Mislocalization of Cancer-associated Thyroid Hormone Receptor Mutants

Abstract

The thyroid hormone receptor (TR) is essential for the proper regulation of metabolism and development, as it regulates gene expression in response to thyroid hormone. Nuclear localization signals (NLSs) and nuclear export signals (NESs) allow for TR transport into and out of the nucleus, respectively. Previous research suggests that nuclear import, nuclear retention, and nuclear export of TR are associated with modulation of gene expression, the alteration of which can contribute to various diseases. Here, we examined the impact of cancer-associated mutations on TR localization patterns as a way of analyzing key structural components of TR and to further explore the correlation between TR trafficking, misfolding, and disease. Through mammalian cell transfection of expression plasmids for green fluorescent protein (GFP) and mCherry-tagged TRα1 and quantitative fluorescence microscopy, we examined particular groups of TRα1 mutations that were observed in patients with hepatocellular carcinoma, renal cell carcinoma, and thyroid cancer, and are associated with NLSs and NESs of TRα1. We also investigated structural alterations of the mutants by in silico modeling. Our results show striking shifts towards a more cytoplasmic localization for many of the mutants and an increased tendency to form cytosolic and nuclear aggregates.

1. Introduction

The thyroid hormone receptors (TRs) are members of the nuclear receptor superfamily that play a key role in human health. By mediating thyroid hormone (T3) action in numerous tissues, they regulate important physiological, developmental, and cellular processes; including regulation of cell proliferation, differentiation, and apoptosis. Their most well-characterized role is as T3-dependent transcription factors; TRs bind thyroid hormone response elements (TREs) in the presence or absence of T3 to facilitate the expression of target genes, often as heterodimers with the retinoid X receptor (RXR) [1–9]. On positive TREs, corepressors, such as nuclear receptor corepressor 1 (NCoR1) and histone deacetylase, are bound to TR in the absence of T3, leading to repression of target gene expression. Upon T3 binding, TR undergoes a conformational change, resulting in binding of coactivators such as SRC-1, histone acetyltransferase, and MED1, a subunit of the Mediator complex that functionally bridges TR with the general transcription apparatus. Interactions of TR with coactivators lead to changes in chromatin structure and the subsequent transcription of the target gene [10–13].

The TRs are highly dynamic proteins. Precise targeting from synthesis in the cytosol to their ultimate destination in the cell is essential for receptor function [14–17]. Although they primarily reside in the nucleus, we and others have shown that the major TR subtypes, TRα1 and TRβ1, shuttle rapidly between the nucleus and cytoplasm [18, 19], suggesting that the balance between nuclear import and export of TRs provides another level of control for modulating T3-responsive gene expression. Proteins destined for the nucleus cross the nuclear envelope through the nuclear pore complexes (NPCs), facilitated by transport proteins [20, 21]. We have identified two nuclear localization signals (NLSs) and multiple nuclear export signal (NESs) within TRα1 and TRβ1 that interact with importins and exportins, respectively, to mediate translocation through the NPCs [22–26]. TR mutations have been identified in patients with hepatocellular carcinoma, renal clear cell carcinoma, and cancers of the breast, pituitary, and thyroid [5, 27–34], raising the question of whether mislocalization of TR correlates with pathogenesis. In addition, we have observed the propensity of TR variants to form cytoplasmic aggregates, exemplified by the oncoprotein v-ErbA, a dominant negative mutant of TRα1 [15, 16, 18, 35]. Further, we have shown that two synthetic dominant negative mutants of TRβ1, with defects in DNA binding and transactivation, accumulate in aggregates in the cytoplasm at steady state, illustrating that even single amino acid changes in functional domains may lead to a marked shift in the intracellular distribution of TR [18].

Here, we sought to ascertain the impact of mutations of TR associated with hepatocellular carcinoma, renal cell carcinoma, and papillary thyroid cancer on the receptor’s localization patterns. To this end, a panel of TR mutants was generated, based on sequences from patient samples reported in the literature. To explore, in particular, the correlation between TR trafficking, misfolding, and disease-association, we selected mutations that clustered within NLS and NES motifs. The effects of these mutations on intracellular localization were assessed using HeLa (human) cell transfection as a model cell system, quantitative fluorescence microscopy, and in silico modeling to analyze key residues of TR.

2. Materials and Methods

2.1. Plasmids

Human wild-type TRα1, mutant coding regions, and NES domain plasmids were designed using GeneArt® gene synthesis services (Life Technologies) and ligated into GFP or mCherry expression vectors (Clontech), or a maltose binding protein (MBP)-mCherry-Hinge expression vector (for the NES constructs). The Hinge domain contains NLS-1 of TRα1. GFP-v-ErbA, GFP-250, and GFP170* plasmids were as previously described [16].

2.2. Cell Culture and Transfection

HeLa cells (ATCC® CCL2^™) were grown in Gibco^™ Minimum Essential Medium (MEM) with 10% fetal bovine serum (FBS) at 37°C, 5% CO₂, and 98% humidity. Cells were seeded at a density of 2.0–2.5 × 10⁵ cells per well onto coverslips in a six-well plate overnight, and transfected with expression plasmids using either Lipofectamine 2000 [23] or Lipofectamine 3000, according to the manufacturer’s instructions.

2.3. Analysis of Nucleocytoplasmic Distribution

Approximately 18–24 h post-transfection, cells were fixed in 3.7% formaldehyde and coverslips were mounted in Fluoro-Gel II with DAPI (Electron Microscopy Sciences) [23]. Slides (blinded by another lab member) were scored on an inverted Nikon fluorescence microscope. For semi-quantitative analysis (NES characterization), cells were scored based on observation of their red fluorescence distribution patterns and assigned a score of whole cell/cytoplasmic or primarily nuclear (n=200 cells per independent replicate). The percentage of cells with a particular distribution pattern per replicate was calculated and the mean of these percentages was used. Nuclear export function of NESs was measured as a significant decrease in the average percent of cells with a primarily nuclear distribution relative to the MBP-mCherry-Hinge control.

For quantitative analysis (cancer-associated mutations), Nikon® NIS-Elements microscope imaging software was used. Two identical Regions of Interest (ROIs) were drawn in order to measure the fluorescence intensity of two areas: one in the nucleus and one in the cytoplasm. Cells with aggregates were not excluded; instead, ROIs were positioned such that they provided fluorescent intensities representative of the whole cell. Nuclear-to-cytoplasmic fluorescence ratios (N/C) were calculated and normalized to baseline conditions for corresponding biological replicates. A normalized N/C greater than 1.0 was interpreted as having a more nuclear distribution than wild-type TR, while a normalized N/C less than 1.0 indicated a greater distribution of TR in the cytoplasm. In addition, the presence of cytoplasmic or nuclear aggregates was assessed qualitatively, to determine the percentage of cells with or without aggregates.

Four main phenotypes were defined as categories for scoring aggregates: 1) “Smooth” refers to a homogeneous distribution of TR within a cell, without aggregates; 2) “Aggregates in both the nucleus and cytoplasm,” describes cells with aggregates in both cellular compartments; 3) “Nuclear aggregates” describes cells which have aggregates clearly in the nucleus either by fluorescent delineation or by overlap with the DNA stain DAPI; and 4) “Cytoplasmic aggregates” encompasses a whole cell distribution of aggregates.

For each mutant, at least three separate, biologically independent replicates were completed with a total of 100 cells scored per slide. A student’s t-test was used to calculate p values and determine statistical significance. p<0.05 was considered significant.

2.4. Analysis of Colocalization by Confocal Microscopy

Confocal microscopy was used to image aggregates and aggresomes in three dimensions. The channel lock was engaged so that any bleed-through between channels would not be captured. The lasers for each respective channel were optimized so that pixel oversaturation did not occur. The pinhole was set to 1 Airy unit. These image planes were then combined and analyzed using Nikon Elements Colocalization Tool. All data and images were collected using Nikon Elements. Pearson’s Correlation Coefficient (PCC) was used to assess the relationship between the fluorescence intensities of the two fluorophores. A standard threshold of significant correlation, r ≥ 0.5, was used as a measure of significant colocalization [36]. ROI’s were manually generated in merged GFP and mCherry images. Three separate, biologically independent replicates were completed with 30 cells per slide imaged. For aggregate analysis, each cell contained approximately five aggregates that were scored.

2.5. Structural Analysis In Silico

Amino acid sequences of wild type and mutant TRα1 were entered into the I-TASSER server [37–39], to calculate the normalized B-factors, which predict the stability of each amino acid position [40]. The sequences were then grouped based on domain and an average normalized B-factor was calculated for each domain, which gave relative stability of the domain. To determine whether stability was significantly affected, a paired student’s t-test was performed to compare each mutant domain to the wild type.

In addition, the mutants and wild type receptors were visualized using UCSF Chimera [41]. The LBD is the most well-characterized domain of TRα1. A homology study was performed with Chimera, by utilizing X-ray crystallographic LBD structures (PDB:1NAV). The conserved DNA-binding domain (DBD) with two zinc fingers was modeled based on homology with other members of the nuclear receptor superfamily (RXR, VDR, and GR). This template model (T M) was uploaded to the interactive threading assembly refinement (I-TASSER) server [37–39] and the best-fit model was calculated for wild-type TRα1. Network analysis was conducted to assess the interaction between the NES and NLS and other residues within TRα1. Confidence (C) and template modeling (T M) scores were calculated to determine the accuracy and quality of the predicted structures.

3. Results and Discussion

3.1. Characterization of NES-H3 and NES-H6 in TRα1

As a first step to analyzing cancer-associated mutations in TRα1’s NLS and NES motifs, we needed to more precisely characterize regions with export activity in the ligand-binding domain (LBD). In addition to the two NLSs in TRα1, two characterized nuclear export signals, NES-H12, located in helix 12 within the LBD, and NES-H3/H6, located in the helix 3 to helix 6 region of the LBD, have significant effects on nucleocytoplasmic shuttling of TRα1 [25] (Fig. 1A). Previously, we characterized a region in helix 3–6,spanning residues 209–265, that is necessary and sufficient for TR nuclear export [25]; however, this region is much larger than known classical NESs and NES-H12. This size disparity led us to investigate whether this region contains a single NES, or multiple NES motifs, in addition to identifying a minimal sequence sufficient for export of a heterologous nuclear protein. While our previous work showed that helices 3 and 6 do contain some export functionality necessary for TR export, the regions of interest examined in these studies possessed overlapping residues, and did not establish minimal independent regions sufficient for export activity [25].

Figure 1.

NES-H3 and NES-H6 of TRα1 are independently-acting NES motifs. A) Diagram (not to scale) of domains of TRα1: DBD, DNA-binding domain; LBD, ligand-binding domain. The inset shows the non-overlapping regions of NES-H3 and NES-H6 that were added to the MBP-mCherry-Hinge nuclear fusion protein to test for their export activity. B) Cells were transfected with expression plasmids for MBP-mCherry-tagged fusion proteins as indicated, fixed, stained with DAPI to visualize the nucleus, and scored for intracellular distribution by fluorescence microscopy. Scale bar = 10 μm. C, D) Bars indicate the percentage of cells with primarily nuclear distributions. Error bars represent ±SEM (n=4–5 independent, biologically separate replicate experiments, with >200 cells scored per replicate); *p<0.05; n.s., p>0.05.

Expression plasmids were designed containing maltose binding protein (MBP)-mCherry, to create a fusion protein too large to diffuse through the nuclear pore complexes; the TRα1 hinge region NLS-1, to allow nuclear import; and each NES motif of interest (Fig. 1A). The motifs of each helix were designed to exclude the proline and methionine residues located between the helix 3 and helix 6 regions, and create two non-overlapping fusion proteins containing their respective putative NESs. Expression plasmids were then transfected into HeLa cells and the intracellular fluorescence distribution patterns of the MBP-mCherry-Hinge-NES fusion proteins were compared with the MBP-mCherry-Hinge fusion protein. As a control, to ensure that addition of LBD residues did not simply interfere with nuclear import, we also analyzed the distribution of a fusion protein containing residues 188 to 206, adjacent to NES-H3 (MBP-mCherry-Hinge-LBD-188-206).

As expected, MBP-mCherry-Hinge and the control, MBP-mCherry-Hinge-LBD-188-206, were localized primarily to the nucleus by NLS-1, although the distribution of the LBD construct was more variable. In contrast, fusion proteins with NES-H3 or NES-H6 showed a whole cell distribution (Fig. 1B). There was a significant decrease in the percent of cells with a primarily nuclear distribution for both NES-H3 (p=0.028) and NES-H6 (p=0.030), when compared with the nuclear-localized Hinge control (Fig. 1C); whereas, the percent of cells with a primarily nuclear distribution of LBD-188-206 was not significantly different from the Hinge control (p>0.05) (Fig. 1D). These data indicate that both NES-H3 and NES-H6 are sufficient to individually target a heterologous nuclear protein to the cytoplasm, albeit inefficiently. Based on these findings, along with our prior studies showing that the H3-H6 region is necessary for TR export [25], we conclude that there are two independent, non-overlapping motifs within the helix 3 to helix 6 region with nuclear export activity. Having further refined characterization of the NESs in the LBD of TRα1, we then sought to ascertain whether cancer-associated mutations in regions housing localization signals alter the intracellular distribution patterns of TRα1.

3.2. Cancer-associated Mutants of TRα1 with Wild-type Localization

Aberrant expression of TRα1 has been observed in tumor tissue of patients with papillary thyroid carcinoma [42]. Along with altered levels of both mRNA and protein, there also exists a high incidence of somatic mutations that alter the amino acid sequence of TRα1 [42]. Here, we focused on two papillary thyroid carcinoma-associated mutations that cluster within NES-H3 in the LBD of TRα1. One patient-derived mutant contained three amino acid substitutions: serine to asparagine at position 183 (S183N), histidine to glutamine at position 184 (H184Q), and arginine to histidine at position 228 (R228H) within NES-H3 (Fig. 2A, B). The other papillary thyroid carcinoma mutant examined contained a substitution of glutamate to aspartate at position 213 (E213D), also within NES-H3. We used our well-validated approach of transient transfection assays in HeLa (human) cells to compare localization patterns of GFP-tagged wild-type and mutant TRα1. Our prior studies show that expression of exogenous wild-type TRα1 in HeLa cells does not saturate the nuclear import process; wild-type TRα1 is predominantly nuclear over a wide range of expression levels in transfected cells. Further, overexpressed GFP-TRα1 does not typically form protein aggregates, and the fluorescent protein tag does not alter distribution patterns [15, 16, 18, 23, 43].

Figure 2.

Thyroid cancer-associated mutants TRα1(S183N, H184Q, R228H) and TRα1(E213D) are primarily localized to the nucleus. A) TRα1 functional domains and mutation sites. B) I-TASSER protein prediction modelling, visualized in UCSF Chimera showing the structural domains of TRα1, for residues 52–410. The unstructured N-terminal A/B domain is not shown. NLS-1, NES-H3/H6, and NES-H12 are colored in red. C, F) HeLa cells transfected with expression plasmids for wild-type GFP-TRα1, GFP-TRα1(S183N, H184Q, R228H), or GFP-TRα1(E213D) as indicated, were analyzed by quantitative fluorescence microscopy after fixation and staining with DAPI to visualize the nucleus. Scale bar = 10 μm. D, G) Bars indicate relative N/C for mutant TRα1, normalized to the N/C ratio of wild-type. Error bars indicate ±SEM (n=3 independent, biologically separate replicate experiments, with 100 cells scored per replicate); n.s., p>0.05. E) Alignment of wild type TRα1 with mutant TRα1(S183N, H184Q, R28H). H) TRα1(E213D) mutant (blue) shows a distinct perturbation of the α-helix starting at I221, with two stabilizing contacts with L276 and M280. Wild-type (white) only interacts with L276 (not modelled).

Here, HeLa cells were transfected with GFP-tagged wild-type TRα1, GFP-TRα1(S183N, H184Q, R228H) (also termed tc15-TRα1 [42]), and GFP-TRα1(E213D), and the relative nuclear to cytoplasmic fluorescence ratio (N/C) was quantified by fluorescence microscopy. As expected, wild-type TRα1 showed a primarily nuclear localization with a smooth distribution (Fig. 2C, F). The distribution patterns of TRα1(S183N, H184Q, R228H), and TRα1(E213D) were comparable to wild-type TRα1, although there was variability between replicates. On average, both mutants showed a primarily nuclear localization (normalized N/C of approximately 1.0; p=0.2966 and p=0.9011, respectively) (Fig. 2D, G). A structural examination of TRα1(S183N, H184Q, R228H) revealed nearly identical alignment of the NLS-1, NES-H3/H6 and NES-H12 regions compared to the wild-type (Fig. 2E), with some minor deviation in unstructured regions (C-score = 0.57; T M-score = 0.79), reinforcing how structure dictates function.

While E213D is located in NES-H3, both glutamate and aspartate are acidic with negatively charged side chains, suggesting that the structure of the NES may not be affected enough to change its function. However, structural analysis reveals that TRα1(E213D), which is primarily nuclear-localized, facilitates structural changes within the NES-H3 through destabilizing properties, despite no interactions with nearby residues (Fig. 2H). There is a distinct perturbation of the α-helix starting at residue I221. Further analysis of position I221 in the mutant shows two stabilizing contacts with residues L276 and M280, while the wild type only has an interaction with L276 (not modelled) (C-score = 0.57; T M-score = 0.79).

Similarly, S183N represents substitution of one polar amino acid for another and, although serine residues may be subject to post-translational modification, S183 is not a known phosphorylation site in TRα1 [44]. The R228H substitution represents a change from one basic amino acid for another and occurs within the solvent-exposed internal position of NES-H3 within the α-helix (see Fig. 2E). Although the histidine residue in the cancer-associated mutant makes no contacts with other nearby residues, it forms hydrogen bonds with the polypeptide backbone, indicating that there is local unfolding of the region near position 228. Finally, although the H184Q substitution is a switch from a positively charged side chain to a polar side chain, glutamine would maintain the ability to form hydrogen bonds, suggesting there would be little impact of these mutations on overall structure near NES-H3.

Functionally impaired TRs have been proposed to play an important role in tumorigenesis, through both aberrant expression and somatic mutations. When Puzianowska-Kuznicka et al. [42] tested a panel of TRα1 mutants derived from papillary thyroid carcinoma tumor tissue for their transcriptional activity in a TRE-luciferase reporter assay, they observed weak activity for all mutants. Of relevance for the present study, TRα1(S183N, H184Q, R228H) had ~20% of the activity of wild-type TRα1, while TRα1(E213D) had ~40% of the transactivation ability relative to wild-type TRα1 [42]. Here, we showed that these mutants have wild-type nuclear localization, however, indicating that in this case, receptor mislocalization is not a contributing factor to impaired T3-mediated gene regulation and cancer progression.

3.3. Cancer-associated Mutations in TRα1 NLS-2 Lead to a More Cytoplasmic Localization

TRα1 mutations with impaired T3 and DNA binding are prevalent in tumors of patients with hepatocellular carcinoma, suggesting that TRα1 mutants play a role in liver carcinogenesis [45]. We thus examined a TRα1 mutant originally isolated from a patient with hepatocellular carcinoma that contained three amino acid substitutions: glycine to glutamate at position 24 (G24E) in NLS-2, methionine to valine at position 256 (M256V) in NES-H6, glutamate to alanine at position 343 (E343A), and proline to leucine at position 269 (P269L) [45] (Fig. 3A). To investigate further the impact of mutations in NLS-2 and NES-H6, we also constructed single mutants for G24E and M256V.

Figure 3.

The distribution pattern of hepatocellular carcinoma-associated mutant TRα1(G24E, M256V, E343A, P269L) is shifted towards the cytoplasm. A) TRα1 functional domains and mutation sites. B) HeLa cells transfected with expression plasmids for wild-type GFP-TRα1, GFP-TRα1(G24E, M256V, E343A, P269L), GFP-TRα1(G24E), or GFP-TRα1(M246V) as indicated, were analyzed by quantitative fluorescence microscopy after fixation and staining with DAPI to visualize the nucleus. Scale bar = 10 μm. C) Bars indicate relative N/C for mutant TRα1, normalized to the N/C of wild-type. Error bars indicate ±SEM (n=3 independent, biologically separate replicate experiments, with 100 cells scored per replicate); *p<0.05; n.s., p>0.05.

There was a statistically significant difference in the localization pattern of GFP-TRα1(G24E, M256V, E343A, P269L) compared with wild-type GFP-TRα1. On average, the mutant had a greater cytoplasmic distribution relative to wild-type (normalized N/C < 1.0; p=0.011) (Fig. 3B, C). In addition, we found that single-mutant TRα1(G24E) had a greater cytoplasmic distribution, compared to wild-type TRα1 (normalized N/C < 1.0; p=0.0350). In contrast, GFP-TRα1(M256V), with a substitution of one nonpolar amino acid for another, was not significantly different (normalized N/C of approximately 1.0; p=0.4415) (Fig. 3B, C). Finally, there was no significant increase in aggregate formation; wild-type and mutants showed a smooth distribution in the nucleus and cytoplasm.

We know from our prior study that flanking sequence context is key to NLS-2 activity [25], so it is not surprising that G24E, a change from a nonpolar to acidic side chain, has a negative impact on nuclear localization. These data suggest that the disruption of NLS-2 is a driver of cytoplasmic localization for the TRα1(G24E, M256V, E343A, P269L) mutant associated with hepatocellular carcinoma. In addition, although the mutant still binds DNA in vitro, the altered residues within the ligand binding domain have been shown to interfere with T3 binding [45]. Given the important role of TRα1 in transcriptional regulation of T3-mediated genes involved in metabolism, cell differentiation, proliferation, and apoptosis; a combination of a lack of ligand responsiveness and a greater cytoplasmic localization would be detrimental to receptor function, interfering with appropriate regulation of T3-responsive gene expression. Data presented here provide further support for a model in which receptor mislocalization contributes to impaired function of cancer-associated TRα1 mutants [14–16].

3.4. Cancer-associated Mutations at Position 225 in NES-H3 of TRα1 Lead to a Shift to a More Cytoplasmic Localization and Aggregate Formation

We noted that a number of cancer-associated mutations occur at position 225 in TRα1, and thus sought to ascertain if this residue is of particular importance for TRα1 localization. First, we examined a TRα1 mutant originally isolated from a patient with hepatocellular carcinoma [45] that contains three amino acid substitutions: alanine to glycine at position 225 (A225G) in NES-H3, aspartate to asparagine at position 246 (D246N) in NES-H6, and glutamate to lysine at position 350 (E350K) in the LBD (Fig. 4A).

Figure 4.

The distribution pattern of hepatocellular carcinoma-associated mutant TRα1(A225G, D245N, E350K) is shifted towards the cytoplasm with aggregate formation. A) TRα1 functional domains and mutation sites. B) HeLa cells transfected with expression plasmids for wild-type GFP-TRα1 or GFP-TRα1(A225G, D245N, E350K), as indicated, were analyzed by quantitative fluorescence microscopy after fixation and staining with DAPI to visualize the nucleus. Scale bar = 10 μm. C) Bars indicated relative N/C for mutant TRα1, normalized to the N/C of wild-type. Error bars indicate ±SEM (n=3 independent, biological separate replicate experiments, with 100 cells scored per replicate); ***p<0.001. D) Bars indicate the percentage of cells with no aggregates (smooth), aggregates in both the nucleus and cytoplasm, nuclear aggregates, or cytoplasmic aggregates, as indicated. Error bars indicate ±SEM (n=3 independent, biologically separate replicate experiments, with 100 cells scored per replicate); ***p<0.001. E) (i) TRα1(A225G, D246N, E350K). NLS-1 of the mutant (blue) shows significant structural perturbation compared to the wild-type (white) as indicated by the red arrows. Wild-type residues D246 (ii) and E350 (iii) (orange) interact with L333 and W364, respectively. Mutant residue K350 interacts, possibly through charge repulsion, with R157 and W364 (iv), and residue N246 (v) interacts with L242 and P243 in NES-H6 (blue).

We found that the TRα1 (A225G, D246N, E350K) mutant showed a significantly greater cytoplasmic localization compared with wild-type TRα1 (normalized N/C < 1.0; p<0.0001) (Fig. 4B, C). In addition, TRα1(A225G, D246N, E350K) showed a propensity to form aggregates, with a significantly greater number of aggregates in both the nucleus and cytoplasm, compared with wild-type TRα1 (Fig. 4D; p<0.0005). Protein aggregation is a process in which proteins self-associate into non-native and less soluble structures [46]. The observation of no change in the propensity of other mutants (see Figs. 2 and 3) to form aggregates provides further evidence that aggregate formation is not simply a result of expression of exogenous fluorescent fusion proteins.

Although the change from alanine (a small nonpolar amino acid) to glycine (another small nonpolar amino acid) may seem inconsequential, there is evidence suggesting that the stability of a helix with alanine versus glycine varies based on the location of the helix within a protein and on the surrounding amino acids. More specifically, glycine leads to a more stable helix when the helix is at an N-terminal or C-terminal location, but alanine allows for a more stable helix when the helix is located internally within the protein [47]. Since the mutation A225G occurs internal to TRα1, it may be destabilizing to TRα1, contributing to mislocalization and formation of aggregates. The change at position 245 is from aspartic acid (a negatively charged amino acid) with a helix stability score of −2.02 (kcal/mol) to asparagine (a polar amino acid) with a helix stability score of −0.86 (kcal/mol), where negative value indicates the free energy of unfolding associated with stability [47]. This substitution, from a negatively charged amino acid to a polar amino acid, occurs near the solvent-exposed side of the N-terminus of the NES-H6 (Fig. 4E), and is likely to contribute to the altered cellular distribution of TRα1 and promote aggregate formation, since the structure and possible interactions of these amino acids are different. The conversion of glutamate (negatively charged amino acid) to lysine (a positively charged amino acid) at position 350 in the LBD could also be contributing to the formation of aggregates.

TRα1(A225G, D246N, E350K), which tends to form cytoplasmic aggregates, demonstrates an interaction between the LBD and the hinge domain that is facilitated by the K350 amino acid substitution (Fig. 4E). Within the hinge domain, NLS-1 of the mutant shows significant structural perturbation compared to the wild-type. Residues E139 and R157 form polar interactions facilitated by a positive and negative R-group ionization (C-score = 0.80; T M-score = 0.82). In wild type TRα1, residues D246 and E350 make one interaction with residue L333 and residue W364, respectively. In the mutant, residue K350 makes two interactions with residue R157 and residue W364, and residue N246 makes two interactions with residues L242 and P243, resulting in more stable interactions with NES-H3/H6 residues compared to the wild type. Residue 225 (not shown) does not have interactions with NES-H3/H6 in either the wild-type or the cancer mutant (C-score = 0.80; TM-score = 0.82). Since there are interactions between the LBD and NLS-1, this reinforces the idea that point mutants within any domain can influence folding or unfolding of other domains. Notably, position E350K has interactions in both the wild-type and the cancer-associated mutant, although these interactions differ. One possible explanation for this observation is the presence of conserved nearby residues W364 and R157, which could make interactions with amino acid substitutions that have smaller or larger R-groups. In combination, these three mutations have been shown to inhibit binding of T3, although TRα1(A225G, D246N, E350K) retains DNA binding capability in vitro [45]. Our data suggest that mislocalization of TRα1(A225G, D246N, E350K) to the cytoplasm further exacerbates the misregulation of T3-mediated gene expression by a receptor that is also unresponsive to ligand.

We next examined a dominant-negative TRα1 mutant that was originally isolated from patients with renal clear cell carcinoma that contains three amino acid substitutions: isoleucine to asparagine at position 116 (I116N) in the DBD, alanine to threonine at position 225 (A225T) in NES-H3, and methionine to isoleucine at position 388 (M388I) in the LBD [48] (Fig. 5A). TRα1(I116N, A225T, M388I) had a statistically significant cytoplasmic shift compared to wild-type TRα1 (normalized N/C < 1.0; p=0.0025) (Fig. 5B, C). In approximately 60% of cells, TRα1(I116N, A225T, M388I) showed many aggregates which formed multiple patterns, including both cytoplasmic and nuclear aggregates (Fig. 5B). In combination, these mutations have been shown to interfere with T3 binding, although TRα1(I116N, A225T, M388I) retains the ability to bind DNA in vitro [48].

Figure 5.

The distribution pattern of hepatocellular carcinoma-associated mutant TRα1(I116N, A225T, M388I) is shifted towards the cytoplasm with aggregate formation. A) TRα1 functional domains and mutation sites. B) HeLa cells transfected with expression plasmids for wild-type GFP-TRα1 or GFP-TRα1(I116N, A225T, M388I), or single mutations, as indicated, were analyzed by quantitative fluorescence microscopy after fixation and staining with DAPI to visualize the nucleus. Scale bar = 10 μm. C) Bars indicate relative N/C for mutant TRα1, normalized to the N/C of wild-type. Error bars indicate ±SEM (n=3 independent, biologically separate replicate experiments, with 100 cells scored per replicate); *p<0.05. D) HeLa cells transfected with expression plasmids for wild-type GFP-TRα1, GFP-TRα1(I116N), GFP-TRα1(A225T), or GFP-TRα1(M388I), as indicated. Right-hand panel summarizes the relative N/C normalized to the N/C of wild-type, and p-values for the single mutants. Images are representative of commonly observed phenotypes with cytoplasmic and nuclear aggregates. Scale bar = 10 μm. E) TRα1 (A225T). Wild-type A225 interacts with M259 (i), while the mutant T225 (ii) facilitates nearby interactions. (iii) Alignment of mutant and wild-type A225T shows only a slight perturbation (red arrows).

Based on these data, we sought to ascertain which, if any, of these point mutation(s) would have an impact on TRα1 localization individually. HeLa cells were transfected with expression plasmids for GFP-tagged TRα1(I116N), TRα1(A225T), TRα1(M388I), and wild-type TRα1 (Fig. 5D). Interestingly, TRα1(I116N) and TRα1(A225T) showed a greater cytoplasmic distribution compared with wild-type TRα1 and TRα1(M388I) (p<0.05), suggesting that mutations in the DBD and altered nuclear export activity are key drivers of localization. Both TRα1(I116N) and TRα1(A225T) had greater numbers of aggregates compared to wild-type TRα1 or TRα1(M388I), so much so that they were indistinguishable from the triple mutant TRα1(I116N, A225T, M388I) (Fig. 5D).

Aggregates generally form from exposed hydrophobic amino acids; however, none of the three mutations of TRα1 (I116N, A225T, M388I) involve a direct substitution to a more hydrophobic amino acid. For I116N, the amide group on the R chain of asparagine is capable of accepting and donating hydrogen bonds, a feature that isoleucine lacks. This may indicate that I116N, which is capable of inducing aggregates, is creating a conformational change in TRα1 that exposes a hydrophobic region. For A225T, the hydroxyl group of the R chain of threonine is capable of donating a hydrogen bond, a characteristic not shared by alanine. TRα1 (A225T), a mutant prone to nuclear and cytoplasmic aggregation, stabilizes NES-H3 by facilitating interactions between nearby residues. Residues shown in Fig. 5E are within a 5Å sphere centered on position 225. In wild type TRα1 there is an interaction between A225 and residue M259, while the mutant (T225) further facilitates nearby interactions. Although the alignment of the wild-type and the mutant only shows slight differences, the perturbation is enough to facilitate a greater interaction within NES-H3 (C-score = −0.05; T M-score = 0.71).

A225T, which is also capable of forming aggregates, may also induce a conformational change in TRα1 by a similar mechanism to I116N. In contrast, M388I, in which isoleucine carries no charge and is incapable of forming hydrogen bonds, does not consistently form aggregates. Although there could be steric changes that induce conformational change, it appears unlikely that M388I produces an appreciable change in structure on its own. In combination, these mutations have been shown to interfere with T3 binding, although TRα1(I116N, A225T, M388I) retains the ability to bind DNA in vitro [48]. Mislocalization to the cytoplasm and a tendency to form aggregates would further compound the misregulation of TR-mediated gene expression.

3.5. The Intracellular Distribution Pattern of TRα1(I116N, A225T, M388I) Resembles the Oncoprotein v-ErbA

As part of our prior studies, we documented that v-ErbA, a retroviral dominant negative oncogenic variant of TRα1 has a greater cytoplasmic localization compared with the wild-type receptor and often forms aggregates that are recruited by a microtubule-dependent mechanism to the perinuclear aggresome [15, 16, 18, 35]. The amino acid sequence changes which contribute to its oncogenic properties include fusion of a portion of avian erythroblastosis virus (AEV) Gag to its N-terminus, N- and C-terminal deletions, and 13 amino acid substitutions (Fig. 6A). The cytoplasmic shift of TRα1(I116N, A225T, M388I) and the substantial presence of aggregates phenotypically resembles the distribution pattern of the oncoprotein v-ErbA (Fig. 6B), and is also reminiscent of the tumor suppressor protein p53 [49–51]. To test for colocalization of the TR mutant with v-ErbA, intracellular distribution patterns were quantified by confocal microscopy. mCherry-TRα1(I116N, A225T, M388I) and GFP-v-ErbA were observed to colocalize in aggregates (Fig. 6C; Pearson’s Correlation Coefficient r=0.92±0.00). This significant colocalization between TRα1(I116N, A225T, M388I) and v-ErbA warranted further investigation of the nature of the aggregates forming in TRα1(I116N, A225T, M388I).

Figure 6.

Colocalization of TRα1(I116N, A225T, M388I) with v-ErbA and aggresome markers. A) Diagram showing domains of v-ErbA. Black dots indicate the locations of the 13 amino acid substitutions that distinguish v-ErbA from TRα1 (not to scale). NES-CRM1, CRM1-dependent NES in the viral Gag region of v-ErbA; NES-H3/NES-H6 predicted, based on the sequence identity with NES-H3 and NES-H6 in TRα1. B-E) HeLa cells were cotransfected with expression plasmids, as indicated. Each cell scored contained multiple aggregates and/or one perinuclear aggresome. Scale bars = 10 μm. B) Representative images of the intracellular distribution of GFP-v-ErbA. C) Colocalization of mCherry-tagged TRα1(I116N, A225T, M388I) with GFP-v-ErbA. D) Colocalization of mCherry-TRα1(I116N, A225T, M388I) with aggresome marker GFP-250. E) Colocalization of mCherry-TRα1(I116N, A225T, M388I) with the aggresome marker GFP170* (n=3 replicates; 60 cells scored per replicate).

To determine whether TRα1(I116N, A225T, M388I) was recruited to bona fide aggresomes, HeLa cells were cotransfected with expression plasmids for mCherry-TRα1(I116N, A225T, M388I), GFP-250 (Fig. 6D), and GFP170* (Fig. 6E). GFP-250 and GFP170* are well-established cytoplasmic and nuclear aggresome markers, respectively [16]. The high correlation between TRα1(I116N, A225T, M388I) and GFP-250 (r=0.91±0.02) and GFP170* (r=0.74±0.03) further highlights the similarity in phenotype between TRα1(I116N, A225T, M388I) and v-ErbA, and confirms recruitment of the cancer-associated mutant to aggresomes. Whether the increase in cytoplasmic localization is solely the product of recruitment to the aggresome requires further study, but it seems likely that aggresome formation contributes significantly to this phenotype.

The aggresome is a dynamic structure in which aberrant proteins, at least in some cases, still maintain mobility and thus can shuttle to and from the aggresome [16]. Shuttling to and from cytoplasmic or nuclear aggresomes adds an alternate pathway for TRα1(I116N, A225T, M388I) to travel, thus decreasing the amount available to shuttle into and out of the nucleus. The dynamic trafficking of TRα1(I116N, A225T, M388I) between the nucleus, cytoplasm, and aggregates/aggresomes could potentially have critical effects on gene regulation. TRα1(I116N, A225T, M388I) is a dominant negative inhibitor of wild-type TRα1 and therefore may be recruiting wild-type TRα1 into the aggresome-shuttling pathway, thereby decreasing the amount of nuclear wild-type TRα1.

TRα1 mutants associated with hepatocellular carcinoma have been shown to regulate only a fraction of the genes targeted by wild-type TRα1, but have gained the ability to regulate many other unique targets, including genes involved in glycolysis and energy metabolism [34]. For example, TRα1 (I116N, A225T, M388I) has an increased repertoire of gene targets, compared to the wild-type, that are either upregulated or downregulated [34]. This means that TRα1(I116N, A225T, M388I), when present in the nucleus, is abnormally altering gene expression regardless of the dominant negative effects it exhibits on wild-type TRα1. As shown here, TRα1(I116N, A225T, M388I) has a high colocalization with v-ErbA, and thus may follow a similar pathway that contributes to pathogenesis. It is likely that cytoplasmic localization of this mutant and sequestration in aggregates contribute to the aberrant patterns of TR-mediated gene expression found in hepatocellular carcinoma.

3.6. Mutations of TRα1 in Other Regions of Helices 3, 6 and 12 that Alter Distribution Patterns

Next, we examined a TRα1 mutant originally isolated from a patient with hepatocellular carcinoma containing four amino acid substitutions: serine to threonine at position 40 (S40T) in the A/B domain, lysine to arginine at position 136 (K136R) in NLS-1, leucine to proline at position 251 (L251P) in NES-H6, and valine to alanine at position 390 (V390A) in NES-H12 [45] (Fig. 7A). Both TRα1(S40T, K136R, L251P, V390A) and the single mutant TRα1(L251P) showed a significant shift towards a more cytoplasmic distribution, relative to wild-type TRα1 (normalized N/C < 1.0; p<0.0001) (Fig. 7B, C), suggesting that L251P plays a major role in the mislocalization of TRα1. In addition, TRα1(S40T, K136R, L251P, V390A) formed a significantly greater number of cytoplasmic aggregates. The percentage of cells with a smooth distribution was significantly higher among cells expressing wild-type TRα1 compared with mutant TRα1 (p<0.0005) (Fig. 7D). In addition, the single mutant, TRα1 (V390A), had a distribution more similar to TRα1(S40T, K136R, L251P, V390A) than wild-type, with a significantly greater cytoplasmic localization and formation of aggregates (p<0.01; Fig. 7E, F). In another study, we showed no significant difference in TRα1 localization for TRα1 (K136R) [52], so this single mutant was not analyzed again here.

Figure 7.

The distribution pattern of hepatocellular carcinoma-associated mutant TRα1(S40T, K136R, L251P, V390A) is shifted towards the cytoplasm with aggregate formation. A) TRα1 functional domains and mutation sites. B) HeLa cells transfected with expression plasmids for wild-type GFP-TRα1 or GFP-TRα1(S40T, K136R, L251P, V390A), or single mutations, as indicated, were analyzed by quantitative fluorescence microscopy after fixation and staining with DAPI to visualize the nucleus. Scale bar = 10 μm. C) Bars indicate relative N/C for mutant TRα1, normalized to the N/C of wild-type. Error bars indicate ±SEM (n=3 independent, biologically separate replicate experiments, with 100 cells scored per replicate. n.s., p>0.05; *p<0.05; **p<0.01; ***p<0.001. D) Bars indicate the percentage of cells with no aggregates (smooth), cytoplasmic aggregates, nuclear aggregates, or aggregates in both the nucleus and cytoplasm, as indicated. Error bars indicate ±SEM (n=3 independent, biologically separate replicate experiments, with 100 cells scored per replicate). E) HeLa cells transfected with expression plasmids for wild-type GFP-TRα1 or GFP-TRα1(V390A), as indicated, were analyzed by quantitative fluorescence microscopy. F) Bars indicate the percentage of cells in which TR was primarily nuclear with no aggregates (Nuclear: smooth), primarily nuclear with aggregates (Nuclear Aggregates), both nuclear and cytoplasmic with a smooth distribution, or both nuclear and cytoplasmic with aggregates, as indicated. Error bars indicate ±SEM (n=3 independent, biologically separate replicate experiments, with 100 cells scored per replicate).

It is striking that even single point mutations within NESs of TRα1, such as L251P and V390A (within NES-H3 and NES-H12, respectively), can markedly alter its localization. Such a difference in localization may occur in the L251P mutant because, unlike leucine, proline forms a nitrogen-containing ring and introduces kinks into α-helices. It has been previously shown that TRα1(S40T, K136R, L251P, V390A) still binds DNA in vitro, but does not bind T3 [45]. Taking these functional characteristics into account, the formation of kinks in the LBD may be disrupting the binding of T3 to TRα1 and interfering with target gene transcription. These kinks could also be contributing to the formation of aggregates, in part through interaction of proline with hydrophobic moieties [53], which would further augment the negative impact on gene regulation.

Since the NLSs and NESs in TR are conserved among vertebrate species, the mutations in the NES region may be leading to decreased affinity for exportins. However, reduced affinity for exportins does not explain why the TRα1 mutants are showing a decreased nuclear-to-cytoplasmic ratio compared to wild-type TRα1. Exportins transport TRα1 out of the nucleus, and reduced affinity for exportins would be expected to lead to a higher nuclear-to-cytoplasmic ratio of mutant TRα1 compared with wild-type TRα1. One possibility for the observed more cytoplasmic distribution may be that the mutations in the NES-H3 and NES-H12 regions interact with other amino acids in TRα1. For NES-H3, the mutations are close enough to the hinge domain that they could be altering the structure of the NLSs of TRα1, disrupting the binding of importins to the NLSs. Such disruption would prevent TRα1 from entering the nucleus, leading to the observed decrease in its nuclear-to-cytoplasmic ratio. Another possibility is that the mutations in the NES region may be leading to increased affinity for exportins. In this case, the exportins would more efficiently transport the TRα1 mutant out of the nucleus, leading to a decrease in nuclear-to-cytoplasmic ratio. However, increased affinity is unlikely because, although NES consensus sequences are not well-defined, it has been repeatedly observed that they tend to be hydrophobic [54]; the TRα1 mutations L251P and V390A probably do not increase hydrophobicity of the NES-H3 and NES-H12 regions because proline is not hydrophobic, and valine and alanine are both hydrophobic. The bottom line is that structural changes which promote cytoplasmic localization of TRα1(S40T, K136R, L251P, V390A) and sequestration in aggregates likely contribute to the aberrant patterns of TR-mediated gene expression associated with hepatocellular carcinoma.

3.7. A Dominant-negative TRα1 Mutant Promotes Aggregation of Wild-type TRα1

As noted previously, tumors of patients with hepatocellular carcinoma display a high incidence of mutant TRs, and one such mutant of TRα1 contains two amino acid substitutions: lysine to glutamate at position 74 (K74E) in the DBD, and alanine to valine at position 264 (A264V) [45] in NES-H6 (Fig. 8A). To determine whether these mutations impact receptor localization, HeLa cells were transfected with expression plasmids for GFP-tagged wild-type TRα1 or TRα1(K74E, A264V). Elevated protein aggregation was noted, in direct contrast with the smooth distribution of wild-type TRα1 (Fig. 8B). On average, 98% of cells expressing wild-type GFP-TRα1 displayed a smooth distribution of fluorescence, with the remaining 2% showing nuclear or cytoplasmic aggregates (Fig. 8C). On the other hand, for cells expressing GFP-TRα1(K74E, A264V), the number of cells with aggregates was significantly greater than for cells expressing wild-type GFP-TRα1. On average, only 77% of cells expressing GFP-TRα1(K74E, A264V) had a smooth distribution (p<0.0001), while 13.5% of cells had nuclear aggregates (p=0.004), and 9.5% of cells had cytoplasmic aggregates (p=0.002). In contrast, the intracellular distribution pattern of a mutant TR constructed to only have the K74E mutation was not significantly different from that of wild-type TRα1 (p=0.5464) (Fig. 8D). TRα1(K74E) was primarily localized to the nucleus, and there were no aggregates, suggesting, by default, that the driver mutation of the K74E/A264V phenotype is the A264V substitution within helix 6.

Figure 8.

GFP-TRα1(K74E, A264V) is primarily nuclear but tends to form aggregates. A) TRα1 functional domains and mutation sites. B) HeLa cells transfected with expression plasmids for wild-type GFP-TRα1 or GFP-TRα1(K74E, A264V), were analyzed for aggregate formation by fluorescence microscopy after fixation. Representative examples of a smooth distribution, nuclear aggregates, and cytoplasmic aggregates are shown. Scale bar = 10 μm. C) Bars indicate the percentage of transfected cells with no aggregates (smooth), nuclear aggregates, or cytoplasmic aggregates, as indicated. Error bars indicate ±SEM (n=5 independent, biologically separate replicate experiments, with 200 cells scored per replicate); ***p<0.001. D) HeLa cells transfected with expression plasmids for wild-type GFP-TRα1 or GFP-TRα1(K74E), as indicated, were analyzed by quantitative fluorescence microscopy. Bars indicate relative N/C for mutant TRα1, normalized to the N/C of wild-type. Error bars indicate ±SEM (n=3 independent, biologically separate replicate experiments, with 100 cells scored per replicate); n.s., p>0.05. E, F) HeLa cells were transfected with expression plasmids for mCherry-GFP-TRα1 and GFP-TRα1(K74E, 264V), either together or singly, as indicated, and the distribution of the mutant (E) or wild-type (F) was analyzed by quantitative fluorescence microscopy. Bars indicate the percentage of cotransfected or singly transfected cells with no aggregates (smooth), nuclear aggregates, or cytoplasmic aggregates, as indicated. Error bars indicate ±SEM (n=3 independent, biologically separate replicate experiments, with 100 cells scored per replicate).

After demonstrating that TRα1(K74E, A264V) forms aggregates, we next sought to ascertain whether the mutant TR aggregates could interfere with the localization of wild type TRα1 through dominant negative activity. The distribution of GFP-TRα1(K74E, A264V) was unchanged in the presence of mCherry-TRα1 (Fig. 8E). In contrast, the percentage of cells with aggregates of wild-type TRα1 was significantly greater in the presence of TRα1(K74E, A264V), compared to wild-type TRα1 expressed alone. On average, only 91% of cells coexpressing wild-type TRα1 with twice the amount of TRα1(K74E, A264V) plasmid displayed an even distribution (p<0.0001), while 6% (p<0.0001) and 3% (p=0.09) displayed nuclear aggregates, and cytoplasmic aggregates, respectively (Fig. 8F). These data provide support for dominant-negative sequestration of wild-type TRα1 into aggregates by TRα1(K74E, A264V), comparable to sequestration of wild-type TRα1 by the oncoprotein v-ErbA [15, 16].

Our data also show that TRα1(K74E, A264V) displays an altered localization by way of aggregate formation and a capability to induce wild-type TRα1 aggregate formation. K74 is a conserved residue in the DBD, which acts as an “allosteric sensor,” regulating transcriptional activity in response to DNA binding [29]. In this mutant, the positively-charged lysine is exchanged for the negatively-charged glutamate, effecting a substantial conformational shift in a crucial area of the protein. This shape change causes the mutant to display an altered target gene repertoire leading to the misregulation of tumor suppressor genes, oncogenes, and genes involved with cell migration and metastasis [55]. Additionally, the A264 residue in the LBD is mutated to the bulkier valine, resulting in a delayed corepressor release and reduced sensitivity to T3 [33].

Aggregates are generally associated with misfolded proteins, so the significant levels of aggregation in the mutant TR suggest that A264V changes the conformation of either or both of the mutation-containing domains. This could increase the likelihood of overall misfolding within the cell. Given the often complete aggregation within cells shown here, misfolded TRs might interact with other TRs to promote further misfolding. TRα1(K74E, A264V) has not only acquired dominant negative function as a result of allosteric sensing [29], but also by its propensity to induce aggregate formation and cytoplasmic mislocalization. Taken together, the impaired functionalities of this mutant likely contribute to the transcriptional misregulation of TR-mediated gene expression and progression of hepatocellular carcinoma.

4. Conclusion

TRα1 is a transcription factor that regulates genes involved in cell proliferation, differentiation, and apoptosis. Given that TRα1 is the cellular homolog of the dominant negative, viral oncoprotein v-ErbA, it is logical to speculate that impaired function of TRα1 could contribute to cancer progression. Indeed, the correlation of somatic mutations of TRα1 with human cancers supports the hypothesis that the loss of normal function of TR leads to uncontrolled proliferation and loss of cell differentiation [5]. It is likely, however, that other factors drive the initial process of tumorigenesis, and that somatic mutations in TR exacerbate cancer progression [48]. By analyzing natural TRα1 variants associated with hepatocellular carcinoma, renal clear cell carcinoma and papillary thyroid carcinoma, we have gained greater insight into structural changes that impact TR intracellular localization. Table 1 summarizes the distribution patterns of the mutants reported on in the present study, and what is known about their other functional characteristics from the literature, including T3 binding, DNA binding, transcriptional activity, and dominant negative activity.

Table 1.

Intracellular Localization Patterns of Cancer-Associated Mutations in TRα1

Type of Cancer	NES or NLS	Mutations*	Known Impaired Activity (Ref)	Localization Pattern (this study)
Hepatocellular carcinoma	NES-H3, NES-H6	A225G, D246N, E350K	T3 binding: No DNA binding: Yes [45]	Cytoplasmic aggregates
	NLS-1, NES- H6, NES-H12	S40T, K136R, L251P, V390A K136R L251P V390A	T3 binding: No DNA binding: Yes [45]	Mostly cytoplasmic with aggregates Nuclear Mostly cytoplasmic with aggregates Nuclear/cytoplasmic; some aggregates
	NES-H6	K74E, A264V K74E	T3 binding: No DNA binding: No [45]	Nuclear/cytoplasmic; some aggregates Nuclear
	NLS-2, NES-H6	G24E, M256V, E343A, P269L G24E M256V	T3 binding: No DNA binding: Yes [45] Dominant negative activity: Yes [29]	Nuclear/cytoplasmic; no aggregates Nuclear/cytoplasmic; no aggregates Nuclear
Renal cell carcinoma	NES-H3	I116N, A225T, M388I I116N A225T M388I	T3 binding: No DNA binding: Yes Dominant negative activity: No [48]	Nuclear/cytoplasmic; some aggregates Nuclear/cytoplasmic; some aggregates Nuclear/cytoplasmic, some aggregates Nuclear
Papillary thyroid carcinoma	NES-H3	E213D	Transcription: Yes, but weak Dominant negative activity: Yes [42]	Nuclear
	NES-H3	S183N, H184Q, R228H	Transcription: Yes, but weak Dominant negative activity: Yes [42]	Nuclear Nuclear

^*Bold indicates mutations that fall within TRα1 NESs or NLSs.

For cancer-associated mutants with significantly altered localization patterns, structural stability was measured with normalized B-factor prediction [40] and applied to each TRα1 domain. Table 2 summarizes their predicted differential stability. Overall, mutations increased the stability of the ligand-binding domain (LBD) compared to wild-type. The most perturbed structures were found in TRα1(G24E, M256V, E343A, P269L), correlating with disruption of the T3 binding site. T3-dependent transcriptional regulation by TR is modulated by interactions between all domains of the protein [48], which makes dissecting the exact functional consequences of TR mutations complex. Taken together, our results provide further insight into structure-function relationships in TRα1, highlighting the importance of key single amino acid residues in the ligand-binding domain and A/B domain of TR. In particular, mutations in the NLS and NES regions of TRα1 often lead to mislocalization and aggregation of TRα1, emphasizing that the fine balance of nuclear import, nuclear retention, and nuclear export plays a critical role in modulation of T3-responsive gene regulation. In addition to altered TRα1 trafficking, protein folding, transcriptional regulation, and hormone responsiveness; the functional consequences of cancer-associated mutations in TRα1 may be further accentuated by dominant negative activity.

Table 2.

Predicted Differential Stability of TRα1 Cancer-Associated Mutants^*

	DBD	Hinge	LBD	NLS-1	NES-H3/H6	NES-H12
Cancer Mutant/Residue Number	52–127	128–189	190–410	130–147	209–265	390–407
A225G, D246N, E350K	Destabilized	-	Stabilized	-	Stabilized	-
S40T, K136R, L251P, V390A	-	-	Stabilized	-	Stabilized	Stabilized
K136R	-	Stabilized	Stabilized	-	-	Destabilized
L251P	Destabilized	-	Stabilized	-	Stabilized	Stabilized
V390A	-	-	Stabilized	-	Stabilized	-
K74E, A264V	-	-	Stabilized	-	Stabilized	-
G24E, M256V, E343A, P269L	Stabilized	Destabilized	Stabilized	Destabilized	-	-
M256V	Destabilized	-	-	Stabilized	-	Destabilized
I116N, A225T, M388I	Stabilized	-	Stabilized	Stabilized	Stabilized	Stabilized
I116N	Stabilized	-	Destabilized	-	Destabilized	-
A225T	-	Destabilized	Stabilized	-	Stabilized	-
M388I	-	-	-	Destabilized	Stabilized	-
E213D	-	Stabilized	Stabilized	-	Stabilized	-
S183N, H184Q, R228H	-	-	Stabilized	Stabilized	Stabilized	Destabilized
R228H	-	-	Stabilized	-	-	-

^*Normalized B-factor analysis. Values with statistically significant differences (p<0.05) compared with wild-type TRα1 are indicated as either destabilized or stabilized (values below zero). No significant difference is indicated by a dash.