.6 ± 4.6) were compensated by higher cell death rates as determined by TUNEL staining (1.1 ± 0.9 vs. 0.4 ± 0.4) in Braf thyr-V600E ;Pten thyr-Δ/+ mice as compared with Braf thyr-V600E mice (Fig 7), which likely explains the non-significant difference in tumor sizes between the 2 mouse groups (Fig 2). Although the staining intensity seemed stronger in Braf thyr-V600E ;Pten thyr-Δ/+ than Braf thyr-V600E  Mouse models of thyroid cancer with BRAF V600E and/or PTEN haploinsufficiency Macroscopic lung nodules were observed in 2 of 9 Braf thyr-V600E and 6 of 9 Braf thyr-V600E ; Pten thyr-Δ/+ mice. BRAF V600E expression in these nodules (Fig 6) excluded the possibility of the spontaneously arisen primary lung tumors, but negative staining for TG and PAX (data not shown) did not provide convincing evidence that these nodules were metastases. Although Ad-TgP-Cre-mediated BRAF V600E expression was very unlikely even if adenovirus had disseminated systemically, because the Tg promoter we used in this study is exclusively thyroidspecific and has been widely and successfully used for many genetically engineered mice (e.g., Tg-Braf V600E ) [2], we found that these nodules were positive for surfactant protein-A (Fig 9), which is reportedly expressed in BRAF V600E -induced lung adenomas [6,16]. A spontaneously developed rat lung tumor [22] also stained positive.
Finally, despite the absence of tumor development in Pten thyr-Δ/+ mice, most Pten Δ/+ mice developed thyroid hyperplasia/adenoma by the age of 6 to 8 months (Table 1, Fig 10). These mice were sacrificed during this time period because tumor had developed in other organs.

Discussion
Although we previously reported the insufficiency of postnatal expression of BRAF V600E for thyroid cancer development in mice [3], in the present study, we re-evaluated this issue using a different genetically engineered mouse model (i.e., Braf CA ). As BRAF V600E is frequently found in sporadic thyroid cancers in euthyroid subjects, BRAF V600E should be expressed in a small fraction of thyroid cells (ideally in a single cell, but it is currently not possible experimentally) after birth under physiologic TSH levels. In this regard, our experimental design-that is, intrathyroidal injection of Ad-TgP-Cre into one side of the thyroid lobes of genetically engineered mice harboring the loxP sequences-is likely ideal. The feasibility of adenovirus-mediated Cre gene transfer to temporally and spatially control Cre expression has been well demonstrated [23,24]. In the present study, we clearly showed that thyroid cancers did develop in Ad-TgP-Cre-injected Braf CA mice, indicating that postnatal expression of BRAF V600E alone Our previous failure with Tg(LNL-Braf V600E ) mice [3] appeared to be attributable to a lower efficiency of Cre-mediated DNA recombination, although we cannot exclude the other possibilities that the different genetic backgrounds (B6C3 in Tg(LNL-Braf V600E ) vs. B6 in Braf CA ) and/or different promoters (CAG promoter vs. the endogenous Braf promoter) could have affected our previous results. Different recombination frequencies of distinct alleles have been reported [25]. Presumably, the frequency of transformation of BRAF V600E -expressing normal, differentiated (i.e., TG-expressing) thyroid cells into malignant cells is extremely low.
The Braf CA ;TPOCreER mouse model with tamoxifen reported by McFadden et al. may also be ideal, although the TSH levels increased slightly (<10 fold) [10]. However, thyroid cancers developed several weeks after administration of tamoxifen in their model, in a sharp contrast to the present study, in which thyroid cancers were only detectable 1 year (not 6 months) after adenovirus injection. It is unclear whether the slight increase in TSH promoted tumorigenesis in their model. In this regard, fine dose-response experiments may be necessary to find the appropriate concentration of tamoxifen to induce thyroid cancer on one hand while maintaining physiologic TSH levels on the other.
Significant increases in TSH levels (up to 500 fold) have been noted in other models [2,4,9,10]. As elevated TSH is known to induce thyroid enlargement and sometimes promote tumorigenesis by itself [26], there is no doubt that elevated TSH has substantially affected the results obtained with the above-mentioned mouse models of thyroid cancer with marked TSH elevation. However, the significance of low TSH levels for thyroid tumorigenesis is controversial. On one hand, Tg-Braf V600E ;Tshr -/mice [27] and LSL-Braf V600E ;TPO-Cre;Tshr -/mice [4], both of which are unresponsive to TSH stimulation due to a lack of TSH receptor expression, can develop thyroid cancers, albeit less aggressive, but, on the other hand, transplantation of thyroid cancers developed in LSL-Braf V600E ;TPO-Cre mice (with high TSH levels) into nude or syngeneic immuno-competent mice (with normal TSH levels) leads to regression and senescence [28].
Regarding the question as to how many mutations are required for full development of differentiated thyroid cancer, recent studies using human samples show that number of non-synonymous mutations in exomes is~0.4/Mb [29][30][31], and the number of mutations among 341 cancer-related genes in PTC is reportedly 1 ± 1 (median ± interquartile range) [30,32]. Thus, similar to pediatric cancer and leukemia, thyroid cancer is associated with a very low number of mutations, suggesting that a single or perhaps only a few mutations are sufficient for thyroid cancer to develop. In our model, however, the possibility cannot be excluded that other mutations occurred during the 1-year observation period.
BRAF V600E was first discovered in malignant melanoma, but later also found to be present in benign nevi, which seldom progress to melanoma unless additional mutations occur [33]. In accordance with this observation, in mouse experiments, BRAF V600E alone cannot induce melanoma, but it can in combination with PTEN loss or activating PI3KCA mutations [16,34]. Concurrent mutations in BRAF and diminished PTEN expression are common in human melanomas [34]. Similar data were also reported in lung adenocarcinoma and prostate cancer in genetically engineered mice [17,35]. Of interest, in contrast to thyroid cancer, melanoma and lung cancer are among cancers with a high number of mutations [29,31].
The combination of BRAF V600E and reduced PTEN expression tended to induce larger and more undifferentiated thyroid cancers in our study, and these data were similar to those in LSL-Braf V600E ;Pten f/f ;TPO-Cre mice in which PTC rapidly progressed to poorly differentiated thyroid cancers as compared with LSL-Braf V600E ;TPO-Cre mice [36] and also to those in Thyro::CreER;Braf CA/+ ;Pik3ca lat-1047R/+ mice, which developed anaplastic cancers as compared with Thyro::CreER;Braf CA/+ mice [37]. Although the mutations in Pten gene are not common [38], reduced expression of PTEN due to hypermethylation is frequently detected even in differentiated thyroid cancers [39].
Tumorigenesis associated with PTEN loss by itself is well known in human Cowden syndrome, in which a germline loss-of-function mutation in the PTEN gene induces thyroid multinodular goiter and adenoma [40]. Experimentally, the tumorigenesis of prenatal PTEN loss in the mouse thyroid gland was clearly shown by Yeager et al. using Pten L/L ;TPO-Cre mice [15]. Thus, similar to the Pten Δ/+ mice used in our study, the majority of mice in the 129Sv genetic background developed well-circumscribed follicular adenomas and nodular hyperplasia, often characterized by increased cellularity and mitotic figures at 8 to 10 months of age. However, no thyroid tumors were observed in Pten f/+ mice injected with Ad-TgP-Cre in our study. These data clearly indicate that the tumorigenic potential of reduced PTEN expression differs between the prenatal and postnatal periods.
We interpret our data on lung tumors as showing that adenovirus injected into the thyroid lobes leaked, disseminated systemically, and reached the lung, where BRAF V600E was expressed aberrantly from the Tg promoter, even when the volume of adenovirus injected was low (1 μl) and highly thyroid specific Tg promoter was used. Thus, one of the limitations of our study is the leakiness of locally injected adenovirus as well as leakiness of the Tg promoter. Our model is therefore not suitable for study of metastasis. Only 2 reports of lung metastasis have been reported, one by Rusinek et al. using transgenic Tg-2HA-Braf V600E mice [19] and the other by McFadden using TPOCreER;Braf CA/+ ;p53 LSL-R270H/+ mice [10]. Another limitation is that we cannot completely exclude the possible effect of adenovirus-induced inflammation and/or disruption of local tissue architecture on cancer development in our experimental setting.
In conclusion, using our mouse model with Ad-TgP-Cre, we show that postnatal expression of BRAF V600E alone under physiologic TSH levels is sufficient for development of thyroid cancer and that simultaneous reduced expression of PTEN tends to promote tumor growth and de-differentiation. It will be of interest in the future to compare the differences/similarities of thyroid cancers associated with postnatal vs. prenatal expression of BRAF V600E . Our data also indicate that the effects of BRAF V600E expression and reduced PTEN expression differ between the prenatal vs. postnatal periods. Thus, unlike BRAF V600E , the tumorigenic potential of PTEN depends on a prenatal reduction in expression.