Pten Loss in the Mouse Thyroid Causes Goiter and Follicular Adenomas: Insights into Thyroid Function and Cowden Disease Pathogenesis

The thyroid gland is characterized by an extremely low proliferation index (1). However, and rather paradoxically, the thyroid is also extremely sensitive to alterations of the equilibrium governing its homeostasis: hyperplastic disorders of this rather quiescent organ affect up to 67% of the U.S. population (2). Increased stimulation of thyroid growth is causally linked to a variety of pathologic conditions, ranging from nodular hyperplasia (goiter) to neoplastic transformation (3).

Thyroid growth is primarily induced by the pituitary-derived thyroid-stimulating hormone (TSH), whose levels are regulated by the thyroid hormone through a negative feedback mechanism (4). TSH binding to its receptor (TSHR) activates the cyclic AMP (cAMP)/protein kinase A (PKA)–dependent mitogenic cascade, leading to both the induction of cell cycle progression and the expression of differentiation markers (5). Increasing evidence from in vitro models indicates that the mitogenic activity of TSH necessitates the cooperation of peptide growth factors, such as insulin-like growth factor-I (IGF-I), epidermal growth factor, insulin, etc. (6). However, the often contradictory results obtained in different cell culture systems have not yet clarified the relative roles and contribution of TSHR- and receptor tyrosine kinase (RTK)–initiated signal transduction pathways to the mitogenic process (7).

Phosphatidylinositol-3-kinase (PI3K) is a central mediator of all RTK-initiated signaling cascades, catalyzing the conversion of phosphatidylinositol (4,5)-biphosphate (PIP-2) into phosphatidylinositol (3,4,5)-triphosphate (PIP-3). The major effector of PI3K is the Akt kinase, which is activated upon PIP-3–mediated membrane recruitment and, in turn, phosphorylates an ever-growing list of _target proteins regulating key processes, such as proliferation, survival, cell size, and mRNA translation (8). This process is counteracted by the PTEN tumor suppressor, which opposes PI3K activity by dephosphorylating PIP-3 to PIP-2 (9). In vitro studies have shown that the PI3K/Akt/PTEN signaling pathway is involved in the implementation of the growth factor–dependent proliferative signals in thyroid cells (10, 11). In addition, deregulation of this cascade, through activation of PI3K and Akt and loss of PTEN expression, is frequently found in thyroid cancer (11–17). Finally, heterozygous mutation of PTEN causes Cowden disease, a dominant genetic syndrome whose characteristics include thyroid benign disorders, such as multinodular goiter and adenoma, and a 10% lifetime risk for developing thyroid cancer, mostly of the follicular type (18–20).

Despite many correlative data suggest that this signaling cascade plays a central role in the control of normal thyroid function and that its deregulation is linked to thyroid disease, direct in vivo evidence supporting this hypothesis is still missing. Moreover, until now, there has been no genetic model that could be exploited to tease out the relative contributions of the two major growth-stimulating signals in the thyroid, TSH and growth factors, and to dissect in a physiologically relevant setting the molecular pathways that are altered in thyroid proliferative lesions.

To address these issues, we have generated a mouse strain, in which the Pten gene is selectively deleted in the thyroid follicular cells, thus constitutively activating the PI3K/Akt pathway and reproducing the genetic events taking place in the nodular lesions developing in Cowden disease patients. Our data show that Akt activation is sufficient, in vivo, to induce thyroid hyperplasia and diffuse colloid goiter in young mice by increasing the thyroid mitotic index and to create fertile ground for neoplastic transformation in older mutants.

Animals and treatments. Generation of Pten^+/−, Pten^L/L−, and thyroid peroxidase (TPO)-Cre mice has been described (21–23). Pten^L/L mice in the 129Sv background were initially bred to TPO-Cre mice in the FVB/N background and then backcrossed in the 129Sv background. At different time points, mice were euthanized and weighed. The thyroid was dissected and weighed; one lobe was fixed for pathologic analysis, and the other was frozen in liquid nitrogen. Hypothyroidism was induced at 8 weeks of age by administration of 0.5% sodium perchlorate and 0.05% methimazole (both from Sigma Chemical Co., St. Louis, MO), given in drinking water for 4 weeks. Mice were weighed and killed, and samples were collected as described above.

Real-time PCR. The recombination frequency for the Pten-floxed allele was determined as described previously (24). Briefly, two sets of primers were designed to distinguish the floxed allele from the recombined allele (Fig. 1A). Reactions with the two primer pairs (D/E and F/G) were done in separate wells of the same 96-well reaction plate using SYBR Green Master Mix (Applied Biosystems, Foster City, CA) and the results were analyzed according to the following formula: % recombined allele = [1 / (1 + 2^ΔCT)] × 100, where ΔC_T = recombined C_T − floxed C_T.

Hormone measurements. Blood was collected by cardiac puncture, and serum was stored at −80°C until analysis. TSH and total T4 were measured by radioimmunoassay at the Harbor-University of California at Los Angeles Medical Center (Torrance, CA).

Immunohistochemical analysis. The following rabbit polyclonal antibodies were used: Ki-67 (Vector Laboratories, Burlingame, CA), cyclin D1 (Lab Vision, Fremont, CA), phosphorylated p70S6K, phosphorylated S6, phosphorylated Foxo1, and phosphorylated estrogen receptor α (P-ERα) S167 (Cell Signaling, Danvers, MA); for PTEN and phosphorylated Akt (pAkt) S473, rabbit monoclonal antibodies were used. Tissues were fixed, embedded in paraffin, and sectioned at 6 μm. Sections were subjected to antigen retrieval in 0.1 mol/L sodium citrate and counterstained with hematoxylin.

Morphometric analysis. The H&E-stained sections were photographed at ×100 and ×400 magnification and analyzed using the ImageJ software. The area of at least 100 follicles was determined by measuring the luminal surface. Follicle density was determined by counting different areas to cover 2 to 5 mm². Cell density was calculated counting the number of cells in different areas to cover 0.2 to 0.5 mm² (2,500–5,000 cells were counted per section, per mouse).

Generation of Pten^L/L;TPO-Cre mice. To generate a model of Pten loss in the thyroid follicular cells, we crossed Pten^L/L mice, harboring two loxP sites flanking Pten exons 4 and 5 (23), with transgenic mice expressing the Cre recombinase under the control of the human TPO promoter (Fig. 1A; ref. 22). TPO-Cre mice express the recombinase in most thyrocytes, starting between 14.5 and 16.6 d.p.c., when the mouse thyroid enters the final steps of differentiation expressing thyroglobulin, thyroid peroxidase, and sodium-iodide symporter (25).

PCR analysis of different tissues, using primers selectively amplifying the unrecombined or the recombined allele, showed that Pten was specifically deleted in the thyroids of mice expressing Cre (Fig. 1B; data not shown). To quantitate the deletion efficiency in the Pten^L/L;TPO-Cre thyroids, we compared the amount of the intact Pten allele with that of the recombined allele by quantitative PCR (24). In agreement with previous reports (24), no recombination was detected in other tissues (data not shown). Approximately 70% and 90% of the floxed alleles had undergone recombination in 10- and 45-week-old mice, respectively (Fig. 1C). We did not detect any recombined product in Cre-negative samples, whereas Pten deletion reached, as expected, 100% in T-cell lymphomas from Pten^L/L mice that express Cre in the thymus (Fig. 1C). Taking into account the presence of unrecombined alleles deriving from the thyroid endothelium and from C cells, we can assume that the thyrocyte recombination efficiency in this model approaches 100%.

Pten^L/L;TPO-Cre mice were born at the expected Mendelian ratios and developed normally, achieving sexual maturity without displaying obvious physical or behavioral abnormalities. No significant differences in body weight gain between wild-type (WT) and mutant mice were found in a 20-week follow-up (Fig. 1D; data not shown).

Goiter development in Pten^L/L;TPO-Cre mice. WT and Pten^L/L;TPO-Cre littermates were sacrificed at different time points, starting at 10 weeks of age, to assess the effect of Pten deletion on the thyroid gland. At all time points, the mutant thyroids, irrespective of gender, were strikingly enlarged compared with WT glands (Figs. 1E and 2A and B). At 10 weeks of age, the average weight of mutant thyroids, adjusted for the body weight, was already increased five times compared with controls (Fig. 1E). In addition, whereas the growth of WT thyroids slows down drastically after ∼10 weeks from birth, Pten^L/L;TPO-Cre glands continued to increase in size and their normalized weight, at 40 weeks of age, was on average 12 times larger than controls (Fig. 1E).

The analysis of H&E-stained sections from 10- to 18-week-old mice showed that the increased gland size was associated with a widespread follicular enlargement and that the follicles were homogeneously filled with colloid (Fig. 2C and D). To verify that Pten expression had been completely abrogated, we did immunohistochemical detection of Pten and of the active form of Akt (pAkt), which is increased as a result of Pten deletion. WT glands showed strong Pten reactivity in both the cytoplasm and the nucleus of the follicular cells, which was completely abolished in the mutant glands. C cells and endothelial cells were still immunoreactive, thus providing additional evidence of the thyrocyte specificity of Pten ablation (Fig. 2E and F). pAkt was expressed at low levels in the nucleus and cytoplasm of control thyrocytes. However, loss of Pten resulted in a dramatic increase in the pAkt immunoreactivity of these cells (Fig. 2G and H). These findings are in sharp contrast with the normal thyroid size and structure that characterizes Pten^+/− mice at the same age (data not shown). Thus, activation of Akt in the thyroid follicular cells, as a consequence of complete Pten ablation, results in the early development of diffuse colloid goiter.

To determine whether the goiter developing in Pten^L/L;TPO-Cre mice is associated with an altered hormonal milieu, we measured serum levels of TSH and T4 by radioimmunoassay in a group of 10- to 13-week-old WT and Pten^L/L;TPO-Cre mice (n = 10). As expected, TSH levels in the female mice were about half of those in the males; however, no significant differences in the levels of TSH and T4 were detected between control and mutant mice (Fig. 3A and B). These data show that Pten^L/L;TPO-Cre mice are euthyroid and suggest that the defects leading to goiter development do not result in an increase of thyroid hormone synthesis by the follicular cells, thus preserving the negative feedback loop regulating TSH production from the pituitary.

To analyze in detail the morphologic changes induced by loss of Pten, we determined the average number of follicles per square millimeter in a cohort of age-matched (10–12 weeks) WT and mutant mice. Irrespective of their sex, Pten^L/L;TPO-Cre mice displayed a striking decrease (>50%) in follicle density compared with controls (Fig. 3C). Reduced follicle density was primarily caused by generalized follicle enlargement, again irrespective of the mouse gender: mutant follicles were on average three times as large as their WT counterparts (Fig. 3D).

Increased proliferation in Pten^L/L;TPO-Cre thyrocytes. Microscopic examination of WT and mutant thyroids revealed a subtle but reproducible and statistically significant phenotypic difference between male and female mutant mice. In fact, the cell density of mutant males (follicular cells per square millimeter) was reduced ∼30% compared with control males (Figs. 3E and G and 4A), reflecting the fact that, in the mutant mice, the follicular lumen is increased. Conversely, the female thyroids had the same cell density as their WT counterparts, despite the fact that male and female mutant thyroids have similarly enlarged follicles. Thus, this difference in cell density must reflect a net increase in total follicular cell number in the females (Figs. 3F and H and 4A).

To start understanding the mechanism behind this differential increase in cellularity between genders, we measured the proliferation of follicular cells in WT and Pten^L/L;TPO-Cre thyroids at 12 to 15 weeks of age using an antibody against the proliferation marker Ki-67. WT mice have, at this age, a very low proliferation index, irrespective of gender. However, the proliferation of mutant follicular cells was strikingly different between male and female thyroids. Male Pten^L/L;TPO-Cre mice had on average a 3-fold increase in follicular proliferation, whereas female mutants displayed a 6-fold increase (Fig. 4B and C). Taken together, these data show that the enlarged follicles developing as a consequence of Pten loss are in part due to increased thyrocyte proliferation and that this effect is more marked in the females.

Signaling cascades activated by Pten loss. The development of thyroid hyperplasia with full penetrance in young Pten^L/L;TPO-Cre mice makes it possible to analyze in vivo, in a genetically defined model, the pathways that are altered as a result of Pten loss and that thus contribute to goiter pathogenesis and increased proliferation rate. We used immunohistochemistry to investigate the expression levels and phosphorylation status of several proteins acting in pathways directly affected by Akt activation and found that both p70S6K1 and its major _target, ribosomal protein S6, were highly phosphorylated and thus activated in mutant follicular cells compared with controls (Fig. 4D). In addition, the transcription factor Foxo1, a direct _target of Akt, was phosphorylated and delocalized to the cytoplasm in mutant thyroids (Fig. 4D). Pten^L/L;TPO-Cre thyroids also displayed increased nuclear levels of cyclin D1, thus establishing a causative link with the increased proliferation rates observed in the mutant thyroids (Fig. 4D). Western blot analysis revealed a strong increase in the expression levels of cyclin D3, whereas the expression of p27 remained substantially unaltered (Fig. 4E).

To gain insight into the mechanism(s) behind the gender differences in cellularity and proliferation detected in mutant mice, we stained control and mutant thyroids with antibodies recognizing ERα and its Ser¹⁶⁷-phosphorylated form, which is a direct _target of Akt and has increased transcriptional activity compared with nonphosphorylated receptor, thus potentially contributing to the differences in proliferation between male and female mutants. Immunoreactivity for ERα and P-ERα was restricted to scattered follicles and patches of thyrocytes in WT mice. However, Pten^L/L;TPO-Cre thyroids showed a striking increase in the number of follicular cells positive for both ERα and P-ERα compared with controls (Fig. 4D), suggesting that Pten loss might result in the expansion of an ERα⁺ population or, less likely, in an increase of ERα expression. These data show that Pten loss in the mouse thyroid activates several pathways that can cooperatively contribute to the increased proliferation and overall organ growth characterizing Pten^L/L;TPO-Cre thyroids.

Elevation of TSH levels minimally increases the proliferation and growth of Pten^L/L;TPO-Cre thyroids. TSH constitutes the major growth signal for thyroid cells. Activation of TSHR primarily initiates the cAMP signaling cascade. However, several conflicting data exist on the involvement of other pathways, including the PI3K/Akt cascade, downstream of activated TSHR, and on the extent of the contribution to thyrocyte proliferation of these pathways. We reasoned that if the contribution of PI3K/Akt to proliferation was additive to the other pathways initiated by TSH, such as cAMP/PKA and Ras/MAPK, then supraphysiologic TSH levels should lead to a further increase in the proliferation of the Pten mutant cells.

To boost TSH-dependent thyroid stimulation, control and mutant mice were subjected to a 4-week treatment with the goitrogens methimazole and sodium perchlorate, which increase TSH secretion by decreasing thyroid hormone synthesis. As expected, the weight of treated thyroids was increased 3-fold in WT mice, irrespective of gender, compared with untreated controls. However, unexpectedly, goitrogen treatment did not further augment the weight of the Pten^L/L;TPO-Cre thyroids over the 5- to 6-fold increase deriving from Pten ablation (Fig. 5A,, B, and E). WT goitrous thyroids showed a striking reduction of the follicle area with almost complete disappearance of the follicular structure (Fig. 5C). However, the thyroids of Pten^L/L;TPO-Cre mice still displayed rather large follicular lumina despite goitrogen treatment (Fig. 5D). In addition, remarkable thyrocyte hypertrophy characterized both genotypes.

In agreement with the thyroid weight data, the proliferation index of control goitrogen-stimulated glands was increased almost 7-fold compared with WT follicular cells, with no gender difference. Thyroids from treated male mutants had a 4-fold higher mitotic index compared with untreated male mutants and reached the same levels as treated females. However, the proliferation index of goitrogen-treated female Pten^L/L;TPO-Cre mice was only minimally increased compared with untreated mutant females (Fig. 5F). These data suggest that the PI3K/Akt cascade conveys a major proliferative signal in thyroid follicular cells, even under strong TSH stimulation, and that unrestrained PI3K activity is sufficient to obtain near-maximal proliferation in female mice.

Pten^L/L;TPO-Cre mice develop thyroid adenomas. To determine whether loss of Pten is sufficient to induce the development of neoplastic lesions, a cohort of mice was aged to 8 to 10 months and then sacrificed. At this age, the mutant thyroids were still growing (Fig. 1E) and still characterized by diffuse goiter with enlarged colloid-filled follicles (Fig. 6A–C). Additionally, all mice now showed focal hyperplasia, small nonencapsulated areas of hypercellularity with solid and/or microfollicular patterns, variable nuclear atypia, and little or no colloid among the dilated follicles of the colloid goiter (Fig. 6A–C). In addition, 70% of the female mice developed well-circumscribed follicular adenomas, often encapsulated, characterized by increased cellularity, severe reduction of the follicular areas resulting in a microfollicular and solid pattern, and the presence of mild nuclear atypia and several mitotic figures (Fig. 6A–F). The follicular nature of these lesions was confirmed by absence of calcitonin staining (data not shown).

The proliferation index of the nonneoplastic thyroid areas was, as in younger mice, significantly increased compared with control mice, again with a more severe increase in the females (Fig. 6G and H). Strikingly, the hyperplastic and adenomatous areas showed a further 5-fold proliferation increase, irrespective of sex, and with a wide variability between individual lesions, even within the same gland (Fig. 6H). In comparison, the thyroids from age-matched Pten^+/− mice displayed only scattered hyperplastic or small nodular lesions, characterized by complete loss of Pten expression, in an otherwise normal background (Fig. 6I; Table 1; data not shown). These results suggest that Pten loss creates a fertile environment for the development of highly proliferative neoplastic lesions, as a result of focal, clonal genetic alterations that act in cooperation with the activation of Akt.

The identification of the central signaling nodes in the cascades involved in thyrocyte growth and division is vital to understand the pathogenetic mechanisms of thyroid proliferative disorders, from simple goiter to thyroid carcinoma. In vitro and in vivo approaches have helped establish that TSH is the major driving force for thyroid growth and function (7). However, the relative contribution of the different TSH-initiated cascades to the various aspects of thyroid physiology and growth is still largely unknown. In fact, the different cell models used to dissect these pathways have often led to contradictory results (7).

One major unresolved issue is the extent of the contribution of PI3K/Akt pathway to thyrocyte proliferation, as a direct or indirect downstream effector of TSH. This question is particularly relevant in view of some striking divergences between clinicopathologic data and several in vitro studies.

Numerous clinical data point to the PI3K/Akt pathway as a major player in thyroid neoplastic transformation. Inherited PTEN mutations lead to Cowden disease, characterized by thyroid multinodular goiter and adenoma, and increased risk for thyroid cancer (18–20); in addition, activation of Akt (14) and gain-of-function mutations of PIK3CA, encoding the catalytic subunit of PI3K (13), have been correlated to thyroid carcinoma progression. PTEN subcellular distribution seems to shift from nuclear to cytoplasmic during neoplastic progression, before the dramatic reduction that characterizes advanced thyroid tumors (26) and, in follicular tumors, the localization of activated Akt is inversely correlated to the presence of PTEN in the nucleus (27).

However, whereas rat thyroid cells in culture seem to require PI3K activation downstream TSH stimulation for proliferation (7), TSH effectors, cAMP and PKA, cannot activate PI3K or Akt in dog primary thyrocytes, a system in which the PI3K/Akt pathway does not seem to be involved in TSH-mediated thyrocyte proliferation (11).

In this study, we have used a genetic model of Pten loss and constitutive Akt activation in thyroid follicular cells to investigate the role of the PI3K/Akt cascade in thyrocyte proliferation and in neoplastic transformation. The Pten^L/L;TPO-Cre mice can be considered both as a model of Cowden disease, in which loss of heterozygosity at the WT locus results in goiter and adenomas (28), and as a tool to investigate the connection between PI3K activation and sporadic goiter pathogenesis. Although more in-depth studies are necessary to define the Pten^L/L;TPO-Cre mice as a faithful model of human disease, they represent the first genetically defined model designed to examine the role of increased PI3K signaling in thyroid disease without resorting to nonphysiologic overexpression. Studies conducted using transgenic mice overexpressing members of the IGF-I pathway (29, 30) have resulted in mild phenotypes, and overexpression of members of the TSHR/cAMP pathway have reinforced the notion that TSH delivers, in vivo, the main proliferative signal for thyroid follicular cells (31–33).

Our data strongly suggest that activation of PI3K, in the absence of any alteration of TSH levels, is sufficient to induce a high rate of thyrocyte proliferation and the development of goiter. Several PI3K/Akt effectors were found activated in mutant thyroids. The p70S6K1/S6 axis has been suggested to play an essential role in thyroid proliferation and activity downstream TSH (34). Indeed, Pten^L/L;TPO-Cre thyroids were characterized by highly phosphorylated p70S6K1 and S6 proteins, suggesting that PI3K activation is sufficient to stimulate this pathway. We also found that PI3K/Akt activation was associated with increased levels of both cyclin D1 and cyclin D3, consistent with previous data supporting their role in thyrocyte proliferation and human thyroid proliferative disorders (35, 36).

Interestingly, activation of the PI3K signaling cascade was not associated to a reduction of p27Kip1 levels. Although this is in contrast with previous data linking Akt activation to p27Kip1 expression (37), it is tempting to speculate that safeguard mechanisms are in place to retain p27Kip1 levels to constrain the proliferative signals triggered by Akt activation. Later, on the other hand, loss of p27Kip1 expression might shift this compromised equilibrium toward proliferation, leading to malignant transformation as shown in human thyroid disease (38) and in a Pten^+/−;p27Kip1^−/− mouse model (39).

Strikingly, the increased proliferation characterizing Pten^L/L;TPO-Cre thyrocytes is more pronounced in females than males, mimicking the increased prevalence of thyroid disorders among women (40). Although more in-depth studies will be necessary to elucidate the exact mechanism responsible for the greater proliferative response in the mutant females, and its relevance to human disease, our data suggest that the activation of the PI3K/Akt/ERα axis, in the presence of physiologic levels of estrogen, may contribute to this phenotype, extending our recent findings obtained in an endometrial cancer model (41). It is also noteworthy that mutant thyroids show phosphorylation and delocalization of Foxo1, which has been reported to act as an ERα repressor (42).

It has been proposed that TSH can induce local IGF-I production and thus activate IGF-I receptor (IGF-IR) and its downstream cascades (43). In fact, several data in the literature point to a role of IGF-I in thyroid growth, either in parallel to or downstream TSH (7). In addition, thyroid enlargement is present in the majority of patients with acromegaly (44), and transgenic mice coexpressing IGF-I and IGF-IR develop diffuse goiter (29). Our data show that constitutive activation of one major IGF-IR effector, PI3K, induces a phenotype that is qualitatively similar, although quantitatively much more dramatic, to IGF-I/IGF-IR mice, underlining the relevance of this pathway to thyroid disease. Furthermore, we found that pathologically increased levels of TSH could only slightly increment the mitotic index of mutant thyrocytes, with no increase in the mutant gland weight, strongly suggesting that a conspicuous part of the proliferation signal induced by TSH is funneled through the PI3K/Akt cascade.

Despite the proliferative disorder in the thyroids of young Pten^L/L;TPO-Cre mice and the development of adenomas in several older mice, none of the mutants had developed invasive lesions by 11 months of age, showing that activation of the PI3K/Akt pathway is not sufficient for thyroid malignant transformation. Current studies are aimed at defining the identity of the genetic lesions that can cooperate with Pten loss to promote invasive tumor formation.

In summary, our data show that the PI3K/Akt cascade conveys a major proliferative signal in thyroid follicular cells that is sufficient to obtain near-maximal proliferation, at least in the females, and to induce the development of goiter and nodular thyroid lesions.

Grant support: Commonwealth of Pennsylvania to the Fox Chase Cancer Center (FCCC) and a National Cancer Institute core grant (FCCC).

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 Dr. Maureen Murphy for critical reading of the manuscript and FCCC for the Transgenic, DNA Synthesis, Laboratory Animal, and Biomarker facilities.