Involvement of mTOR and Regulation by AMPK in Early Iodine Deficiency-Induced Thyroid Microvascular Activation

Iodine deficiency (ID) induces TSH-independent microvascular activation in the thyroid via the reactive oxygen species/nitric oxide-hypoxia-inducible factor-1α/vascular endothelial growth factor (VEGF) pathway. We hypothesized the additional involvement of mammalian target of rapamycin (mTOR) as a positive regulator of this pathway and AMP-activated protein kinase (AMPK) as a negative feedback regulator to explain the transient nature of ID-induced microvascular changes under nonmalignant conditions. mTOR and AMPK involvement was investigated using an in vitro model (human thyrocytes in primary cultures) and 2 murine models of goitrogenesis (normal NMRI and RET-PTC mice [a papillary thyroid cancer model]). In NMRI mice, ID had no effect on the phosphorylation of ribosomal S6 kinase (p70S6K), a downstream target of mTOR. However, rapamycin inhibited ID-induced thyroid blood flow and VEGF protein expression. In the RET-PTC model, ID strongly increased the phosphorylation of p70S6K, whereas rapamycin completely inhibited the ID-induced increase in p70S6K phosphorylation, thyroid blood flow, and VEGF-A expression. In vitro, although ID increased p70S6K phosphorylation, the ID-stimulated hypoxia-inducible factor/VEGF pathway was inhibited by rapamycin. Activation of AMPK by metformin inhibited ID effects both in vivo and in vitro. In AMPK-α1 knockout mice, the ID-induced increase in thyroid blood flow and VEGF-A protein expression persisted throughout the treatment, whereas both parameters returned to control values in wild-type mice after 4 days of ID. In conclusion, mTOR is required for early ID-induced thyroid microvascular activation. AMPK negatively regulates this pathway, which may account for the transient nature of ID-induced TSH-independent vascular effects under benign conditions.

Iodine deficiency (ID) remains a global sanitary problem that still affects 1.88 billion people worldwide (1, 2). ID is associated with many thyroid disorders. To accommodate these disorders, the thyroid gland adapts endlessly to moderate ID independently of TSH via several mechanisms, including the preferential synthesis of T₃ over T₄, the recycling of intracellular iodide, and the peripheral conversion of T₄ into T₃ (3–5). Local iodine clearance can also be improved by an increase in the local blood flow. This was clearly demonstrated in animal models of goitrogenesis where an early TSH-independent phase starts as soon as 1 day after ID onset (6). Thus, in response to ID, a quick reaction occurs within thyrocytes, leading to an increase in thyroid blood flow and endothelial and pericyte activation under the control of vascular endothelial growth factor (VEGF)-A. Upon ID, the redox status of the cell changes immediately, which promptly increases intracellular levels of reactive oxygen species (ROS). This contributes to the stabilization of the α-subunit of hypoxia-inducible factor (HIF)-1, which then interacts with HIF-1β, thereby triggering VEGF-A gene transcription (6, 7). ROS encompass also reactive nitrogen species (RNS), including nitric oxide (NO), which is synthesized by NO synthase 3 via a pathway involving calcium and ryanodine receptors (8).

ID also induces VEGF-A release in RET receptor/papillary thyroid carcinoma 3 (RET/PTC3) mice, a mouse model of papillary thyroid cancer, but in contrast to nonmalignant models, the ID-induced VEGF-A mRNA increase is not ROS dependent, and HIF-1α plays a marginal role compared with its role in normal cells (9).

Iodine uptake by thyrocytes may also be influenced by the mammalian target of rapamycin (mTOR), as suggested by the decrease in iodine uptake when mTOR is stimulated (10). mTOR is a Serine (Ser)/Threonine (Thr) kinase involved in cell metabolism, proliferation, and survival in response to growth factors, hormones, nutrients, energy levels, hypoxia, and stress signals (11, 12). One of the main mTOR downstream cellular targets is the p70 ribosomal S6 kinase (p70S6K), a Ser/Thr kinase that is phosphorylated by mTOR at Thr389. pThr389-p70S6K induces ribosome biosynthesis by phosphorylating several proteins, including the ribosomal protein S6 and the eukaryotic initiation factor 4B (13). In addition to its pleiotropic effects on cell metabolism, mTOR plays also a role in regulating the transcription, translation, and stability of HIF-1α (14–16). mTOR is deregulated in many human malignancies, including those of the breast, lung, and kidney, and is now a target for the treatment of several cancers, including renal clear cell cancer, acute myeloid leukemia, and metastatic breast cancer (17–19). Increasing evidence suggests that mTOR plays also a role in thyroid cancers (20). On the basis of these observations, it seemed worthwhile considering the possibility that mTOR influences ID-induced vascular activation, not only in normal, but also in cancerous thyrocytes.

mTOR activity is negatively regulated by AMP-activated protein kinase (AMPK), a key energy-sensing kinase (21). AMPK is activated when the AMP to ATP ratio increases, as occurs during hypoxia. An increase in the AMP to ATP ratio induces the phosphorylation of AMPK by liver kinase 1 resulting in AMPK activation, which then inhibits mTOR (22, 23). Phosphorylated AMPK may also directly interact with mTOR to limit its activity and phosphorylates the tuberous sclerosis factor 2, which also contributes to the down-regulation of mTOR activity (24). Because early ID-induced TSH-independent microvascular activation in the thyroid is transient, this suggests the existence of a negative feedback mechanism. Based on the aforementioned arguments, we looked to see whether AMPK could act as a negative regulator of microvascular activation.

Materials and Methods

Human primary cultures

Human paranodular thyroid tissues were obtained from patients operated on for multinodular goiters (10 women and 2 men) after obtaining informed consent. Thyrocytes were isolated as described before (25). The cells were cultured in MEM medium containing 5% newborn calf serum (Invitrogen), 2.4mM glutamine (Sigma), 1-mU/mL TSH (Sigma), 100-U/mL penicillin-streptomycin (Invitrogen), 2.5-μg/mL fungizone (Invitrogen), and 10⁻⁸M NaI for 7 days. The day of the experiment, the culture media were replaced by fresh media with (controls) or without (ID) NaI. Rapamycin (10nM; LC Laboratories) or metformin (1mM; Sigma) was added to inhibit mTOR and activate AMPK, respectively.

Animals and treatments

All in vivo experiments were carried out with the approval of Université Catholique de Louvain Comité d'Ethique pour l'Expérimentation Animale and according to national and European animal care regulations. Mice were kept in the animal facility under a 12-hour light, 12-hour dark cycle and received tap water and an AO3 mouse breeding diet with normal iodine content (0.4 mg/kg; Scientific Animal Food and Engineering). For the mTOR study, 6-week-old Naval Medical Research Institute (NMRI) mice, as well as 6-month-old RET/PTC3 transgenic mice (RET receptor/papillary thyroid carcinoma 3) (a gift from Dr Hanane Derradji, SCK.CEN, Mol, Belgium) were used. For the AMPK study, 6-week-old AMPK-α1 knockout (KO) mice (a gift from Dr Benoit Viollet, Département d'Endocrinologie, Métabolisme et Cancer, Inserm, Institut Cochin, Paris, France) were used (26). Only female mice were used, except for transgenic RET/PTC3 mice, which were used irrespective of their gender. To induce ID, mice were fed a low-iodine diet (LID) (0.1 μg/kg; Animalabo) supplemented with NaClO₄ (1%), a sodium-iodide symporter inhibitor, in drinking water for 1, 2, 4, or 6 days. To inhibit mTOR, rapamycin (4 mg/kg · d; VWR) was injected ip, whereas metformin (250 mg/kg · d; Sigma) was injected ip to activate AMPK. Control animals were injected ip with the vehicle solution (4% ethanol, 0.2% sodium carboxymethylcellulose, and 0.25% Tween 80 for rapamycin; 5% polyethylene glycol-400 and 5% Tween 80 for temserolimus, or saline for metformin). Experiments were done with a minimum of 5 mice per group.

Laser Doppler blood flow measurements and tissue sample preparation

A laser Doppler imager (Moor Instruments) was used to measure changes in thyroid blood flow. Mice were anesthetized with 85-mg/kg anesketin (Eurovet) and 10-mg/kg Rompun (Bayer HealthCare) and the thyroid blood flow was measured as previously described (6). After measurement, mice were euthanized using pentothal and the thyroid lobes were dissected out. One thyroid lobe was fixed in 4% buffered-formalin (prepared in PBS) and embedded in paraffin, whereas the other was frozen in liquid nitrogen and kept at −80°C until used for Western blotting (WB) or quantitative real-time PCR (qPCR).

RNA isolation, reverse transcription, standard PCR, and qPCR

Cells and thyroid lysates were suspended in TriPure isolation reagent (Roche Diagnostics GmbH). Total RNA was purified according to the manufacturer's protocol and resuspended in ribonuclease-free water. Two micrograms of extracted RNA were used for the first-strand cDNA synthesis reverse transcription reaction as reported (6). AMPK-α1 and AMPK-α2 mRNAs were detected by standard PCR as follows: cDNAs (2.5 μL) were mixed with Go Taq green master mix (Promega) and 10μM each appropriate primer pair (Supplemental Table 1). Reaction mixtures were incubated for an initial activation at 94°C for 3 minutes, then 30 cycles of denaturation (94°C for 2 min), annealing (specific hybridization temperature [see Supplemental Table 1] for 2 min), and extension (72°C for 2 min), followed by a final extension step at 72°C for 10 minutes. PCR products were separated by agarose gel (1%) electrophoresis. Adipose tissue and heart from NMRI control mice were used as references for AMPK-α1 and AMPK-α2 expression, respectively. Digital imaging of ethidium bromide-stained agarose gels served for quantification by densitometry using the Scion ImageJ Software (National Institutes of Health). Results were expressed as AMPK-α1 to AMPK-α2 ratio.

VEGF-A mRNA (Supplemental Table 1) was analyzed by qPCR as reported previously (8).

WB analysis and antibodies

p70S6K, pThr389-p70S6K, Raptor, pSer92Raptor, AMPK-α, pThr172 AMPK-α, and HIF-1α protein expression was detected by WB in extracts of thyroid lobes and human thyrocytes from primary cultures. Cultured thyrocytes were suspended in Laemmli buffer (50mM Tris-HCL [pH 6.8], 2% sodium dodecyl sulfate, and 10% glycerol) containing a protease inhibitor cocktail (Sigma). The thyroid lobes were crushed in the same buffer. All samples were sonicated for 30 seconds. A BCA Protein Assay kit (Pierce) was used to measure the protein concentration of each sample according to the manufacturer's protocol and WB was performed as previously described (8). For protein detection, membranes were incubated with the primary antibody (see Table 1) followed by a second antirabbit antibody coupled to horseradish peroxidase (1:5000; 32260; Thermo Scientific), or an antimouse biotinylated secondary antibody (1:5000, BA9200; Labconsult) followed by avidin-biotin-peroxidase (ABC Perox kit; Vector Laboratories). The signal was detected using the Supersignal West Pico/Femto chemiluminescence kit (Pierce). WBs were scanned, and protein bands were quantified using ImageJ. Levels of all proteins (phosphorylated and total) were normalized to β-actin level, and the levels of phosphorylated proteins were further normalized to total protein levels.

Table 1

Antibody Table

Peptide/Protein Target	Antigen Sequence (if Known)	Name of Antibody	Manufacturer, Catalog Number, and/or Name of Individual Providing the Antibody	Species Raised in; Monoclonal or Polyclonal	Dilution Used
HIF-1α (human)	/	HIF-1A	BD Transduction, 610959	Mouse; monoclonal antibody	1:500
pThr389 p70S6K	/	Phospho-p70 S6 kinase (Thr389) (108D2)	Cell Signaling, 9234	Rabbit; mAb	1:1000
Total p70S6K	/	p70 S6 kinase (49D7)	Cell Signaling, 2708	Rabbit; mAb	1:1000
pThr172 AMPK-α	/	Phospho-AMPKα (Thr172) (40H9)	Cell Signaling, 2535	Rabbit; mAb	1:1000/1:50 (IHC)
Total AMPK-α	/	AMPKα antibody	Cell Signaling, 2532	Rabbit; mAb	1:1000
pSer792 Raptor	/	Phospho-Raptor (Ser792)	Cell Signaling, 2083	Rabbit; mAb	1:1000
VEGF-A	/	VEGF (VG-1)	Santa Cruz Biotechnology, Inc, SC-53462	Mouse; monoclonal antibody	1:25
β-Actin	N-GFAGDDAPRAVFPSK	Antiactin (20-33)	Sigma, A5060	Rabbit; polyclonal antibody	1:4000
AMPK-α1	CTSPPDSFLDDHHLTR	AMPK-α1	Kinase source AB140	Sheep; polyclonal antibody	1:1000
AMPK-α2	CMDDSAMHIPPGLKPH	AMPK-α2	Kinase source AB141	Sheep; polyclonal antibody	1:2000
Total Raptor	/	Total Raptor	Cell Signaling, 4978	Rabbit; polyclonal antibody	1:1000

Immunohistochemistry

VEGF-A, pThr172 AMPK-α, AMPK-α1, and AMPK-α2 were detected in paraffin-embedded tissue sections (5 μm thick). For VEGF-A and pThr172 AMPK-α assessment, sections were first pretreated in citrate buffer (0.01 mol/L) or in Tris-EDTA buffer (10mM Tris base and 1mM EDTA solution; pH 9) in a microwave oven, as previously described (6). Sections were washed with PBS supplemented with 1% BSA (PBS-BSA) and incubated with nonimmune goat serum (1:50; Vector Laboratories) for 30 minutes at room temperature. The slides were then incubated with the primary antibody (see Table 1) at room temperature overnight (VEGF-A and pThr172 AMPK-α) or for 3 hours (AMPK-α1 and AMPK-α2), washed with PBS-BSA, and incubated with the appropriate secondary antibody (60 min). The peroxidase reaction was visualized by incubating sections for 5 minutes with diaminobenzidine (Dako) and then counterstaining with hematoxylin. Negative controls were performed by omitting the primary antibody.

Statistical analysis

All data are expressed as the mean ± SEM of the data of at least 3 independent experiments (n ≥ 3) in vitro or of 5 mice (n = 5) in 1 representative experiment. Statistically significant differences (P < .05) were determined using two-way ANOVA followed by Tukey-Kramer post hoc test or using unpaired Student's t test when appropriate (GraphPad InStat).

Results

mTOR is involved in ID-induced early thyroid microvascular changes

mTOR activation was assessed by analyzing the phosphorylation state of one of its main substrates, p70S6K. In vivo, a band corresponding to the phosphorylated p70S6K protein was observed, indicating basal mTOR activity in control animals. Although the level of phosphorylated p70S6K was not affected by goitrogen treatment, mTOR inhibition by rapamycin prevented the phosphorylation of p70S6K (Figure 1A), as well as the ID-induced increase in thyroid blood flow (Figure 1B). In addition, as shown by immunohistochemistry, the increased VEGF-A protein expression was blocked by rapamycin (Figure 1C). In human thyrocytes in primary cultures, ID increased the level of phosphorylated p70S6K, as well as HIF-1α and VEGF-A expression in contrast with cells incubated with rapamycin (Figure 1, D–F), where HIF-1α and VEGF-A expression was similar to this in control cells. In control cells, although rapamycin decreased p70S6K phosphorylation, it had no effect on HIF-1α and VEGF-A expression. These results strongly suggest that mTOR is required for the activation of the HIF/VEGF pathway by ID and for ID-induced vascular changes occurring during the early phase of goiter development.

Figure 1

mTOR is involved in ID-induced microvascular changes in normal thyroids and human thyrocytes in primary cultures. Six-week-old NMRI mice were treated with LID+ClO4⁻ (1%) for 1 or 2 days and with rapamycin (rapa) (4 mg/kg · d) or vehicle solution (4% ethanol, 0.2% sodium carboxymethylcellulose, and 0.25% Tween 80) during the same period. Control mice were treated with rapamycin or vehicle solution over 2 days. pThr389 p70S6K (A) was detected by Western blotting. Thyroid blood flow (expressed as a percentage of the control) was measured by laser Doppler (B), and VEGF-A protein expression was detected by immunohistochemistry (C). Scale bars = 10 μm. Western blot densitometric values were normalized against β-actin and total p70S6K protein levels. Data are expressed as the mean ± SEM of the results of 5 mice (n = 5) from 1 representative experiment. *, P < .05; **, P < .01; and ***, P < .001 compared with the control (ctrl); +, P < .05 and ++, P < .01 compared with goitrogen treatment (LID+ClO4⁻). Primary cultures of human thyroid cells were iodide deprived in the presence or absence of rapamycin. pThr389 p70S6K (D) and HIF-1α (E) protein expression was detected by Western blotting. Densitometric values were normalized against β-actin and total p70S6K protein levels. VEGF-A mRNA expression was measured by RT-qPCR and normalized against β-actin levels (F). Data are expressed as the mean ± SEM of at least 3 experiments (n ≥ 3). *, P < .05; **, P < .01; and ***, P < .001 compared with the control (ctrl); +, P < .05 and ++, P < .01 compared with ID.

ID activates the mTOR pathway in RET-PTC3 mice

Because the mTOR pathway is deregulated in many malignancies, we looked at the possible involvement of mTOR in ID-induced microvascular changes in transgenic RET/PTC3 mice, a mouse model of papillary thyroid cancer. In contrast to what can be observed in NMRI mice, the level of Thr389 phosphorylated p70S6K was significantly increased by ID. This increase was fully inhibited by rapamycin as it was observed in NMRI mice despite the high degree of stimulation observed in RET/PTC3 mice (Figure 2A). The concomitant significant increase in the thyroid blood flow was also significantly decreased by rapamycin to the level observed in control animals (Figure 2B). Although VEGF-A protein was barely detected by immunohistochemistry in cells of papilla, VEGF-A staining was higher at day 2 of goitrogen treatment. Nevertheless, heterogeneity in the staining was observed in control mice where, although only a weak staining was observed in the large follicles developing papillae, some smaller follicular exhibiting a stronger staining. On the other hand, all follicles showed a stronger staining after 2 days of goitrogen treatment, including all papillae. The increase in VEGF-A staining was blocked by rapamycin and only a very weak staining was observed after rapamycin treatment (Figure 2C). These results suggest that mTOR is activated by ID in RET/PTC3 mice and is required for VEGF-A-dependent activation of the microvasculature.

Figure 2

mTOR is involved in ID-induced microvascular changes in RET-PTC mice. Six-month-old RET-PTC mice were treated with LID+ClO4⁻ (1%) for 2 days. They also received rapamycin (rapa) (4 mg/kg · d) or vehicle solution (4% ethanol, 0.2% sodium carboxymethylcellulose, and 0.25% Tween 80) during the same time period. Phosphorylated p70S6K (A) was detected by WB, thyroid blood flow (expressed as percentage of control) was measured using a laser Doppler (B), and VEGF-A protein expression was detected by immunohistochemistry (C). Scale bars, 10 μm. Data are expressed as the mean ± SEM of 5 mice (n = 5) from 1 representative experiment. **, P < .01 and ***, P < .001 compared with the control (ctrl); +, P < .05 and +++, P < .001compared with goitrogen treatment (LID+ClO4⁻).

AMPK exerts a negative feedback on ID-induced early thyroid microvascular changes

Because ID-induced microvascular activation is transient (6, 27), there must exist a negative feedback mechanism that turns off microvascular activation. Bearing in mind the role of mTOR, we hypothesized that AMPK may act as a negative regulator of mTOR. First, we sought to examine the expression of the 2 catalytic AMPK-α subunit isoforms. Both were detected by immunohistochemistry in normal murine thyroids (Figure 3A), although AMPK-α2 was detected weakly. Their respective mRNAs were also detected in murine thyroids and in human thyrocytes in primary cultures (Figure 3B). The AMPK-α1 to AMPK-α2 mRNA ratio observed in the thyroid (ratio of 2) was similar to that in the adipose tissue (ratio, 2.5), but different from that in the heart (ratio, 0.3), indicating that AMPK-α1 was the most predominant isoform (Figure 3B). At day 4 of goitrogen treatment, total AMPK-α expression was significantly increased (Figure 3C). The level of phosphorylated AMPK-α significantly increased from day 2 to day 6 after a decrease at day 1 (Figure 3D). To assess the activity of AMPK, we examined the phosphorylation status of Raptor, a downstream target of AMPK implicated in the regulation of mTOR (24). AMPK-mediated Raptor phosphorylation induces Raptor inactivation and reduces mTOR activity. After a decrease at day 1, no significant change in total Raptor protein expression was observed over the experimental period (Figure 3E). The phosphorylation state of Raptor was increased by 3-fold at day 4 of goitrogen treatment and remained high at day 6 (Figure 3F). These results indicate that goitrogen treatment activates AMPK, which is functional in our model from day 4 of goitrogen treatment.

Figure 3

AMPK-α is activated by ID in the thyroid gland. AMPK-α1 and AMPK-α 2 protein expression was detected by immunohistochemistry in thyroid sections of NMRI control mice (A). The mRNA expression of both isoforms was measured by standard RT-PCR and normalized against β-actin levels. Adipose tissue and heart were used as positive controls (B). Six-week-old NMRI mice were treated with LID+ClO4⁻ (1%) for 1, 2, 4, or 6 days. Total AMPK-α (C), pThr172 AMPK-α (D), Raptor (E), and pSer792 Raptor (F) were detected by WB. Densitometric values were normalized against β-actin. The results are expressed as the mean ± SEM of 8 mice (n = 8) from 2 experiment. *, P < .05; **, P < .01; and ***, P < .001 compared with the control (ctrl).

We used both pharmacological and genetic approaches to test the hypothesis that AMPK negatively regulates mTOR and ID-induced effects. We first used metformin, a drug commonly used in diabetes treatment, to activate AMPK both in vivo and in vitro and then evaluated the effects of ID in AMPK-α1 KO mice. Metformin increased the cytoplasmic level of phosphorylated AMPK-α (Figure 4A), confirming its stimulating effect on AMPK. This stimulating effect was also observed when administered together with the goitrogen treatment where an increased level of phosphorylated AMPK-α was observed, whereas total AMPK remained unchanged (Figure 4, B and C). When administered together with goitrogen, metformin blocked ID-induced blood flow, whereas it had no effect alone (Figure 5A). In addition, metformin inhibited ID-induced VEGF-A protein and mRNA expression at days 1 and 2 of goitrogen treatment (Figure 5, B and E). ID-induced HIF1α protein expression was decreased by metformin treatment to reach an expression level significantly lower than that in control animals (Figure 5C). As a result of AMPK activation, phosphorylated p70S6K, which was not increased by ID treatment, was decreased below the control level by metformin (Figure 5D). Both ratio of phosphorylated on total protein of AMPK-α (Figure 6A) and of Raptor (Figure 6B) were higher in human thyrocytes in primary cultures incubated with metformin alone for 4 hours. Metformin inhibited the ID-induced increase in the level of phosphorylated p70S6K, as well as that of VEGF-A mRNA. Metformin alone decreased the level of phosphorylated p70S6K, but unexpectedly increased VEGF-A mRNA expression (Figure 6, C and D). In AMPK-α1 KO mice, goitrogen treatment induced a significant increase in thyroid blood flow that was similar to that observed in wild type (WT) mice after 2 days. However, after 4 days of treatment, thyroid blood flow remained significantly high in KO mice (Figure 7A) whereas it returned to the control level in WT mice, indicating that, in contrast to WT mice, microvascular activation was not transient in KO mice. Similarly, VEGF-A protein expression detected by immunohistochemistry after 4 days of goitrogen treatment was higher in AMPK-α1 KO mice than in WT mice (Figure 7B). Taken together, these results suggest that AMPK acts as a negative feedback regulator of thyroid microvascular activation via a mechanism involving ID-induced activation of the HIF-1α/VEGF-A pathway.

Figure 4

Metformin activates AMPK-α in vivo. Six-week-old NMRI mice were treated with LID+ClO4⁻ (1%) for 1 or 2 days. They also received metformin (metf) (250 mg/kg · d) or saline solution. pThr172 AMPK was detected by immunohistochemistry (A). Scale bars, 10 μm. Total AMPK-α (B) and pThr172 AMPK-α (C) were detected by WB. Densitometric values were normalized against β-actin. The results are expressed as the mean ± SEM of 5 mice (n = 5) from 1 representative experiment. *, P < .05, compared with the control (ctrl); +, P < .05 compared with goitrogen treatment (LID+ClO4⁻).

Figure 5

AMPK-α activation by metformin inhibits ID-induced effects on thyroid blood flow, HIF-1α and pThr389 p70S6K protein, and on VEGF-A mRNA and protein expression. Six-week-old NMRI mice were treated with LID+ClO4⁻ (1%) for 1 or 2 days. They also received metformin (250 mg/kg · d) or saline solution. The thyroid blood flow (expressed as percentage of control) was measured using a laser Doppler (A). VEGF-A mRNA expression was measured by RT-qPCR and normalized against β-actin levels (B). HIF-1α protein expression was detected by WB. Densitometric values were normalized against β-actin (C). pThr389 p70S6K (D) was detected by WB. Densitometric values were normalized against β-actin and total p70S6K protein levels. VEGF-A protein expression was analyzed by immunohistochemistry (E). Scale bars, 10 μm. The results are expressed as the mean ± SEM of 5 mice (n = 5) from 1 representative experiment. *, P < .05; **, P < .01; and ***, P < .001 compared with the control (ctrl); +, P < .05 compared with goitrogen treatment (LID+ClO4⁻).

Figure 6

Metformin blocks ID-induced effects on pThr389 p70S6K protein and VEGF-A mRNA expression in vitro. Primary cultures of human thyroid cells were iodide deprived in the presence or absence of metformin (metf) (1mM). pThr172 AMPK-α (A) and pSer792 Raptor (B) were detected by WB. Densitometric values were normalized against β-actin and total AMPK-α protein levels. pThr389 p70S6K (C) was detected by WB. Densitometric values were normalized against β-actin and total p70S6K protein levels. VEGF-A mRNA expression was measured by RT-qPCR and normalized against β-actin levels (D). Data are expressed as the mean ± SEM of at least 3 experiments (n ≥ 3). *, P < .05 and **, P < .01 compared with the control (ctrl); +, P < .05 and ++, P < .01 compared with goitrogen treatment (LID+ClO4⁻). $, P < .05 compared with metformin treatment alone.

Figure 7

ID-induced effects on thyroid blood flow and VEGF protein expression are not transient in AMPK-α1 knockout mice. Six-week-old wild-type and AMPK-α1 knockout mice were treated with LID+ClO4⁻ (1%) for 1, 2, or 4 days. The thyroid blood flow (expressed as percentage of control) was measured using a laser Doppler (A) and VEGF-A protein expression was detected by immunohistochemistry (B). Scale bars, 10 μm. Data are expressed as the mean ± SEM of 5 mice (n = 5) from 1 representative experiment. *, P < .05 and **, P < .01 compared with the control (ctrl); ++, P < .01 compared with goitrogen treatment (LID+ClO4⁻). $, P < .05 compared with AMPK-α1 knockout mice.

Discussion

Although ID is still a common problem worldwide, little is known about the impact of mild or short-term ID in thyrocytes (2). In both healthy and cancerous thyroids, ID induces a rapid TSH-independent activation of the microvasculature via a VEGF-dependent pathway (6, 7, 9, 27). In healthy thyroid, this mechanism depends on HIF-1α stabilization and ROS/NO production whereas in cancerous thyrocytes, it is only partially dependent on HIF and independent of ROS (6, 9). This led us to hypothesize that additional pathways might be involved (6, 8). mTOR, which is known for its role in the regulation of cell metabolism, as well as in proliferation and cell survival, plays also a role in angiogenesis by regulating the HIF-1/VEGF pathway (28–30). Our results show that mTOR activity is required for HIF-1/VEGF-dependent microvascular activation induced by ID. This was clearly demonstrated by the inhibitory effect of rapamycin, which fully blocked both the ID-induced HIF/VEGF pathway and thyroid blood flow. mTOR is well known as a factor involved in cell energy metabolism, which it regulates by increasing glycolytic flux via activation of HIF-1 at transcriptional and translational levels (12). There is also growing evidence to suggest that mTOR participates in the activation of the HIF/VEGF pathway and, in angiogenesis, a common feature of many cancers (31). For example, in a murine melanoma cell line, the inhibition of mTOR decreased VEGF-A production both in cell cultures and xenografts and reduced the number of mature and immature blood vessels (32). In addition, mTOR inhibition by rapamycin effectively combats tumor growth by reducing VEGF-A protein expression and by inhibiting HIF-1α activity in ovarian cancers and myeloma cells (33, 34). Less is known about the role of mTOR in vascular regulation under benign conditions, although rapamycin was shown to inhibit the stimulation of VEGF by nerve growth factor in a human keratinocyte cell line (35). We showed both in vivo and in vitro that a link exists between mTOR and the ID-induced HIF-1/VEGF pathway. Hence, these results support the hypothesis for a role of mTOR in ID-induced microvascular activation (Figure 8).

Figure 8

Schematic representation of the involvement of mTOR and its regulation by AMPK in ID-induced intracellular HIF-VEGF pathway. In thyrocytes, mTOR is required for the ID-induced stabilization of HIF-1α, which then associates with HIF-1β to form the active transcription factor HIF-1. Upon activation, HIF-1 binds to the hypoxia response element (HRE) regulatory unit in the promoter region of the VEGF-A gene, thereby initiating VEGF transcription and promoting VEGF-A protein production. Once released, VEGF-A acts as a paracrine factor to activate adjacent endothelial cells and promote microvascular thyroid reshaping. Thereafter, a negative feedback mechanism dependent on AMPK-α is initiated. When activated, AMPK-α inhibits the mTOR-induced HIF-VEGF pathway. This negative regulatory mechanism limits the duration of TSH-independent ID-induced activation of the microvascular bed directly adjacent to thyrocytes. The mechanism could be lost in precancerous or cancerous conditions, thereby resulting in unrestrained microvascular expansion. NIS, sodium-iodide symporter; TJ, tight junctions.

Nevertheless, although rapamycin clearly inhibited ID-induced VEGF expression and thyroid blood flow, as well as p70S6K phosphorylation in vivo, no effect of ID on p70S6K activation was observed. This suggests that although mTOR is required for ID-induced vascular activation, it may not be directly stimulated by ID in normal thyrocytes. Although p70S6K is one of the main targets of mTOR and is used as an indicator of mTOR activation, it is not the only mTOR target acting upstream of HIF-1 (36). mTOR may actually act on HIF-1α either directly or indirectly via various intermediates, such as eukaryotic initiation factor 4E (eIF4E)-binding protein-1, whose inactivation by mTOR phosphorylation leads to the activation of eIF4E. eIF4E is involved in the regulation of HIF-1α target genes (37, 38). Alternatively, mTOR may act directly on HIF-1α by interacting with the oxygen-dependent degradation domain of HIF-1α. Indeed, the oxygen-dependent degradation domain is a serine- and threonine-rich site with Ser/Thr-Pro motifs (39) that are potential targets of the mTOR kinase domain (28, 40). Further investigations are needed to determine precisely which pathway links mTOR to HIF.

In RET/PTC mice, the ID-induced increase in VEGF-A protein expression and thyroid blood flow was associated with strong activation of p70S6K, indicating that goitrogen treatment induces mTOR in this model. ROS/RNS are not directly involved in ID-induced vascular activation in cancerous thyrocytes, as previously reported (7). The fact that mTOR inhibition by rapamycin completely inhibited ID-induced increase in p70S6K phosphorylation, thyroid blood flow, and VEGF-A expression suggests that, in contrast to normal thyrocytes where, in addition to mTOR, the ROS/RNS pathway is involved, the mTOR-VEGF pathway plays a predominant role in ID-induced microvascular changes in this specific model of thyroid papillary carcinoma. This important role of mTOR in RET/PTC mice is in line with observations made in human papillary thyroid carcinoma where mTOR is overexpressed (41) and with the well-known role of the phosphoinositide 3-kinase (PI3K)/protein kinase B (AKT)/mTOR pathway as one of the regulators of tumor-associated angiogenesis (38).

Because the activation of microcirculation by ID is transient in normal thyroids (6), it is worth considering the possibility that a negative feedback mechanism is involved in restraining the microvascular activation induced by ID. Because AMPK is a well-known negative regulator of mTOR, it appears to be a good candidate for this task. Both catalytic isoforms of AMPK-α are expressed in thyrocytes, in agreement with results obtained previously with FRTL5 cells (42). Involvement of AMPK was first suggested by data showing an ID-induced increase in the expression and phosphorylation of AMPK-α and in the phosphorylation of an AMPK-α substrate Raptor. In this study, the role of AMPK as a negative feedback regulator of microvascular activation was confirmed by using a pharmacological approach (metformin) and a model of gene silencing (the AMPK-α1 KO mouse model). Metformin is known for its effects on the thyroid gland; it not only reduces the incidence of nodular goiters, but it is also associated with a higher remission rate of thyroid cancers (43, 44). Likewise, experimental studies show that metformin impairs the growth and migration of thyroid cancer cell lines (45). In this study, metformin convincingly activated AMPK and blocked ID-induced mTOR activation, thyroid blood flow, and VEGF-A protein expression both in vivo and in vitro. It is therefore reasonable to propose that AMPK, from day 4 of goitrogen treatment, acts to counteract ID-induced vascular effects in normal thyrocytes, which may account for the transient nature of the vascular changes induced by ID. This hypothesis is reinforced by the results obtained in AMPK-α1 KO mice where ID-induced thyroid blood flow and VEGF-A protein expression continued to increase without restraint, in contrast to WT mice where both parameters returned to control values. Taken together, these results strongly suggest that AMPK is a negative feedback regulator that limits ID-induced microvascular regulation in the normal thyroid. How AMPK is activated remains to be investigated. It could occur via an increase in NO, which is released under these conditions (8). A role for NO in the activation of AMPK has been proposed in pancreatic islets (46), but whether or not it has a role in the thyroid remains to be determined.

Another issue that remains to be addressed is the unexpected increase in VEGF expression in metformin-treated human thyrocytes. Indeed, although metformin inhibited ID-induced VEGF expression by inhibiting mTOR, metformin alone increased VEGF expression in control animals. Although puzzling at first sight, these observations can be explained as follows. In several cell types (hepatocytes, skeletal muscle cells, endothelial cells, and pancreatic β-cells) metformin inhibits the mitochondrial respiratory chain complex 1, which results in a transient lowering of cellular energy status and the AMP to ATP ratio. This in turn increases the activity of AMPK (47, 48), which can then activate the HIF-VEGF pathway (49, 50). Alternatively, and besides the effect of metformin on the AMP to ATP ratio, the inhibition of the mitochondrial complex 1 by metformin may also result in increased ROS and nitrogen species production, which also activates AMPK (51). In addition to their action on AMPK, ROS produced by mitochondria may, by themselves, directly induce the HIF-1-VEGF pathway (52).

In conclusion, this study adds to our knowledge of how thyroid microcirculation is activated during the early steps of ID and how this activation is regulated to avoid unrestricted vascular expansion (Figure 8). Under ID conditions, the increase in the cytoplasmic level of ROS/NO stabilizes HIF-1α, resulting in an increase in the VEGF-A protein level. mTOR is required for the increase in the VEGF-A protein level. Conversely, AMPK activation, which occurs shortly after mTOR activation, acts as a negative regulator of mTOR and turns off VEGF-A expression. This succession of events explains the transient nature of TSH-independent ID-induced microvascular activation. In malignant RET/PTC3 cells, even though biochemical pathways other than mTOR are likely involved in the microvascular regulation, the dysregulation of the 2 specific pathways considered in the present study could lead to unrestrained VEGF activation.

Acknowledgments

We thank Professor M. Mourad, a surgeon at Saint-Luc Academic Hospital, Brussels and Dr N. Abbes, a surgeon at Centre Hospitalier Régional Mons-Hainaut, for providing human paranodular thyroid tissues; Dr B. Viollet (Département d'Endocrinologie, Métabolisme et Cancer, Inserm, Institut Cochin Paris, France) for providing the AMPK-α1 knockout mice; and Christine de Ville de Goyet and Marc de Bournonville for their expert technical assistance. L.B., S.H., and P.S. are research associates at Fonds de la Recherche Scientifique-Fond national de la recherche scientifique (Belgium).

Present address for I.M.C. and A.-C.G.: Service d'Endocrino-Diabétologie, Centre Hospitalier Régional, Mons-Hainaut 7000, Belgium.

This work was supported by the Fond national de la recherche scientifique Grant J.0112.13 and by Le Fond pour le Cancer from Université Catholique de Louvain. P.S. was supported by the Interuniversity Attraction Pole Grant UP7-03 (Belspo) and by an Action de Recherche Concertée from the Communauté Française de Belgique (ARC 14/19-058).

Disclosure Summary: The authors have nothing to disclose.

Abbreviations

 
• AMPK

  AMP-activated protein kinase

 
• eIF4E

  eukaryotic initiation factor 4E

 
• HIF

  hypoxia-inducible factor

 
• ID

  iodine deficiency

 
• KO

  knockout

 
• LID

  low-iodine diet

 
• mTOR

  mammalian target of rapamycin

 
• NO

  nitric oxide

 
• p70S6K

  ribosomal S6 kinase

 
• qPCR

  quantitative real-time PCR

 
• RET/PTC3

  RET receptor/papillary thyroid carcinoma 3

 
• RNS

  reactive nitrogen species

 
• ROS

  reactive oxygen species

 
• Ser

  Serine

 
• Thr

  Threonine

 
• VEGF

  vascular endothelial growth factor

 
• WB

  Western blotting

 
• WT

  wild type.

Author notes

^*

I.M.C. and A.-C.G. contributed equally to this work.