Activating BRAF and PIK3CA Mutations Cooperate to Promote Anaplastic Thyroid Carcinogenesis

This article is featured in Highlights of This Issue, p. 965

Thyroid cancers derived from follicular epithelial cells are histologically classified into papillary thyroid carcinoma (PTC), follicular or anaplastic thyroid carcinoma (ATC; ref. 1). Although PTC is indolent and readily managed by surgical resection combined with radioiodine therapy, ATC is highly aggressive with more than 50% of ATC patients dying within 1 year after diagnosis (2). Although metastasis is observed in 10% to 20% of patients with ATC, most patients with ATC succumb to locally invasive, inoperable disease that is largely refractory to conventional chemotherapy (3).

Mutationally activated BRAF (commonly T1799→A in exon 15) encoding BRAF^V600E is detected in approximately 40% of PTC and 25% of ATC (4). BRAF^V600E is a constitutively active protein kinase that activates the ERK1/2 MAPK pathway (5). The importance of mutated BRAF in thyroid cancer maintenance is suggested by responses of patients with thyroid cancer to vemurafenib, a pharmacologic inhibitor of BRAF^V600E (6). Moreover, conditional, thyrocyte-specific expression of BRAF^V600E in genetically engineered mouse (GEM) models results in PTC (7). However, as in humans, PTC in this model is indolent and does not routinely result in progressively lethal disease.

Human ATC displays multiple cooperating mutational events in tumor suppressors and oncogenes such as TP53 (70%–80%), PTEN (10%–20%), BRAF (25%), H-, N- or KRAS (20%–30%), PIK3CA (15%–25%), and CTNNB1 (60%–65%; ref. 8). Hence, by analogy to other cancer types, it is likely that progression to more aggressive disease is due to cooperative interactions between these various genetic abnormalities. To test this, we generated mice with thyrocyte-specific expression of BRAF^V600E in conjunction with expression of mutationally activated PIK3CA^H1047R, a constitutively activated form of the p110 catalytic subunit of PI3′-kinase α (9). Expression of PIK3CA^H1047R, which is detected in many cancer types, is predicted to promote elevated PI3′-lipid production leading to activation of AKT protein kinases and other PI3′-lipid effectors in the cell (10). In brief, whereas adult-onset, thyrocyte-specific expression of PIK3CA^H1047R had no detectable effect on the thyroid, it cooperated dramatically with BRAF^V600E such that mice developed rapidly lethal ATC. Similar observations were also made with thyrocyte-specific expression of BRAF^V600E combined with PTEN silencing. Using cultured human thyroid cancer cell lines, we demonstrated that these pathways cooperate to regulate the activity of mTOR and the phosphorylation of 4E-BP1. Hence, we propose that this GEM model of ATC, which recapitulates key features of the human disease, will be useful in understanding thyroid cancer progression and modeling the effects of pathway-targeted therapy in the preclinical setting.

BRaf^CA, Pik3ca^lat-1047R, Pten^lox, and Thyroglobulin::CreER^T2 mice were described previously (7, 11; 9, 12). BRaf^CA mice have been backcrossed in FVB/N in the laboratory for more than 10 generations; all the others have been obtained in C57BL/6 F129–mixed background and crossed in FVB/N since obtained. All the mice considered here are predominantly FVB/N. Thyrocyte-specific activation of CreER^T2 activity was achieved by intraperitoneal injection of 1 mg of tamoxifen dissolved in peanut oil into 4-week-old mice.

The 8505c line was cultured as directed in RPMI complemented with 10% FCS [and validated by short tandem repeat (STR) profiling, performed by Microsynth]. Ocut-2 in DMEM complemented with 10% FCS and nonessential amino acids, STR profiles showed that this cell line was not presenting mouse or human contamination and was of female origin as expected from the literature. The STR profile of Ocut-2 did not present any relevant similarities to any registered cell lines of the American Type Culture Collection.

Animal experiments were carried out in accordance with protocols approved by the University of California, San Francisco (San Francisco, CA) Institutional Animal Care and Use Committee (IACUC). Mice were anesthetized by intraperitoneal injection before tissue dissection, and then euthanized by section of the abdominal aorta. Thyroids were removed, rinsed in ice-cold PBS and fixed for 4 hours in Z-Fix (Anatech). Four- to 5-μm sections of formalin-fixed, paraffin-embedded tissues were stained with hematoxylin and eosin or processed for immunofluorescence. Epitope unmasking was performed by boiling slides for 10 minutes in a buffer comprising 10 mmol/L Tris, 0.5 mmol/L EGTA pH 9.0. Primary antibodies were obtained from the listed commercial sources and diluted as follow: anti–TTF-1 (1:200; Santa Cruz Biotechnology), anti-Ki67 (1:300; Abcam), anti-CK19 (1:300, TROMA-III, Hybridoma bank, University of Iowa, Iowa, IA), Vimentin (Cell Signaling Technology; 1-200), and anti–Galectin-3 (Abcam; 1:200). Antigen–antibody complexes were detected using either goat anti-rabbit Alexa-488 (1:500) or goat anti-rat Alexa-488 (1:500; Molecular Probes) with slides counterstained with DAPI to visualize nuclear DNA.

Mice were anesthetized using 2%(v/v) isofluorane at which time fur around the neck was removed using Veet depilatory cream. Ultrasound images were collected weekly using the Vevo770 system (VisualSonics). Thyroid size was assessed by counting pixels at the largest diameter of the thyroid and converting pixel count into area (mm²) using scaling software internal to the device in combination with ImageJ (NIH, Bethesda, Maryland, USA).

Pik3ca^lat-1047R (Pik3ca^Lat hereafter) mice carry a conditional allele that expresses normal PIK3CA before Cre-mediated recombination after which mutationally activated PIK3CA^H1047R is expressed from the normal chromosomal locus at physiologically relevant levels of expression (9). To test the effects of thyrocyte-specific expression of PIK3CA^H1047R, we treated compound adult Thyro::CreER; Pik3ca^Lat mice with tamoxifen and monitored prospectively for palpable goiter. Even 1 year after tamoxifen administration, mice with thyrocyte-specific expression of PIK3CA^H1047R displayed no obvious thyroid abnormalities. Hence, unlike BRAF^V600E, mutationally activated PIK3CA^H1047R is insufficient to initiate thyroid hypertrophy or tumorigenesis (7).

To test whether PIK3CA^H1047R would influence BRAF^V600E-induced PTC, we initiated tumorigenesis in adult Thyro::CreER; BRaf^CA/+; Pik3ca^Lat/+ and monitored them prospectively for disease (Fig. 1A). At 6.5 months, 2 of 14 mice became emaciated, displayed labored breathing, and had readily palpable thyroid tumors that were ≥1 cm³ requiring euthanasia. Over the next 2 months 71% of mice (10/14) presented developed similar end-stage disease requiring euthanasia. The experiment was stopped 12.5 months after tumor initiation, a time at which 85% (12/14) of the Thyro::CreER; BRaf^CA/+; Pik3ca^Lat/+ mice had been retired. In contrast, only 37% (4/11) of an independent cohort of initiated Thyro::CreER; BRaf^CA/+ mice reached end-stage and required euthanasia (tumors ≥1 cm³). Kaplan–Meier survival analysis clearly demonstrated the shortened survival of mice bearing BRAF^V600E/PIK3CA^H1047R versus BRAF^V600E-expressing thyroid tumors (Fig. 1A, P < 0.01). To further evaluate BRAF^V600E/PIK3CA^H1047R cooperation in thyroid tumorigenesis, tumor growth was measured in two additional cohorts of Thyro::CreER; BRaf^CA/+; Pik3ca^Lat/+ or Thyro::CreER; BRaf^CA/+ mice using ultrasound imaging 2 months after initiation (Fig. 1B). Mice bearing BRAF^V600E/PIK3CA^H1047R thyroid tumors displayed an approximately 60% greater tumor burden compared with BRAF^V600E alone tumors.

To assess the effects of oncogene expression on tumor histology, thyroid specimens were prepared from normal mice or mice with thyrocyte-specific expression of PIK3CA^H1047R, BRAF^V600E, or combined BRAF^V600E/PIK3CA^H1047R (Fig. 1C and D). Both normal- and PIK3CA^H1047R-expressing thyroid (Fig. 1C) tissue displayed a characteristic follicular architecture composed of a monolayered cuboidal epithelium of thyrocytes. Indeed, even 12 months after PIK3CA^H1047R expression, no alteration in thyroid architecture was observed (data not shown). BRAF^V600E expression elicited PTC displaying the characteristic papillary thyroid architecture at 6 to 9 months after initiation as described previously (Fig. 1C, middle; ref. 7). In contrast, when BRAF^V600E and PIK3CA^H1047R were coexpressed, we observed clear evidence of PTC within 3 to 6 months that extended throughout the entire thyroid (Fig. 1D, left). Moreover, these PTC lesions displayed aggressive characteristics with evidence of invasion into surrounding tissues, most notably the trachea (Fig. 1D, middle). Such invasion resulted in reduced airway diameter that was observed in 70% of the animals represented in the Kaplan–Meier survival curve (Fig. 1A). In addition, the thyroid tumors in 80% of these animals displayed polygonal giant cells and/or spindle-like cells (Fig. 1D, right), which is a characteristic of human ATC. Such features (local invasion and aberrant tumor cytology) were not observed in PTCs induced by expression of BRAF^V600E alone (7).

To further characterize BRAF^V600E/PIK3CA^H1047R collaboration in this model, we generated Thyro::CreER; BRaf^CA mice that were either wild-type, heterozygous, or homozygous for the conditional Pik3ca^Lat allele. Of note, 2.5 months after initiation, Thyro::CreER; BRafCA; Pik3ca^Lat/+ mice displayed the expected increase in tumor size compared with mice lacking the conditional Pik3ca^Lat allele. As expected, Thyro::CreER; BRafCA; Pik3ca^Lat/Lat mice also displayed an increase in tumor size (∼37%) compared with mice lacking the Pik3ca^Lat allele (Fig. 2A). However, although the difference in tumor size between mice heterozygous or homozygous for the Pik3ca^Lat allele was not statistically significant (Fig. 2B), mice homozygous for Pik3ca^Lat/Lat died more rapidly than their heterozygous counterparts (Fig. 2A) most likely due to an earlier onset of the tracheal invasion.

Human thyroid malignancies are characterized by alterations in marker expression such as galactin-3 (Gal-3) and CK-19 (13). We, therefore, stained BRAF^V600E/PIK3CA^H1047R–expressing thyroid tumors for expression of these proteins (Fig. 2C). First, we noted that BRAF^V600E/PIK3CA^H1047R–expressing thyroid tumors displayed areas of PTC and ATC. The regions of PTC were generally positive for Gal-3 and CK-19 as described previously (7). In contrast, regions of ATC retained Gal-3 but lacked CK-19 expression. In addition, ATC-like lesions displayed low/no expression of TTF-1 or E-cadherin, increased Vimentin expression and a higher proliferative index (Ki67).

As an alternative strategy to confirm collaboration of BRAF^V600E with alterations in the PI3K pathway, we generated Thyro::CreER; BRaf^CA; Pten^lox/lox mice to initiate BRAF^V600E/PTEN^null thyroid tumors. Compared with BRAF^V600E alone, these mice displayed increased thyroid tumor burden as soon as 1.5 months after initiation (Fig. 3B). This required complete silencing of PTEN expression because the enhanced tumorigenesis was not noted in Thyro::CreER; BRaf^CA mice heterozygous for Pten^lox (Fig. 3B). As anticipated, mice with BRAF^V600E/PTEN^null thyroid tumors developed end-stage disease more rapidly than mice with BRAF^V600E expression alone (Fig. 3A, P < 0.01). It was notable that Thyro::CreER; BRaf^CA mice heterozygous for Pten^lox tended to reach endpoint more rapidly than mice with BRAF^V600E expression alone, presumably because of loss-of-heterozygosity of the remaining Pten^WT allele (Fig. 3B).

At the histologic level, BRAF^V600E/PTEN^null thyroid tumors displayed similar characteristics as BRAF^V600E/PIK3CA^H1047R–expressing thyroid tumors. Although BRAF^V600E/PTEN^WT thyroids presented expected PTC features after 9.5 months (Fig. 3C), BRAF^V600E/PTEN^Null thyroid tumors displayed a clear PTC phenotype as early as 3.5 months after initiation (Fig. 3D, left). In addition, we also detected evidence of tracheal invasion (Fig. 3D, middle) and progression to ATC (Fig. 3D, right) of BRAF^V600E/PTEN^null tumors at time points 3.5 to 8 months after tamoxifen induction.

To explore biochemical collaboration between BRAF^V600E and PI3′-kinase signaling in more mechanistic detail, we used two human BRAF-mutated ATC cell lines: 8505c and Ocut2 (14). Whereas Ocut2 expresses both BRAF^V600E and PIK3CA^H1047R, 8505c is BRAF^V600E/TP53^null. Cells were treated with inhibitors of either MEK1/2 (PD325901; ref. 15) or class I PI3′-kinases (GDC-0941; ref. 16) either alone or in combination with effects on cell proliferation assessed using Crystal Violet staining (Fig. 4A). Although both PD325901 and GDC-0941 displayed antiproliferative effects, in general, MEK1/2 inhibition had a more potent inhibitory effect on cell proliferation than inhibition of class I PI3′-kinases. Moreover, combined blockade of both pathways had a more striking effect on cell proliferation compared with the antiproliferative activity of the single agents.

In parallel with the cell proliferation assays, cells were treated with PD325901 or GDC-0941, either alone or in combination, for 4 or 24 hours at which time cell extracts were analyzed by immunoblotting (Fig. 4B). At both time points, pathway blockade was highly selective in that PD325901-inhibited pERK1/2 with no effect on pAKT- and GDC-0941–inhibited pAKT with no effect on pERK1/2. Because these pathways are reported to coordinately regulate the machinery that regulates the initiation of protein synthesis, we assessed the effects of PD325901 and GDC-0941 on the phosphorylation of ribosomal protein S6 (rpS6) and 4E-BP1, that latter of which directly regulates Cap-dependent protein translation (17).

Phosphorylation of rpS6 and 4E-BP1 seemed more sensitive to PI3′-kinase inhibition at 4 or 24 hours in Ocut2 cells compared with 8505c cells, which is likely a reflection of the fact that Ocut2 cells express PIK3CA^H1047R. Moreover, p-RPS6 and p-4E-BP1 were sensitive to MEK1/2 inhibition in both cell types but this was not observed until 24 hours after inhibitor addition. However, in Ocut2 cells, we detected cooperative effects of combined MEK1/2 plus PI3′-kinase inhibition on p-RPS6 at 4 and 24 hours after inhibitor addition. This was also true for p-4E-BP1 24 hours after inhibitor addition but not at 4 hours. We also detected evidence of cooperative inhibitory effects on p-RPS6 and p-4E-BP1 in 8505c cells 24 hours after inhibitor addition, but the effects were not as striking as those observed in Ocut2 cells. Overall, these data suggest that BRAF^V600E and PI3′-kinase signaling cooperate for sustained thyroid cancer cell proliferation and that mutational activation of PIK3CA may predict for more potent antiproliferative and signaling effects following inhibition of class I PI3′-kinases.

Anaplastic thyroid carcinoma is a difficult disease to treat because it is generally inoperable and resistant to current regimens of chemotherapy. Although there are provocative hints from small numbers of patients from the phase I vemurafenib clinical trial, it remains to be unequivocally demonstrated that BRAF mutational status is prognostic for therapeutic benefit from BRAF^V600E inhibitors. However, the potential clinical benefit of vemurafenib might be influenced by altered signaling through ERBB receptors for EGF ligands as demonstrated by others (18). Furthermore, it was recently demonstrated that blockade of BRAF^V600E signaling with a MEK1/2 inhibitor (AZD6244) can restore the expression of the sodium-iodide symporter (NIS), and thereby enhance the efficacy of radioiodine therapy in both GEM models and in patients (19, 20). These data, therefore, suggest that combined pathway-targeted and conventional radiotherapy may be a successful strategy for treating late-stage thyroid cancer.

There are a large number of inhibitors of BRAF^V600E and PI3′-kinase signaling being developed for the treatment of a wide range of human cancers. However, it remains unclear whether mutational activation of BRAF, PIK3CA, or silencing of PTEN will serve as strong prognostic biomarkers for thyroid cancer patient responses to pathway-targeted therapy, in part because of effects of these agents on intracellular feedback signaling (18). Indeed, although BRAF mutation is prognostic of response in melanoma, that is not true for patients with colorectal cancer (21, 22). The availability of GEM models of a wide range of thyroid cancer types, including the new model of ATC based on a foundation of mutational activation of BRAF plus either mutation of PIK3CA or silencing of PTEN described here, will allow preclinical modeling of combination pathway-targeted, conventional or immunomodulatory cancer therapy. It was notable that expression of mutationally activated PIK3CA^H1047R in thyrocytes was largely without effect in the mouse but that it cooperated strongly with mutationally activated BRAF^V600E. This is not without precedent because PIK3CA^H1047R expression fails to initiate tumorigenesis in melanocytes and in the lung and pancreatic epithelium (23–25). In contrast, mutationally activated BRAF^V600E is sufficient to initiate tumorigenesis in all of the above target tissues. Hence, we propose that mutationally activated BRAF^V600E is a potent initiator of tumorigenesis that is also frequently required for tumor maintenance. In contrast, mutationally activated PIK3CA^H1047R is a weak initiator of tumorigenesis but greatly promotes tumorigenesis initiated by other oncogenic events (25). This may explain why, to date, pathway-targeted inhibition of PI3′-kinase signaling has not proven successful in patients, even those whose tumors are silenced for PTEN or mutated for PIK3CA.

TRP53 is also frequently mutated in anaplastic thyroid carcinoma (50% of the cases). Indeed, separately, McFadden and colleagues have demonstrated that combined expression of BRAF^V600E with silencing of TP53 also elicits anaplastic thyroid carcinoma (26). It has also been showed that the combination of TRP53 and PTEN knockouts could lead to ATC in another mouse model (27). This model has a similar latency to the model described here even if it using an embryonic onset of the cre-recombinase (TPO-Cre). This shows that even if cannot exclude p53 alteration in the BRAF^V600E PIK3CA^H1047R model, subsequent p53 alteration resulting in ATC progression is very unlikely.

The cooperation observed between BRAF^V600E and PI3′-kinase signaling in GEM models was recapitulated, at least in part, by analysis of bona fide human ATC-derived cell lines. It was clear that combined inhibition of both BRAF^V600E and PI3′-kinase signaling had more potent antiproliferative effects than that of single-agent inhibition. Moreover, regulation of key nodes in protein translation, most notably the phosphorylation of 4E-BP1 was clearly under the dual control of both pathways, suggesting that efficient Cap-dependent protein synthesis requires inputs from at least two pathways in ATC-derived cell lines. Although there was a suggestion that the PIK3CA^H1047R-expressing Ocut2 cells were more sensitive to PI3′-kinase, an extended analysis of a larger panel of thyroid cancer cell lines will be required to confirm if there is a genotype–drug response phenotype in this disease.

In conclusion, we describe a new mouse model of anaplastic thyroid carcinoma that recapitulates key features of the genetics and pathobiology of the cognate human disease. We propose that this GEM model will be useful for future preclinical studies and to understand the mechanisms by which these pathways cooperate to promote progression of thyroid cancer from an indolent to a lethal disease

M. McMahon has received commercial research grant from Novartis, Glaxo-Smith-Kline, and Plexxicon, and is a consultant/advisory board member of Abbvie, Sutro Biopharma, and Igenica Inc. No potential conflicts of interest were disclosed by the other authors.

Conception and design: R.-P. Charles, M. McMahon

Development of methodology: R.-P. Charles

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): R.-P. Charles, J. Silva, W.A. Phillips

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): R.-P. Charles, J. Silva, G. Iezza, M. McMahon

Writing, review, and/or revision of the manuscript: R.-P. Charles, M. McMahon

Study supervision: M. McMahon

The authors thank all of the members of the McMahon laboratory, especially Christy Trejo, Teruyuki Muraguchi, Victoria Marsh, Shon Green, Anny Shai, Allison Landman, Eric Collisson, Joseph Juan, and J. Edward van Veen for advice, guidance, and support throughout this project. The authors thank the support for mouse husbandry from Allen Villanueva. The authors also thank Bill Hyun and Jane Gordon from the UCSF Laboratory for Cell Analysis for assistance with microscopy, Cynthia Cowdrey of the BTRC Tissue Core for processing of formalin-fixed samples and Byron Hann and colleagues of the UCSF Preclinical Therapeutics Core for assistance with ultrasound imaging. The authors thank Dr. Lori Friedman (Genentech) for GDC-0941, Prof. J.A. Fagin (MSKCC) for providing the 8505c and Ocut-2 cells and Drs. David McFadden and Tyler Jacks (MIT) for ongoing discussion and communicating results prior to publication.

R.-P. Charles was supported by a Swiss National Science Foundation Fellowship, J. Silva was supported by Institutional Research and Academic Career Development Award (IRACDA), W.A. Phillips was supported by project grants from the National Health and Medical Research Council of Australia, and M. McMahon was supported by the National Cancer Institute (CA131261).

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.