The thyroid hormones (TH) triiodothyronine (T3) and thyroxine (T4) are important
regulators of biological functions. Lack of TH action can lead to developmental,
metabolic, and cardiovascular diseases. At the cellular level, nuclear receptors
called “thyroid hormone receptors” (THRs) mediate TH function [1 ]. Heritable syndromes of impaired TH
sensitivity include, among others, resistance to thyroid hormone (RTH), which is
caused by mutations in genes encoding for THRs, in particular the two genes
THRA and THRB
[2 ]. In
general, patients’ symptoms depend on the expression pattern of the affected
gene and the resulting functional defect. Refetoff et al. coined the acronym RTH
when they described the first patient in 1967 [3 ], who, 20 years later, was found to have a homozygous deletion in the
THRB gene encoding TRβ [4 ]. During this time, additional patients with RTH were identified, most
of whom carried a heterozygous missense mutation in THRB
[5 ]
[6 ]. Patients with RTH have elevated T3 and T4 serum concentrations, with
some having elevated, but never suppressed thyrotropin (TSH) levels, making the
association with TH-dependent disease relatively clear. The phenotype is variable,
and few patients show severe symptoms such as attention-deficit hyperactivity
disorder (ADHD), tachycardia, or goiter [7 ]
[8 ]. TRβ is mainly
responsible for the negative feedback loop regulating the
hypothalamus-pituitary-thyroid axis [9 ]
[10 ], which explains why TSH
is not suppressed despite high T3 and T4 levels.
The two genes encoding the TH receptors are TRα (THRA ) and TRβ
(THRB ), therefore, the differences in the phenotype of patients with TH
receptor gene mutations are likely due to different expression patterns of the two
isoforms. With the development of new technologies such as whole-exome sequencing
(WES), the first THRA mutation was identified in 2012 [11 ], followed shortly by a second case in
the same year identified by Sanger sequencing [12 ]. Since then, the number of THRA missense mutations has been
steadily increasing. These patients lack the changes in blood TH concentration that
make other conditions with RTH so distinct, as the hypothalamus-pituitary-thyroid
axis is only regulated by TRβ, whereas TRα plays no role in this
feedback loop. Therefore, the T3 and T4 serum levels are mainly normal in these
patients. It appears that the T3 levels are in the higher normal range while T4 is
rather low-normal resulting in a shifted T3/T4 ratio; however, so far, no
reference values are available for the T3/4 ratio leaving the relevance of
this shift open. The normal T3 and T4 values in most cases make the phenotype even
more striking, as THRA mutation carriers have an even more severe phenotype
than TRHB mutation carriers with mainly hypothyroid symptoms. These include
growth retardation, mild to moderate mental retardation, mild skeletal dysplasia,
severe constipation, broad facial features, and bradycardia. Interestingly, most of
the identified mutations result in partial or complete loss of function that
inhibits gene regulation in a dominant-negative manner [11 ]
[12 ]
[13 ]
[14 ]
[15 ]
[16 ]
[17 ]
[18 ]
[19 ]
[20 ]
[21 ]
[22 ].
Recently, we described a novel heterozygous point mutation that enhances the function
of both TRα isoforms, TRα1 and TRα2 leading to increased
T3-activation of TRα1 and increased antagonism of TRα2 [23 ]. This novel mechanism in RTH due to
THRA mutations brings the TRα2 splice variant into focus and
raises further interest in its physiological function.
The TRα encoding gene THRA (17q21.1) was previously described as a
proto-oncogene “c-erb-A” and was isolated from embryonic chicken,
human placental, and rat brain libraries [24 ]
[25 ]. This gene encodes
proteins that share structural features with other nuclear receptors, consisting of
a regulatory A/B-domain, a DNA-binding domain (DBD), a hinge region, and a
ligand-binding domain (LBD) ([Fig.
1a ] ) [26 ]
[27 ]. C-erb-A was identified as
TRα1 by nuclear localization, ability to specifically bind TH and
transcriptional regulation of TH-responsive genes [24 ]
[25 ]. Briefly, TRα1 can interact with TH responsive elements (TRE)
located in the promoter regions of TH-regulated genes via the two
C4 -zinc fingers in the DBD and regulate transcription ([Fig. 2a ] ) [28 ]. Even in an unliganded state, TRs
occupy TREs in the function of transcriptional regulators [29 ].
Fig. 1 (a ) Functional domains of TRα1 and TRα2
isoforms. Published disease-causing variants are marked on the protein
domain structure. (b ) TRα isoforms resulting from alternative
splicing, or different transcription start points, leading to proteins of
different molecular weight. (A/B: a regulatory domain; DBD: DNA-binding
domain; LBD: ligand-binding domain)
Fig. 2 (a ) The DNA binding domains (DBD) of the unliganded
TRα1 homodimer bind to responsive elements of the thyroid hormone
receptor (TRE) in the promoter region of target genes. The ligand-binding
domains (LBD) are in complex with co-repressors (CoRs) and target gene
expression is inhibited. (b ) Upon binding of thyroid hormone T3, CoRs
are exchanged for co-activators (CoA), and the target genes are then
expressed. (c ) The weak antagonistic effect of TRα2 is
probably due to competition for TREs between both isoforms without
interaction with CoRs.
TREs typically consist of a 5´-AGGTCA-3´motif arranged in repeats,
either as palindromic, inverted palindromic, or direct repeats spaced by four
nucleotides (DR4) [30 ]. TRs can be
positive or negative regulators being able to initiate or inhibit transcription
depending on which TRE is bound [31 ]. The
C-terminal LBD is crucial not only for regulating activity through ligand binding
but also for interacting with cofactors and dimerizing with other nuclear receptors.
Upon binding to T3, the LBD undergoes conformational changes that lead to the
replacement of co-repressors (CoRs) by co-activators (CoAs) ([Fig. 2b ]) [32 ]. TRs are known to exist as monomers,
but can also form homodimers [33 ] as well
as heterodimers with other nuclear receptors, the most common being the retinoid X
receptor (RXR) [34 ]
[35 ]. Interestingly, RXR has been shown to
significantly inhibit transcriptional activity in vitro , suggesting a
regulatory role for this heterodimer [36 ].
In addition to the canonical function as transcription regulators, non-canonical TH
signaling has a more rapid effect on the target cell. Flamant et al. proposed a
classification of TH action into four subtypes [37 ]: Type 1 corresponds to the canonical model of TR as a
transcription factor by direct binding to DNA, as described above, with TH signaling
in mitochondria via the shorter isoform (as described below) also belonging
to this type. Type 2 includes signaling via indirect binding to DNA,
e. g., by binding to other transcription factors. Type 3 includes
signaling independent of DNA binding, such as direct activation of the
phosphoinositide 3-kinase/protein kinase B (PI3K/AKT) pathway, which
is described in particular for the plasma membrane-bound isoform p30 (see below).
Type 4 summarizes TH signaling independent of TRs, e. g.,
integrin αVβ3, which has been proposed as a membrane receptor for T3
and T4.
Shortly after the discovery of TRα1, the TRα2- isoform was identified
resulting from an alternative splice site at exon 9 and transcription of an
additional exon 10 [38 ]
[39 ] ([Fig. 1b ]). Interestingly, this isoform is
unable to bind TH due to the extended C-terminal LBD [40 ] leading to an antagonistic effect on
TRα1 and TRβ. However, TRα2 has only a weak antagonistic
effect on TRα1 and TRβ, as it needs high expression levels to
inhibit transcriptional activity ([Fig.
2c ]) [41 ]
[42 ]
[43 ]
[44 ] suggesting a rather
minor physiological role for TRα2. Several mechanisms have been suggested to
explain this phenomenon: (i ) competition for TRE binding sites, (ii )
interaction with RXR as the preferred dimerization partner of TRα1 on
specific TREs such as DR4 [45 ], or
(iii ) a DNA-independent mechanism such as interaction with basal
transcriptional factors [41 ]
[46 ]. Moreover, the phosphorylation state
of the elongated C-tail has been shown to influence the antagonistic effect of
TRα2 [47 ]. Interestingly, the
inhibitory effect seems to occur only on positive TREs [48 ], and not on negative TREs. The rather
weak effect can be explained by the lack of interaction between TRα2 and
CoRs [42 ]
[43 ].
Nevertheless, protein studies in mouse models and post mortem human brains
indicate a high expression ratio of TRα2 to TRα1 [49 ]
[50 ]. This suggests a T3-independent regulatory role for the physiology of
TRα2. Important for this role is that TRα2 has a functional DBD,
which still binds to TREs as homo- or heterodimer with RXR, albeit with lower
affinity as compared to TRα1 [42 ]
[45 ]. Therefore, as long as
TRα2 is highly expressed, it forms homodimers and occupies TREs without the
capability to be activated by T3 and thus acts as an antagonist to TRα1. A
direct heterodimer with TRα1, which would result in an even more direct
TRα1-antagonism, was not observed on any tested TRE. However, since most
dimerization studies of TRs examined the formation of dimers on TREs indirectly by
using DNA mobility shift assays, it cannot be excluded that heterodimerization
between TRα1 and TRα2 may occur independently of DNA-binding, which
was suggested by Katz et al. 1995 [47 ].
Over time, other TRα isoforms were discovered ([Fig. 1b ]), including a suspected third
splicing variant, TRα3 that originates from another splicing event in exon
9. Similar to TRα2, this isoform presumably has an elongated C-tail encoded
by the sequence of exon 10 (448 amino acids), which is also unable to bind to TH.
As
only one study was able to detect this isoform, [38 ] it is vastly understudied, but the
predicted structure anticipates the same function as TRα2. Yet another
alternative splicing event is responsible for the TRα1-ΔE6 isoform
that carries an exchange of exon 6 for the micro-exon 6b. The resulting protein
lacks T3-binding capacity but can reduce the transcription-enhancing activity of
TRα1. Its proposed role is also regulatory for TRα2 and seems to be
important for myocardial development, though the expression pattern of
TRα1-ΔE6 suggests additional roles in other tissues [51 ]. Although it has not been
investigated, a TRα2-ΔE6 isoform likely exists as well.
Further truncated isoforms, ΔTRα1 and ΔTRα2, have
been observed, which are the result of transcription from an internal promoter in
intron 7 and thus are missing the N-terminal domain, but otherwise resemble
TRα1 and TRα2 [52 ].
Additionally, alternative translational start points on the TRα1 mRNA can
result in shorter isoforms (p43, p30, p33, and p28) that can be activated by T3 but
lack the ability to bind to DNA. They are thought to be bound to the plasma membrane
(p30 and p33) [53 ]
[54 ] or located in the mitochondria (p28
and p43) [55 ] and maybe responsible for
more immediate non-canonical signaling via the MAPK pathway.
Since the first description of a loss-of-function THRA mutation in 2012, about
26 other missense mutations have been identified. One-half of the identified
mutations are positioned in the TRα1-specific region of the LBD [11 ]
[12 ]
[14 ]
[16 ]
[19 ]
[55 ]
[56 ]
[57 ]
[58 ]
[59 ]
[60 ], the other half is located in the coding exons that are shared
between TRα1 and TRα2 [13 ]
[15 ]
[16 ]
[17 ]
[18 ]
[19 ]
[20 ]
[21 ]
[22 ]
[61 ]
(
[Fig. 1 ]). The latter ones could
also disturb the function of TRα2, so the phenotypes seen in these patients
might provide information about the physiological role of TRα2.
Although several studies in mice have addressed TRα1 and its function [62 ]
[63 ]
[64 ]
[65 ]
[66 ], the relative contributions of both isoforms to the overall phenotype
have proven difficult to dissect. As mentioned earlier, a comparison of protein
levels of TR isoforms in mice revealed organ-specific expression, particularly a
unique expression pattern of high TRα2 abundance in the central nervous
system [49 ], suggesting an important
regulatory role. The specific knockout of TRα2 by adding a strong
polyadenylation site that follows the stop codon of TRα1 and transcriptional
stop codon resulted in overexpression of TRα1 and a mixed phenotype with
hyper- and hypothyroid tissue states [67 ]. Although this model provides valuable information to the field, it could
not fully explain the physiological role of TRα2. Another interesting model
is the
Pax8
-/-
TRα
0/0
compound mouse. Here Pax8 , a differentiation factor for thyroid cells was
knocked out, which on its own results in the absence of thyroid cells and
consequently the complete absence of TH. Without T4 treatment, this defect leads to
an early death around weaning time [68 ].
This model was combined with a TR
0/0
model,
harboring a complete deletion of all known TRα isoforms, which on its own
was viable, but exhibited reduced growth, delayed bone maturation, moderate
hyperthermia, and reduced intestinal mucosal thickness [69 ]. The
Pax8
-/-
TRα
0/0
compound model survived without TH-treatment and partially rescued the lethal
phenotype of Pax8
-/-
mice, but growth was delayed
[70 ]. This study helped to understand
how the unliganded receptors might have a physiological function and TH is required
to relieve these effects during the postnatal stage. In contrast,
Pax8
-/-
TRα1
-/-
compound models, which still express TRα2 and ΔTRα2
isoforms, have a similar lethal phenotype to
Pax8
-/-
mice, probably due to an intact
TRα2 isoform that could affect the activity of other THRs such TRβ
[71 ]. When comparing these two
studies, a possible physiological role for TRα2 is to modulate survival,
especially in the first weeks of postnatal development.
Most mutations found in patients were studied in vitro using reporter gene
assays based on TRE-dependent luciferase expression and interaction with DNA or with
cofactors to show how they inhibit TRα function. For mutations jointly
affecting TRα1 and TRα2, some but not all studies have also examined
the effects in both splicing isoforms. However, when TRα2 function was
tested, most mutations had no measurable effect on antagonistic function.
Interestingly, for two mutations (p.A263S and p.N359Y) the inhibitory effect of
TRα2 on TRα1 was slightly reduced [15 ]
[16 ], suggesting a decrease in the dominant-negative effect.
In contrast to all other studies, we recently reported a THRA mutation that
resulted in a gain-of-function in both isoforms [23 ]. A mutated glutamate-to-glycine
residue in the first helix of the LBD had a promoting effect on T3-inducible
TRα1 activity but also resulted in a gain-of-antagonistic effect for
TRα2. Based on the computational model of the LBD, we suspect altered
dimerization interphase, although an altered interaction with TRE or cofactors
cannot be excluded. Nonetheless, this mutation is the first to enhance TRα2
function by increasing its antagonistic capacity, at least in vitro . At the
same time, TRα1 function was increased as well, leading to a pronounced T3
effect. Given this strong gain of function effect of TRα1 in vitro ,
one would expect a hyperthyroid phenotype of the patients, but this was only the
case in patients with mild tachycardia. In fact, we observed more hypothyroid
symptoms such as low IQ and global developmental delay, severe constipation, and
obesity. Matching these symptoms with our in vitro results suggests that the
activated antagonistic effect of the mutant TRα2 was able to counteract the
increased activity of the mutant TRα1. Since in most brain regions the
TRα2: TRα1 ratio is high [49 ]
[50 ], the mutant
TRα2 appears to significantly suppress the activation of the mutant
TRα1, eventually leading to the neuronal hypothyroidism of patients with the
THRAp.(E173G) mutation. In other tissues with predominant TRα1 expression,
such as cardiomyocytes, the gain-of-function mutation of TRα1 without
TRα2 antagonism results in hyperthyroidism-like symptoms. These particular
findings of the p.(E173G)-mutant, leading simultaneously to activation of
TRα1 and enhanced antagonism of TRα2, suggests that the
physiological function of TRα2 is antagonistic to TRα1 function,
which appears to be important for the tissue-specific fine-tuning of TH action in
target cells.
Overall, 33 years after the discovery of TRα2 and almost 10 years after the
first description of patients with mutations in THRA, the first evidence for
a physiological effect of TRα2 was found only recently in particular
patients carrying a new TRα-mutation. So far, the finding is limited to a
single case report and in principle other -potentially genetic- effects can
influence the patient’s wide phenotype. However, the obvious antagonistic
effect of TRα2, and its increase through this p.(E173G) mutation, proposes a
novel mechanism in RTH due to THRA mutations and argues that TRα2
indeed plays a role in controlling the local response of target cells to circulating
T3. It is tempting to speculate that any mechanism that increases TRα2
expression relative to TRα1 will decrease the cell response to T3. Moreover,
even in tissues with low T3 availability, high levels of TRα2, or mechanisms
that increase the DNA-binding of TRα2, are more likely to suppress
T3-responsive genes. TRα2 was discovered in the 1980s but few publications
on this isoform have appeared in recent decades, thus, it is now time to unravel the
physiological role of TRα2 at different developmental time points and in
different tissues more thoroughly. Here, special attention must be paid to a clear
distinction between the isoforms. Most likely, the potential of single-cell
sequencing will stimulate this process and could lead to new and surprising
discoveries for the other TRα isoforms that have been little studied so far.
Unfortunately, the lack of suitable TRα antibodies, let alone
isoform-specific antibodies, prevents the generation of genome-wide chromatin
immunoprecipitation sequencing data (ChIP-Seq) of any species. For now, a lot of
knowledge regarding TRα isoforms remains to be uncovered.
