Metformin blocks progression of obesity-activated thyroid cancer in a mouse model

Compelling epidemiologic evidence indicates that obesity is associated with a high risk of human malignancies, including thyroid cancer. We previously demonstrated that a high fat diet (HFD) effectively induces the obese phenotype in a mouse model of aggressive follicular thyroid cancer (ThrbPV/PVPten+/−mice). We showed that HFD promotes cancer progression through aberrant activation of the leptin-JAK2-STAT3 signaling pathway. HFD-promoted thyroid cancer progression allowed us to test other molecular targets for therapeutic opportunity for obesity-induced thyroid cancer. Metformin is a widely used drug to treat patients with type II diabetes. It has been shown to reduce incidences of neoplastic diseases and cancer mortality in type II diabetes patients. The present study aimed to test whether metformin could be a therapeutic for obesity-activated thyroid cancer. ThrbPV/PVPten+/−mice were fed HFD together with metformin or vehicle-only, as controls, for 20 weeks. While HFD-ThrbPV/PVPten+/−mice had shorter survival than LFD-treated mice, metformin had no effects on the survival of HFD-ThrbPV/PVPten+/−mice. Remarkably, metformin markedly decreased occurrence of capsular invasion and completely blocked vascular invasion and anaplasia in HFD-ThrbPV/PVPten+/−mice without affecting thyroid tumor growth. The impeded cancer progression was due to the inhibitory effect of metformin on STAT3-ERK-vimentin and fibronectin-integrin signaling to decrease tumor cell invasion and de-differentiation. The present studies provide additional molecular evidence to support the link between obesity and thyroid cancer risk. Importantly, our findings suggest that metformin could be used as an adjuvant in combination with antiproliferative modalities to improve the outcome of patients with obesity-activated thyroid cancer.


INTRODUCTION
The incidence of thyroid cancer, the most common malignancy in endocrine organs, has been increasing rapidly in the past decades [1,2]. At the same time, the rates of obesity and metabolic syndrome have also risen. Recent epidemiologic studies have shown a positive association of obesity with thyroid cancer incidence [3][4][5][6]. Many retrospective studies of patients with papillary thyroid cancer (PTC) show that a higher body mass index is correlated with a more aggressive PTC phenotype, such as increased tumor size, extrathyroidal invasion, and advanced tumor, node, metastasis (TNM) stage independent of age, sex, and other confounding factors [7]. These compelling epidemiologic data on the positive correlation of obesity with the risk of thyroid cancer prompted us to explore the molecular basis underpinning such a correlation.
We used a mouse model of follicular thyroid cancer (Thrb PV/PV Pten +/− mice) to elucidate the underlying mechanisms. Thrb PV/PV Pten +/− mice express a potent dominantly negative thyroid hormone receptor β (TRβPV) and haplo insufficiency in the Pten gene (phosphatase and tensin homologue deleted from chromosome 10) [8]. We fed Thrb PV/PV Pten +/− mice a high fat diet (HFD) to induce obesity marked by increased body weight, enlarged fat cells, and elevated serum leptin levels [9]. Biochemical and histopathologic analyses showed that the obese Thrb PV/ PV Pten +/− mice exhibit more aggressive tumor progression with increased tumor cell proliferation, shortened survival, and frequent occurrence of anaplasia [9]. Moreover, we also identified leptin-JAK2-STAT3 signaling as one pathway that mediates the obesity-induced aggressive tumor progression. Thus, these findings not only provide direct molecular evidence to support the link between obesity and thyroid cancer risk, but also open the possibility of using Thrb PV/PV Pten +/− mice to test potential molecular targets for treatment of obesity-induced thyroid cancer.
More recently, we treated Thrb PV/PV Pten +/− mice with a STAT3-specific inhibitor, S3I-201, aiming to block the STAT3-downstream signaling to delay obesity-exacerbated thyroid cancer progression [10]. We found that S3I-201 effectively inhibits HFD-induced aberrant activation of STAT3 and its downstream targets to markedly inhibit thyroid tumor growth and prolong survival. S3I-201 also acts to decrease expression of the key regulators of the epithelial-mesenchymal-transition, i.e., vimentin and matrix metallo proteinases, to block anaplasia and lung metastasis [10]. Thus, using HFD-Thrb PV/PV Pten +/− mice, we have shown that inhibition of the STAT3 activity would be a novel treatment strategy for obesity-induced thyroid cancer.
With the availability of HFD-Thrb PV/PV Pten +/− mice as a preclinical mouse model, we expanded the search for other potential treatment modalities for obesityinduced thyroid cancer. We considered metformin (1,1-dimethylbiguanide hydrochloride), the most widely used antihyperglycemic drug for treatment of type II diabetes patient sowing to its effectiveness, safety profile, and affordability [11,12]. In addition to its anti-diabetic effect, epidemiologic evidence suggests that metformin may lower cancer risk, increase healthy life span and improve outcomes among diabetic patients [13][14][15][16][17][18][19]. Numerous studies have shown that metformin could reduce the risk of developing solid tumors [20], such as colorectal, liver, pancreatic, stomach, breast, and thyroid cancer [20][21][22][23][24]. Recently, several studies have reported that metformin inhibits cell proliferation in thyroid cancer cells including medullary, anaplastic, and PTC cell lines [20,[25][26][27]. Still lacking, however, is direct molecular evidence to demonstrate the effectiveness of metformin in the treatment of obesity-induced thyroid cancer in vivo. In the present studies, we treated HFD-Thrb PV/PV Pten +/− mice with metformin and evaluated its effects on survival, tumor growth, tumor cell invasion, and occurrence of anaplasia. We found that metformin markedly decreased occurrence of capsular invasion and completely blocked vascular invasion and anaplasia in HFD-Thrb PV/PV Pten +/− mice. These results suggest that metformin could be beneficial for patients with obesity-activated thyroid cancer.
We further evaluated the effect of metformin on thyroid tumor progression by comparing histopathologic characteristics of Thrb PV/PV Pten +/− mice with different treatments at the same age. Consistent with our previous observations, HFD promoted tumor progression from extensive hyperplasia ( Figure 2B in that no occurrences of vascular invasion and anaplasia were detected in HFD-Thrb PV/PV Pten +/− mice treated with metformin (compare bar 4 with bar 3 in panels b and c). These results indicate that metformin treatment was effective in blocking tumor progression.

Metformin inhibits the STAT3 signaling pathway in HFD-Thrb PV/PV Pten +/− mice
Previously, we elucidated that the activation of leptin-JAK2-STAT3 signaling accounts for the HFDinduced promotion of thyroid tumor progression in Thrb PV/ PV Pten +/− mice [9]. To understand how metformin blocked HFD-induced tumor progression, we first evaluated the changes brought about by metformin in the key regulators of the Leptin-JAK2-STAT3 signaling. As shown in Figure  3A-I, metform in lowered p-STAT3 (Y705) protein levels in thyroid tumors of HFD-Thrb PV/PV Pten +/− mice (compare lanes 10-12 to lanes 4-6) without changing the total STAT3 protein levels (panel b). The quantitative analysis of the ratios of p-STAT3 (Y705) versus total STAT3 indicated that the HFD-induced activation of STAT3 signaling ( Figure 3A-II, bar 3) was attenuated by metformin ( Figure  3A-II, bar 4). We further carried out immunohistochemical analysis to determine the protein abundance of p-STAT3   Figure 3B-II, bar 2 versus bar 1). Taken together, these data indicate that metformin acted to inhibit the activation of STAT3.
Leptin mediates its effects not only via STAT3, but also via extracellular signal-regulated kinase (ERK) [28]. We therefore evaluated the activity of ERK by examining p-ERK protein levels. HFD elevated p-ERK (T202/204) without significant changes in total ERK protein levels ( Figure 4A-I-a and 4A-I-b, lanes 4-6 versus lanes 1-3; also see the quantitative data: Figure 4A-II-a, bar 3 versus bar 1). Metformin treatment reduced p-ERK protein levels ( Figure 4A-I, lanes 10-12 versus lanes 4-6; also see the quantitative data: Figure 4A-II-a, bar 4 versus bar 3). Therefore, leptin-mediated activation of STAT3 and ERK pathways was attenuated by metformin.
Since metformin blocked thyroid tumor progression of HFD-Thrb PV/PV Pten +/− mice, we next focused on the analysis of regulators affecting cytoskeletal structure, cell motility, and migration. We first examined whether vimentin protein levels were affected by metformin.
Vimentin is a type III intermediate filament protein and is a major cytoskeletal component in mesenchymal cells. Vimentin is often used as a marker for cells undergoing epithelial-mesenchymal-transition (EMT) during metastatic progression. Vimentin is positively regulated by STAT3 [29,30]. Moreover, biochemical analyses demonstrated direct interaction of vimentin with ERK, which promoted ERK activation and enhanced vimentin transcription [31,32]. In line with these findings, we found that elevated vimentin in thyroid tumors of HFD-Thrb PV/PV Pten +/− mice was inhibited by metformin, thereby decreasing the extent of EMT ( Figure 4A-I, panel c, lanes 10-12 versus lanes 4-6; also quantitative data: Figure  4A-II-b, bar 4 versus bar 3) to suppress cell invasion. We next evaluated whether extracellular matrix components were affected by metformin in thyroid tumor cells of HFD-Thrb PV/PV Pten +/− mice. Fibronectin (FN) is a high molecular weight protein of the extracellular matrix that binds to membrane-spanning receptor proteins, known as integrins. FN plays a major role in cell adhesion, migration, and metastasis [33,34]. Recent studies suggest that metformin treatment could reduce tumor cell invasion in pancreatic cancer [35]. Metformin treatment of diabetetic patients is associated with low recurrence of cervical lymph node metastasis of differentiated thyroid cancer [36]. Indeed, we found that FN protein abundance was higher in thyroid tumors of HFD-Thrb PV/PV Pten +/− mice than in LFD-Thrb PV/PV Pten +/− mice ( Figure 4B-I, panel a, lanes 4-6 versus lanes 1-3; quantitative data: Figure  4B-II-a, bar 3 versus bar 1). Metformin treatment lowered FN protein abundance in thyroid tumor cells of HFD-Thrb PV/PV Pten +/− mice ( Figure 4B-I, panel a, lanes 10-12 versus lanes 4-6; quantitative data: Figure 4B-II-a, bar 4 versus 2).
Increased integrin expression levels contribute to earlier metastatic potential of thyroid cancer [37]. For example, integrin α6 has been reported to mediate progression of papillary thyroid cancer [38]. Therefore, we further evaluated the protein levels of FN receptors such as integrins α6, β1, and β3 in thyroid tumors of LFDor HFD-Thrb PV/PV Pten +/− mice. We found that integrin α6 ( Figure 4B

DISCUSSION
In the present studies, we used the preclinical model of Thrb PV/PV Pten +/− mice fed with HFD to test the effect of metformin on obesity-activated thyroid cancer progression. Treatment of HFD-Thrb PV/PV Pten +/− mice with metformin for 20 weeks had no effect on the survival or the thyroid tumor growth of these mice. Remarkably, however, the treatment reduced the occurrence of capsular invasion and abrogated the development of vascular invasion and anaplasia. Since the endpoint of metformin treatment was 20 weeks, which did not allow sufficient time for metastasis to occur, we were unable to evaluate the effect of metformin on the development of metastasis. Using molecular and biochemical analyses, we found that metformin attenuated the activity of STAT3-ERK-vimentin and FN-integrin signaling to reduce the occurrence of capsular invasion and blocked vascular invasion and anaplasia. However, the altered FN-integrin signaling may not be the only pathway that led to the reduced tumor cell invasion by metformin. We found that metformin also suppressed the occurrence of thyroid tumor cell invasion in LFD-treated Thrb PV/PV Pten +/− mice, but no changes in the FN-integrin signaling. These observations suggested that there were additional pathways affected by metformin that could reduce the tumor cells invasion. The identification of such pathways would await future studies.
The findings that metformin blocked vascular invasion of thyroid tumors of Thrb PV/PV Pten +/− mice are in line with the reports in which the anti-angiogenic effects of metformin were described in colon cancer [39], HER2+ tumor cells [40] and breast tumors [41,42]. In colon cancer, metformin was shown to down regulate tumor angiogenesis and augment the antitumor effect oxaliplatin [39]. In HER2+ tumor cells, metformin treatment decreased microvessel-induced inhibition of tumor angiogenesis [40]. In breast tumors, metformin reduced tumor microvessel density and attenuated tumor angiogenesis [41,42]. While metformin was shown to have anti-neoplastic effects in diabetic patients with differentiated thyroid cancer [25], whether metformin has anti-angiogenic effects in thyroid cancer progression remains to be elucidated.
That metformin could block cancer cell invasion and anaplasia, but not tumor growth, suggested that the molecular pathways of anti-diabetic and anticancer actions of metformin could differ in vivo. Although the detailed molecular basis underlying the metabolic effects of metformin are not completely understood, the primary molecular mechanism mediating this effect appears to be the activation of AMP-activated protein kinase (AMPK) and the subsequent inhibition of mammalian targets of rapamycin (mTOR) [43][44][45]. Intriguingly, no significant changes in the AMPK-mTOR-p70 S6K and AMPK-mTOR-4EBP1 signaling pathways were detected in thyroid tumors of HFD-Thrb PV/PV Pten +/− mice (data not shown). These results suggest that metformin treatment had not led to the changes of protein synthesis and lipid metabolism during thyroid carcinogenesis of HFD-Thrb PV/ PV Pten +/− mice as expected for its anti-diabetic effects. This notion is consistent with our findings that metformin treatment did not affect tumor growth ( Figure 1B-b).
We have shown previously that in the thyroid of Thrb PV/ PV Pten +/− mice, PI3K-AKT is highly activated to drive aggressive tumor growth [8]. Knowing that AKT is also a downstream target of activated insulin signaling [46], we speculate that highly activated AKT signaling could interfere with the AMPK signaling via cross talk such that the metformin-mediated inhibition of AMPK could be blunted in the thyroid of HFD-Thrb PV/PV Pten +/− mice. This conjecture raises the possibility that the effectiveness of metformin as an antiproliferative drug would depend on the genetic abnormalities and altered signaling pathways of thyroid cancer. That AMPK sensitivity to metformin inhibition could be modulated by cellular context is not without precedent. It has been shown that p53 is involved in mediating the energy-conserving response to AMPK activation and that loss of p53 heightens the metformininduced energy stress on cancer cells, implying that metformin may have increased efficacy in p53-deficient tumors like ovarian cancer [47]. In addition, recent studies showed that cancer cell lines with mutations in mitochondrial DNA genes have an increased response to metformin [48]. In line with these observations, an ongoing clinical trial is currently evaluating the effect of metformin on dosing of levothyroxine (T4) required for TSH suppression in patients with differentiated thyroid cancer (ClinicalTrials.gov; Identifier: NCT01341886). The outcome of this trial should uncover how the sensitivity of TSH response to T4 is modulated by metformin and could further support the notion that the effectiveness of metformin is subject to modulation by cellular regulators.
The present studies showed that metformin failed to exert antiproliferative effects to inhibit HFD-induced tumor growth in Thrb PV/PV Pten +/− mice. Still, metformin was effective in delaying thyroid cancer progression by reducing the frequency of capsular invasion and abrogating the development of vascular invasion and anaplasia. These findings suggest that metformin would be useful in preventing the metastatic spread of thyroid cancer. These findings also suggest that metformin would be useful as an adjuvant in combination treatment with other antiproliferative drugs. In view of metformin's safety and affordability, we would expect that clinical trials could soon be extended to test the effectiveness of metformin to treat differentiated and de-differentiated thyroid cancer. In addition to Thrb PV/PV Pten +/− mice, other mouse models have recently been developed to study thyroid cancer [49][50][51][52][53][54], and it would be possible to use different mouse models mimicking different types of thyroid cancer to assess the effectiveness of metformin in those models. Furthermore, these preclinical studies might also lead to the identification of genes that could potentially enhance metformin's effects, as well as genes that could render metformin insensitive. The findings of these proposed preclinical studies would be very helpful in the consideration of using metformin to improve the outcome of patients with thyroid cancer.

Mice and treatment
The National Cancer Institute Animal Care and Use Committee approved the protocols for animal care and handling in the present study. Mice harboring the ThrbPV gene (Thrb PV/PV mice) were prepared via homologous recombination, and genotyping was carried out using the PCR method, as previously described [8]. Pten +/− mice were kindly provided by Dr. Ramon Parsons (Columbia University, NY, USA). Thrb PV/PV Pten +/− mice were obtained by first crossing Pten +/− mice with Thrb PV/PV mice followed by further crossing the heterozygous offspring Thrb PV/+ Pten +/− with Thrb PV/+ Pten +/− mice. The high fat diet (HFD) (60 % Kcal from fat) was purchased from Research Diets (New Brunswick, NJ, USA). The mice, housed in SPF (specific pathogens free) animal facility, were administrated HFD diet from the age of 6 weeks until the end of the study. Metformin (cat#M1566, Spectrum, Gardena, CA, USA) was diluted in drinking water (0.5 mg/ml) [55], and the solution was changed weekly. They were monitored until they reached the age of 21 weeks, or they became moribund with rapid weight loss, hunched posture, and labored breathing. After the mice were euthanized, the thyroids were dissected for weighing, histologic analysis, and biochemical studies.

Histopathologic analysis
Thyroid glands, lungs and inguinal fat were dissected and fixed in 10% neutral-buffered formalin (Sigma-Aldrich, St. Louis, MO, USA) and subsequently embedded in paraffin. Five-micrometer-thick sections were prepared and stained with hematoxylin and eosin. For each animal, single random sections of thyroid were examined. For thyroids, morphologic evidence of hyperplasia, capsular invasion, and vascular invasion was routinely examined in that single section.
Immunohistochemistry (IHC) was conducted as previously described with some modifications [50]. For the antigen retrieval step, slides were heated in 0.05% citraconic anhydride solution (pH 7.4; Sigma-Aldrich, St. Louis, MO, USA) at 98°C for 60 minutes followed by treatment with rabbit anti-p-STAT3 antibody (1:100 dilution, Cell Signaling, Denver, MA, USA) and anticleaved caspase 3 antibody (dilution 1:300, cat. 9661, Cell Signaling) at 4°C overnight. The antigen signals were detected by treatment with the peroxidase substrate diaminobenzidine, followed by counterstaining with Gill's hematoxylin (Electron Microscopy Sciences, Hatfield, PA, USA). Relative positive cell ratio was quantified by using NIH IMAGE software (Image J 1.47).

Statistical analysis
All data are expressed as mean ± standard errors, and Student's t test was used to compare continuous variables accordingly. The Kaplan-Meier method with log-rank test was used to compare survival in each treatment group. Statistical significance was set at p<0.05. GraphPad Prism 6.0 (GraphPad Software, La Jolla, CA, USA) was used to draw graphs.

ACKNOWLEDGMENTS
The present research was supported by the Intramural Research Program at the Center for Cancer Research, National Cancer Institute, National Institutes of Health.