Thyroid hormone receptor phosphorylation regulates acute fasting-induced suppression of the hypothalamic–pituitary–thyroid axis

Significance

Thyroid hormones (T₄ and T₃, TH) and fasting suppress the hypothalamic–pituitary–thyroid (HPT) axis. A mechanism for acute down-regulation of the HPT axis is described, involving phosphorylation of a thyroid hormone receptor (THR) B2 isoform at serine 101. S101 is a major THRB2 phosphorylation site not found on other THR isoforms but is present in other members of the nuclear hormone receptor superfamily. S101 is phosphorylated after T₃ binding and increases HPT axis sensitivity to T₃ inhibition. S101A is also phosphorylated by a fasting responsive pathway, yielding HPT axis suppression. A physiologically relevant THR phosphorylation function is described, permitting nutritional deprivation and TH to use a common mechanism for acute suppression of the HPT axis.

Abstract

Fasting induces profound changes in the hypothalamic–pituitary–thyroid (HPT) axis. After binding thyroid hormone (TH), the TH receptor beta 2 isoform (THRB2) represses Trh and Tsh subunit genes and is the principle negative regulator of the HPT axis. Using mass spectrometry, we identified a major phosphorylation site in the AF-1 domain of THRB2 (serine 101, S101), which is conserved among many members of the nuclear hormone receptor superfamily. More than 50% of THRB2 is phosphorylated at S101 in cultured thyrotrophs (TαT1.1) and in the mouse pituitary. All other THR isoforms lack this site and exhibit limited overall levels of phosphorylation. To determine the importance of THRB2 S101 phosphorylation, we used the TαT1.1 cell line and S101A mutant knock-in mice (Thrb2^S101A). We found that TH promoted S101 THRB2 phosphorylation and was essential for repression of the axis at physiologic TH concentrations. In mice, THRB2 phosphorylation was also increased by fasting and mimicked Trh and Tshb repression by TH. In vitro studies demonstrated that a master metabolic sensor, AMP-activated kinase (AMPK) induced phosphorylation at the same site and caused Tshb repression independent of TH. Furthermore, we identified cyclin-dependent kinase 2 (CDK2) as a direct kinase phosphorylating THRB2 S101 and propose that AMPK or TH increase S101 phosphorylation through the activity of CDK2. This study provides a physiologically relevant function for THR phosphorylation, which permits nutritional deprivation and TH to use a common mechanism for acute suppression of the HPT axis.

The thyroid hormones (THs) T₃ and T₄ influence vertebrate development and facilitate metabolic responses to environmental cues such as temperature change, stress, inflammation, and nutrient deficiency (1). TH production in the thyroid gland is regulated by the hypothalamic–pituitary–thyroid (HPT) axis through coordinated release of thyrotropin-releasing hormone (TRH) from the paraventricular nucleus (PVN) of the hypothalamus and thyroid-stimulating hormone (TSH) from anterior pituitary thyrotrophs.

Peripheral and central actions of THs are mediated by three related thyroid hormone receptors (THRA1, THRB1, and THRB2) encoded by two genes (Thrb and Thra, and for review, see refs. 1 and 2). All isoforms are capable of either activating or repressing target gene expression in a TH-dependent manner (3–5). THRB2 is unique, however, due to its tissue-limited yet robust expression in the anterior pituitary and TRH neurons, where it functions as a potent negative regulator of the Trh and Tsh subunit genes (5–7).

Similar to the effect of TH, a reduced nutritional status down-regulates the HPT axis both in rodents and in humans (8–12). These changes are presumed to be an adaptive mechanism to conserve energy during times of food shortage. Fasting alters the central HPT axis (reviewed in refs. 1 and 12–14), resulting in reduced levels of TH (especially T₃), which in part is due to a reduction in TRH and TSH levels (11, 15, 16). Prolonged fasting also affects peripheral TH metabolism and cooperatively reduces THs in the circulation (1, 16, 17). However, low serum THs during fasting do not elevate Trh and Tshb expression in hypothalamus and pituitary, which suggests that TH-negative feedback has been disrupted and the HPT axis set point altered.

Several mechanisms have been proposed to explain nutritional regulation of the HPT axis, and most are centered on leptin acting directly on the TRH neuron or indirectly through neuropeptide secretion from the arcuate nucleus (18–22). In addition, prolonged nutrient deprivation has a significant effect on gene expression in the pituitary and hypothalamus (23, 24). After prolonged fasting in rodents, for example, hypothalamic Dio2 is up-regulated (25–27), local T₃ production increases, and Trh repression is noted. In the anterior pituitary, in contrast, longer fasting reduces Dio2 and Thrb2 messenger RNA (mRNA) expression (28), and these changes tend to oppose a reduction in Tshb expression.

While multiple central and peripheral factors contribute to HPT repression during fasting, it remains unknown how a lower HPT set point is actually achieved and how TH-negative feedback is suppressed. Our group has previously demonstrated that THRB2 is the main mediator of TH-negative feedback in the hypothalamic TRH neuron and pituitary thyrotroph (5, 6, 29), which correlates the relatively high expression of THRB2 versus other THRs in the central HPT axis (6, 30, 31). In contrast, the role of THRB2 in the fasting response is less clear, although PVH Trh suppression in Thrb2^−/− mice is partially impaired during fasting (6).

THRB2 is also unique among the THR isoforms in displaying ligand-independent action on Trh and Tsh subunit genes (32, 33). THRB2’s ligand-independent property was mapped to amino acids 89 to 120 in the AF-1 domain. While exploring posttranslational modification differences among the THR isoforms using a proteomics approach, we found a major phosphorylation site on THRB2 located at serine 101 in mice (S102 in human, Fig. 1A). We determined that S101 in the AF-1 domain of THRB2 is highly phosphorylated in both a cultured pituitary thyrotroph (TαT1.1) cell line (5) and in the mouse pituitary and identified CDK2 as a key enzyme responsible for THRB2 phosphorylation in vitro and cultured cells.

Fig. 1.

Serine 101 is a major phosphorylation site in THRB2. (A) AF-1 domains in several nuclear hormone receptors carry a conserved site for proline-directed kinases (serine or threonine followed by proline, S/TP, red), N-terminal to the DNA-binding domain (green). The position of the site and respective kinases are shown on Right. (B) Phos-tag PAGE separation of HA-THRs expressed in TαT1.1 cells followed by Western blotting with an anti-HA antibody. THRB2 showed two major bands (phosphorylated to nonphosphorylated band intensity ratio > 0.9). The phosphorylated band disappeared upon protein phosphatase (PP) treatment (THRB2 + PP). Bands representing potential phosphorylation of THRB1 and THRA1 (high exposure used for detection, red arrows) showed about 0.1 and 0.02 intensity, respectively, compared to nonphosphorylated isoforms (black arrows). (C) MS/MS spectrum of the 74DSDLDMALSQSSQPAHLPEEKPFPQVQpSPPHSQK107 peptide with y (C-terminal) ion and b (N-terminal) ion peaks (phosphorylated m/z 3+). (D) Phos-tag PAGE separation of HA-THRB2 and HA-THRB2^S101A mutant proteins expressed in TαT1.1 cells, showing that THRB2 carries a major phosphorylation site at position 101. (E) T₃ treatment (1.0 and 10 nM) caused a significant increase of phosphorylated to nonphosphorylated protein intensity ratio (pTHRB2/nTHRB2) between 3 to 10 h after treatment. pTHRB2/nTHRB2 ratio from several experiments presented as a bar graph (n > 3). (F) Representative Western blot of HA-THRB2^WT protein expressed in TαT1.1 cells and separated in Phos-tag PAGE. Increasing T₃ concentrations (5-h treatment) caused increased THRB2 phosphorylation. Regular PAGE (4 to 20%) was used to for β-Actin Western blotting. pTHRB2/nTHRB2 ratio shown on the bottom of the blot. (G) Protein extracts from Thrb2^HA/HA mice pituitary (30) separated by Phos-tag PAGE showed two bands matching the migration position of pTHRB2 and nTHRB2 of HA-THRB2 expressed in TαT1.1 cells. The pTHRB2/nTHRB2 ratio in mouse pituitary was increased after T₃ injection (5 h). The pTHRB2/nTHRB2 ratio is shown on the bottom of the blot. (H) Data from several experiments, 5 and 24 h after treatment, summarized in the bar graph, numbers of animals shown on the bars. ANOVA used for statistical analysis, values are mean ± SEM, ****P < 0.0001, **P < 0.01, *P < 0.05.

While the S101 phosphorylation site of THRB2 is unique among THRs, several other nuclear receptors contain phosphorylation sites in the N-terminal AF-1 domain. For instance, similar sites (Fig. 1A) in retinoid acid receptor alpha and gamma (RARA and RARG), progesterone receptor beta, peroxisome proliferator-activated receptor gamma (PPARG), and estrogen receptor alpha (ESR1) have been shown to be phosphorylated by CDKs (34–39). Phosphorylation of these sites modulates functions of nuclear receptors in a ligand-dependent and ligand-independent manner, allowing an additional layer of regulation beyond ligand binding.

We found that phosphorylation of THRB S101 also integrated both TH and nonhormonal cues. Using Thrb2^S101A mutant mice and TαT1.1 cells, we established that THRB2 S101 phosphorylation was essential for acute suppression of the HPT axis during fasting. We demonstrated that both TH and a key cellular energy sensor, AMP-activated kinase (AMPK), phosphorylated THRB S101 and repressed Tshb expression in physiologically relevant TαT1.1 cells. We propose a convergence of nutritional and TH signaling pathways on the THRB2 AF-1 domain at this phosphorylation site, providing a mechanism to adjust the HPT axis set point under a variety of metabolic states.

Results

Identification of a Novel Phosphorylation Site in THRB2.

Although not extensively studied, THRs, like many other nuclear hormone receptors, are phosphoproteins (40–45), but phosphorylation is rarely detected on a Western blot, given that THR isoforms migrate on sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE) as single bands (5, 30). Therefore, we used Phos-tag acrylamide in SDS-PAGE (46) to separate phosphorylated and nonphosphorylated proteins. Phos-tag generally causes a unique band retardation for each phosphorylated site and/or combination of sites. HA-tagged THRs were expressed in TαT1.1 cells (5). After Phos-tag separation, we found that HA-THRB2 migrated as two major bands (Fig. 1B). The upper band (red arrow) was eliminated after the phosphatase treatment, indicating the upper band was pTHRB2 (Fig. 1D). Phos-tag separation of THRB1 and THRA1 also demonstrated retarded bands (red arrows, Fig. 1B), but <10% of THRB1 and <2% of THRA1 proteins were estimated to be phosphorylated.

To identify phosphorylation sites and other posttranslational modifications in THRB2 that might alter function, we expressed HA-tagged THRB2 in TαT1.1 cells both in the presence and in absence of T₃. Two combined samples of THRB2 protein were then purified using anti-HA magnetic beads and analyzed by in-gel tryptic digestion and nanoscale liquid chromatography coupled to tandem mass spectrometry (nano LC-MS/MS). The LC-MS/MS data yielded sequence coverage of ∼80% for THRB2 (SI Appendix, Fig. S1A). From both data sets, one monophosphorylated peptide was identified (Fig. 1C), and MS/MS spectra determined that serine 101 was phosphorylated (Fig. 1C). Quantitative analysis of phosphorylated and nonphosphorylated (D74-K107) peptides showed over 75% of THRB2 was phosphorylated under these conditions (SI Appendix, Fig. S1B).

To prove that the upper band represented S101 phosphorylation, we expressed a HA-THRB2 S101A (serine to alanine) mutant in TαT1.1 cells. As expected, the HA-THRB2 S101A mutant protein migrated as a single band on a Phos-tag gel and at the same position as the nonphosphorylated form of wild-type (WT) HA-THRB2 (Fig. 1D, nTHRB2). Thus, S101 phosphorylation of THRB2 is abundant in TαT1.1 cells and can be unambiguously detected using Phos-tag SDS-PAGE.

T₃-Dependent Regulation of THRB2 Phosphorylation.

To test if the level of THRB2 phosphorylation changes in presence of its ligand (T₃), we expressed HA-THRB2 in TαT1.1 cells and treated these cells with different concentrations of T₃ for time points from 1 to 24 h (Fig. 1 E and F). We observed an increase in phosphorylation as early as 1 h after treatment, which was significantly increased between 3 and 10 h of T₃ treatment. After 24 h, the effect of T₃ on THRB2 S101 phosphorylation was lost (Fig. 1E). The extent of phosphorylation was dependent on T₃ concentration in which 10 nM T₃ treatment resulted in the highest THRB2 phosphorylation (Fig. 1F).

Since T₃ is known to promote degradation of THRB2 (47), we also determined if a lack of phosphorylation at S101 affected THRB2 levels in TαT1.1 cells. While a reduction in THRB2 protein levels was observed after 24 h of T₃ treatment (SI Appendix, Fig. S1C), there was no difference between WT and S101A mutant protein levels, suggesting that THRB2 degradation and phosphorylation were mediated by different mechanisms. T₃ is also known to affect the binding of THRB2 to corepressors and coactivators (45, 48, 49). In coimmunoprecipitation (co-IP) experiments (SI Appendix, Fig. S1D), both WT and S101A mutant proteins demonstrated similar binding to SMRT (silencing mediator of retinoid and TH receptor) and SRC-1 (steroid receptor coactivator 1). Upon T₃ treatment, the binding of SMRT to both WT and mutant THRB2 was reduced, and interaction with SRC-1 was increased, suggesting that THRB2 S101 phosphorylation does not affect interaction with these major coregulators.

To determine the role of THRB2 phosphorylation in vivo, we used Thrb2^HA/HA mice that express endogenous HA-THRB2 (SI Appendix, Fig. S2 A and C (30). HA-THRB2 from pituitary extracts runs as a single band of ∼57 kDa on standard SDS-PAGE (30) but as two bands on Phos-tag SDS-PAGE, reproducing the same phosphorylation pattern as HA-THRB2^WT–expressed TαT1.1 cells (Fig. 1G). Average basal phosphorylation levels of THRB2 in the pituitary of euthyroid, free-fed mice are slightly higher than in TαT1.1 cells cultured without TH (pTHRB2/nTHRB2 ∼1.3 versus 1.0) and similar to phosphorylation levels of THRB2 after 0.1 nM T₃ treatment in TαT1.1 cells (Fig. 1 E–H). We compared HA-THRB2 proteins from the pituitary of T₃-treated (2 µg/100 g of body weight, intraperitoneally (i.p.) and saline-injected mice using Phos-tag SDS-PAGE and Western blotting with an anti-HA antibody. After 5 h of T₃ treatment, we detected a significant increase in phosphorylation (approximately threefold, Fig. 1 G and H), but after 24 h, the proportion of phosphorylated THRB2 returned to control levels (Fig. 1H). Therefore, in both cultured thyrotroph cells and mouse pituitaries, THRB2 demonstrated high monophosphorylation (∼50%), which undergoes a strong transient increase in response to TH treatment.

CDK2 Phosphorylation of THRB2 S101.

THRB2 S101 is located within the N-terminal AF-1 domain, in close proximity to the DNA-binding domain and is similar in location and amino acid consensus to previously identified serine/threonine Proline (S/TP) phosphorylation sites in other nuclear hormone receptors (Fig. 1A). The kinases reported to phosphorylate these sites include numbers of CDKs, MAPK, and JNK (34–39, 50–52). We used NetPhos3.1 (53) to identify predicted kinases for THRB2 S101 with scores above 0.5 (CDKs, GSK3, MAPK, and RSK) for further screening using kinase inhibitors. In addition, we tested IKK, AMPK, PI3K inhibitors (SI Appendix, Table S1). We found that the pan-CDK inhibitor flavopiridol caused a significant reduction in THRB2 phosphorylation both in the absence and presence of T₃ (SI Appendix, Fig. S3A and Table S1). We next focused on CDK inhibitors and found that NU6027, an inhibitor of CDK1, CDK2, ATR and DNA-PK, dramatically reduced THRB2 phosphorylation at 100 µM and significantly reduced phosphorylation at 10 µM. At 100 µM, NU6027 increases HA-THRB2 expression levels in TαT1.1 cells, likely by the mechanisms unrelated to THRB2 phosphorylation (SI Appendix, Fig. S3A).

To avoid nonspecific effects of the kinase inhibitor on THRB2 expression levels and to establish a cell-free system to perform a knock-down (KD) screen, we used in vitro–translated HA-THRB2 protein and protein extract from NIH 3T3 cells. In vitro–translated HA-THRB2 ran as a single nonphosphorylated band on Phos-tag SDS-PAGE (SI Appendix, Fig. S3 B and C). The treatment of the in vitro–translated protein with cell extracts from NIH 3T3 in presence of ATP resulted in 25 to 45% phosphorylation of THRB2 (pTHRB2/nTHRB2 ratio of 0.4 to 0.9, SI Appendix, Fig. S3 B and C). Phosphorylation was inhibited by increasing concentrations of flavopiridol and NU6027 (SI Appendix, Fig. S3B). Using small interfering RNAs (siRNAs), we performed CDK KDs in NIH 3T3 cells and used protein extracts from these cells to phosphorylate in vitro–translated HA-THRB2. CDK2 was identified as the most likely candidate for THRB2 phosphorylation (SI Appendix, Fig. S3C).

To demonstrate a direct action of CDK2 on THRB2, we performed an in vitro phosphorylation assay using a purified CDK2/CycA complex. When synthesized in a coupled in vitro transcription/translation system in rabbit reticulocyte extract, both WT and S101A proteins were nonphosphorylated (Fig. 2A). Nearly 100% of in vitro–translated HA-THRB2^WT was phosphorylated within 30 min after the addition of CDK2/CycA/ATP, while the HA-THRB2^S101A mutant showed no related phosphorylation shift on Phos-tag gel, confirming that CDK2 specifically and directly phosphorylated S101 (Fig. 2A).

Fig. 2.

CDK2/CycA-dependent phosphorylation of THRB2 S101 and its effect on DNA binding and transcription. (A) In vitro transcription/translation proteins from rabbit reticulocyte lysate extracts were treated with CDK2/CycA in the presence of ATP and analyzed by Phos-tag PAGE/Western blotting and an anti-HA antibody. Nearly 100% of WT protein showed a phosphorylation shift, while the migration position of THRB2^S101A protein was not altered. (B) Binding of WT and S101A mutant proteins to DR4 element was tested by EMSA. WT and mutant homodimers gradually dissociated with increasing T₃ concentrations. CDK2/CycA treatment significantly reduced homodimer formation for WT but not for mutant protein. (C) Bar graph summarizing data from three experiments. Band intensity was normalized to the level of untreated WT homodimer. (D) In presence of RXRA, both WT and S101A mutant proteins formed heterodimers on DR4 (heterodimer bands migrate higher and ∼10 to 15 times stronger in intensity than homodimer) both in absence and presence of T₃ (50 nM). The addition of CDK2/CycA significantly increased binding of WT protein but did not affect S101A heterodimer DNA binding. (E) Bar graph with data from three experiments, band intensity normalized to WT untreated homodimer in each experiment. WT protein binding intensity shown in blue and S101A mutant protein in red. (F) HA-THRB2^WT and HA-THRB2^S101A proteins expressed in Thrb2 KD TαT1.1 cells were tested by Western blotting (4 to 20% PAGE) with anti-HA and β-Actin antibodies. (G) After Thrb KD in TαT1.1 cells, re-expression of HA-Thrb2^WT (Thrb2^WT, blue) but not Thrb KD HA-Thrb2^S101A (Thrb2^S101A, red) significantly repressed Tshb at 0.2 nM of T₃. The numbers of replicates shown on the bar graph. (H) Gene repression (Tshb) and activation (Rab27b, Sema3c) were tested in TαT1.1 cells across a wide range of T₃ concentrations. Data collected from three experiments, normalized to expression levels (100%) at 0 nM T₃, and presented as % activation/repression, n > 3 in each experiment. Relative expression of T₃-regulated genes were measured by qRT-PCR and normalized to Actb and Rpl13a mRNAs. Values are mean ± SEM, ****P < 0.0001, ***P < 0.001, **P < 0.01, *P < 0.05, ns for not significant.

In human, the putative phosphorylation S/TP site (S102) is surrounded by slightly different amino acids (Fig. 1A). Similar to mouse proteins, we synthesized human WT and S102A proteins in a coupled in vitro transcription/translation system and treated it with CDK2/CycA/ATP (SI Appendix, Fig. S4A). WT protein showed about 60% phosphorylation, while the HA-THRB2^S102A mutant showed no related phosphorylation shift on Phos-tag gel, confirming that CDK2 was able to phosphorylate S102 in human THRB2. Furthermore, we expressed both proteins in HEK293 cells. WT protein showed ∼30% phosphorylation (pTHRB2/nTHRB2 ∼0.4), which was significantly reduced by CDK inhibitor flavopiridol (pTHRB2/nTHRB2 ∼0.2, SI Appendix, Fig. S4 B and C). The pTHRB2 band was not detected in the THRB2S102^A protein (SI Appendix, Fig. S4B), proving that human THRB2 is highly phosphorylated at the equivalent serine residue and targeted by CDKs.

CDK2/CycA-Mediated Phosphorylation of S101 Affects THRB2 DNA Binding.

THRs interact with DNA (DR4 elements) as both a homodimer or a heterodimer with Retinoid X Receptor (RXR) (54–56). Using an electromobility gel-shift assay (EMSA), we tested the binding characteristics of WT and S101A THRB2 proteins in the absence and in presence of CDK2/CycA-dependent phosphorylation. Nonphosphorylated WT and S101A THRB2 proteins were able to form homodimers on the DR4 element, which were dissociated in a T₃ concentration-dependent manner as previously reported (Fig. 2B) (5). The addition of CDK2/CycA induced phosphorylation of WT THRB2 (Fig. 2A) and significantly reduced homodimerization on the DR4 element (Fig. 2 B and C). In contrast, CDK2/CycA treatment did not affect THRB2^S101A homodimer binding to DR4.

An EMSA of the DR4 element in the presence of RXRA and THRB2 showed strong heterodimer binding (Fig. 2 D and E), which was further increased after CDK2/CycA/ATP treatment. No change in heterodimer binding was observed for the S101A mutant protein. Thus, S101 phosphorylation of THRB2 reduced homodimeric and enhanced heterodimeric binding with RXRA on a DR4 element, indicating that the S101 phosphorylation regulated THRB2 DNA binding by a T₃-independent mechanism.

Effect of S101 Phosphorylation on T₃ Regulation of the HPT Axis.

To test if the mutation of S101 affected TH-dependent gene expression in vitro, we knocked down endogenous Thrb in a thyrotroph cell line (TαT1.1) and re-expressed HA-THRB2^WT and HA-THRB2^S101A at equal levels using a previously established method (Fig. 2F) (5). T₃-dependent expression of target genes (Tshb, Rab27b, and Sema3c) were measured using qRT-PCR (Fig. 2 G and H).

HA-THRB2^WT gradually repressed Tshb expression at T₃ concentrations ranging from 0.1 to 10 nM and showed significant reduction of Tshb mRNA at 0.2 nM. HA-THRB2^S101A was less effective in reducing Tshb expression at low T₃ concentrations (Fig. 2 G and H) but equally effective at reducing Tshb expression at concentrations of 0.5 nM and higher (Fig. 2 G and H). In contrast, HA-THRB2^WT and HA-THRB2^S101A functioned similarly in T₃-dependent positive regulation on at least two genes (Fig. 2H). These results suggest that the S101A mutation selectively affected gene expression of Tshb at low to euthyroid levels of TH.

In order to assess the effect of THRB2 phosphorylation in control of the HPT axis in vivo, we generated Thrb2^S101A knock-in mice (SI Appendix, Fig. S2A, asterisk). These mice developed normally, and litter sizes and body weights of Thrb2^S101A/S101A mice on a regular chow diet were similar to that of WT. The mutation did not affect expression levels of Thrb2, Thrb1, or Thra1 in either the pituitary or hypothalamus (SI Appendix, Fig. S2B). Similar to data from TαT1.1 cells (SI Appendix, Fig. S1C), the S101A mutation did not affect protein level or stability of THRB2 in the mouse pituitary (SI Appendix, Fig. S2 C and D).

Serum free T₄ (fT₄), free T₃ (fT₃), and TSH levels in Thrb2^S101A/S101A mice were not significantly different from that in WT controls in free-fed (FF) conditions (Fig. 3 A–C). We next placed mice on a low iodine/propylthiouracil (LoI/PTU) diet to achieve hypothyroidism followed by a TSH suppression test using increasing concentrations of T_3. During the hypothyroid phase, the TSH response was similar between WT and Thrb2^S101A/S101A (SI Appendix, Fig. S2E). However, the Thrb2^S101A/S101A response during T₃ suppression appeared to be less robust, although lacking a statistically significant difference, at the 0.5 µg/100 g body weight (BW) dose, which is a dose that, in combination with a LoI/PTU diet, mimics the euthyroid condition in mice (SI Appendix, Fig. S2F).

Fig. 3.

Effect of THRB2 S101 phosphorylation on TH levels and HPT set point during fasting. (A–C) Serum hormones were measured in FF and 12-h and 24-h fasted state. (D) Tshb relative expression in mouse pituitary was measured by qRT-PCR and normalized to Rpl13a. In WT animals (blue), fasting caused a reduction of serum fT₄ (A), serum fT₃ (B), serum TSH (C), and pituitary Tshb mRNA levels (D). Thrb2^S101A/S101A animals (red) in contrast showed no significant down-regulation of serum fT₃, TSH, and Tshb after 12 h of fasting. After 24 h of fasting, serum fT₃ was unchanged and TSH and Tshb were suppressed but less than WT mice. (E) Bar graph summarizing measurements of integrated density of TRH in individual animals. In WT (blue), PVH TRH decreased after fasting, but no change was observed in mutant PVNs (red). (F) Representative anti-TRH staining (immunofluorescence) showing the response of TRH in the rostral PVN to 24 h fasting in WT and Thrb^S101A/S101A animals. (Scale bar, 200 μm.) (G) Western blot (Phos-tag PAGE, anti-HA antibody) showing an increase in HA-THRB2 phosphorylation (Thrb2^HA/HA mice, pituitary) after 5 h of fasting. Phosphorylated to nonphosphorylated protein intensity ratio (pTHRB2/nTHRB2) shown on the bottom of the blot. (H) Bar graph summarizing data from FF and 5-h and 24-h fasted animals. (I) Bar graph showing increase in THRB2 phosphorylation in response to AICAR injection (i.p., 400 mg/kg BW, 5 h). The numbers of mice are shown on the bar graphs, and values are mean ± SEM, ****P < 0.0001, ***P < 0.001, **P < 0.01, *P < 0.05, ns for not significant.

The S101A Mutation in Mice Thrb2 Reduced Responsiveness of HPT Axis to Fasting.

We next tested the response of the HPT axis to short- and long-term fasting. We compared serum fT₄, fT₃, and TSH levels (Fig. 3 A–C) in Thrb2^WT and Thrb2^S101A/S101A FF and fasted animals. In WT animals, TSH decreased significantly after 12 and 24 h of fasting (Fig. 3C), and the levels of fT₄ and fT₃ declined, respectively (Fig. 3 A and B). In contrast, Thrb2^S101A/S101A mice showed no change in fasting serum fT₃ levels and a blunted response in TSH levels, although fT₄ levels were reduced similar to that in WT (Fig. 3 A–C). While serum fT₃ levels appear to depend on the central HPT axis, changes in circulating fT₄ are consistent with a peripheral increase in DIO3 expression and activity (17, 57).

To determine if differences in serum TSH between WT and Thrb2^S101A/S101A mice resulted from differences in Tshb expression, we compared Tshb mRNA levels in pituitaries of FF and fasted mice. In WT animals, Tshb mRNA levels were significantly reduced after 12 h of fasting and were further decreased after 24 h to about 50% of FF levels. In Thrb2^S101A/S101A mice, no difference in Tshb expression was observed in 12-h fasted mice compared to FF mice, although 24-h fasted mice had a ∼30% reduction of Tshb mRNA (Fig. 3D).

Given the localization of THRB2 in both anterior pituitary thyrotrophs and the paraventricular nucleus of the hypothalamus (PVN), we next examined PVN TRH levels in WT and Thrb2^S101A/S101A FF and 24-h fasted mice. As previously reported, fasting reduced TRH mRNA levels in WT mice (6), and these data on TRH levels were consistent (Fig. 3 E and F). In contrast, TRH levels were unchanged in 24-h fasted Thrb2^S101A/S101A mice. In summary, the central HPT axis (TRH neuron and thyrotroph) showed resistance to the effects of short-term fasting in S101A mutant mice.

Fasting and AICAR Treatment Increase THRB2 Phosphorylation in the Mouse Pituitary.

THRB2 protein levels in PVN are insufficient for the Phos-tag assay, but we were able to observe a significant increase in pTHRB2/nTHRB2 in the pituitary (2.5-fold, Fig. 3 G and H) after 5 h of fasting. Interestingly, THRB2 phosphorylation returned to basal levels after 24 h of fasting (Fig. 3H), which was similar to the phosphorylation response after T₃ treatment (Fig. 1 G and H).

Fasting-induced suppression of the HPT axis is controlled by leptin (58) and potentially by other nutritional and hormonal signals, including ghrelin, insulin, GLP-1, glucagon, many of which function through AMPK activation (59, 60). Fasting also reduces cellular energy and increase the AMP-ADP/ATP ratio, leading to AMPK activation (61). Although the extracellular mediator of the acute fasting response of the HPT central axis is presently unknown, we tested if i.p. injection of the AMP analog, 5-aminoimidazole-4-carboxamide ribonucleotide (AICAR), could induce THRB2 phosphorylation. Interestingly, a single injection of AICAR caused a moderate but statistically significant increase in THRB2 phosphorylation in vivo (Fig. 3I).

Role of AMPK in THRB2 Phosphorylation and Control of the HPT Axis.

The fasting and activation of AMPK affect metabolism through action on the central axis (hypothalamus and anterior pituitary), thyroid, autonomic nervous system, and peripheral tissues. To test the direct and cell autonomous effect of reduced cellular energy and AMPK activation on THRB2 phosphorylation, we employed the TαT1.1 cell line since it is the only available in vitro physiologically relevant model of the HPT axis. The TαT1.1 cell model separates the effects of T₃ and AMPK activation on the thyrotroph from hypothalamic TRH stimulation and peripheral signals. We tested AMPK activators, AICAR and PF-06409577, on phosphorylation of HA-THRB2 in TαT1.1 cells. Both activators triggered a short-term increase in HA-THRB2 phosphorylation (∼1 h, SI Appendix, Fig. S5), but selective allosteric activator PF-06409577 caused the most stable AMPK activation and THRB2 phosphorylation and was used for further experiments (Fig. 4).

Fig. 4.

Effect of AMPK on THRB2 S101 phosphorylation and Tshb expression in TαT1.1 cells. (A) Schematic representation of the action of AMPK activator (PF-06409577) and inhibitor (Compound C). (B) Western blot of TαT1.1 cell extracts probed with anti-pAMPK and anti-AMPK antibodies, demonstrating activation of phosphorylation of AMPK (pAMPK) by PF-06409577 and inhibition by Compound C. Ponceau S staining and total AMPK served as loading controls. (C) Representative Phos-tag PAGE Western blot of HA-THRB2 protein expressed in TαT1.1 cells treated with PF-06409577 (2.5 µM), T₃ (1 nM), and Compound C (5 µM). Intensity ratio of phosphorylated to nonphosphorylated protein (pTHRB2/nTHRB2) shown on the bottom of the blot. (D) Bar graph represents the pTHRB2/nTHRB2 ratio from several experiments with total number of replicates on each bar. (E) Bar graph showing reduction in phosphorylation (pTHRB2/nTHRB2 ratio) upon treatment with CDK1/2 inhibitor NU6027 (10 µM). (F) Schematic representation of the action of a NU6027 on a pAMPK–CDK1/2–pTHRB2 pathway. The inhibition of CDK1/2 reduced both basal and AMPK-dependent THRB2 phosphorylation. (G) PF-064095077 (2 µM) caused a significant reduction of Tshb expression independent of T₃ in TαT1.1 cells (scr short hairpin RNA (shRNA) control, gray). (H) Thrb KD (green) abolished both T₃- and PF-064095077/AMPK-dependent Tshb repression. Tshb (G, H, K, and L) and Thrb (I) relative expression was measured by qRT-PCR and normalized to Rpl13a. (J) HA-THRB2^WT and HA-THRB2^S101A proteins were re-expressed in Thrb KD TαT1.1 cells and tested by Western blotting (4 to 20% PAGE) with anti-HA antibody and β-Actin for loading control. (K) Re-expression of WT THRB2 (Thrb2 KD HA-Thrb2^WT) TαT1.1 cells (blue) rescued both T₃-dependent and PF-064095077/AMPK-dependent Tshb repression. (L) THRB2^S101A (Thrb2 KD HA-Thrb2^S101A) restored T₃-dependent Tshb repression at 2 nM of T_3, but it did not rescue low-dose T₃ or AMPK-dependent Tshb regulation. Bar graph values are mean ± SEM, n = 4, ****P < 0.0001, ***P < 0.001, **P < 0.01, *P < 0.05, ns for not significant. ANOVA used for statistical analysis. (M) Model, illustrating effect of fasting on THRB2 S101 phosphorylation, and HPT axis (AMPK/CDK2 regulation inferred from biochemical and in vitro experiments).

To assure specificity of PF-06409577 for AMPK activation, we also used an AMPK competitive inhibitor, Dorsomorphin/Compound C (Fig. 4 A–D). Compound C suppressed both AMPK activation (Fig. 4B) and additional phosphorylation of HA-THRB2 mediated by PF06409577 and T₃ (Fig. 4 C and D). Because AMPK is not the kinase that directly phosphorylates THRB2, we tested if AMPK-mediated THRB2 phosphorylation would be blocked by a CDK inhibitor. Indeed, the CDK1/2 inhibitor NU6027 blocked PF-06409577/AMPK-induced phosphorylation (Fig. 4E).

Furthermore, we found that PF-06409577 treatment of TαT1.1 cells caused a reduction in Tshb expression (Fig. 4G). This effect was completely lost in Thrb KD cells (Fig. 4 H and I), confirming that Thrb was required for both T₃- and AMPK-dependent repression of Tshb. To test the role of THRB2 phosphorylation on Tshb repression, we re-expressed HA-THRB2^WT and HA-THRB2^S101A at equal levels in TαT1.1 cells (Fig. 4J). AMPK-dependent Tshb repression was restored by re-expression of HA-THRB2^WT but not by HA-THRB2^S101A (Fig. 4 K and L). Consistent with previous data (Fig. 2), HA-THRB2^S101A was unable to repress Tshb at 0.2 nM T₃ but exhibited normal repression at 2.0 nM (Fig. 4L). Unlike T₃ (SI Appendix, Fig. S1C), PF-06409577 treatment did not cause THRB2 degradation or any significant change in protein abundance (Fig. 4J). Taken together, our data demonstrate that S101 phosphorylation is increased by either AMPK or T₃ treatment. The increase of THRB2 phosphorylation resulted in repression of Tshb in either case, indicating a ligand-independent effect of THRB2 phosphorylation on gene expression.

Discussion

Effects of Fasting and Role of AMPK Activation on THRB2 Phosphorylation and HPT.

The HPT axis is exquisitely sensitive to TH-negative feedback, allowing serum TSH-dependent regulation of TH levels near a set point established for each individual (Fig. 4M) (1, 2). In particular, T₃ bound to THRB2 is known to suppress both Trh and Tsh subunit gene expression (5, 6, 31). Like TH, nutrient deficiency also down-regulates the HPT axis (8, 10–12); however, little is known about mechanisms establishing the HPT set point and whether the HPT negative feedback mechanisms of TH and fasting are related.

The relationship between nutrient and TH regulation of the HPT axis has been investigated by several laboratories. After the discovery of leptin, Ahima et al. showed that fasting-induced suppression of the HPT axis was mediated in part by a reduction in leptin secretion (58). Subsequently, other pathways were implicated in the fasting response of the HPT axis (reviewed in refs. 1 and 62). For example, leptin feedback was defined as both activating the PVN TRH neuron directly and indirectly via neuropeptides from the arcuate nucleus (16, 19).

In this study, we provide a plausible mechanism implicating THRB2 N-terminal phosphorylation in set-point regulation by both TH and nutritional status. Using Thrb2^S101A/S101A mice, we show that phosphorylation of THRB2 is required for robust down-regulation of TRH and TSH during fasting (Fig. 3). The HPT axis, however, is regulated at multiple levels (hypothalamus, anterior pituitary, and thyroid), which complicates the study of the HPT set point in vivo. We therefore used a permanent thyrotroph cell line (TαT1.1 cells) that expresses endogenous TSH and is regulated by T₃ (5). In cultured thyrotrophs and pituitary, we found little or no leptin effect on THRB2 phosphorylation and Tshb expression, likely because of low levels of LepRb expression in these cells. Instead, we discovered that a universal energy sensor, AMPK, activated THRB2 phosphorylation and mediated Tshb repression (Figs. 3 and 4). PF-06409577–mediated AMPK activation, for example, caused a reduction in Tshb mRNA levels in TαT1.1 cells independent of TH. This effect required THRB2 and an intact serine at position 101. The S101A mutation in THRB2 not only abolished suppression of Tshb mRNA levels by PF-06409577 but also blunted Tshb repression at low concentrations of T₃ (Fig. 2 G and L). These in vitro data are consistent with phenotypes of Thrb2^S101A/S101A mice (Fig. 3), indicating that THRB2 S101 phosphorylation during fasting functions to repress Trh and Tshb subunits and increases sensitivity of the HPT axis to low concentrations of THs (Fig. 4M).

AMPK integrates extracellular nutritional and hormonal signals mediated by leptin, ghrelin, insulin, GLP-1, glucagon, and TH signaling pathways (59, 60, 63) but also functions as a direct intracellular sensor of energy charge (reviewed in refs. 61 and 64). The relationship between AMPK and THs has been studied in the brain, where T₃ administration appears to decrease the activity of hypothalamic AMPK (63). A direct effect of AMPK on the HPT axis has not been well studied, although its role can be inferred from data showing that metformin (a known AMPK activator) decreases serum TSH and TH levels (65–67).

Role of CDK2-Mediated Phosphorylation in THRB2 DNA Binding and Transcriptional Control of Tshb.

THRs are known to be phosphorylated at several sites based on in vitro studies (40–45, 68). Unfortunately, none of these studies compared relative phosphorylation among the various THRs, provided in vivo evidence of a physiological effect of phosphorylation, or evaluated the THRB2 isoform. Indeed, when we compared HA-THRs expressed in TαT1.1 cells, THRA1 and THRB1 were weakly phosphorylated (<10% of total receptor protein is phosphorylated, Fig. 1B). In contrast, a single major phosphorylation site in THRB2 was found, which was 30 to 85% phosphorylated depending on conditions. THRB2 S101 (S102 in human) is located in the most divergent AF-1 domain of THRs and is a part of the S/TP motif, typically phosphorylated by CDKs and MAPK. This site is absent in THRB1 and THRA1/2 but found in several nuclear receptors (Fig. 1A), including S118 of ESR1 and S112 of PPARG (50, 69, 70).

Using a combination of kinase inhibitors and siRNA (SI Appendix), we identified CDK2 as the most potent kinase acting on S101. A CDK2/CycA purified complex was able to phosphorylate nearly 100% of in vitro–translated THRB2 protein, proving a direct action of this kinase on the S101 site (Fig. 2A). CDK inhibitors (flavopiridol/NU6027) significantly reduced but did not eliminate THRB2 S101 and S102 phosphorylation in TαT1.1 and in HE293 cells (Fig. 4E and SI Appendix, Figs. S2A and S4B), leaving open the possibility that other kinases may target the same site. Consistent with this finding, the S/TP motif is often targeted by multiple kinases. For example, PPARG S118 is phosphorylated by MAPK, CDK5, CDK7, and CDK9 (34, 35, 50, 52, 70) and ESR1 S112 by CDK7 and MAPK (69). However, in our studies, three MAPK inhibitors failed to reduce basal or T₃-mediated THRB2 phosphorylation (SI Appendix, Table S1).

Both AMPK and T₃ mediate S101 phosphorylation in THRB2, resulting in Tshb repression in thyrotrophs, and appear to have an additive effect at least in vitro (Fig. 4 D and G). We suggest that both AMPK activation and T₃ result in CDK2-dependent phosphorylation of S101 (Fig. 4 E, F, and M and SI Appendix, Fig. S3). Interestingly, T₃ has been previously reported to increase CDK2 levels (71). How AMPK activates CDK2 function remains unclear, but Short et al. suggested that AMPK signaling may result in cytoplasmic sequestration of p27, which would then elevate CDK2 nuclear activity (72). In EMSA, CDK2-induced THRB2 S101 phosphorylation correlated with increased THRB2/RXRA heterodimer and reduced THRB2 homodimer binding to a DR4 element (Fig. 2). Our laboratory has previously associated THRB2 homodimer formation in EMSA with TH-negative regulation (5, 73). It is conceivable that S101 phosphorylation selectively affects negative regulation of transcription by altering THRB2 DNA binding.

Phosphorylation of the AF-1 S/TP site in other nuclear receptors (Fig. 1A) also affects DNA binding, and two major mechanisms have been proposed: conformational change by serine-proline bond isomerization (74) and the recruitment of CDKs and other coregulators to DNA (75, 76). For example, S/TP sites in PR are suggested to recruit a CDK2/CycA complex to promoters, where it facilitates phosphorylation of coactivators such as SRC-1 (75). CDK2 recruitment to promoters may also induce phosphorylation of poly-(ADP)-ribose polymerase (PARP-1), cause local displacement of H1 linker histones, and induce chromatin remodeling (76). Phosphorylation of ESR1 S118 enhances DNA binding and increases transcriptional activity through recruitment of peptidyl prolyl cis/trans isomerase Pin1 (74). Finally, and most similar to THRB2, phosphorylation of RARA Ser-77 by CDK7/CycH (Fig. 1A) increases efficiency of RARA/RXR heterodimer binding in EMSA (39). Taken together, the S/TP motifs found in the AF-1 domain of THRB2, and several other nuclear receptors, may utilize common mechanisms to alter DNA binding and change transcription. Similar to other nuclear receptors, THRB2 AF-1 phosphorylation has a selective effect on target gene expression, repressing Tshb and possibly Trh but not altering expression of genes like Rab27b or Sema3c.

By promoting S101 THRB2 phosphorylation, fasting utilizes the same pathway as T₃ to suppress the HPT axis. This establishes a new lower set point in the HPT axis, which is resistant to the effects of lower TH levels. In humans, significant changes in HPT axis set point and circulating TH concentrations are also found in the nonthyroidal illness syndrome, which is due to a broad range of disorders such as trauma and surgery. Given that suppression of the HPT axis appears to happen sooner in rodents than humans, extrapolation to the human disease may be difficult. Nonetheless, human THRB2 carries the same site (S102, Fig.1A) targeted by CDKs in vitro (SI Appendix, Fig. S4); and importantly, AMPK activation by metformin reduces serum TSH and TH levels in men (62–64). Therefore, S101/102 phosphorylation may represent the convergence of nutritional and TH signaling pathways at THRB2 in both mice and humans and a pathway to regulate the HPT axis.

Materials and Methods

Functional assays in TαT1.1 cells, Thrb KD, and HA-THRB2 rescue experiments were performed as previously described (5). For phosphorylation analysis, proteins were separated using 9% SDS-PAGE containing 100 μM MnCl and 50 μM Phos-tag acrylamide (Phos-tag, AAL-107, Wako Chemicals). Full details for cell culture experiments, qRT-PCR, oligonucleotides, antibody information, in vitro assays, cell-free assays, and mass spectrometry are available in SI Appendix, Supplemental Materials and Methods.

All animal experiments were approved by the Institutional Animal Care and Use Committee of Rutgers, The State University of New Jersey. All mice had a C57BL/6J genetic background and were maintained on a standard diet (no. 5008, LabDiet) with water ad libitum in a temperature-controlled facility with a 12-h light/12-h dark schedule. Unless otherwise indicated, 3- to 3.5-mo-old adult males were used for serum and tissue collection. Thrb2^HA mice were previously characterized, and the difference between levels of THRB2 expression and TSH levels in male and female were noted (30). Therefore, only males were used in this study. Thrb2^S101A mutant mice were generated by Genome Editing Shared Resource, Rutgers University, using the CRISPR-Cas9 technique. The detailed technique, as well as hormone measurements, tissue collection, and immunofluorescence are available in SI Appendix, Supplemental Materials and Methods.

Data Availability

All study data are included in the article and/or SI Appendix.