Thyroid Hormone Signaling Pathways: Time for a More Precise Nomenclature

Abstract

Current literature makes a distinction between two pathways for thyroid hormone signaling: genomic and nongenomic. However, this classification is a source of confusion. We propose a clarification in the nomenclature that may help to avoid unproductive controversies and favor progress in this field of research. Four types of thyroid hormone signaling are defined, and the experimental criteria for classification are discussed.

A classification of thyroid hormone signaling into four distinct types is proposed, to replace the confusing “genomic” vs “nongenomic” designation.

Early studies established that thyroid hormones (THs) exert a host of physiological effects by regulating gene expression (1). After the cloning of two genes, now called THRA and THRB (2, 3) (in humans, or Thra and Thrb in mice) encoding TH nuclear receptors (THRs), these were recognized by sequence and domain homology to belong to the nuclear receptor superfamily. This superfamily now contains 48 human genes, which encode a number of ligand-dependent transcription factors that bind directly to DNA. On the other hand, a number of reports have suggested that THs might act through alternative pathways that do not require direct interaction between THRs and DNA. These observations have led to the proposal that two mechanisms of action exist, respectively called “genomic” and “nongenomic.” However, in the light of recent data, this distinction seems rather simplistic. Accordingly, we propose here to adopt a new nomenclature to classify THs’ action, to indicate more clearly the molecular pathways that are involved.

Current Confusion

It is well established that THs, including 3,3′,5-triiodo-l-thyronine (T3) and thyroxine (T4), act by binding to several THR isoforms and subtypes (TRα1, TRβ1, and TRβ2, encoded by the THRA and THRB genes). T3, which has higher affinity for THRs, is considered the active form of the hormone and T4 mainly its precursor. THRs have a modular structure, with an N-terminal DNA-binding domain and a C-terminal ligand-binding domain. They can bind DNA as homodimers or, most commonly, as heterodimers with retinoid X receptor (RXR). These dimers recognize specific DNA sequences called thyroid response elements (TREs). TREs are composed of two half-sites, most often organized as direct repeats, separated by four nucleotides [consensus DR4: 5′(A/G)GG(A/T)CANNNN(A/G)GG(A/T)CA 3′]. The 5′ half-site of DR4 is bound by RXR and the downstream 3′ half-site by THR. Both unliganded and liganded THRs can bind to TREs. However, T3 binding changes the conformation of the ligand-binding domain, favoring the recruitment of transcription coactivators on chromatin at the expense of corepressors. By directly recruiting other proteins or by generating posttranslational modifications on histone tails such as acetylation and methylation, coactivators and corepressors influence the rate of transcription initiation of proximal target genes.

This canonical model, in which THRs behave as ligand-dependent transcription factors, is supported by abundant experimental evidence and is essentially identical to the mode of action accepted for the other nuclear hormone receptors (4). However, a longstanding body of research suggests that this model is incomplete and supports the existence of additional mechanisms of action. In particular, studies have shown that TH can sometimes act within minutes, a period that seems too rapid for a transcription-mediated response. In addition, T4 can be more potent than T3 for some of these effects. These alternate pathways are usually called “nongenomic pathways,” a term that covers numerous and controversial possibilities. One problem with this designation is that it is a misleading, as it seems to imply that whenever TH acts through a nongenomic pathway, gene expression is not changed. However, over the long term, an initial nontranscriptionally mediated change in cellular physiology could lead to changes in gene expression. Therefore, only the initial action of TH should be called nongenomic. Furthermore, using the term “nongenomic” for the proposed direct action of T3 on mitochondrial transcription (5) is certainly inappropriate, as mitochondria possess a genome. Finally, the rapidity of TH action cannot be easily used to make a distinction between nongenomic and genomic signaling because rapid posttranslational change, such as THR phosphorylation, can quickly influence gene expression mediated by chromatin-bound THRs.

We thus believe that a clarification in the nomenclature will help to avoid unproductive controversies and favor research progress in TH action. Without discussing the reality or physiological relevance of the multitude of alternate signaling pathways, we propose to follow some simple rules to classify the various models of TH signaling. Our goal is not to impose a dogmatic view on how T3 signaling operates but to propose a simple framework that can evolve with new scientific advances in knowledge as well as help nonspecialists to better understand future progress in a promising field.

Proposed TH Signaling Classification

First, we do not consider any mechanism that is only influencing the metabolism of TH or its transport across the plasma membrane or within the cell. Although in some pathological or physiological situations, the synthesis, degradation, and delivery of TH can be limiting steps; they are not part of the signaling process per se. Second, we will only consider T3 and T3 mimics. These can be natural hormones made by the thyroid gland, some close derivatives and metabolites (e.g., triiodothyroacetic acid), or synthetic analogues designed to act in a similar way. Although more distantly related iodinated compounds, such as 3-iodothyronamine, possess biological activity, they appear to act on distinct pathways that are not mediated by THs. Within this field, four possible modes of action can be considered (Fig. 1).

Figure 1.

The four types of thyroid hormone action. Type 1: THRs (blue) bind DNA as monomer, homodimer, or heterodimer, usually with RXR (orange). This requires the presence of the N-terminal DNA-binding domain (D). TH binding in the C-terminal ligand-binding domain (L) changes the conformation of the protein and interactions with transcription cofactors. Type 2: THRs are tethered to DNA by other proteins or multiprotein complexes (green). The capacity of THR to bind DNA is dispensable. Type 3: THRs exert their function without being recruited to chromatin. Localization can be cytoplasmic or nuclear. Type 4: TH acts independently of THRs, binding to other types of proteins.

Type 1: THR-dependent signaling of TH with direct binding to DNA

This corresponds to the canonical model described previously. Although a single pathway is currently described, one may envision other subdivisions here. This could occur if later investigations clearly demonstrate that negative regulation of gene expression by liganded THR, which is well illustrated in vivo by the TSHb gene, involves different modes of action for DNA-bound TRs. Several studies have suggested different types of THR/DNA association: (1) monomer or homodimer binding to response elements, (2) THR binding to enhancer elements in which half-sites are organized as an everted repeat with a six-nucleotide spacer or an inverted repeat without a spacer, and (3) THR binding with other heterodimer partners such as RAR (6). Recent genome-wide analyses of chromatin occupancy (7–9) have confirmed that the THR/RXR/DR4 combination is the predominant combination for THR binding to TREs, without ruling out alternative possibilities. Finally, DNA binding by THR has been proposed to have other consequences than a change of proximal gene transcription. Liganded THR could act as “pioneer” factor, opening chromatin access to other transcription factors (9), or exert a general control of gene expression by influencing the capacity of CTCF to insulate independent chromatin domains (10). Also often called “nongenomic,” the proposed influence of the p43 protein, encoded by THRA, on mitochondrial transcription fits in this type (5).

Type 2: THR-dependent signaling of TH with indirect binding to DNA

TRs can interact with a number of chromatin-associated proteins, which can tether THRs to specific genomic locations, even if THRs do not directly contact DNA. Although direct biochemical evidence for protein-protein interactions is lacking, older data support the view that the T3-bound THRs can modulate the function of AP1 (jun/fos) complexes (11, 12) without binding to DNA. Recent ChipSeq analysis also indicates that beside the DR4 consensus element, consensus for other transcription factors is found where T3 induces chromatin remodeling (9). This raises the possibility for tethering of THRs by other transcription factors, without direct interaction between DNA and THR.

Type 3: THR-dependent signaling of TH without DNA binding

Several studies suggest that cytoplasmic THRs can interact with kinases normally found at the plasma membrane, activate them, and serve as a substrate for phosphorylation. THs and THRs could thus participate in signaling pathways, like the PI3K pathway, independently of direct or indirect binding to DNA (13–16). This role has also been attributed to the p30 protein, which is a transcriptionally incompetent protein that is translated from an internal AUG codon of the TRα1 messenger RNA (17).

Type 4: THR-independent TH signaling

Iconoclastic reports propose that TH can act without binding to THR. Integrin αVβ3 may serve as a membrane receptor of T4 and T3, notably in the context of cancer (18–20). TH can also influence actin in vitro polymerization by some unknown mechanism (21). Another intriguing possibility would be that T3 simply acts as an allosteric regulator of a metabolic enzyme. This has been suggested for Crym (a cytoplasmic protein also called μ-crystallin) NAD phosphate–regulated TH-binding protein, and cytosolic 3,5,3′-T3-binding protein, which was recently found to possess an enzymatic activity, catalyzing the reduction of sulfur-containing cyclic ketimines (22).

Experimental Criteria to Classify the Mode of Action of TH

Full demonstration that TH acts on a given signaling pathway requires the accumulation of convergent evidence. In particular, attributing a specific type of cellular TH response to one of the four types defined previously should combine genetic and biochemical support as well as in vitro and in vivo evidence. In any case, the active compound (usually T3) should be >95% pure and used at a realistic concentration. These parameters should also be carefully considered when studying putative TH disruptors, which are environmental chemicals that interfere with TH signaling.

Type 1

Biochemistry

It is difficult to demonstrate that the transcription of a gene is directly regulated by DNA-bound THR. Electromobility gel shift assays are often used to define TREs. Although these assays are reliable to evaluate the capacity of THR/RXR heterodimers to recognize specific elements in vitro, bioinformatic analyses indicate that there are more than 70,000 putative TREs in the human genome. This largely exceeds the number of sites occupied in chromatin, as defined by ChipSeq analysis, which rarely exceeds a few thousand, depending on protocol sensitivity and cell type. Therefore, the presence of a consensus sequence next to a T3-responsive gene is only a weak indication that the transcription of this gene can be regulated by THRs. Although ChipSeq analysis provides key information, interpreting this type of data is problematic. Because chromatin-bound THRs can act at a long distance, the absence of a THR binding site within 20 or 30 kb of the transcription start site of a T3-regulated gene does not rule out type 1 mode of action. Reciprocally, and more surprisingly, the presence of THR on proximal regulatory sequences is often observed for genes which expression is not sensitive to T3. Therefore THR binding does not always influence the expression of proximal genes. A growing number of other assays have been developed, which can be performed at a genome-wide scale that will help to better understand the direct consequences of THR recruitment to the chromatin: analysis of histone marks and enhancer RNAs (23), genomic run-on (GRO-seq) (24), etc.

Genetics

Transient expression assays are routinely used to demonstrate that an interaction between THR and TRE is required for T3 response of reporter constructs. However, it is noteworthy that transfected DNA is poorly chromatinized, so the physiological relevance of these assays is limited. In vivo, amino acid substitutions have been introduced to prevent THR-DNA interaction (25, 26), which are expected to fully inhibit type 1 response. Whether these mutations only prevent DNA binding or alter the properties of THRs in some unexpected manner, changing some uncharacterized protein-protein interaction, remains to be determined. Mutating a TRE by CRISPR/Cas9 genome editing is currently the best way to assess the functionality of TRE in a genomic context (27).

Type 2

Biochemistry

Indirect recruitment of THR to chromatin, by binding to other DNA-bound proteins, should be detectable by ChipSeq analysis because the method involves proteins cross-linking. Such recruitment is more likely to occur at genomic binding sites where motif search cannot identify a consensus TRE-like sequence. In contrast to formaldehyde, the use of ultraviolet crosslinking, which induces covalent protein–DNA adducts and does not crosslink proteins, might be useful to distinguish between direct and indirect interaction of THR with DNA (28). The protein complexes involved in tethering could be identified by coimmunoprecipitation, ligand proximity assay, or other biochemical means.

Genetics

Type 2 response should be maintained in mice in which THR has been mutated to impair DNA binding and reduced if amino acid residues necessary for nuclear localization are mutated. Knocking out or knocking down genes encoding interacting factors should also inhibit this type of response. In vitro interaction studies may identify critical residues on both THR and its tethering partner, which could be substituted for a complete genetic demonstration that the identified interaction is of importance. There is currently no compelling evidence for the existence of type 2 response.

Type 3

Biochemistry

PI3K inhibitors have been used to uncouple type 1 and type 3 response (15). Most hypotheses that support the possibility for THR acting without binding to DNA are based on the presence of a substantial fraction of the protein in the cytoplasm or at the plasma membrane. Thus far, THRs have been convincingly demonstrated at these locations by immunocytochemistry, fluorescence microscopy using GFP-THR fusion proteins (29), or biochemical fractionation. Because this has only been done in cells transfected to overexpress THR, results should be interpreted with caution. Overexpressed THRs could saturate the capacity of cells to transport proteins into the nucleus and lead to accumulation in compartments outside the nucleus. Detecting endogenous THRA- and THRB-encoded proteins that are less abundant in the cells appears to be challenging. Because different antibodies have given conflicting results, it is prudent to use cells from knockout mice as controls to ascertain signal specificity (30).

Genetics

According to Martin et al. (14), the TRβ^Y147F mutation prevents the class 3 response and preserves the class 1 response. Indeed, TRβ^Y147F acts like wild-type TRβ on a DR4 TRE in vitro, whereas the circulating level of thyroid-stimulating hormone is normal in TRβ^Y147F/+ mice. However, the mutated tyrosine is located in the C-terminal zinc finger of the DNA-binding domain. This raises the possibility that, although still present, class 1 response could be modified by the mutation. This possibility could be addressed by transcriptome analysis.

Type 4

Biochemistry

The key observations that support the existence of the type 1 pathway are (1) the high affinity binding of TH to purified THR, later confirmed by X-ray analysis of the ligand-binding structure, which revealed the presence of the ligand-binding pocket, and (2) a similar hierarchy for the affinity of various ligands for TR (triiodothyroacetic acid >T3 > T4 > reverse T3) and their biological activity. The existence of the type 4 pathway should be supported by similar biochemical evidence, although the ligand hierarchy could be different. Affinities are expected to be low; otherwise, receptors responsible for type 4 response might have been identified in the precloning era, but this could be compensated by protein abundance. TH actually displays weak binding affinity for several cytoplasmic proteins. It has been difficult to determine whether these proteins serve for transport and storage or initiate signaling events, as exemplified by Crym. TH has also been found to interact with proteins located on the outer cellular membrane. The main experimental evidence supporting the possibility that TH can act independently of THR through integrin αVβ3 is as follows: (1) T4 has similar potency as T3 in some cases; (2) beads coated with T4, which do not enter into cells, can still trigger a response; (3) antibodies raised against αVβ3 RGD peptide inhibit the effect of T4 on mitogen-activated protein kinase phosphorylation (31); and (4) small interfering RNA knockdown of messenger RNA encoding integrin αVβ3 inhibits this response. Although small interfering RNA knockdown does not result in full elimination of the integrin and may have side effects, it was preferred to gene knockout, which seems hardly compatible with cell survival.

Genetics

Mice devoid of TR should provide a useful resource to better substantiate type 4 response. In a recent unbiased study of primary neurons prepared from Thra/Thrb double knockout mice, T3 treatment did not affect gene expression (32). This argues against any transcriptional consequence of type 4 response.

Limitations

Determining to which type a TH response belongs can be technically challenging in a number of situations. Moreover, TH response may not be classified with certainty because different modes of action may be coupled. In many instances, keeping ambiguity is acceptable as long as unsubstantiated claims for mechanism of action are not made. We believe that referring to the four types of TH action will nevertheless provide a clear benefit over the simplistic genomic/nongenomic distinction. We also hope that a similar effort will be made for other nuclear receptors because confusion also exists for the signaling pathways activated by ligands for these receptors.

Abbreviations:

RXR

retinoid X receptor

TH

thyroid hormone

THR

thyroid hormone nuclear receptor

TRE

thyroid response element

T3

3,3′,5-triiodo-l-thyronine

T4

thyroxine.