Hypermetabolism in mice caused by the central action of an unliganded thyroid hormone receptor α1

To determine if the unliganded TRα1 affected metabolism, tissue analyses, gene expression profiling and measurements of oxygen consumption were done. Heterozygous male TRα1+m mice (4–7 months old; n=5–6 per group) had a 16% reduction in BW as compared to wild‐type (wt) littermate controls (Figure 1A). Analyses of tissues showed that mutant mice had a 51% reduction in their epididymal WAT (eWAT), a 26% reduction in liver and a 33% reduction in interscapular BAT (iBAT) weight. In contrast, soleus and extensor digitorum longus (EDL) weights were normal (Supplementary data 1), as were the adult skeletal dimensions of TRα1+m mice (Bassett et al, 2007). This shows that the reduced tissue weights are associated with a lean body mass and not dwarfism.

Histological analyses showed that adipocytes in eWAT and lipid vacuoles in iBAT were smaller in mutant as compared to wt mice (Figure 1B), a striking difference that was exacerbated by a 16‐h fast. The WAT capsule surrounding the iBAT showed an increased number of multi‐ocular adipocytes in mutant mice, whereas wt animals had large lipid vacuoles (Figure 1B). To determine if the adipose tissue phenotype was caused by aporeceptor activity of the mutant TRα1, we studied TRα1+m mice also lacking TRβ. Such TRα1+m β−− mice have a 10‐fold increase in serum thyroid hormone levels that activates the mutant TRα1 (Forrest et al, 1996; Tinnikov et al, 2002). Figure 1C shows that, as expected, eWAT and iBAT cell sizes were comparable to that in control animals (TRα1++ β+−) and that fat cell necrosis was present in both TRα1+m β−− mice and control animals. In addition, deletion of TRβ normalized BW. Histological liver analysis showed that mice carrying the TRα1 mutation had lower glycogen content in liver, and that this was normalized by the TRβ‐null allele (Figure 1D). The morphological changes in the mutant mice suggested an increased metabolism that could be resolved through reactivation of the mutant TRα1 by high levels of thyroid hormone.

To elucidate the cause of the tissue weight differences, we determined the metabolic rate: O₂ consumption was increased by approximately 20% over an 18‐h period in the mutant mice (Figure 1E). A control experiment showed that this difference was not due to increased overall locomotor activity (Supplementary data 2). The data thus indicate that the reduction in adipose tissue mass is caused by an increased metabolic rate.

Next, we tested if the BW of the mutant mice could be normalized by increasing their caloric intake through a high‐fat diet (HFD). A 12‐week treatment failed to normalize BW in the mutant mice, despite their increased caloric intake (Figure 2A–C, Supplementary data 3), indicating a resistance to diet‐induced obesity. This was further supported by morphological analysis of iBAT, which showed moderate activation in the TRα1+m mice, whereas wt mice showed typical signs of inactive tissue such as increased adipocyte size and scattered fat cell necrosis in eWAT.

However, activation of the mutant receptor by treating the adult mice with pharmacological doses of T3 in combination with an HFD caused a rapid increase in BW in TRα1+m mice and markedly reduced their resistance to diet‐induced obesity (Figure 2A). In an independent experiment, we measured free T3 levels in mice that received T3 via drinking water, which revealed no difference between the groups (wt 61.6±7.9; TRα1+m 68.2±3.3 pmol/l). The increase in BW upon T3 treatment was paralleled by a partial normalization of caloric intake (Figure 2C). Furthermore, T3 treatment in combination with control diet (CD) normalized white and brown adipocyte morphology in the mutant mice (Figure 2D). We thus conclude that the resistance to diet‐induced obesity and in part also the hyperphagia were caused by the unliganded, mutant TRα1.

To determine if carbohydrate level balancing also was affected in TRα1+m mice, we performed intraperitoneal glucose tolerance tests (ipGTTs). Figure 3A shows an accelerated glucose clearance, while the insulin response was unexpectedly unaltered. HFD treatment increased glucose concentrations in an ipGTT to the level of wt animals on a CD (Figure 3B). Higher substrate utilization was further supported by increased insulin‐mediated glucose uptake in isolated soleus and EDL muscles of TRα1+m mice (Figure 3C). Under fed conditions, glucose levels were comparable in wt and mutant mice, whereas insulin tended to be lower in mutants (Supplementary data 4). The increased insulin requirement that is associated with high thyroid hormone levels was reflected by a strong elevation in insulin levels in T3‐treated wt animals on an HFD, whereas the mutant mice were protected from this effect (Supplementary data 4). These data indicate that the TRα1+m mice have enhanced insulin sensitivity that improves glucose handling, which may be related to their reduced body fat.

Serum parameters were analyzed as a first step to identify the cause of the observed metabolic phenotype. The expected decrease in total T3 (TT3) and total T4 (TT4) in response to a 16‐h fast was present, although TT3 and TT4 levels upon fasting remained somewhat higher in mutant mice (Figure 4A). Increased lipid utilization in mutant mice was reflected by lower serum FFAs, triglycerides and cholesterol, independent of diet and feeding status (Figure 4B). These differences were effectively normalized by T3 treatment (Figure 4B). β‐Hydroxybutyrate levels were normal in TRα1+m mice, but remained lower upon fasting, a difference that was not ameliorated by T3 treatment (Figure 4B). We conclude that these results reflect an increased energy demand in TRα1+m mice.

To identify which tissues contributed to the observed changes in serum parameters, gene expression profiling was performed on eWAT, iBAT, liver and soleus muscle tissues. The results revealed strong induction of genes involved in lipolysis, lipogenesis and glucose handling in eWAT (Figure 5A; Supplementary data 5A and B). The effects were less pronounced in iBAT and liver, and absent in soleus muscle (Figure 5B and C; Supplementary data 5C–H). In eWAT, target gene expression reflected an increase in both lipogenesis and β‐oxidation: ACC1 and fatty acid synthase (FAS) were induced six‐fold and were paralleled by a four‐fold induction of ACC2 (Figure 5A). In concordance with this, PGC1α was three‐fold and PPARα four‐fold increased in TRα1+m mice (Figure 5A), whereas no changes were observed in PPARγ expression (Supplementary data 5A–H). Similar but less dramatic changes were seen in iBAT and liver (Figure 5B and C). In the liver, we found evidence for increased gluconeogenesis; PEPCK showed a three‐fold increase (Figure 5D). GLUT4 was doubled in eWAT, but remained unaffected in iBAT (Figure 5D). A full overview of all genes analyzed and subsequent statistical analyses are available in Supplementary data 5A–H. The differences between adipose tissues and liver may be partially related to local differences in T3 concentrations, since these tissues rely on different T3 sources. eWAT is dependent on circulating T3, whereas in liver a four‐fold induction of type I deiodinase (Dio1) mRNA indicated an increase in local T4 to T3 conversion, suggesting higher local T3 concentrations that would allow partial reactivation of the mutant TRα1 (Figure 5C). Type II deiodinase (Dio2) mRNA levels in the fed status in iBAT were similar in wt and mutant mice, whereas in the fasted status Dio2 mRNA increased approximately 12‐fold in the mutants as compared to four‐fold in the wt mice (Figure 5B). Our data indicate that the TRα1 aporeceptor affects glucose and lipid handling via distinct mechanisms and that local differences in deiodination may contribute to the differences between tissues.

In view of the substantial differences in tissue weights and morphology, we performed ELISA on serum leptin levels to confirm that altered tissue mRNA levels were reflected in relevant serum changes. We confirmed that TRα1+m mice have lower leptin levels, which are consistent with their increased food intake (Figure 5E). Importantly, lipid handling measured in eWAT and iBAT by enzyme activity assays revealed increased ACC activity in both eWAT and iBAT and malonyl CoA decarboxylase (MCD) activity in iBAT (Figure 5F). The increased lipid mobilization was substantiated by the increased β‐oxidation seen specifically in iBAT of the mutant mice, which indicated that this tissue is responsible for the increased energy demand in TRα1+m mice and therefore for their lean phenotype.

Intriguingly, the observed lean phenotype resembles a state of hyper‐ rather than hypothyroidism. We therefore tested the possibility that increased sympathetic signaling would be overriding the effects of the TRα1 aporeceptor at the tissue level. Since iBAT showed the most hallmarks of hypermetabolism, 8‐week‐old mice were acclimated to 30°C to inhibit sympathetic stimulation of facultative thermogenesis. Figure 6A shows that this resulted in normalized eWAT and iBAT weights in mutant mice, although body and liver weights remained lower. Food intake was decreased below wt levels and O₂ consumption was normalized (Figure 6B). The mutant mice gained weight rapidly (Figure 6C), which was associated with normalization of gene expression levels of ACC, CPT1β, PGC1α, PPARα and GLUT4 in eWAT (Figure 6D). The adipose tissue morphology of the mutant mice normalized during the acclimation to 30°C (Figure 6E).

To study if the facultative thermogenesis in BAT was responsive to a norepinephrine (NE) challenge, the ability of TRα1+m mice to defend their body temperature was tested. Both in a short‐term (1.5 h) and long‐term (6 h) experiment, TRα1+m mice successfully defended their body temperature (Figure 7A and B). However, the mutant mice had a significant increase in their lower critical temperature (LCT) (which is the ambient temperature below which basal metabolic rate becomes insufficient to balance heat loss), and the defended body temperature was higher (37°C versus 36°C in wt), as calculated by extrapolation of the temperature defense curve (Figure 7A). Furthermore, their body temperature, when kept at 21°C, was approximately 1°C lower than normal (37.3°C versus 38.2°C in wt mice; Figure 7B). This contrasts the subsequent iBAT gene expression analysis. The mutant mice that were exposed to cold for 6 h showed reduced UCP1 and PGC1α mRNA levels, whereas Dio2 and mitochondrial transcription factor 1 (TFAM1) were normal (Figure 7C). This demonstrates a minor but significant impairment in iBAT function of TRα1+m mice, which is in agreement with their lower body temperature despite the increased calculated theoretical defended body temperature.

BAT sensitivity to β‐adrenergic receptor‐induced thermogenesis was studied by measuring O₂ consumption after injection of NE. We found that the increase in O₂ consumption caused by NE was delayed in TRα1+m mice, even though the O₂ consumption during basal thermogenesis and the maximum response to NE were unaltered (Figure 7D, left panel). The NE challenge was subsequently performed in mice acclimated to 30°C to exclude that increased basal sympathetic tone in mutant mice affected the response to acute NE stimulation. Surprisingly, the BAT response to NE was even further impaired in these mutant mice (Figure 7D, right panel). In addition, TRα1+m mice acclimated to 30°C showed lower O₂ consumption during basal thermogenesis, which is in agreement with the lower O₂ consumption in TRα1+m mice above the LCT. Our data thus indicate that impaired sympathetic signaling in the BAT due to the mutant TRα1 is partially compensated through increased basal sympathetic tone.

Obligatory thermogenesis is sufficient to sustain core temperature over a limited range of environmental temperatures, below which facultative thermogenesis is required to stay warm. Hence, cold exposure elicits a complex response marked by increases in thermogenic capacity of BAT, oxygen consumption and food intake. The thermogenic response in BAT is induced by sympathetic stimulation resulting in NE release, which acts synergistically with T3 (Silva, 2006). Increased sympathetic activity causes increased local T3 concentrations via induction of Dio2, brown adipocyte proliferation, increased UCP1 expression and activity and mitochondrial biogenesis, a process known as recruitment (Cannon and Nedergaard, 2004). Morphological analyses of adipose tissues of TRα1+m mice indicated interconversion of WAT into BAT, which was in accordance with BAT activation reflected by increased PGC1α, LPL and Dio2 mRNA expression upon fasting. However, there were no differences in UCP1 expression, or in TFAM or NRF1, which are involved in mitochondrial biogenesis. These data are in line with observations in mice lacking all thyroid hormone receptors that are able to recruit BAT and have normal UCP1 expression (Golozoubova et al, 2004).

Fatty acid synthesis and oxidation occur in separate cell compartments to prevent immediate oxidation of newly synthesized fatty acids. ACC1 converts acetyl CoA into a malonyl CoA pool, which serves as a substrate for fatty acid elongation by FAS, whereas the malonyl CoA produced by ACC2 serves as a potent inhibitor of CPT1 and thus of β‐oxidation (Abu‐Elheiga et al, 2001). Upregulation of both fatty acid synthesis and β‐oxidation is stimulated by sympathetic signaling and occurs during cold exposure, when fatty acid synthesis secures fat stores to provide fuel for heating the body during continued periods of increased β‐oxidation. The underlying mechanism may be explained by distinct intracellular malonyl CoA pools and selective increased malonyl CoA turnover caused by increased MCD activity, resulting in increased CPT1 activity (Yu et al, 2002). The increased levels of ACC1, ACC2 and FAS mRNA and increased MCD activity and β‐oxidation in TRα1+m mice are in accordance with such a state of increased sympathetic signaling.

At thermoneutrality, when obligatory thermogenesis is able to meet the body requirements to maintain body temperature, TRα1+m mice have a lower metabolic rate (Figure 7), which is congruent with a receptor‐mediated hypothyroidism. However, at lower ambient temperatures, TRα1+m mice had an increased metabolic rate, which necessitated a study of their thermoregulation. Previously, TR isoform‐specific actions were identified in facultative thermogenesis: transcriptional induction of UCP1 requires TRβ, whereas TRα is essential for the synergistic effect between sympathetic signaling and thyroid hormone action (Ribeiro et al, 2001). Upon a 6‐h cold exposure, the TRα1+m mice failed to increase UCP1 and PGC1α expression levels to the same extent as wt mice. This may be due to interference by the mutant TRα1 with TRβ function in BAT. The ability of TRα1+m mice to successfully defend their body temperature, despite impaired UCP1 stimulation, indicates that the TRα1+m mice rely in part on different mechanisms to maintain body temperature, possibly shivering thermogenesis.

The defective sympathetic stimulation of facultative thermogenesis observed in hypothyroid mice (Ribeiro et al, 2001) was also present in TRα1+m mice, as was reflected by their delayed increase in O₂ consumption in response to an NE challenge. The O₂ response was even further impaired in mice acclimated to 30°C, suggesting that increased sympathetic outflow at 21°C is part of a compensatory mechanism to keep body temperature sufficiently high. In addition, the lower body temperature observed at 21°C was in contrast with the higher theoretical defended body temperature. This discrepancy suggests that facultative thermogenesis is unable to produce enough heat through uncoupling to reach the defended body temperature. This is likely caused by a BAT‐specific effect of the mutant TRα1, causing increased hypothalamic outflow that would stimulate thermogenesis through alternative mechanisms and causing energy expenditure through a futile cycle.

More than 300 patient families with resistance to thyroid hormone syndrome, RTH, have been found. The patients have an inherited mutation in the TRβ gene (Weiss and Refetoff, 1996, 2000). However, no patients harboring an equivalent mutation in the TRα gene have been described. This may be related to the absence of obvious aberrancies in thyroid hormone levels, unlike those found in the RTH patients.

Based on the results presented here, a hypothesis regarding the metabolic characteristics of patients with a TRα1 mutation indicates hypermetabolism. However, the phenotypes of two other mouse strains with mutant TRα1 genes differ from what is described here: dwarfism accompanied by reduction in WAT by the TRα1PV allele (Liu et al, 2003; Ying et al, 2007), or obesity in combination with impaired catecholamine‐stimulated lipolysis by the TRα1P398H mutant (Liu et al, 2003; Ying et al, 2007). In cultured WAT cells, the TRα1PV mutant was described to repress the ability of PPARγ to activate its target genes, and the expression of ACC, FAS and PPARγ was found to be reduced in the mutant animals, in contrast to what was seen with the TRα1R384C allele in our study. The TRα1P398H mutation strongly inhibits liver PPARα expression and reduces expression of genes involved in fatty acid oxidation by interfering with PPARα action (Liu et al, 2007), whereas TRα1R384C mutation led to a strong induction of PPARα gene expression, no change in PPARγ levels and overexpression of genes involved in lipid handling. Taken together, this indicates that the position of a mutation may determine how the mutant receptor interacts with different hormonal response elements, other nuclear receptors and/or their coregulators.

Nevertheless, our results as well as those by Liu et al (2007) support the concept that the TRα1R384C receptor is akin to a wt aporeceptor: it confers on the CPT1α and ACO response elements the same moderate PPARα‐interfering properties as wt TRα1 (Liu et al, 2007), and its activation by high levels of T3 leads to normalization of the hypermetabolism as well as most other phenotypic aberrancies (Tinnikov et al, 2002; Venero et al, 2005; Bassett et al, 2007). The normalization of metabolism by acclimation to 30°C furthermore argues against mechanisms involving, for example, constitutive binding by TRα1R384C to the coactivator PCG1α or the corepressor RIP140, events that could lead to a phenotype very similar to the one we have described.

The dramatic improvement of the metabolic phenotype by functional denervation of sympathetic signaling to the BAT points to the importance of the hypothalamus in regulating metabolic rate and the ability of central signaling to override peripheral effects. Intriguingly, increased sympathetic outflow resembles a state of hyperthyroidism, rather than of hypothyroidism. Discrepancies between peripheral and central thyroid hormone signaling have also been reported during critical illness in humans and fasting in rodents (Lechan and Fekete, 2004, 2006; Fliers et al, 2006). During severe illness, thyroid hormone levels decrease without giving rise to high levels of TSH or TRH, and the effects are mediated by the hypothalamus (Fliers et al, 1997; Wiersinga, 2000). During fasting and in experimental models of critical illness, a relative state of hyperthyroidism is present in the hypothalamus as a result of increased local deiodination (Diano et al, 1998; Boelen et al, 2004; Fekete et al, 2004; Coppola et al, 2005). It is feasible that the mutant TRα1 is involved in regulating hypothalamic T3 levels under these conditions and thereby affects sympathetic outflow to BAT and thus could contribute to the hypermetabolic phenotype. The possibility that altered hypothalamic signaling also can contribute to metabolic wasting, such as that observed in cancer cachexia, merits further study.

The mouse strain carrying the dominant‐negative R384C mutation in TRα1 and the combination with a TRβ‐null allele have been described previously (Tinnikov et al, 2002).

The TRα1R384C mice used in the experiments here had been backcrossed to C57BL/6NCrl for 3–4 or 8–10 generations. Experiments done in both cohorts produced similar results. Littermate male mutant and wt mice aged 4–7 months were kept at 21°C on a 12 h light/12 h dark cycle. For thermoneutrality studies, mice were transferred to 30°C at the age of 2 months and kept at this temperature for at least 4 weeks. Control and HFD mice were obtained from Research Diets (New Brunswick, NJ) (D12450B: 3.85 kcal/g, 10% kcal fat; D12451: 4.73 kcal/g, 45% kcal fat). Animal care procedures were in accordance with the guidelines set by the European Community Council Directives (86/609/EEC). Required animal permissions were obtained from the local ethical committees.

For determination of serum parameters, fed or overnight fasted (16 h) animals were killed by decapitation (after 11 weeks on diet, when applicable), after which trunk blood was collected and tissues dissected for further analyses. Serum was obtained after centrifugation of blood samples and stored at −80°C until assayed for serum parameters.

Animals were placed on control or HFD, and BW and food intake were determined weekly. Animals received T3 for 12 days via drinking water (0.01% albumin, T3 concentration 0.5 μg/ml). TT3 and TT4 were measured by radioimmuno assay (TKT31 and TKT41; Diagnostic Products Corporation, Los Angeles, CA). FFA, triglycerides, cholesterol, β‐hydroxybutyrate and insulin were assayed according to the manufacturer's instructions (FFA: Wako Chemicals GmbH, Neuss, Germany; triglycerides: Sigma Diagnostics Inc., St Louis, MO; cholesterol and β‐hydroxybutyrate: STANBIO Laboratory, Boerne, TX; insulin: Mercodia AB, Uppsala, Sweden).

O₂ consumption was measured using the Oxymax System (Columbus Instruments, Columbus, OH) or the Somedic Inca System (Somedic Sales AB, Hörby, Sweden). To determine the thermoregulatory metabolism in the mutant TRα1 mice, O₂ consumption was studied as a function of ambient temperature. Measurements were performed at temperatures ranging between 7 and 34°C during a 1.5‐h period. Metabolic rates at the different temperatures were defined as the lowest, stable metabolic rate observed for at least 4 min. A minimum interval of 3 days was present between measurements at different temperatures. For the NE challenge studies, mice were anesthetized with sodium pentobarbital (70 mg/kg BW) and injected with (NE 1 mg/kg BW). Defense of body temperature over a 6‐h period was tested by measurements using a rectal probe with a 1‐h interval (n=6 mice per group).

2‐Deoxyglucose uptake was measured on isolated EDL (fast‐twitch glycolytic) and soleus (slow‐twitch oxidative) muscles as described elsewhere (Shashkin et al, 1995). For ipGTT, glucose was injected (2 g/kg BW: 20% solution) after a 16‐h fast. For determination of glucose levels, blood samples were obtained from the tail at 0, 15, 30, 60 and 120 min after injection and analyzed using an Accu‐Check Sensor glucose meter (Roche Diagnostics). For insulin levels, blood samples were taken at 0, 30, 60 and 120 min after glucose injection.

Tissues were fixed in 10% formalin rinsed in PBS, dehydrated through increasing concentrations of ethanol, cleared and embedded in paraffin, sectioned and stained with hematoxylin and eosin (H&E). To demonstrate intracellular lipids, formalin‐fixed tissue was cryoprotected with sucrose, and frozen sections were stained with Oil Red O.

RNA was isolated from snap‐frozen tissues using an RNeasy mini kit or RNeasy lipid kit (Qiagen, Sweden) according to the manufacturer's instructions. cDNA was obtained after reverse transcription and used for real‐time PCR using the ABI 7300 system and the ABI Prism 7000 (Applied Biosystems, Sweden). Quantification was performed using a standard curve and HPRT was used as a reference gene. Primer sequences are listed in Supplementary data 6.

Fat tissue (eWAT and iBAT) was incubated in low‐glucose DMEM (Gibco, Sweden) containing 2% (w/v) fatty acid‐free BSA, 0.30 mM l‐carnitine and ³H‐palmitic acid (3 μCi per well). Excess palmitic acid was removed by trichloroacetic acid precipitation. After extraction with chloroform/methanol (2:1), ³H₂O production was determined (Wang et al, 2003). ACC activity was determined by an NADH‐coupled assay and normalized to protein content (Wagner et al, 1998). Background phosphatase activity was determined in samples without acetyl CoA and used for correction. MCD activity was determined using a carnitine acetyltransferase‐linked assay (Antinozzi et al, 1998).

Prism 4 for Macintosh and In Stat 3 for Macintosh software was used for statistical analysis. Data were analyzed using the Student's t‐test or two‐way ANOVA followed by a Bonferroni or Tukey test to compare between groups. Differences were considered significant if P<0.05. All data are represented as their mean value±s.e.m.

Supplementary data are available at The EMBO Journal Online (http://www.embojournal.org).