Thyroid Hormone, Thyromimetics, and Metabolic Efficiency

Abstract

Thyroid hormone (TH) has long been recognized as a major modulator of metabolic efficiency, energy expenditure, and thermogenesis. TH effects in regulating metabolic efficiency are transduced by controlling the coupling of mitochondrial oxidative phosphorylation and the cycling of extramitochondrial substrate/futile cycles. However, despite our present understanding of the genomic and nongenomic modes of action of TH, its control of mitochondrial coupling still remains elusive. This review summarizes historical and up-to-date findings concerned with TH regulation of metabolic energetics, while integrating its genomic and mitochondrial activities. It underscores the role played by TH-induced gating of the mitochondrial permeability transition pore (PTP) in controlling metabolic efficiency. PTP gating may offer a unified target for some TH pleiotropic activities and may serve as a novel target for synthetic functional thyromimetics designed to modulate metabolic efficiency. PTP gating by long-chain fatty acid analogs may serve as a model for such strategy.

• I.

  Introduction

  • A.

    Metabolic efficiency

  • B.

    Mitochondrial energetics

  • C.

    Futile cycling

• II.

  TH and Metabolic Efficiency

  • A.

    Genomic/nongenomic actions of TH

  • B.

    TH-induced mitochondrial uncoupling

  • C.

    Gating of mitochondrial uncoupling proteins (UCPs) by TH

  • D.

    Mitochondrial permeability transition pore

  • E.

    Mitochondrial PTP gating by TH

  • F.

    Activities of TH transduced by PTP gating

  • G.

    TH-induced futile cycles

  • H.

    TH-induced mitochondrial biogenesis

• III.

  Thyromimetics and Metabolic Efficiency

  • A.

    Thyromimetic agents

  • B.

    Mitochondrial uncoupling by free fatty acids

  • C.

    Mitochondrial uncoupling by long-chain fatty acid analogs

• IV.

  Conclusion

I. Introduction

One of the most pronounced effects of thyroid hormone (TH; T₃, T₄) is modulation of energy expenditure/calorigenesis, metabolic efficiency, and heat production/thermogenesis. Thus, hypothyroidism results in an increase in metabolic efficiency, resulting in decreased energy expenditure and heat production, whereas hyperthyroidism results in a decrease in metabolic efficiency, accompanied by thermogenesis and weight loss despite increased energy intake. The role of TH in modulating metabolic efficiency has been realized for over a century, but its cellular mode of action remained elusive. This review summarizes historical and up-to-date findings concerned with TH regulation of metabolic energetics. It underscores the role played by TH-induced gating of the mitochondrial permeability transition pore (PTP) in controlling metabolic efficiency. PTP gating may offer a unified target for some TH pleiotropic activities and may serve as a novel target for thyromimetics designed to modulate metabolic efficiency.

A. Metabolic efficiency

The chemical bond energy of nutrients is transformed into work performance by means of essentially two consecutive metabolic processes. The first one consists of mitochondrial substrate oxidation, whereby carbohydrate, lipid, and protein nutrients are oxidized to yield ATP that serves as biological work currency. The second one consists of ATP utilization, mostly extramitochondrial, whereby the chemical bond energy of ATP is transformed into task performance, like basal housekeeping, growth, movement, etc. In line with the second law of thermodynamics, both processes involve heat production. Metabolic efficiency refers to the proportion of work (ΔG) vs heat (ΔS) derived from a given change in enthalpy (ΔH). In the mitochondrial context, enthalpy reflects the chemical bond energy of respective nutrients, whereas ΔG reflects net ATP production. In the extramitochondrial context, enthalpy is represented by ATP availability, whereas ΔG represents actual task performance. Low metabolic efficiency in the mitochondrial context implies an increase in heat production at the expense of ATP production per given nutrient enthalpy, whereas low extramitochondrial metabolic efficiency implies increased heat production at the expense of task performance per given ATP enthalpy. The efficiency in carrying out specific processes by work input is determined by their inherent resistance and reversibility. Processes of high resistance imply inherent reversibility, making it possible to extract maximal work performance while minimizing heat production. In contrast, nonreversible low-resistant processes are exothermic and of low efficiency. Overall, decrease in metabolic efficiency, due to mitochondrial or extramitochondrial processes, implies a requirement for an increase in nutrient enthalpy—namely, an increase in substrate utilization/calorigenesis for carrying out a given task. TH controls metabolic efficiency of all work tasks that comprise total body energy expenditure (eg, basal metabolism, thermic effect of food, thermic effect of exercise).

Metabolic efficiency and heat production are classically discussed within two distinct categories—namely, obligatory or adaptive/facultative thermogenesis (1, 2). Obligatory thermogenesis indicates heat production required for maintaining “normal” body temperature under conditions of environmental thermoneutrality, being dependent on the specific concerned species (24°C for humans). In contrast, adaptive thermogenesis refers to heat production required for maintaining normal body temperature against environmental temperatures below thermoneutrality. Adaptive thermogenesis may also refer to “diet-induced thermogenesis,” namely, the heat produced in maintaining body weight “set point” under conditions of excess energy intake. Despite the apparent similar characteristics of cold- and diet-induced thermogenesis, these two adaptive modes may be transduced by different transduction pathways (3). However, TH control of metabolic efficiency entails obligatory thermogenesis as well as the various adaptive thermogenic modes.

B. Mitochondrial energetics (Figure 1)

Figure 1

Modes of mitochondrial uncoupling. Oxygen reduction to H₂O, carried out by passing NADH electrons down through the respiratory chain, results in proton pumping from the mitochondrial matrix to the cytoplasmic side of the inner mitochondrial membrane, generating a PMF that consists of a membrane potential (ΔΨ) and a pH gradient (ΔpH), positive outside. ATP synthesis from its Pi and ADP precursors (state 3 respiration) is driven by driving protons in the direction of the proton gradient back into the mitochondrial matrix through the F0F1-ATP synthase. When mitochondrial PMF is dissipated by the ATP synthase only, the coupling between substrate oxidation and ATP synthesis is maximized. However, if protons leak back across the IMM down the proton gradient through alternative proton conductance channels (eg, UCP1, PTP), coupling efficiency is decreased, resulting in heat production at the expense of ATP synthesis.

Mitochondrial oxidative phosphorylation serves as the main biological machine for converting the chemical bond energy of variable nutrients into ATP. Thus, oxidizable substrates, such as glucose, fatty acids, or amino acids, are first oxidized to yield reduced NADH and FADH, followed by reducing oxygen to water at the expense of oxidizing the respective nucleotides. Oxygen reduction by reduced nucleotides is carried out by passing the electrons down through consecutive intermediates within the electron transport chain of the inner mitochondrial membrane (IMM). In the course of passing down electrons through the chain, protons are pumped out across the IMM by three proton pump complexes, generating across the IMM a proton motive force (PMF) that consists of a membrane potential and a pH gradient, positive outside. ATP synthesis from its Pi and ADP precursors (state 3 respiration) is driven by driving protons in the direction of the proton gradient back into the mitochondrial matrix through the F₀F₁ATP synthase. When mitochondrial PMF is only dissipated by the ATP synthase, the coupling between substrate oxidation and ATP synthesis is maximized. However, if protons leak back across the IMM down the proton gradient through alternative proton conductance pathways, coupling efficiency is decreased, resulting in heat production at the expense of ATP synthesis. Metabolic efficiency in synthesizing ATP at the expense of oxygen reduction is maintained by the reversibility of the electron transport chain and of the ATP synthase. Indeed, ATP hydrolysis by the ATP synthase may generate PMF across the IMM, followed by driving electrons up the chain by reverse electron transport at the expense of dissipating mitochondrial PMF, resulting in reducing NADH at the expense of hydrolyzing ATP.

Coupling electron transport to ATP synthesis is due to the IMM being impermeable to protons (4). The limited permeability of the IMM to protons is best exemplified in mitochondria respiring oxidizable substrates in the absence of ADP (state 4 respiration). Because the proton gradient resists proton pumping across the IMM in proportion to the PMF, state 4 respiration is restrained, and substrate oxidation rate would be minimized or blocked under conditions of absolute impermeability of the IMM to protons. However, some oxygen consumption still prevails under conditions of state 4 respiration, implying an inherent basal leak of the IMM to protons (4). Basal leak to protons may similarly account for the limited substrate oxidation in the presence of added ADP, under conditions of inhibiting the ATP synthase by oligomycin (5). About 95% of mitochondrial proton leak in intact mitochondria is mediated by IMM proteins, including the adenine nucleotide translocase (ANT), the uncoupling proteins (UCPs), and the PTP. ANT may account for 50–70% of basal mitochondrial proton leak, independently of its activity in catalyzing ADP/ATP exchange across the IMM (6), being complemented by the glutamate/aspartate carrier (7). However, the mode of ANT involvement in enabling proton leakage through the IMM still remains elusive (8). Proton leak may further be affected by the UCP family of uncoupling proteins (9) and the mitochondrial PTP (10–15), to be discussed in Section II.D. Applying an inherent basal proton leak to state 3 respiration may account for an estimated 15–20% decrease in metabolic efficiency in synthesizing ATP in rats (16). Hence, an estimated 15–20% of total body energy expenditure in rats may drive heat production by operating a futile cycle of proton pumping and proton leak across the IMM. It is noteworthy, however, that basal proton leak verified under conditions of maximized PMF of state 4 respiration may not necessarily apply to state 3 respiration, whereby PMF is reduced in the presence of functional ATP synthesis.

Proton leak down the proton gradient may also be artificially applied by uncoupling protonophores (eg, 2,4-dinitrophenol [2,4-DNP], CCCP, FCCP). Proton transfer across the IMM mediated by uncouplers results in collapsing the proton gradient and in bypassing the ATP synthase. The nonreversible flux of protons across the IMM in the direction of the gradient results in dissipating the PMF as heat.

In addition to mitochondrial metabolic efficiency being affected by the proton permeability of the IMM, the efficiency of mitochondrial ATP synthesis may depend on the mode of feeding reducing equivalents to the electron transport chain. Thus, feeding reduced NADH electrons into complex I of the chain results in synthesizing about 3 mol of ATP per mole atom of respired oxygen. In contrast, reduced FADH feeds electrons into complex II of the chain, resulting in synthesizing about 2 mol ATP per mole atom of oxygen (17). Because the mitochondrial membrane is impermeable to nucleotides, reduced NADH formed in the cytosol has to be transported into the mitochondrial matrix by nucleotide shuttles. The two major ones consist of the malate-aspartate and the glycerol-3-phosphate (G3P) shuttles (1). The malate-aspartate shuttle catalyzes an apparent exchange of cytosolic-mitochondrial reduced NADH, whereas the G3P shuttle catalyzes an apparent exchange of cytosolic NADH for mitochondrial FADH. The exchange of cytosolic NADH for mitochondrial FADH by the G3P shuttle is enabled by cross talk between the NADH-dependent cytosolic G3P dehydrogenase (G3PDH) and the FADH-dependent mitochondrial G3PDH (mG3PDH). Because feeding FADH electrons to complex II results in only 2 mol of ATP per mole atom of respired oxygen, the predominant use of the G3P shuttle implies a thermogenic function at the expense of ATP production.

C. Futile cycling

Extramitochondrial metabolic efficiency—namely, the efficiency of ATP utilization in carrying out a given task—is mediated by the operation of substrate/futile cycles (18). A substrate cycle consists of opposing nonequilibrium reactions, catalyzed by distinct opposing enzymes that operate simultaneously without generating net product, but with a net result of heat production due to hydrolysis of ATP. The hexokinase/glucose-6-phosphatase or the phosphofructokinase/fructose 1,6-diphosphatse pairs may serve as examples for futile cycles that consist of two opposing reactions (19). Similarly, enzyme-catalyzed ATP-dependent ion pumping in one direction combined with enhanced ion leakage or ion pumping in the opposite direction may comprise a futile cycle resulting in net ATP hydrolysis. Substrate cycles may further comprise opposing pathways, rather than two opposing discrete reactions. Thus, glycolysis/gluconeogenesis (20), lipogenesis/fatty acid oxidation, lipolysis/fatty acid re-esterification, or protein turnover may represent futile cycles that consist of opposing metabolic pathways, whereby the overall heat produced depends on the total ATP hydrolyzed within the concerned cycle. Hence, whereas mitochondrial uncoupling decreases metabolic efficiency at the stage of mitochondrial ATP production, substrate/futile cycles allow for regular mitochondrial ATP synthesis while decreasing metabolic efficiency at the stage of ATP utilization.

II. TH and Metabolic Efficiency

The first description of TH-induced calorigenesis dates to 1895, when Magnus-Levy (21) reported an increase in metabolic rate in subjects eating TH extract. That initial report has been followed by exhaustive data focusing on the phenomenology of TH action in hyper- and hypothyroidism. Thus, high levels of TH in mammals increase oxygen consumption and heat production, resulting in pronounced body weight loss despite increased energy intake, whereas low levels of TH are associated with a decrease in energy expenditure, cold sensitivity, and pronounced increase in body weight (22–25). These observations have implied that TH may play a role in controlling metabolic efficiency, being decreased by hyperthyroidism while increasing with hypothyroidism. However, its cellular mode of action remained elusive. Upon realizing that TH activities are transduced by transcriptional modulation (26, 27), TH research has focused on understanding TH mode of action in controlling gene transcription as well as in defining downstream TH targets involved in modulating metabolic efficiency.

A. Genomic/nongenomic actions of TH

TH transcriptional activities are species, tissue, and cell-type dependent and include genomic and nongenomic actions. These are extensively reviewed by Davis and colleagues (28, 29) and will be only briefly summarized here to the extent required for integrating TH action in controlling gene transcription with its activity in modulating metabolic efficiency.

The genomic activity of TH involves transcriptional regulation of hundreds of TH-target genes that code for the expression of about 8% of total liver proteins. These are involved in transducing carbolipid and protein metabolism, energy metabolism, tissue biogenesis, cell cycle, apoptosis, and others (30). Most TH-target genes are induced, whereas some are suppressed by TH (31). TH effects on target genes are determined by time frame, whereby the expression of some is modulated during the first 6 hours of exposure to TH, whereas that of others is modulated within the next 24–48 hours.

TH genomic action is mediated by binding of the hormone to nuclear TH receptors (THRs) (32). Binding affinity is higher for the T₃ hormone than its T₄ precursor. THRs are members of the family of nuclear receptors that regulate the expression of genes as a function of ligand binding (33). THRs are encoded by the α and β c-erbA genes located on chromosomes 17 and 3, respectively, and are expressed as several spliced isoforms (34, 35). The c-erbAα gene encodes the TH-binding receptor THRα1 and two spliced variants that do not bind hormone (THRα2 and THRα3), whereas the c-erbAβ gene encodes the THRβ1, THRβ2, and THRβ3 isoforms (36, 37). The three β isoforms share high sequence homology in their DNA and ligand binding domains but differ in their amino-terminal domains. Both THRα1 and THRβ1 are expressed ubiquitously. However, THRα1 has its highest expression in skeletal and cardiac muscle, bone, and brown fat, whereas THRβ1 is highly expressed in liver, brain, and kidney (38).

Transcriptional regulation by TH involves binding of respective nuclear THRs to TH-response elements (TREs) in the promoter of respective TH-target genes. THRs may bind either as homodimers or as heterodimers with the retinoic-X receptor. In the absence of ligand, THRs bound to respective TREs suppress target gene transcription due to the binding of transcriptional corepressors (eg, NCor, SMRT) to the transcriptional initiation complex. Upon binding of TH to THR, corepressors are displaced, being replaced by transcriptional coactivators (eg, SRC), resulting in transcriptional modulation of concerned genes (28, 31). Only a limited number of TH-responsive genes consist of TREs in their promoters, thus being directly regulated by TH/THR (39). The expression of these genes is usually increased during the first 6 hours of exposure to TH (eg, peroxisome proliferator-activated receptor [PPAR] γ coactivator-1α [PGC-1α]). Instead, most TH-responsive genes lack TREs, and their transcription is usually affected within 24–48 hours after exposure, implying the involvement of intermediate transcription factors induced by TH, and which may modulate transcription through their responsive promoter elements. These may include transcription factors (eg, PPARγ; the nuclear respiratory factors, NRF1 and NRF2; coactivators [eg, PGC-1α, SRC-1]; or corepressors [eg, RIP140, hairless]) (30). Thermoregulation by TH is abrogated by the absence of THRs (40), implying their obligatory role in mediating TH metabolic activity.

TH genomic action may be complemented by TH nongenomic activities (29, 41). TH nongenomic activities may require high-dose TH, are rapid, and are initiated at the level of the plasma membrane, cytoplasm, or intracellular organelles (eg, mitochondria), being mediated by cytoplasmic THR (eg, THRβ1, THRα1), truncated mitochondrial THR, or plasma membrane receptors (eg, ανβ3 integrin) (28, 42, 43). Nongenomic TH actions are involved in modulating kinase/phosphatase activities (eg, protein kinase C [PKC], calcineurin, phosphatidylinositol-3-kinase, protein kinase B/Akt, mammalian target of rapamycin, p70S6K, ERK1/2, c-Src), intracellular calcium levels (via inositol 1,4,5-trisphosphate receptor [IP3R], sarcoplasmic/endoplasmic Ca⁺²-ATPase [SERCA], Ca²⁺-ATPase among others), ion pumps and channels (eg, Na⁺/H⁺ exchanger, Na,K-ATPase), intracellular protein trafficking (eg, Bcl-2 proteins, THRs), and transcription factors (eg, THR, STAT1,3, p53, HIF-1α). The interplay between genomic and nongenomic activities of TH in modulating metabolic efficiency is exemplified by TH-induced mitochondrial biogenesis, where TH genomic, nongenomic, and indirect activities do converge (30). Because genomic and nongenomic actions of TH/THR may modulate each other and converge to the same targets, the contributions made by these activity modes are dose- and context-dependent.

B. TH-induced mitochondrial uncoupling

Upon realizing that TH stimulates calorigenesis and thermogenesis by affecting respiration, exhaustive attempts were made by the scientific community throughout the 20th century to verify its putative mitochondrial target(s) (5, 8).

Studies by Lardy and Feldott (44) and Hess and Martius (45) have pointed out during the 1950s that the respiratory control ratio of isolated mitochondria, namely, the ratio between ATP production and oxygen consumption, was robustly decreased in the presence of added TH, implying uncoupling of mitochondrial oxidative phosphorylation. However, the high TH doses used in those studies, the short response time instead of the latent period preceding the physiological response, and the similar effects displayed by biologically inactive TH analogs were suspected of possible nonphysiological activity rather than authentic TH-induced mitochondrial uncoupling. Later evidence in support of mitochondrial uncoupling has, however, indicated an increase in oxygen consumption and mitochondrial proton leak of isolated liver mitochondria derived from hyperthyroid rats, with a concomitant decrease in mitochondrial phosphate potential and IMM potential (5, 46, 47). These observations were further corroborated by an increase in liver oxidizing capacity of hyperthyroid rats, accompanied by a decrease in phosphate and cytosolic redox potential (48), whereas opposite effects were reported in livers of hypothyroid rats (49). Similarly, hepatocytes isolated from TH-treated rats show higher oxygen consumption and lower IMM potential as compared with nontreated control (50–52). Specifically, in transition from hypothyroid to euthyroid as well as from euthyroid to hyperthyroid, about half of the respective increase in oxygen consumption was accounted for by TH-induced mitochondrial proton leak (52). Also, a decrease in IMM potential has been reported in TH-treated human lymphocytes or those derived from hyperthyroid patients (53). Uncoupling of mitochondrial oxidative phosphorylation in human skeletal muscle by TH has also been verified by the disproportionate increase in tricarboxylic acid cycle flux compared with ATP synthesis, measured by C13 and P31 NMR, respectively (54, 55). Overall, these findings suggested that TH indeed induces uncoupling of mitochondrial oxidative phosphorylation and that uncoupling may account for the cellular mode of action of TH in modulating metabolic efficiency in vivo.

Concomitantly with efforts made in resolving TH targets involved in promoting mitochondrial proton leaks, studies by Silva and colleagues (1) have pointed out that TH induces the activity of mG3PDH, the rate-limiting enzyme of the G3P shuttle, implying its putative role in TH-induced thermogenesis. However, total body oxygen consumption was reduced by only 7–10% in mice lacking mG3PDH, being quite limited in accounting for the 30–40% increase in oxygen consumption in transition from hypothyroid to euthyroid (56). Also, treating the mice with TH resulted in overcoming the decrease in oxygen consumption to an extent similar to wild-type controls (56), implying that the contribution of the G3P shuttle to TH-induced thermogenesis is limited. Indeed, because nutrient oxidation that yields reduced NADH is mostly carried out by mitochondrial tricarboxylic acid cycle or β- oxidation, rather than extramitochondrial metabolism, heat production by predominant G3P shuttling must be quite limited. This reservation may not apply to skeletal muscle that generates cytosolic reducing equivalents during activity, and where the G3P shuttle may allow for the rapid generation of ATP, but with reduced efficiency.

C. Gating of mitochondrial uncoupling proteins (UCPs) by TH

Several candidates were pursued in searching for TH target(s) involved in promoting mitochondrial proton leaks. Thus, TH treatment appeared to enrich cardiolipin (57) and to modulate the composition of n-3 and n-6 long-chain fatty acids (LCFAs) (58) and phospholipids (59) in mitochondrial membranes, resulting in increased fluidity. Similarly, TH appeared to induce mitochondrial oxidative stress with an increase in lipid peroxides in skeletal muscle and heart, implying a putative role of oxidative stress in TH-induced proton leak (60). However, the induced changes in IMM lipid composition or in reactive oxygen species (ROS) production, while being correlated with modulating proton leak, did not offer an explicit mode of action for TH in inducing mitochondrial uncoupling.

With the discovery of mitochondrial UCP1 in brown adipose tissue (BAT), extensive efforts were invested in verifying the putative role of UCP-family proteins in mediating the thermogenic effect of TH. In fact, the UCP-coding genes have TREs in their promoters, and their expression level is increased by TH treatment, implying their putative role in mediating TH-induced thermogenesis (61). Mitochondrial uncoupling by UCP is mediated either by proton transport in the direction of the proton gradient mediated by UCP-associated anionic carboxylates (eg, free fatty acids [FFAs]) or by transport of protonated carboxylic acids into the matrix, followed by UCP-catalyzed transport of the carboxylate anions in the opposite direction (9, 62).

UCP1 is specifically expressed in BAT that specializes in adapting rodents to cold or excessive food intake (cafeteria diet) (63). UCP1 ablation abrogates cold-induced thermogenesis in mice (64, 65). Adaptive thermogenesis by BAT is driven by the interplay of TH and adrenergic stimulation, resulting in BAT hyperplasia, enrichment of BAT UCP1 content, and increase in substrate availability, mainly fatty acids, for mitochondrial oxidation (66). Both TH and adrenergic stimulation are required for maximal adaptive thermogenesis by BAT, being abrogated by blocking either one (66–68). The TH/adrenergic interplay in inducing BAT adaptive thermogenesis is accounted for by TH-induced central nervous system sympathetic activity, resulting in up-regulating BAT thermogenic markers. The central nervous system effect of TH is claimed to be transduced by activating the lipogenic pathway in the ventromedial nucleus of the hypothalamus (VMH), due to suppressing hypothalamic AMP-activated protein kinase activity (69). Inhibition of THRs in the VMH or suppressing VMH lipogenesis abrogates BAT adaptive thermogenesis induced by high-dose TH. In addition, the TH/adrenergic cross talk in inducing UCP1 expression is further mediated by an enhancer element located at −2.2/−2.5 kb upstream of the transcriptional initiation point in the rat UCP1 gene promoter, and which consists of several cAMP-response elements juxtaposed to TREs and retinoic acid-responsive promoter elements (70, 71). Because LCFAs serve as the main substrate being oxidized in BAT during adaptive thermogenesis, TH is further involved in controlling substrate availability for BAT oxidation by inducing lipogenesis as well as by activating lipolysis of BAT triglycerides (72–74). BAT adaptive thermogenesis induced by TH is transduced by the interplay of both the α and β isoforms of TH nuclear receptors. Thus, adaptive thermogenesis, but not UCP1 level, becomes defective in mice lacking all THRα isoforms (75), whereas UCP1 expression and adaptive thermogenesis are abrogated in mice with a dominant-negative knock-in mutation in the THRβ gene (the PV mutation), implying a role for THRβ in inducing UCP1 expression (76).

Although BAT UCP1 may indeed account for adaptive thermogenesis in rodents, it is noteworthy that BAT becomes absent or sparse in the adult human. Recent findings point to the presence of cervical and lumbar brown adipose islets in adult humans (77–82); however, their putative impact on total body energy expenditure still remains to be resolved (83). Hence, TH-induced BAT thermogenesis is of questionable relevance to human metabolic efficiency and its control by TH. This understanding has resulted in pursuing the thermogenic role played by UCP2 and UCP3, sharing 60% sequence homology with UCP1 (9).

UCP2 is ubiquitously expressed and may catalyze proton transport in the direction of the proton gradient when activated by ROS or lipid peroxidation products (84). In contrast to UCP2, UCP3 is tissue specific, being abundant in skeletal muscle, but less in BAT and heart (85). In light of the role of muscle in contributing to total body oxygen consumption, UCP3 abundance in skeletal muscle could in principle drive TH-induced calorigenesis. Indeed, overexpression of UCP3 in mice results in a decrease in body weight gain despite increased caloric intake (86, 87). However, UCP3 knockout mice show normal response to TH (88), implying that TH-induced thermogenesis is not due to TH-induced UCP3 expression. Also, the expression of liver UCP2/UCP3 proteins is restricted to Kupffer cells (89), implying that the uncoupling effect of TH in liver parenchymal cells is not due to UCPs.

D. Mitochondrial permeability transition pore

In analogy to UCPs, mitochondria consist of PTPs (11–15, 90) located at the contact sites of the IMM and outer mitochondrial membrane (OMM). Mitochondrial PTP is conserved through evolution, being reported in mammalian and nonmammalian eukaryotes. The molecular composition and structure of mitochondrial PTP is still elusive and remains to be resolved. The current working model of PTP consists of several interacting proteins, including the ANT (in the IMM), the voltage-dependent anion channel (VDAC; in the OMM), cyclophilin D (CypD; in the mitochondrial matrix), the phosphate carrier (in the IMM), the Bcl2 family of proteins (in the OMM), cytosolic hexokinase, and intermembrane creatine kinase. This working model has been supported by yeast two-hybrid screening and coimmunoprecipitation studies, showing dynamic complexes containing various combinations of PTP components (91–95). However, recent studies, using genetic approaches, seem to question the current model (96–100). These studies may imply redundancy in PTP components or respective isoforms, variability in composition of PTP species as a function of their variable functions, or a regulatory role for PTP components previously considered integral components of the channel.

Irrespective of PTP structural identity, biochemical and biophysical evidence has indicated that PTP gating may present itself in two modes, namely, high-conductance PTP (HC-PTP) and low-conductance PTP (LC-PTP), differing in electrical conductance, reversibility, and synchronization. Definitive, synchronized HC-PTP gating is induced by intramitochondrial Ca⁺² load (101), being enhanced by oxidative stress, depletion of adenine nucleotides, increase in inorganic phosphate, increase in matrix pH, and depolarization of the IMM (102–104). PTP effectors may either modify the IMM potential as such or affect the threshold voltage (“gating potential”) at which opening occurs (105, 106). Thus, pore inducers may shift the apparent gating potential to more negative values, thereby favoring pore opening, whereas pore inhibitors may increase the gating potential and favor PTP closure. HC-PTP results in extensive depolarization of the IMM (>70% decrease in IMM potential), rapid passage of ions and solutes of less than 1500 Da across the IMM, and mitochondrial swelling. These may lead to rupture of the OMM, release of mitochondrial proapoptotic proteins (eg, cytochrome c, apoptotic intrinsic factor), followed by programmed cell death/apoptosis (107). Alternatively, nonsynchronized, transient/flickering LC-PTP gating, due to cyclic opening and closure of individual PTP channels, may result in limited depolarization of the IMM, moderate decrease in PMF, and passage of solutes of less than 300 Da (eg, H⁺, Ca²⁺, K⁺), accompanied by mitochondrial contraction rather than swelling (108–115). Most importantly, in contrast to the irreversible proapoptotic depolarization inflicted by HC-PTP, LC-PTP gating is innocuous and reversible, implying a physiological role of mitochondrial PTP in regulating metabolism (116). Specifically, LC-PTP gating may perhaps account for the basal proton leak previously ascribed to ANT (6).

Both modes of PTP gating are blocked by cyclosporin A (CsA) or by Sanglifehrin A (SfA) binding to CypD, resulting in interfering with CypD interaction with ANT within the PTP complex (117, 118). Hence, inhibition by CsA or SfA may indicate PTP involvement. SfA seems to be more specific because CsA may as well inhibit protein phosphatase 2B (PP2B; calcineurin) activity.

The two modes of PTP gating and their respective cross talk may be exemplified by the pathophysiological set-up of cardiac or hepatic ischemia/reperfusion (I/R) damage and ischemic pre/post conditioning (IPC) (119–121). I/R damage refers to apoptotic, necrotic, or autophagy injury inflicted by reperfusion that follows prolonged ischemia. I/R damage is induced by HC-PTP gating due to massive accumulation of mitochondrial Ca²⁺, being enhanced by oxidative stress, ROS-induced ROS-release (122), depletion of adenine nucleotides, and increase in matrix pH. I/R damage is abrogated by added CsA (123) or in CypD knockouts (98), implying PTP involvement.

In contrast to the pathological activity of mitochondrial HC-PTP in transducing I/R damage, IPC attenuates I/R damage by promoting LC-PTP (119–121). Ischemic preconditioning may be induced by applying cycles of short I/R before subjecting the concerned organ to extended ischemia. Similarly, ischemic postconditioning may be induced by subjecting the ischemic organ to short I/R cycles before allowing for full-blown reperfusion. Both protocols result in a moderate decrease in IMM potential, accompanied by significant attenuation of I/R damage in animal models as well as in humans. Most importantly, protection by IPC may be abrogated by CypD deficiency or added CsA, SfA, and antioxidants (eg, N-methylpropionylglycine), implying the involvement of mitochondrial PTP and ROS (113). It still remains to be investigated whether ROS production may result in oxidation of one or more PTP components, resulting in promoting its opening in the low-conductance flickering mode (114). Indeed, ANT and CypD are particularly sensitive to oxidative modification, with direct impact on PTP opening susceptibility (124, 125). Most importantly, attenuation of I/R damage by IPC-induced LC-PTP implies cross talk between the two PTP modes, whereby prior LC-PTP may suppress I/R-induced HC-PTP. Because LC-PTP enables mitochondrial Ca²⁺ rapid efflux through the PTP channel, LC-PTP gating during IPC may result in minimizing the impact of mitochondrial Ca²⁺ accumulation under conditions of I/R-induced Ca²⁺ overload (126).

The apoptotic outcome of HC-PTP, as contrasted with the physiological function of LC-PTP, calls for strict regulation of PTP gating. Indeed, in addition to controlling ROS, cellular pH, Pi, and other PTP activators/inhibitors, several signal transduction pathways converge to prevent definitive HC-PTP gating. Thus, IPC or pharmacological conditioning is transduced by a group of reperfusion injury salvage kinases that include Akt, Erk1/2, PKG, PKCε, and others, capable of phosphorylating and inactivating mitochondrial GSK3β. Inactivating GSK3β abrogates the phosphorylation of yet unresolved PTP components (eg, CypD, ANT, VDAC), resulting in increasing the threshold for PTP gating (127–129).

E. Mitochondrial PTP gating by TH

In testing the role played by PTP gating in TH action, a straightforward approach would be to examine whether TH-induced uncoupling of mitochondrial oxidative phosphorylation is inhibited by PTP suppressors. Indeed, TH-induced depolarization of mitochondrial PMF is abrogated by added CsA, pointing to PTP involvement (130). In addition, liver mitochondria of hypothyroid rats show a decrease in mitochondrial Ca⁺² efflux and swelling, being restored by TH treatment (131–133). Furthermore, TH treatment of Jurkat cells induces CsA-sensitive LC-PTP gating (134), implying that mitochondrial PTP may serve as a target for TH in modulating metabolic efficiency. As outlined in Sections II.E.4.–E.6. TH activity in gating PTP is transduced by regulating the mitochondrial activity of Bcl2-family proteins, rather than by modulating the expression of mitochondrial PTP components. The concerned transduction pathway is initiated by TH-induced efflux of endoplasmic reticulum (ER) Ca⁺² ([Ca⁺²]_ER), followed by dephosphorylation of mitochondrial Bcl2(S70) by [Ca⁺²]-activated PP2B/calcineurin (Figure 2).

Figure 2

PTP gating by thyroid hormone.

1. Adenine nucleotide translocase

ANT function in the PTP context has been verified by its direct association with CypD and VDAC (91), as well as by PTP gating being activated or inhibited by the ANT ligands atractylate or bongkrekic acid, respectively (135). Moreover, ANT/CypD/VDAC-reconstituted liposomes show PTP characteristics in terms of sensitivity to Ca⁺², CsA, and ANT ligands (136, 137) (see also Ref. 96). Also, overexpression of ANT isoforms (ANT1, ANT3) promotes apoptosis, being inhibited by CsA or by overexpressed CypD (138, 139). Moreover, ANT expression levels affect mitochondrial IMM potential, with high ANT levels resulting in IMM depolarization and proton leak (6, 134, 138). Hence, in light of ANT structural and regulatory functions in the PTP context and because the expression level of ANT2, the only ANT isoform expressed in liver, is increased in hyperthyroidism and decreased by hypothyroidism (140), TH-induced ANT expression could apparently account for liver TH-induced PTP gating. However, overexpression of ANT2 in the HeLa cell line or in rat primary hepatocytes resulted in extensive mitochondrial depolarization that was not inhibited by CsA (134), implying the formation of CsA-insensitive ANT channels (6, 141), rather than authentic PTP gating.

2. Cyclophilin D

CypD is a member of the family of peptidyl-prolyl cis-trans isomerases (PPIases) (142). The CypD protein contains a mitochondrial-targeting sequence that directs it specifically to the mitochondrial matrix. The link between CypD and PTP has been verified by CypD direct association with ANT (91) as well as by CsA inhibition of PTP gating, being accounted for by its interaction and inhibition of CypD activity (117). Moreover, PTP opening by Ca⁺² and oxidative stress was enhanced in isolated mitochondria of neurons overexpressing CypD (143), whereas CsA-sensitive PTP opening was abrogated in isolated mitochondria of CypD knockout mice (100). These CypD characteristics may indicate that CypD could apparently serve as a protein target of TH in inducing PTP opening and mitochondrial uncoupling. Indeed, liver mitochondria of hyperthyroid rats show increased expression of CypD as well as increased PPIase activity, whereas the opposite prevails in hypothyroid rats (134). However, overexpression of CypD in the HeLa cell line or in rat primary hepatocytes resulted in mitochondrial hyperpolarization rather than PTP opening (134, 144). Moreover, overexpressed CypD was found to desensitize cells to apoptotic stimuli or to protect cells from mitochondrial depolarization and apoptosis induced by overexpression of ANT1 (138, 139). Hence, TH-induced CypD expression may not account for TH-induced PTP gating and thermogenesis. CypD induction by TH may reflect TH activity in inducing PPIase activity, rather than PTP opening (145).

3. Voltage-dependent anion channel

VDAC is a highly abundant protein of the OMM. Its primary function consists of exchanging anions between the cytosol and the intermembrane mitochondrial space (146). Previous findings have indicated its putative role in gating mitochondrial PTP (92, 147). However, its expression level is not affected by in vivo TH treatment (134), excluding VDAC from being a direct target of TH in inducing mitochondrial uncoupling.

4. Bcl2-family proteins

The family of Bcl2 proteins consists of more than 20 proteins that were extensively studied in terms of their role in cell death (148). The Bcl2 family is grouped into two main subfamilies: proapoptotic proteins (eg, Bax, Bak, and others) and antiapoptotic proteins (eg, Bcl2), which promote or inhibit PTP gating, respectively (149). Bcl2-family proteins may directly interact with PTP components such as ANT (95, 150, 151) or VDAC (152), and when overexpressed or added to isolated mitochondria may specifically induce (eg, Bax and Bak) or antagonize (eg, Bcl2) PTP gating (150–153). Similarly, depletion of proapoptotic Bax or Bak results in failure of PTP gating (95, 153, 154), whereas Bcl2 inactivation results in definitive PTP gating triggered by oxidative stress (155). Thus, mitochondrial Bcl2-family proteins and their respective heterodimers (eg, Bax/Bcl2, Bad/Bcl2) may apparently serve as candidate targets of TH in inducing PTP gating (156). Indeed, TH-induced PTP gating is accompanied by an increase in mitochondrial Bax and Bak, together with a decrease in mitochondrial Bcl2 content, whereas hypothyroidism results in opposite effects that are reversed by TH (134). Modulation of the mitochondrial content of Bcl2 proteins by TH is due to their specific translocation in/out of mitochondria, rather than reflecting modulation of their expression and total cellular content. Amplifying the ratio of mitochondrial proapoptotic vs antiapoptotic proteins by TH results in a robust decrease in mitochondrial Bax/Bcl2 heterodimer and concomitant increase in free Bax (134), resulting in PTP gating induced by free mitochondrial Bax (157, 158). Indeed, overexpression of Bcl2 protects against TH-induced mitochondrial PTP gating (134), implying that depletion of mitochondrial Bcl2 by TH may account for TH-induced mitochondrial uncoupling.

5. Ca⁺²/Bcl2 cross talk

Because Bcl2-Bax heterodimerization may depend on the Bcl2(S70) phosphorylation state (159), mitochondrial Bcl2 depletion by T₃ has been further pursued in terms of Bcl2(S70) phosphorylation profile. Indeed, concomitantly with a decrease in mitochondrial Bcl2, TH treatment results in decreased phosphorylation of monomeric mitochondrial Bcl2(S70) as well as of Bcl2(S70)-Bax heterodimer (160), indicating that mitochondrial Bcl2 depletion may reflect Bcl2(S70) dephosphorylation by TH. In pursuing kinases (eg, protein kinase A [PKA], PKC) or phosphatases (eg, PP2A, PP2B/calcineurin) reported to be involved in Bcl2(S70) phosphorylation, neither PKA, nor PKC, nor PP2A was found to mediate phosphorylation/dephosphorylation of Bcl2(S70) by TH (160). However, dephosphorylation of Bcl2(S70) by TH and its mitochondrial depletion were both reversed by the FK506 inhibitor of PP2B, indicating that the TH effect may be mediated by activation of PP2B (160). Indeed, Ca⁺²-activated PP2B has previously been reported to bind and dephosphorylate Bcl2(S70) (161, 162). Furthermore, added FK506 blocked TH-induced PTP gating, indicating that dephosphorylation of Bcl2(S70) and its mitochondrial depletion by TH-activated PP2B may account for mitochondrial PTP opening by TH. TH-induced PP2B activation was not accompanied by an increase in PP2B expression. However, TH treatment resulted in a pronounced increase in cytosolic Ca⁺², whereas Ca⁺² chelation by BAPTA resulted in abrogating PTP gating by TH (160), implying that TH-induced PTP gating involved modulation of cytosolic Ca⁺² by TH.

6. Inositol 1,4,5-trisphosphate receptor

The dynamic equilibrium between cytosolic Ca⁺² ([Ca⁺²]c) and [Ca⁺²]_ER is maintained by an interplay of Ca⁺² efflux, mediated by the endoplasmic IP3R1 and Ca⁺² influx catalyzed by the SERCA (163). IP3R1 is activated by binding of the IP3 ligand and may further be modulated by its phosphorylation by PKA, PKC, or CaMKII, by its dephosphorylation by PP2B, or by its association with one or more of about 50 proteins, including FKBP12 or Bcl2 (164, 165). Indeed, PTP opening, dephosphorylation of mitochondrial Bcl2(S70), and depletion of mitochondrial Bcl2 are all abrogated in cells lacking IP3R1, indicating that IP3R1 is obligatory for TH- induced mitochondrial uncoupling (160). Similarly, TH fails to increase [Ca⁺²]c upon inhibition of IP3R1 by 2-aminoethoxydiphenyl borate, indicating a specific requirement for IP3R1 activity in modulating [Ca⁺²]c by TH. TH-induced gating of IP3R1 is mediated by an increase in IP3R1 expression, complemented by IP3R1 truncation into channel-only isoforms (160). Truncated IP3R1 isoforms have been reported to serve as channel-only peptides capable of carrying out [Ca⁺²]_ER efflux in the absence of added IP3 (166–170). IP3R1 truncation by TH may reflect TH activation of IP3R1 proteases that remain to be identified. PTP gating induced by constitutive PP2B prevails under conditions of suppressing IP3R1 expression by small interfering RNA (171), implying that [Ca⁺²]c-activated PP2B is acting downstream to TH-induced IP3R1 and is obligatory as well as sufficient in mediating PTP gating.

Overall, TH-induced expression of the endoplasmic IP3R1 channel, accompanied by its truncation, is proposed to result in [Ca⁺²]_ER efflux, an increase in [Ca⁺²]c and [Ca⁺²]c-activated PP2B, followed by dephosphorylation of mitochondrial Bcl2(S70), a decrease in mitochondrial Bcl2 protein levels, and an increase in mitochondrial free Bax (Figure 2). The decrease in mitochondrial Bcl2 and/or the respective increase in mitochondrial free Bax may initiate and promote variable PTP gating, resulting in LC-PTP or HC-PTP gating, depending on prevailing mitochondrial conditions. Because LC-PTP may drift to HC-PTP gating, culminating in apoptosis, the TH transduction pathway is proposed to be constrained by negative feedback at its IP3R1 target (Figure 2). Negative feedback may be due to inhibition of IP3R1 opening by an increase in [Ca²⁺]c, binding of dephosphorylated Bcl2(S70) to IP3R1, or dephosphorylation of IP3R1 by activated PP2B (164). Conditions and biological context that favor LC-PTP drift to HC-PTP gating by TH still remain to be investigated.

F. Activities of TH transduced by PTP gating

PTP gating may offer a unified TH target for some TH pleiotropic activities. In addition to TH action in modulating metabolic efficiency, PTP gating may account for TH-induced apoptosis and TH-induced pre/post conditioning (PC) of I/R damage.

1. Metabolic efficiency

TH action in controlling obligatory thermogenesis is proposed to involve LC-PTP gating. Because mitochondrial PTP is ubiquitously distributed, an increase in total body energy expenditure induced by TH may reflect the sum total of body LC-PTP gating driven by various tissues, mostly muscle and liver, which account for about 45–60% of total body resting oxygen consumption (172). LC-PTP gating by TH of brain, testes, or spleen, where TH does not increase oxygen consumption, still remains to be verified. LC-PTP gating in muscle and liver may further complement BAT UCP1 in mediating adaptive thermogenesis, whereby each may drive specific adaptive thermogenic modes due to their distinct characteristics. Thus, in light of the 40-fold difference in organ mass between muscle and cold-acclimated BAT, the contribution of muscle LC-PTP gating to total body uncoupling capacity may far exceed that of BAT UCP1. However, because blood flow through cold-acclimated BAT after norepinephrine administration accounts for 33% of cardiac output, as compared with only 8% for muscle (173), the thermogenic output of BAT may exceed that of muscle, despite its limited mass. Hence, BAT UCP1 may drive low-capacity high-output adaptive thermogenesis and therefore may better serve an emergency call for heat production during cold exposure. In contrast, muscle and liver LC-PTP may drive high-capacity low-output adaptation of metabolic efficiency to variable energy intake. This distinction between adaptation to cold and to energy intake has been noted first by Kozak (3, 65), pointing out that UCP1 ablation resulted in sensitivity to cold exposure but was paradoxically accompanied by resistance to diet-induced obesity. Similar observations have been reported for other modes of abrogating BAT thermogenesis, eg, by introducing a homozygous aP2-UCP1 transgene (174), resulting in apoptosis of BAT cells due to excess UCP1 activity; by THRα1–3 (THRα0/0) ablation (175, 176), resulting in abrogating BAT TH activity; by ablation of type 2 deiodinase (177), resulting in BAT TH deficiency; or by adipose ablation of both the IGF-I and insulin receptors (FIGIRKO) (178) whereby BAT development is abrogated. All result in the inability to maintain body temperature when exposed to cold, while becoming resistant to diet-induced obesity, implying a non-BAT alternative mode of adaptive thermogenesis. The concerned alternative thermogenic mode that copes with diet-induced obesity in rodents is proposed to be driven by LC-PTP gating in muscle and liver. It is tempting to speculate that the LC-PTP thermogenic mode induced in rodents by BAT ablation may represent that of the adult human, implying a potential target for treating human obesity.

2. Apoptosis

TH controls cell differentiation, cell growth, and cell death (179, 180). TH involvement in controlling cell growth and death is most dramatically exemplified by amphibian metamorphosis, whereby tadpole organs are transformed into those of the adult frog (181). The metamorphosis process combines selective elimination of unwanted cells via apoptosis, together with new cell growth and development. Both processes are controlled by TH. Amphibian organ development may result either in entire tissue loss (eg, tadpole tail and gills) or in tissue remodeling (eg, brain, intestine, pancreas, and skin) (42, 181–184). TH-induced apoptosis has been similarly reported in other systems as well, eg, apoptosis of promyeloleukemic HL-60 cells (185), of differentiating hematopoietic cells (186), or of human T lymphocytes in vitro, exemplified in vivo in patients with Graves' disease (53).

The mechanism by which TH promotes apoptosis is not completely resolved. However, because HC-PTP gating initiates the mitochondrial apoptotic pathway and because TH-induced apoptosis is inhibited by CsA (187), TH-induced apoptosis is proposed to be partly mediated by HC-PTP gating. Indeed, TH has been reported to suppress the antiapoptotic protein Bcl-2 in T lymphocytes in vivo (53), while modulating the activities of Bax, Bid, and xR11 (homolog of the mammalian Bcl-X_L) in the course of tail apoptosis and brain differentiation of Xenopus laevis (182, 188). Although apoptosis may be important for the mechanism of TH action in amphibian metamorphosis and morphogenesis in mammals, it is probably of no relevance for the physiological thermogenic effect of TH.

3. I/R damage and PC

The activity of TH in inducing both modes of PTP gating is exemplified well by the double-edge action of TH in promoting or protecting I/R damage. Thus, acute or chronic TH treatment is reported to protect the heart against I/R damage, in terms of cardiac left ventricular diastolic pressure recovery, infarct size, lactate dehydrogenase release, and activation of stress kinases (189, 190), being transduced by THRα1 (191). Similarly to IPC, preconditioning by TH is abrogated by atractylate, indicating putative LC-PTP gating. Cardiac preconditioning by TH is additive to IPC, implying its prospective use in handling I/R damage during elective coronary artery bypass graft. In line with cardiac preconditioning, TH treatment is reported to protect liver I/R damage in terms of histology, liver function tests, and acute phase response (192, 193), implying its prospective use in handling I/R damage during elective liver transplantation or abdominal surgery requiring transient hepatic vascular occlusion (194). It is worth noting, however, that in contrast to reports pointing to conditioning by TH, others have reported increased I/R damage induced by hyperthyroidism (195–197), and tolerance to cardiac, liver, or kidney I/R damage by hypothyroidism, whereby PTP gating is abrogated despite Ca⁺² overload (132, 198–201). These apparent conflicting data may indicate that LC-PTP-induced conditioning by euthyroid TH levels may drift to HC-PTP gating and I/R damage under conditions of thyrotoxic TH levels or increased sensitivity to TH.

G. TH-induced futile cycles

As pointed out in Section I.A., mitochondrial uncoupling decreases metabolic efficiency at the stage of mitochondrial ATP production, whereas substrate/futile cycles allow for regular mitochondrial ATP synthesis while decreasing metabolic efficiency at the stage of ATP utilization. Thus, concomitantly with the mitochondrial uncoupling paradigm, others have proposed that TH-induced calorigenesis is due to TH induction of a variety of futile cycles, mostly extramitochondrial. The mitochondrial uncoupling and futile cycle paradigms are not mutually exclusive, and both may complement each other in modulating metabolic efficiency by TH.

1. Glycolysis/gluconeogenesis

Several studies, summarized by Muller and Seitz (202), have indeed shown increased endogenous production and utilization of glucose in hyperthyroid rats and humans. Similarly, tracer studies have shown that the Cori cycle (glucose/lactate/glucose) and the glucose/glucose-6-phosphate and fructose-6-phosphate/fructose-1,6-diphosphatase substrate cycles are increased in hyperthyroidism and decreased in hypothyroidism (20, 203, 204), implying TH-induced futile cycling of glucose production and its utilization. The quantitative contribution of these cycles to TH-induced thermogenesis has been estimated at 10% and 2% for rats and humans, respectively (26).

2. Lipolysis/lipogenesis/fatty acid oxidation

TH activates rat liver and adipose lipogenesis within a time frame similar to that of a TH-induced increase in oxygen consumption (23, 205). TH-induced lipogenesis is due to the coordinate induction of the expression of genes encoding the enzymes of fatty acid synthesis (eg, ACC, FAS) (206), as well as those responsible for generating the necessary reducing equivalents (eg, glucose-6-phosphate dehydrogenase) (207). Moreover, TH stimulates the β-oxidation of fatty acids by increasing the activity of carnitine palmitoyl transferase, the enzyme responsible for transporting fatty acids into the mitochondria (208, 209). The concomitant increase in lipogenesis and fatty acid oxidation may result in futile cycling of LCFA biosynthesis and its degradation. The energy cost of TH-induced lipogenesis has been estimated at 3–4% of TH-induced thermogenesis (73).

Concomitantly to promoting lipogenesis and fatty acid oxidation, TH activates adipose lipolysis by increasing its sensitivity to catecholamines (210, 211). Because the production of fatty acids by adipose lipolysis exceeds their rate of oxidation in humans (212), the balance is presumably accounted for by re-esterification of lipolysed fatty acids back into adipose or liver triglycerides, implying a futile cycle of adipose fat lipolysis and fatty acid re-esterification. The energy cost of this futile cycle has been estimated at 15% of the hyperthyroid increment in oxygen consumption in humans. The combined activity of TH-induced lipogenesis and lipolysis results in increased availability of LCFAs, accounting for the finding that fatty acids constitute the principal fuel in hyperthyroidism (212).

3. Protein turnover

In addition to increasing the transcription of TH-responsive genes in target tissues, TH has a more generalized stimulatory effect on total RNA synthesis in liver and muscle (28, 213), resulting in overall increase in protein synthesis. Concomitantly, TH activates muscle protein degradation, implying futile protein turnover (214). The energy cost of TH-induced protein turnover is estimated at 5–10% of TH-induced calorigenesis (26).

4. Ion pumping

Maintaining ion gradients across cellular membranes is an energetically costly process. Maintaining Na⁺, K⁺ gradients across the plasma membrane is mainly ascribed to the activity of Na,K-ATPase (215). Indeed, the expression of Na,K-ATPase (216, 217) as well as its translocation to the plasma membrane (218, 219) is induced by TH, being accompanied by an increased leak of sodium and potassium across the plasma membrane (220), thus resulting in futile ion pumping. Whereas TH-induced Na,K-ATPase may generate futile ion cycling at the level of the plasma membrane, TH-induced expression of [Ca⁺²] transporters in the ER may promote an intracellular futile [Ca⁺²] cycling. Thus, TH-induced expression of SERCA (221), which mediates ATP-dependent [Ca⁺²]c pumping into the ER, combined with TH-induced expression of IP3R (160) and/or the ryanodine receptors (222) that mediate [Ca⁺²]_ER efflux into the cytosol, may result in [Ca⁺²] cycling in and out of the endoplasmic/sarcoplasmic reticulum. Control of [Ca⁺²] cycling by TH is further mediated by TH suppression of phospholamban expression, resulting in activating [Ca⁺²]c pumping by SERCA (223). Although cardiac [Ca⁺²] cycling may serve cardiac inotropic performance (224), that of skeletal muscle may be of relevance to overall thermogenesis and temperature homeostasis. The energy cost of futile ion pumping has been estimated at 5–10% of TH-induced thermogenesis (26, 215, 221, 225).

H. TH-induced mitochondrial biogenesis

Independently of the mode of action of TH in modulating metabolic efficiency, a decrease in metabolic efficiency implies a requirement for an increase in substrate utilization for carrying out a given task. Irrespective of whether the concerned substrate is carbohydrate, fat, or protein, nutrient energy is mostly derived by mitochondrial oxidative phosphorylation, implying that a requirement for an increase in substrate utilization must be fulfilled by increased mitochondrial capacity. Hence, TH-induced mitochondrial biogenesis complements its control of metabolic efficiency.

Mitochondrial biogenesis by TH results in an increase in mitochondrial number, membranes, DNA, proteins, respiratory enzyme activities, and oxidative phosphorylation (30, 226). Because mitochondrial DNA encodes only a small fraction of mitochondrial proteins, mitochondrial biogenesis by TH requires the cooperation of both the mitochondrial and the nuclear genomes (30). Indeed, TH induces the expression of nuclear- or mitochondrial-encoded components of the mitochondrial respiratory chain and oxidative phosphorylation (eg, cytochrome c oxidase, cytochrome c, ANT, F0/F1ATP synthase) (Ref. 226 and references therein). Moreover, TH induces the expression of nuclear-encoded transcription factors that control mitochondrial DNA replication and transcription, eg, PGC-1α, NRF-1, mitochondrial transcription factor A (226, 227), resulting in driving the expression of mitochondrial components encoded by the mitochondrial genome. Concomitantly, TH induces protein (eg, CypD) import into the mitochondrial matrix (134, 228) by up-regulating mitochondrial protein importers (eg, heat shock protein 70) (229).

In addition to transcriptional activation by nuclear THR, TH may affect mitochondrial biogenesis directly, independently of nuclear THR. Thus, in vitro TH treatment of liver mitochondria isolated from hypothyroid rat resulted in an increase in mitochondrial RNA polymerase activity and mitochondrial gene transcription (230, 231). Mitochondrial gene transcription, independently of nuclear THR, has been ascribed to mitochondrial high-affinity TH binding sites (42, 232), proposed to consist of truncated THRα1 and THRβ1 (43). These are located in the mitochondrial matrix, may heterodimerize with other mitochondrial transcription factors (eg, retinoic-X receptor, PPAR), and specifically bind to putative TRE promoter sequences within the D-loop domain of the mitochondrial genome (eg, cytochrome c oxidase promoter) (43, 233–238).

It is worth noting that a TH-induced increase in substrate oxidation and oxidative phosphorylation due to mitochondrial biogenesis implies an increase in thermogenesis while maintaining euthyroid proportions of work performance (ΔG) and heat production (ΔS) within the overall increase in enthalpy/energy expenditure (ΔH). In other words, thermogenesis due to TH-induced mitochondrial biogenesis still maintains euthyroid metabolic efficiency. That is in contrast to a decrease in metabolic efficiency due to TH-induced futile cycling or uncoupling of oxidative phosphorylation, whereby the fraction of heat production linked to a given amount of work performance is increased relative to the euthyroid status.

III. Thyromimetics and Metabolic Efficiency

A. Thyromimetic agents

An increase in energy expenditure has long been considered for treating obesity. Thus, 2,4-DNP, a protonophoric uncoupler of mitochondrial oxidative phosphorylation, was introduced in 1933, but despite its efficacy in decreasing body weight, it was abandoned on 1938 due to fatal hyperthermia (239). These failures have been followed later by low-dose 2,4-DNP (240) or by butylated hydroxytoluene proposed to interact with mitochondrial ANT (241). None of these has reached clinical trials.

In line with TH activity in inducing thermogenesis, treating obesity by thyroid extracts was quite popular throughout the 20th century and well into the 1970s, being later abandoned due to severe side effects consisting of cardiac dysrhythmias, osteoporosis, electrolyte disturbances, and loss of lean body mass (242). Thus, a final ruling warning against the use of thyroid preparations for the treatment of obese euthyroid subjects was issued by the US Food and Drug Administration in 1978.

These early attempts were followed by rational drug design of synthetic structural analogs of TH that may avoid the lethal chronotropic inotropic cardiac effects of TH, while maintaining its calorigenic thermogenic activity. In pursuing that objective, and realizing that the cardiac activity of TH is predominantly mediated by THRα isoforms (243, 244), thyromimetics were designed to specifically avoid binding to THRα while maintaining their affinity to THRβ isoforms (245–247). That was accomplished by synthesizing modified TH analogs while mimicking the basic structure of TH—namely, an iodine-substituted biaryl ether skeleton linked to alanine moiety. Modification strategies focused on replacing the polar amino-alanine with a less polar moiety, aiming at decreasing hydrogen bonding with the THR ligand binding domain, and replacing the iodines with other halogens or alkyl groups in order to increase stability (248). Moreover, to further avoid the cardiac activity of TH, selective THRβ analogs were designed to target the liver by adding substituents that promote hepatic first-pass, rather than systemic distribution.

These efforts have resulted in a variety of THRβ-selective thyromimetics that proved efficacy in treating dyslipidemia (249–253). The hypolipidemic (mostly hypocholesterolemic) efficacy is due to transcriptional activation of genes coding for liver low-density lipoprotein receptors, the 7α-cholesterol hydroxylase being the rate-limiting enzyme in bile acid biosynthesis, and the biliary cholesterol transporters ABCG5 and ABCG8. The overall effect results in enhancing hepatic uptake of plasma low-density lipoprotein-cholesterol, followed by conversion of the cholesterol moiety into excreted bile acids and bile cholesterol. Moreover, the hypolipidemic efficacy of liver-specific THRβ-selective thyromimetics is further complemented by induction of the expression of the hepatic scavenger receptor SR-B1 that mediates reverse cholesterol transport.

The clinical benefits of THRβ-selective thyromimetics involve potential harmful risks. Thus, despite their relative selectivity for THRβ, high doses still activate THRα, resulting in positive chronotropic inotropic cardiac effects as well as enhanced bone resorption and muscle catabolism. The limited selectivity is not surprising because most residues that line the ligand binding domain pocket of THR isoforms are conserved (254). Moreover, conformational changes induced by THRβ-selective thyromimetics may result in novel cross talk with TRE promoter sequences, coactivators, or corepressors, resulting in a novel profile of gene expression (252). Conformational changes induced by THRβ-selective thyromimetics may further result in novel nongenomic effects. Hence, the safety of THRβ-selective thyromimetics still remains to be verified, in particular in subjects suffering from congestive heart failure or coronary heart disease (255). Also, because THRβ regulates the feedback loop of hypothalamic TSH, THRβ thyromimetics may suppress the production of endogenous TH, resulting in hepatic hyperthyroidism combined with systemic hypothyroidism due to low levels of endogenous TH.

Most importantly, focusing on THRβ analogs and further restricting their systemic activity by hepatic first-pass implies giving up essentially the prospects of using reasonable doses of THRβ-selective thyromimetics for treating obesity by exploiting the TH mode of action in controlling metabolic efficiency. Indeed, TH-induced total body oxygen consumption of THRα knockout mice kept at thermoneutrality is less than 30% of that induced in wild-type controls (247), indicating that energy expenditure induced by TH is mostly mediated by THRα action. Similarly, oxygen consumption induced by THRβ-selective thyromimetics amounts to about 10% of that induced by T₃, while displaying similar hypocholesterolemic efficacy (247, 256). Moreover, THRα isoforms appear to be required, as well as being sufficient in allowing for TH-induced energy expenditure in humans. Thus, resistance to TH due to mutation in THRα is reported to result in a decrease in basal metabolic rate (257), whereas mutations in THRβ result in an increase in energy expenditure and hyperphagia, comparable to that displayed by hyperthyroidism, being driven by retention of THRα sensitivity to elevated TH levels (258). Hence, THRβ-selective thyromimetics may complement the present drug arsenal for treating dyslipidemia but may have limited prospects in treating human obesity.

In contrast, rather than attempting to achieve selectivity in TH-induced activities while mimicking the essentials of TH structure, LC-PTP gating by TH may point to an alternative strategy—namely, focusing on compounds that may directly target mitochondrial PTP. These may serve as functional thyromimetics, while avoiding the TH/THR transduction pathway altogether. Nonesterified LCFAs/FFAs may fulfill that role.

B. Mitochondrial uncoupling by free fatty acids

LCFAs are mostly esterified, serving as building blocks of membrane phospholipids, triglycerides, or cholesterol-ester deposits. Nonesterified LCFAs/FFAs are present in much lower amounts, mostly bound to specific or nonspecific fatty acid-binding proteins or associated with cellular membranes. Only a minor fraction of FFAs remain essentially free and in equilibrium with bound FFAs. The main reservoir of FFAs in mammals consists of FFAs bound to serum albumin in a variable molar ratio within the range of 1–8. Plasma FFA concentrations may increase significantly under physiological (eg, fasting, physical exercise, high-fat diet) or pathological (eg, diabetes, cerebro-hepato-renal [Zellweger] syndrome, Refsum disease, ischemic and post ischemic brain or heart disease) conditions.

FFAs serve as excellent respiratory substrates in most cell types. After their transport as acyl-carnitines across the IMM, FFAs undergo β-oxidation in the mitochondrial matrix, feeding electrons into the respiratory chain. However, in addition to their substrate role, FFAs may act as mild uncouplers of mitochondrial oxidative phosphorylation (259). The uncoupling activity of FFAs is mediated by their protonophoric activity or by PTP gating (260). The ranking order of FFAs for the two respective activities in terms of FFA chain length and degree of saturation appears to be the same.

Mitochondrial uncoupling due to the protonophoric activity of FFA is mediated by cyclic movement of the protonated FFAs through the IMM, with the release of protons in the alkaline matrix space, and subsequent efflux of the FFA anion mediated by ANT (261, 262), the aspartate/glutamate transporter (263), the dicarboxylate carrier (264), or UCP channels (265). The most studied example is concerned with BAT UCP1 (266).

PTP gating by FFAs may be induced either indirectly by mitochondrial depolarization due to their protonophoric activity or by their direct interaction with PTP components (267). A direct effect of FFAs is supported by the observation that CsA-sensitive mitochondrial swelling was induced by added FFAs but not by the CCCP protonophore, under conditions whereby the initial and induced IMM depolarization caused by both compounds was identical (268). Direct gating of PTP by FFAs has been further confirmed using a partially purified preparation enriched with PTP components reconstituted into phospholipid vesicles (104). PTP gating by FFAs appears to require higher FFA concentrations than those required for their respective protonophoric activity, implying that mitochondrial uncoupling by FFAs may initially be induced by their protonophoric activity, followed by direct PTP gating. Mitochondrial FFA activities in promoting PTP gating may further be complemented by modulating extramitochondrial components that may affect gating sensitivity. Similarly to TH, PTP gating by FFAs may result either in LC-PTP—namely, transient, reversible, limited-size opening, enabling transport of < 300-Da ions—or in HC-PTP, leading to nonreversible cell death. The outcome may depend on FFA chain length and saturation, FFA concentrations, cell type, and the respective cellular context. Thus, HC-PTP gating by FFAs may account for LCFA-induced lipotoxicity in β-cells (112, 269). In contrast, moderate levels of FFA induce LC-PTP gating, moderate drop in ΔΨm, and facilitated mitochondrial Ca²⁺ efflux, without any adverse effect on cell viability (112). LC-PTP gating induced by FFA in pancreatic β-cells further results in activating glucose-stimulated insulin secretion, being blocked by added CsA (112). Hence, LC-PTP gating by FFAs may indicate that FFAs may in principle act as functional thyromimetics in modulating metabolic efficiency. However, the uncoupling activity of FFAs is confounded by their dual role as putative uncouplers of oxidative phosphorylation and as avid substrates for oxidation or esterification. That hurdle may be resolved by nonmetabolized LCFA analogs.

C. Mitochondrial uncoupling by long-chain fatty acid analogs

Methyl-substituted dicarboxylic acid (MEDICA) analogs consist of long-chain dioic acids [HOOC-C(α′)-C(β′)-(CH2)n-C(β)-C(α)-COOH (n = 10–14)] that are substituted in the αα′ or ββ′ carbons (270). MEDICA analogs may be thioesterified endogenously into their respective mono acyl-coenzyme A thioesters (271), but they are not esterified into lipids or β-oxidized, thus dissociating between the substrate role and the mitochondrial uncoupling activity of natural FFA. MEDICA analogs are mostly excreted in bile as respective glucuronides. In mimicking TH, MEDICA analogs induce calorigenesis in animal models. Thus, treatment of obese leptin receptor-deficient rats (eg, Zucker, cp/cp) with MEDICA analogs results in increased oxygen consumption and food consumption together with weight loss, implying a decrease in metabolic efficiency (272, 273). Furthermore, MEDICA treatment results in a decrease in liver mitochondrial phosphate potential and cytosolic redox potential, reflecting mitochondrial uncoupling in vivo (274). In mimicking TH, and similarly to natural FFAs, uncoupling by MEDICA analogs is mediated by targeting BAT UCP1 (275) as well as mitochondrial PTP (276, 277). The role played by the free MEDICA acid as compared with the respective acyl-coenzyme A thioester in targeting BAT UCP1 still remains to be investigated (278). BAT mitochondrial uncoupling by MEDICA partly prevails in BAT UCP1-deficient cells, implying an additional mitochondrial target. Indeed, similarly to TH or natural FFAs, MEDICA analogs induce a CsA-sensitive decrease in mitochondrial PMF, accompanied by atractylate-enhanced, bongkrekic-inhibited activation of mitochondrial Ca²⁺ efflux and implying PTP gating (28, 52, 134, 279). Similarly to TH, PTP gating by MEDICA analogs is mediated by modulating the profile of mitochondrial Bcl2-family proteins, resulting in a decrease in mitochondrial Bcl2-Bax heterodimer with a concomitant increase in mitochondrial free Bax (134, 160, 277). However, different transduction pathways are involved in modulating the mitochondrial content of free Bax by TH or MEDICA analogs. Thus, dissociation of the Bcl2-Bax heterodimer by TH is driven by dephosphorylation of Bcl2(Ser-70) by TH-activated PP2B (160), whereas dissociation of the Bcl2-Bax heterodimer by MEDICA analogs is driven by dephosphorylation of Bad(Ser-112, Ser-155) (277). The decrease in phosphorylated Bad(Ser-112, Ser-155) results in its decreased binding to 14-3-3 followed by its increased association with mitochondrial Bcl2, resulting in Bax displacement and PTP gating. Decrease in phosphorylated Bad by MEDICA analogs is due to suppression of the adenylate cyclase/PKA and the MAPK/RSK1 transduction pathways, and their respective downstream Bad(Ser-155) and Bad(Ser-112) targets (276, 277). Hence, the TH and MEDICA transduction pathways converge at their downstream Bax target but diverge upstream of the Bcl2-Bax heterodimer (Figure 3). LC-PTP gating by MEDICA analogs may account for their thyromimetic calorigenic activity in vivo.

Figure 3

PTP gating by MEDICA analogs.

IV. Conclusion

TH-induced calorigenesis and thermogenesis have long been recognized to reflect uncoupling of mitochondrial oxidative phosphorylation. However, the mode of action of TH in promoting mitochondrial uncoupling remained elusive. PTP gating by TH may offer a unified target of TH in inducing some of its pleiotropic activities. TH-induced LC-PTP gating may account for TH action in decreasing metabolic efficiency and in protecting against I/R damage, whereas TH-induced HC-PTP gating may result in apoptosis. Mitochondrial LC-PTP gating may serve as a novel target for synthetic functional thyromimetics designed to modulate metabolic efficiency. PTP gating by synthetic LCFA analogs may serve as a model for such strategy.

Acknowledgments

This study was funded by the Israeli Science Foundation (ISF).

Disclosure Summary: J.B.-T. is director in SyndromeX, a company that develops prospective drugs for the metabolic syndrome. E.Y.-S. and B.K. have nothing to declare.

Abbreviations

• ANT

  adenine nucleotide translocase

• BAT

  brown adipose tissue

• [Ca⁺²]c

  cytosolic Ca⁺²

• CsA

  cyclosporin A

• CypD

  cyclophilin D

• 2,4-DNP

  2,4-dinitrophenol

• ER

  endoplasmic reticulum

• FFA

  free fatty acid

• G3P

  glycerol-3-phosphate

• G3PDH

  G3P dehydrogenase

• HC-PTP

  high-conductance PTP

• IMM

  inner mitochondrial membrane

• IPC

  ischemic PC

• IP3R1

  inositol 1,4,5-trisphosphate receptor 1

• I/R

  ischemia/reperfusion

• LCFA

  long-chain fatty acid

• LC-PTP

  low-conductance PTP

• MEDICA

  methyl-substituted dicarboxylic acid

• mG3PDH

  mitochondrial G3PDH

• NRF

  nuclear respiratory factor

• OMM

  outer mitochondrial membrane

• PC

  pre/post conditioning

• PGC-1α

  PPAR-γ coactivator-1α

• PKA

  protein kinase A

• PKC

  protein kinase C

• PMF

  proton motive force

• PPAR

  peroxisome proliferator-activated receptor

• PP2B

  protein phosphatase 2B

• PPIase

  peptidyl-prolyl cis-trans isomerase

• PTP

  permeability transition pore

• ROS

  reactive oxygen species

• SERCA

  sarcoplasmic/endoplasmic Ca⁺²-ATPase

• SfA

  Sanglifehrin A

• TH

  thyroid hormone

• THR

  TH receptor

• TRE

  TH-response element

• UCP

  uncoupling protein

• VDAC

  voltage-dependent anion channel

• VMH

  ventromedial nucleus of the hypothalamus.