RXR Ligands Modulate Thyroid Hormone Signaling Competence in Young Xenopus laevis Tadpoles

Abstract

Appropriate thyroid hormone (TH) signaling through thyroid hormone receptors (TRs) is essential for vertebrate development. Amphibian metamorphosis is initiated and sustained through the action of TH on TRs, which are conserved across vertebrates. TRs heterodimerize with retinoid X receptors (RXRs) on thyroid hormone response elements (TREs) in the genome; however, in most cell line and adult animal studies, RXR ligands do not affect expression of TR target genes. We used a quantitative, precocious metamorphosis assay to interrogate the effects of the RXR agonist bexarotene (Bex) and the RXR antagonist UVI 3003 (UVI) on T3-induced resorption phenotypes in Xenopus laevis tadpoles 1 week postfertilization. Bex potentiated gill and tail resorption, and UVI abrogated T3 action. These results held in transgenic tadpoles bearing a TRE-driven luciferase reporter. Therefore, we used poly-A-primed RNA sequencing transcriptomic analysis to determine their effects on T3-induced gene expression. We also assayed the environmental pollutant tributyltin (TBT), which is an RXR agonist. We found that the proteases that carry out resorption were potentiated by Bex and TBT but were not significantly inhibited by UVI. However, several transcription factors from multiple families (sox4, fosl2, mxd1, mafb, nfib) were all inhibited by UVI and potentiated by Bex and TBT. All required T3 for induction. Time course analysis of gene expression showed that although the agonists could potentiate within 12 hours, the antagonist response lagged. These data indicate that the agonists and antagonist are not necessarily functioning through the same mechanism and suggest that RXR liganding may modulate TH competence in metamorphic signaling.

Treatment of Xenopus laevis tadpoles with T3 and RXR ligands in a precocious metamorphosis assay showed RXR agonists potentiating and RXR antagonism inhibiting T3-regulated gene expression.

Appropriate thyroid hormone (TH) signaling through the thyroid hormone receptors (TRs) is critical for vertebrate development across several taxa. Both insufficient (1–3) and excess TH (4, 5) during human development result in a spectrum of adverse outcomes. The complexity inherent in mammalian development, with intrauterine growth and strong maternal effects, makes a model animal system without such complexities more amenable to direct investigation of hormonal effects on development. The African clawed frog, Xenopus laevis, develops externally and is thus minimally affected by maternal influences beyond the egg. In terms of TH biology, TH signaling induces and maintains the process of metamorphosis (6–8), wherein juvenile tadpoles develop into adult frogs, a process that models the perinatal TH signaling surge in human development (9–11). The THs are identical in all vertebrates, and the TRs are highly conserved between human and Xenopus species (12). Furthermore, metamorphosis affects almost every tissue system and cell fate decision, and it encompasses unparalleled morphological changes that facilitate quantifying the effects of THs as well as endocrine disruptors of TH signaling.

TRs are transcription factors of the nuclear receptor (NR) superfamily that bind DNA and regulate transcription in both the presence and the absence of TH (13, 14). T3 is the form of TH that binds to the receptor with the highest affinity (15). Binding of T3 to the TR can cause either activation or repression of the target gene in a gene-specific manner that is not well understood. General models of T3 signaling by TRs focus on T3-bound TRs activating transcription and apo-TRs repressing transcription, because T3-bound TRs recruit coactivator proteins that open the target gene chromatin environment to the general transcription machinery and activate gene expression. By contrast, apo-TRs recruit corepressor proteins that close the chromatin environment to repress transcription (13, 14). TRs can heterodimerize with retinoid X receptors (RXRs) (16), and RXRs greatly increase the affinity of TRs for binding DNA at thyroid hormone response elements (TREs), especially in the presence of T3 (17). RXRs can bind certain vitamin A (VA) metabolites in the ligand-binding pocket, such as 9-cis retinoic acid, and RXRs dimerize with many different NRs (18, 19).

The role of RXR ligand binding differs depending on the other NR of the dimer (18, 19). In permissive heterodimers such as RXR–liver X receptor (LXR), RXR agonists can induce the transcription of LXR target genes in the absence of the LXR agonist, and addition of the LXR agonist results in synergistic activation of target genes. However, RXR-TR behaves as a nonpermissive heterodimer, wherein RXR ligands are not thought to have any effect in either the absence or presence of T3 in most cell types and adult animal tissues. We and others have shown that pituitary cell lines are an exception to this, wherein RXR ligands can affect the expression of TRE-driven reporters and target genes (20, 21). Further linking the two signaling molecules is the finding that RXR agonists repress the hypothalamic-pituitary-thyroid (HPT) axis to the point that >90% of patients taking the RXR agonist bexarotene (Bex) for cancer chemotherapy developed hypothyroidism (22), and a single dose of Bex in healthy individuals reduced TSH expression (23). This finding was also observed in rodents given Bex and other RXR agonists (24, 25). Conversely, concomitant VA and iodine deficiency in both rats and humans increases the incidence of goiter but decreases the incidence of hypothyroidism, the more detrimental of the two conditions. Therefore, RXR agonists overly sensitize the thyrotrope to T3 levels, and VA deficiency makes the thyrotrope refractory to negative feedback by T3 (26–29). These data suggest that VA signaling has evolved to control the level of TH signaling depending on the amount of VA available. Why these two pathways interact to this degree is unknown, as are the molecular mechanisms that control it. Nevertheless, the proposed nonpermissive nature of the RXR-TR heterodimer suggests that RXR ligands should not be able to influence TH signaling during development outside the HPT axis.

On the basis of a previously published qualitative assay for assessing metamorphic program changes in 1-week postfertilization (1wk-PF) tadpoles (30), we developed a quantitative, multitier, precocious metamorphosis assay using 1wk-PF X. laevis tadpoles to assess the role of endocrine disruption on TH signaling (31). At 1wk-PF, tadpoles (NF-48) are very uniform in size, have not started to feed, and are able to take up compounds such as T3 through their rearing water. NF-48 tadpoles do not yet have a functioning thyroid gland, but they are competent to respond to T3 along many metamorphic pathways, although the competence is not complete (30, 32). For example, tail resorption, the final step of metamorphosis, is minimal at NF-48 even when supraphysiological levels of T3 are used that ordinarily would compensate for the elevated type III deiodinase in the tail, showing that TH competence in the tail is not solely dependent on T3 concentration. In an earlier report, we showed that cotreatment of 1wk-PF tadpoles with T3 and the organotin pollutants tributyltin (TBT) and triphenyltin, which can function as RXR agonists, potentiated the action of T3 when resorption phenotypes such as gill and tail resorption were assayed (33). Indeed, TBT induced TH competence in the NF-48 tadpole that is seen ordinarily at NF-50, which occurs at ∼2 weeks of age, and it potentiated gene expression of a panel of genes induced during natural metamorphosis (7, 32). These results raised the question of which gene programs was TBT inducing to “jump-start” TH competence in resorbing tissues, because endocrine disruptors that affect competence could affect developmental outcomes adversely.

TBT, a common pollutant (34–38) used as an antifoulant in marine paints, was shown to be the causative agent of imposex in marine mollusks through the action of binding to mollusk RXR (39–42). However, TBT has also been shown to bind and activate other NRs such as peroxisome proliferator–activated receptor γ (PPARγ) (43), so here we determined that a designed RXR agonist, Bex (44, 45), gave identical results as TBT in our precocious metamorphosis assays. In addition, we showed that the RXR antagonist UVI 3003 (UVI) (45, 46) had the opposite effect, abrogating TH competence at the morphological level. To determine the genetic pathways affected by the RXR ligands, we used poly-A-primed RNA sequencing (Tag-Seq) to interrogate the transcriptomes of T3-regulated genes in NF-48 tadpole tails in the presence and absence of these RXR ligands. We chose tails as the easiest resorptive tissue to isolate, and its diminished competence to respond to T3 at NF-48 suggested that it might show strong changes in response to RXR ligands. Furthermore, competence and gene expression in the tail has been well studied, providing benchmark target genes (32, 47–49). Tag-Seq is an RNA sequencing (RNA-seq) methodology that focuses the sequencing at the 3′-end of the gene, which reduces the necessary sequencing depth per sample, thus allowing more libraries to be run per lane of sequencing (50). This allowed us to sequence 60 different libraries in a single lane of next-generation sequencing with very good results. Furthermore, the allo-tetraploid genome of X. laevis (51) varies most in the untranslated regions of genes, so we expected that sequencing only at the 3′-end of genes might make allele calls more accurately.

We report here that Bex and TBT induced almost identical transcriptomes and that a set of transcription factors was regulated in a manner that accurately predicted the morphological phenotypes. However, other genes that Bex and TBT similarly potentiate are not necessarily the same set of genes that UVI inhibits. The data revealed a core set of target genes that warrant further investigation to determine the role of RXR ligands in competence for TH signaling during development.

Methods

Reagents

T3, TBT, and dimethyl sulfoxide (DMSO) were purchased from MilliporeSigma (St. Louis, MO), Bex and UVI from Tocris Bioscience (Bristol, United Kingdom), and tricaine methanesulfonate (MS-222) from Western Medical Supply (Arcadia, CA).

Animal use and care

The laboratory has an approved University of California (UC), Davis Institutional Animal Care and Use protocol that covers the husbandry, mating, and ligand treatments of both wild-type and transgenic animals.

Tadpole precocious metamorphosis assay

Tadpoles 1wk-PF were raised, treated, and fixed for photography as previously described (31, 33). Morphology photos were taken using a Leica MZLFIII microscope and either a Leica LEI750 camera or a Leica DFC3000 G camera (Leica Biosystems, Buffalo Grove, IL). To compensate for differences in sensor size between the two cameras, each clutch was normalized to vehicle controls. Box and whisker plots were computed in GraphPad Prism 7 (GraphPad Software, La Jolla, CA); boxes represent the 25th to 75th percentiles with the bar at the median; whiskers are maximum and minimum values. For statistical analysis, each animal was treated as an individual (n = 15 to 20) and three to four clutches (five animals per clutch) were assayed independently to control for clutch-to-clutch variability. Two-way ANOVA and one-way ANOVA with a Dunnet or Sidak multiple comparisons test (MCT) were performed using GraphPad Prism 7.

Transgenic tadpole luciferase reporter assay

Tadpoles 1wk-PF were raised, sorted for GFP⁺ expression in the eye lens, treated, and assayed for luciferase (Luc) activity as previously described (31, 33). For statistical analysis, each animal was treated as an individual (n = 15), and 3 clutches (5 animals per clutch) were assayed independently to control for clutch-to-clutch variability. Two-way ANOVA and one-way ANOVA using the Sidak MCT were performed using GraphPad Prism 7.

NF-54 tadpoles were treated for 2 days in 0.1X Marc Modified Ringers (MMR) with compounds as for 1-week-old tadpoles, except there was 50 mL of rearing water per tadpole. After 2 days’ treatment, tadpoles were anesthetized with 0.1% MS-222 (Western Medical) and 0.1% sodium bicarbonate in 0.1X MMR; the posterior one-third of the tail, the brain, and the hind limbs were harvested, with the tails and hind limbs minced on ice before freezing and assayed as described (31, 33). For statistical analysis, each animal was treated as an individual (n = 9 per treatment) and two clutches (4 animals in one clutch and 5 in the other) were assayed independently. One-way ANOVA using the Sidak MCT were performed using GraphPad Prism 7.

Ligand treatments, RNA extraction, and 3′-Tag-Seq

Six independent bioreplicates (clutches) were treated with all ligand treatment groups, and total RNA was isolated from tails, one bioreplicate at a time, as follows. Pools of 15 tadpoles 1wk-PF were treated with ligands in their rearing water [0.1× MMR: 0.5 mM HEPES (pH, 7.4; 10 mM NaCl; 0.2 mM KCl; 0.1 mM MgSO₄; and 0.2 mM CaCl₂)] at 8 mL of rearing water per tadpole. Ligands were suspended in DMSO, and the DMSO concentration in the rearing water was 0.05%. Ligand treatment groups were vehicle (DMSO only), 25 nM T3, 30 nM Bex, 25 nM T3 + 30 nM Bex, 375 nM Bex, 25 nM T3 + 375 nM Bex, 1 nM TBT, 25 nM T3 + 1 nM TBT, 600 nM UVI, and 25 nM T3 + 600 nM UVI. After 2 days of treatment, tadpoles were anesthetized with 0.1% MS-222 and 0.1% sodium bicarbonate in 0.1X MMR and decapitated with brain destruction; tails were isolated into 0.35 mL of Buffer RLT Plus from the RNeasy Plus Mini Kit (Qiagen, Germantown, MD) and stored at −80°C for <3 weeks.

The dissection order was randomized for each bioreplicate. Tail tissue was disrupted and homogenized by bead beating with two 0.125-inch stainless steel beads for 30 seconds at the highest setting in a Mini-BeadBeater (Biospec Products, Bartlesville, OK). Total RNA was extracted using the RNeasy Plus Mini Kit per the manufacturer’s instructions. The extraction order was randomized for each bioreplicate. RNA quality was assessed with the LabChip GX (Caliper Life Sciences, Hopkinton, MA), and RNA quality scores ranged from 8.7 to 9.6. For each sample, 500 ng total RNA was converted into 3′-Tag-Seq libraries by the UC Davis DNA Technologies Core using the QuantSeq 3′ mRNA-Seq Library Prep Kit FWD (Lexogen, Vienna, Austria) per the manufacturer’s instructions. Library quality was assessed with the LabChip GX. All 60 libraries (10 treatments × 6 bioreplicates) were sequenced on one lane of an Illumina HiSeq 4000 using 90-bp single-end reads (Illumina, San Diego, CA).

Tag-Seq transcriptome analysis

Alignment and differentially expressed gene (DEG) analysis were performed by the UC Davis Bioinformatics Core. Sequence reads for each sample were aligned to the X. laevis J strain version 9.1 genome, together with the “XL9_1_v20161019_allTranscripts” annotation (ftp.xenbase.org/pub/Genomics/JGI/Xenla9.1/) using STAR v.2.5.2b (52), which also generated raw counts per gene that were the input into the statistical analysis. Before analysis for DEGs, genes with expression of less than three counts per million reads in all samples were filtered, except for thibz.L, a well-documented, direct TR target gene involved in metamorphosis, which fell below the expression cutoff. The thibz.S allele had not been annotated to the genome at the time of analysis; our quantitative RT-PCR (qRT-PCR) primers amplify both allo-alleles. Differential expression testing was conducted using a single-factor ANOVA model in the limma-voom Bioconductor pipeline (using edgeR version 3.16.5, limma version 3.30.13, in R version 3.3.3). Statistical significance was defined as a Benjamini-Hochberg false-discovery rate adjusted P value <0.05. DEGs were expressed as the log2 fold change (log2FC) in expression compared with vehicle treatment or to T3 treatment. A multidimensional scaling (MDS) plot was generated from the normalized read counts using edgeR (plot.MDS) and was visualized for publication using ggplot2 (version 2.2.1). Two-way Venn diagrams of statistically significant DEGs (adjusted P value <0.05) were generated using the R package VennDiagram (version 1.6.18). For the heat map, DEGs in comparison with vehicle treatment (adjusted P value <0.05) from treatment with T3, T3-UVI, T3-LoBex, or T3-TBT (see Table 1 for treatment designations) were arrayed according to log2FC in expression by T3 treatment. The difference log2FC was computed for T3-UVI, T3-LoBex, and T3-TBT vs T3 and visualized as a heat map using R (version 3.4.3) ggplot2 (version 2.2.1.9). Tag-Seq data were submitted to the Sequence Read Archive with the BioProject ID PRJNA431605.

Table 1.

Treatment Scheme and Nomenclature for Treatment Groups

Abbreviation	Ligand 1	Ligand 2
DMSO	Vehicle	Vehicle
T3	25 nM T3	Vehicle
UVI	600 nM UVI	Vehicle
T3-UVI	25 nM T3	600 nM UVI
TBT	1 nM TBT	Vehicle
T3-TBT	25 nM T3	1 nM TBT
LoBex	30 nM Bex	Vehicle
T3-LoBex	25 nM T3	30 nM Bex
HiBex	375 nM Bex	Vehicle
T3-HiBex	25 nM T3	375 nM Bex

Time course of gene expression by qRT-PCR

Treatments and RNA extractions were carried out as for the Tag-Seq experiment, except that tails from pools of 15 tadpoles were harvested at 12 hours, 2 days, and 4 days. Four different clutches were independently assayed. Total RNA was quantified using a NanoDrop (Thermo Fisher Scientific, Waltham, MA). cDNA was synthesized from 300 ng of total RNA using the High-Capacity cDNA Reverse Transcription Kit (Thermo Fisher Scientific). Quantitative PCR was performed using 0.5 μL of cDNA (from a 20-μL total reaction) in a 10-μL reaction using the LightCycler 480 SYBR Green I Master (MilliporeSigma) in a Roche LightCycler 480. The X. laevis csnk1a1 gene was used as a normalizer. It was chosen from a list of least-changed genes of midlevel expression in the Tag-Seq experiment; the primers performed with 96% efficiency and gave a single peak on a melting curve. Statistics were performed on a log2 transform of fold change results using two-way ANOVA analysis with a Tukey MCT in GraphPad Prism. Sequences for the primers used for quantitative PCR are given in Table 2.

Table 2.

Primer Sequences for Quantitative PCR

Gene	Forward	Reverse	Product Size, bp	Allele Specificity
adamts4	TGACTCAACCTCCGTGTGTG	TGTTGTAGCCGTATCTGGGC	164	Both
csnk1a1	GAGACAACCGGACGAGACAG	TGTTGCCGCCTTTAATCCCT	191	Both
dio3	CTACGGTGCCTACTTTGAGAGACTT	GCCACATCCTCAGTTCAGAGATC	107	l
fap.L	GTTGTAACTCTGCTAATAGTTACTGT	CGGATATATTCGTTCTCTGAAATCCA	152	No S-allele
fosl2.L	GAGGGAGATCACACTCAGCG	GGTCTGGTCTTGTCCAACTCA	127	l
junb.L	CGGATTGCAAAGCACGACTT	AGAGTTGCCCTCCGTTAGGA	166	l
mafb.L	AGCTCCAAAGGGAAAGCTTCA	AGCTGATATCCACCCCCTCA	167	l
mmp11.L	GCCGGCTCATGTTTCTTACC	GCAGTTCTACTGATCCCATTGC	129	l
mmp13l	GAAGAAGCCAGGACCTTGGAT	CAAATTGCAGAGCTCCGTTGA	133	Both
mxd1.L	TTCTTCCAGCAGTCTGTTCCC	AGCAGATTTGGCTGATGGGTT	101	l
sox4.L	GTTCCTGCCCGGCACTGTAA	TGCTGCAACAGTCCCTCTATC	206	l
thibz	TGAACCTCAACCAATGCCTCA	CCAGAAGCACCTCCCTTAAACC	150	Both
thrb	GTGCCAGGAAGGTTTCCTCTT	GGTCGGTGACTTTCATCAGCA	100	Both

Results

RXR agonism potentiated and antagonism inhibited T3-induced morphological changes and a TRE-driven Luc reporter

To determine the effect of RXR ligands on TH-induced metamorphic pathways, we treated X. laevis 1wk-PF tadpoles (NF-48) for 5 days with vehicle or T3 in the presence or absence of various RXR ligands. Figure 1a presents representative dorsal head photos of tadpoles from a single clutch. Comparison of vehicle vs 10 nM of T3 shows that T3 treatment resulted in outgrowth of Meckel cartilage (involved in lower jaw development), increased brain width at the tectum, and resorption of the gills. Tadpoles treated with the RXR agonist Bex or the antagonist UVI in the absence of T3 did not show morphological changes significantly different from those with vehicle (not shown; see quantitation that follows). However, cotreatment of T3 with 30 nM Bex potentiated the morphological changes induced by T3. In contrast, cotreatment of T3 with 600 nM UVI resulted in inhibition of all metamorphic morphological changes. We quantified changes in head area of tadpoles from multiple clutches as a proxy for gill resorption over a range of T3 concentrations (Fig. 1b). In the top panel, 30 nM Bex in the absence of T3 was indistinguishable from vehicle treatment in animals; however, in the presence of T3, Bex cotreatment significantly potentiated the effect of T3 alone. UVI (bottom panel) significantly inhibited the T3-induced decrease in head area to the extent that 25 nM T3 in the presence of UVI was necessary to get a response significantly different from that of vehicle (Dunnet MCT, P = 0.0085). Tail resorption was less sensitive to T3 than gill resorption was, but 1wk-PF tadpoles did significantly resorb tail in response to higher concentrations of T3 (Fig. 1c). In agreement with head area data, 30 nM Bex significantly potentiated tail resorption at all concentrations of T3, although not causing resorption when used alone (Fig. 1c, upper panel). UVI significantly abrogated tail resorption (Fig. 1c, bottom panel).

Figure 1.

Bex potentiated and UVI inhibited the effects of T3 on morphological changes in 1wk-PF tadpoles treated for 5 days. (a) Dorsal head photos of tadpoles from a single clutch treated with the indicated compounds. (b and c) Treatment with T3 caused a decrease in (b) head area or (c) tail length as a function of T3 concentration (open circles). (b) 30 nM Bex (top panel, filled circles) potentiated T3-induced gill resorption (two-way ANOVA: T3, P < 0.0001; Bex, P < 0.0001; interaction, P < 0.0001), and 600 nM UVI (bottom panel, filled circles) inhibited it (two-way ANOVA: T3, P < 0.0001; UVI, P < 0.0001; interaction, P < 0.0001). (c) 0 nM Bex (top panel, filled circles) potentiated T3-induced tail resorption (two-way ANOVA: T3, P < 0.0001; Bex, P < 0.0001; interaction, P < 0.0001), and 600 nM UVI (bottom panel, filled circles) inhibited it (two-way ANOVA: T3, P < 0.0001; UVI, P = 0.0007; interaction, P = 0.0113). Error bars represent 95% CI, and *represents significance from vehicle control (one-way ANOVA with Dunnet MCT: *P < 0.05; **P < 0.01; ***P < 0.001). (d) TRE-Luc reporter activity in 1wk-PF tadpole heads treated for 2 days with the indicated treatments. Bars represent fold activation over vehicle. Black bars, T3-only; slashed bars, T3 + Bex; white bars, T3 + UVI. Error bars represent SEM. *Represents significance from vehicle control (one-way ANOVA with Sidak MCT: *P < 0.05; **P < 0.01; ***P < 0.001), and ^#indicates significance over T3 alone (two-way ANOVA with Sidak MCT: ^###P < 0.001).

In earlier reports, we described transgenic tadpoles that express the firefly Luc gene under the regulation of TREs from the X. laevis thibz gene (53), showing that this reporter construct is a good surrogate for T3-induced gene expression changes by TRs (33). We tested whether the morphological potentiation and inhibition of T3-signaling we found with Bex and UVI, respectively, were maintained in a gene expression‒based end point. The 1wk-PF tadpoles were treated as for morphology experiments, except that treatments lasted 2 days instead of 5 days. Whole heads were assayed for Luc expression. Figure 1d shows that no signal was obtained in the absence of T3 but that all T3 concentrations resulted in significant activation of the Luc reporter. Cotreatment with 30 nM Bex potentiated the T3 signal significantly at all T3 concentrations, whereas cotreatment of T3 with 600 nM UVI abrogated T3-induction to vehicle-treated levels.

We quantified changes in head area and tail resorption over a concentration range of Bex (Fig. 2). In mammalian cell transient transfection assays, Bex (LGD1069) showed an EC₅₀ for RXRs of 20 to 30 nM (54). At 5 nM T3, 1 nM Bex was sufficient to significantly potentiate gill resorption over T3 alone (Fig. 2a). Maximal Bex concentrations potentiated T3 signal of 5 nM T3 to that of 25 nM T3 in the absence of RXR agonist, a result we also observed when using trialkyltins to potentiate T3 signaling (33). At 25 nM T3, 3 nM Bex significantly increased tail resorption compared with T3 alone (Fig. 2b). Over the same concentration curve using the Luc reporter, 10 to 100 nM Bex resulted in significantly higher Luc activity than T3 alone (Fig. 2c). However, at higher concentrations of Bex the results became more variable. At 300 nM of Bex, there was no potentiation beyond the T3-only signal. Cotreatment of 375 nM Bex did potentiate significantly beyond the T3-only signal (Fig. 2c). However, on a clutch-by-clutch basis, the significance of cotreatment with 300 or 375 nM Bex did vary more than the lower Bex dose did (data not shown). In the absence of T3, 375 nM Bex did not activate the reporter. Fig. 2d shows the inhibition of T3-activated Luc over a UVI concentration curve.

Figure 2.

Bex potentiated T3 effects in a concentration-dependent manner. (a and b) Changes in (a) head area or (b) tail length as a function of Bex concentration. See “Methods” section for box-and-whisker plot specifics. *Represents significance from vehicle control (one-way ANOVA with Sidak MCT: **P < 0.01; ***P < 0.001; ****P < 0.0001), and ^#indicates significance over T3 alone (one-way ANOVA with Sidak MCT: ^#P < 0.05; ^##P < 0.01; ^###P < 0.001). (c and d) TRE-Luc reporter activity in 1wk-PF tadpole heads treated for 2 days over a (c) Bex or (d) UVI concentration range. Bars represent fold activation over vehicle treatment, and error bars represent SEM. Statistics are represented as in (a and b).

Transcriptomic analysis of RXR ligand effects of T3-induced differential gene expression in tails

Taken together, our morphology and Luc reporter data suggest that RXR agonists increase the competence of 1wk-PF tadpoles to respond to T3. Conversely, RXR antagonism essentially eliminated T3 competence at this early stage of larval development. To determine the genetic pathways RXR ligands are affecting, we undertook a transcriptomic analysis of differentially expressed genes arising from treatment of 1wk-PF tadpoles with the RXR ligands in both the absence and presence of T3. We pooled groups of 15 tails per treatment per clutch, and we used six different clutches of animals, all raised and treated independently, to generate six bioreplicates. Our previous results using TBT (33) prompted us to include TBT in our experiment. Table 1 outlines our treatment scheme and provides a simple nomenclature for the different treatment groups. The tadpoles take up the compounds from their rearing water; therefore, we do not know their circulating concentrations nor their cellular uptake patterns or concentrations. We chose 25 nM T3 because that concentration routinely caused a significant change in tail resorption in the absence of an RXR agonist at this developmental stage. Although 25 nM T3 is high for a physiological concentration of circulating TH in mammals, the intracellular concentration of active T3 is dependent on the concerted activities of activating and deactivating deiodinases and is not known at the single cell level.

Because our previous results using TBT and triphenyltin suggested that they could function in tadpoles as RXR agonists (33), we wanted to include TBT in our transcriptomic analysis to directly compare it with Bex, a bona fide designed RXR agonist. For TBT, a 1-nM concentration gave a maximal response with no morphologically noticeable general toxicity; 30 nM Bex (designated LoBex), which is in range with the measured Kd for each RXR (54), gave results nearly identical to those of 1 nM TBT in the Luc assay and morphological assays (33). We also chose to use 375 nM Bex (designated HiBex), which gives morphological results very similar to those of 300 nM Bex (not shown) but showed more variable Luc reporter activation than 30 nM Bex (Fig. 2). Further, we had preliminary data that suggested other genes might function differentially between LoBex and HiBex conditions [see iodothyronine deiodinase 3 (dio3) in a later section]. Finally, we chose 600 nM UVI because that concentration was maximally effective at inhibiting T3 action in the absence of any mortality.

Tag-Seq methodology (50) allowed all 60 libraries to be sequenced on one lane of an Illumina HiSeq 4000 using 90-bp single-end reads, which resulted in 340 M total reads and 5.7 M mean reads per library. In all, 93% of reads mapped to the X. laevis J strain version 9.1 genome, and 66% mapped uniquely to a gene. Importantly, because of the allo-tetraploidy of the X. laevis genome, only 4% of reads mapped to more than one gene; 23% mapped to no currently known feature of the genome. Treatment with T3 resulted in 3217 DEGs, which were evenly distributed between activated (48%) and repressed (52%) genes (Fig. 3a; Supplemental Table 1). In the absence of T3, UVI, TBT, and LoBex caused very little change to gene expression (Supplemental Tables 2‒4). Cotreatment of T3 with TBT or LoBex increased the number of DEGs essentially equally (6900 for T3-TBT and 6770 for T3-LoBex) compared with T3-alone (Fig. 3a; Supplemental Tables 5 and 6). The ratio of activated vs repressed genes for each cotreatment remained the same as for T3-alone. In contrast, UVI cotreatment decreased the number of T3-induced DEGs to 2442 (Supplemental Table 7). Therefore, as expected, the morphologically potentiating RXR agonists increased the number of DEGs in a T3-dependent manner, whereas inhibitory UVI decreased T3-induced DEGs.

Figure 3.

Summary of Tag-Seq results. (a) Total numbers of DEGs for each treatment group (see Table 1 for designations) divided by activated (upregulation, red) and repressed (downregulation, blue). Numbers over bars for UVI, TBT, and LoBex indicate the number of genes in each bar. (b) Multidimensional scaling plot showing the results for each clutch labeled by treatment. (c) Venn plots showing the high commonality between T3-LoBex and T3-TBT treatments when compared with vehicle treatment (top, purple) or with T3-alone treatment (bottom, green). (d) Venn plot showing the overlap between genes regulated by T3 alone vs HiBex alone. (e) Tag-Seq results for retinoid degradation enzymes activated by HiBex. ^#Significantly different from HiBex alone. (f) Heat map arraying log2FC compared with T3-alone treatment of T3-UVI, T3-LoBex, and T3-TBT. The genes arrayed were significantly differentially expressed in at least one of the four treatment groups (T3-alone, T3-UVI, T3-LoBex, or T3-TBT). Genes are arrayed from highest T3-alone activation (top) to most repressed by T3-alone (bottom).

MDS (Fig. 3b) showed that the libraries clustered on the basis of the treatment, not the clutch. In the absence of T3, the RXR ligands (with the exception of HiBex) clustered closely with the vehicle-treated group, in accordance with the low number of DEGs for each treatment. T3 treatment created a distinct cluster, and the cotreatment T3-UVI brought the cluster back toward the vehicle cluster, which agrees with the decreased number of DEGs. Interestingly, the clusters for T3-LoBex and T3-TBT overlapped each other, suggesting that they affected gene expression very similarly. A Venn plot (Fig. 3c, top) of shared DEGs shows an 80% overlap between T3-TBT and T3-LoBex when compared with vehicle. Furthermore, the sum of shared upregulated DEGs and downregulated DEGs equaled the total number of DEGs, indicating that TBT and LoBex always affected T3 induction in the same direction. When we compared the same two treatment groups to T3 treatment instead of to vehicle (Fig. 3c, bottom), >20% of the DEGs were statistically different from T3-alone induction, and ∼60% of the DEGs were shared between the two RXR agonists.

Results before embarking upon Tag-Seq indicated that LoBex and HiBex could affect transcription of certain genes differently, as seen with the Luc reporter (Fig. 2; see dio3 in a later section). Therefore, we tested both Bex concentrations by Tag-Seq. The MDS (Fig. 3b; Supplemental Tables 8 and 9) showed that HiBex, in contrast to LoBex, produced a distinct cluster and resulted in 2939 DEGs (Fig. 3a). Figure 3d shows that HiBex and T3 shared 1177 DEGs (37% of T3 DEGs and 40% of HiBex DEGs) (Supplemental Table 10). Interestingly, the highest activated gene by HiBex was cyp26a1.L, which encodes a P450 enzyme that degrades all-trans retinoic acid (55, 56). There are three different cyp26 family members: cyp26a1, cyp26b1, and cyp26c1, all of which were highly induced by HiBex (Fig. 3e). T3 alone did not affect cyp26 expression, except for repressing expression of cyp26b1.S. Cotreatment of T3 with HiBex resulted in significant attenuation of the strong HiBex induction of both alleles of cyp26a1 and cyp26b1. T3 did not significantly affect the induction of either allele of cyp26c1, which showed the lowest induction by HiBex. HiBex also strongly induced both allo-alleles of dhrs3 (Fig. 3e), which has been shown to attenuate retinoic acid signaling in Xenopus embryos (57).

Because the morphological data show inhibition of T3-induced resorption phenotypes by UVI and potentiation of resorption by LoBex and TBT (33), we compared gene expression profiles using a heat map, ranking the genes from most T3-activated to most T3-repressed. All genes that were significantly differentially expressed compared with vehicle in at least one of the four treatment groups (T3, T3-UVI, T3-TBT, or TB-LoBex) were included, and T3-HiBex genes that were in one of those four treatment groups were also used (i.e., for T3-HiBex, genes activated only by HiBex were excluded because we were interested in the effect of T3 induction) (Fig. 3f). A black line separates genes activated (above the line) vs repressed (below the line) by T3. The heat map strikingly shows the opposite effects of UVI vs Bex or TBT on gene expression. Genes activated by T3 are more likely to be repressed compared with T3-alone with cotreatment with UVI (blue) and more highly activated with cotreatment with TBT or Bex (orange). Interestingly, the opposite also holds true: genes repressed by T3 (genes below the black line) are more repressed with TBT or LoBex and are less repressed or even activated by cotreatment with UVI.

RXR agonists potentiated extracellular matrix‒modifying protease expression

Previous work from many groups demonstrated the strong activation of extracellular matrix metalloproteases (MMPs) and other secreted proteases during tail resorption, and our Tag-Seq data are in good agreement with these reports (47, 58–61). Given the potentiated resorption morphologically seen upon RXR agonist cotreatment with T3 and the near-complete inhibition of resorption by UVI, we predicted that UVI would inhibit the transcription of these proteases and that Bex and TBT would enhance their expression. However, the entire model did not hold true. We found that TBT and Bex did strongly enhance the transcription of many proteases known to be involved in tissue remodeling; however, UVI did not significantly inhibit their expression. By Tag-Seq analysis, T3 treatment significantly upregulated six proteases (greater than twofold) that were also affected by all RXR agonists: mmp9, mmp11, mmp13l, mmp19, htra1, and fap (Fig. 4). Because UVI, TBT, and LoBex alone did not significantly change the expression of any of these genes, we omitted them from the figure for clarity. HiBex alone did significantly activate the mmp11.L allele and fap. Cotreatment of T3 with all three RXR agonist conditions significantly potentiated T3-induction (Fig. 4; # over the bar indicates significant difference from T3-alone). However, UVI did not significantly inhibit any of the proteases. This was true for both allo-alleles, with the exception of the mmp9.S allele, which was expressed but did not show a significant fold change even with T3-alone.

Figure 4.

RXR agonists potentiated extracellular matrix proteases, but UVI did not inhibit T3 induction. Tag-Seq results are expressed as the log2FC compared with vehicle. The absence of a bar indicates that under that treatment the gene was not differentially expressed. None of the proteases were differentially expressed by TBT, LoBex, or UVI as the sole treatment, and so they were not included on the graphs. ^#Indicates significance in comparison with T3-alone. Bars are labeled in accordance with Table 1. N.S. FC, no statistical fold change.

To explore the kinetics of RXR ligand potentiation or inhibition, we performed a time course experiment using qRT-PCR and tested selected genes under different treatment conditions. Tag-Seq treatments were for 2 days; for the time course, we repeated that time point, and we also chose 12 hours to test initial responses and 4 days to test when morphological changes were first evident. We did not treat with any of the RXR ligands on their own for two reasons. First, the RXR ligands had not caused any morphological changes after 5 days of treatment (Figs. 1 and 2), so we had no expectation that they would affect gene expression at 4 days of treatment. Second, we were most interested in the effects of the RXR ligands on TH signaling pathways. We cotreated with UVI to see whether a longer incubation resulted in inhibition, and we also cotreated with both LoBex and TBT to compare the pharmacological agonist with the environmental pollutant.

Figure 5 shows the results of the time course, with the 12-hour time point also expanded as a histogram for clarity. Stromelysin-3, a known actor in tissue remodeling and resorption (60), is encoded by the mmp11 gene; primers to the collagenase-3 gene (59) aligned to both allo-alleles of the mmp13l gene (not to mmp13). For both genes after 12 hours of treatment, LoBex had already significantly potentiated expression compared with T3-alone (Fig. 5a and 5b; # indicates significance compared with T3, and * indicates significance compared with vehicle). At 2 days of treatment, the qRT-PCR results were in very good agreement with the Tag-Seq results, for which the average expression of these genes ranged from 0.22 in log2 counts per million reads for mmp13.L to 3.7 for mmp11.L (compare Figs. 4 and 5). This potentiation was maintained at 4 days of treatment; however, the additional treatment time did not result in significant inhibition by UVI. With Tag-Seq, fap expression was potentiated at 2 days by LoBex and TBT; however, during the time course, the apparent potentiation was not statistically significant (Fig. 5c). In comparison, the adamts4 gene—which T3 activated in the Tag-Seq analysis but the RXR ligands did not affect (not shown)—was potentiated at 4 days by both LoBex and TBT (Fig. 5d); however, UVI inhibition was again not significant. Most of these proteases require activation through proteolytic cleavage of a propeptide, although not all the activating proteases are known. However, stromelysin-3, encoded by mmp11, has been shown to be activated by furin (62). Both allo-alleles of furin were expressed in the tadpole tails, and neither was differentially expressed by any of the treatments, nor were any of the timp (tissue inhibitor of metalloproteases) genes differentially expressed.

Figure 5.

RXR agonists maintained potentiation of T3 induction of extracellular matrix proteases over a time course, whereas UVI did not inhibit. Tails from tadpoles 1wk-PF were treated with the indicated compounds for the specified times before extraction of RNA and qRT-PCR. (a‒d) Data are presented as the log2 transform of normalized relative expression. Error bars represent the SEM. For all graphs, symbol labels are in accordance with Table 1. Significance was determined using two-way ANOVA and the Tukey MCT. Histograms show the log2FC for the 12-h time point. *Represents significance from vehicle control: *P < 0.05; **P < 0.01; ***P < 0.001. ^#Represents significance from T3-alone: ^#P < 0.05; ^##P < 0.01; ^###P < 0.001.

We concluded that the potentiation of the T3-induction of extracellular proteases by LoBex and TBT is in good agreement with the observed morphological changes. However, the nonsignificant trend in UVI-induced inhibition of protease expression most likely cannot fully explain the abrogation of morphological change seen with T3-UVI cotreatment. Therefore, we specifically asked which genes more closely followed the ligand-induced phenotypic patterns.

A suite of transcription factor genes closely followed the morphological phenotype

We determined the genes that followed the phenotype of T3-regulated TBT or LoBex potentiation of T3 action and UVI inhibition of T3. For T3-upregulated DEGs, 51 genes were significantly both downregulated by UVI and upregulated by TBT in comparison with T3, and 45 genes were significantly both downregulated by UVI and upregulated by LoBex in comparison with T3 (Supplemental Table 11). Nine of these genes for each RXR ligand encode transcription factors: thrb, mxd1, sox4, fosl2, nr3c2, mafb, nfib, nfic, and cebpd. Figure 6 shows the Tag-Seq results for eight of the transcription factors (thrb is discussed later). With the exception of the cebpd.L allele, which barely missed the P < 0.05 significance cutoff for activation by T3 (see P values in Fig. 6), both allo-alleles of the transcription factors were significantly affected. In the absence of T3, UVI, TBT, and LoBex treatments did not significantly affect the transcription of these genes, so they have not been included in the graphs for clarity. With Tag-Seq, both allo-alleles of all eight genes (with the exception of the l-allele for mxd1, adjusted P = 0.138, and the previously mentioned cebpd.L) were significantly inhibited by UVI cotreatment (# indicates significance compared with T3 induction). Further, all three RXR agonist conditions significantly potentiated T3 induction compared with T3 alone. sox4.L, mafb.L, and nfib.L were all also significantly activated by HiBex in the absence of T3; however, that activation was significantly less than the T3 induction. Of note, HiBex potentiated T3 induction to the same level as LoBex, which is different from the Luc reporter, for which 300 nM Bex (and 375 nM Bex; not shown) actually abolished significant potentiation (Fig. 2c).

Figure 6.

Several transcription factors were activated by T3, and that activation was inhibited by cotreatment with UVI and was potentiated by cotreatment with RXR agonists. Tag-Seq results are expressed as log2FC compared with vehicle. The absence of a bar indicates that under that treatment the gene was not differentially expressed. None of the transcription factors were differentially expressed by TBT, LoBex, or UVI as the sole treatment, and so they have not been included on the graphs. ^#Indicates significance in comparison with T3-alone. Bars are labeled in accordance with Table 1.

We assayed expression of four of these transcription factors by time course to both validate the Tag-Seq results using unique samples and to compare the kinetics of activation among transcription factors of different families (e.g., a Sox family member, sox4, vs a Mad-Max-Myc family member, mxd1). At 12 hours (see histogram expansion of this time point for clarity), T3 induction of mxd1.L was already significant, and TBT and LoBex potentiation were significant compared with T3 (Fig. 7a). The 2-day results are in good agreement with the Tag-Seq data; however, UVI did significantly inhibit T3 induction of the l-allele in the time course, whereas it did not with Tag-Seq. At 4 days of treatment, UVI inhibition and TBT and LoBex potentiation remained significantly different from T3 alone. A similar result was seen when assaying for sox4.L expression over the time course (Fig. 7b). By 12 hours, LoBex and TBT potentiation were already significant. At 2 days of treatment, UVI inhibition was also significant, and this pattern was maintained at 4 days. Time course analysis of fosl2.L expression maintained the inhibition by UVI at 2 and 4 days of treatment; however, potentiation by TBT and LoBex was not significant (Fig. 7c). Unlike mxd1, sox4, and fosl2, mafb was not induced at 12 hours by T3 (Fig. 7d). By 2 days of treatment, potentiation by TBT and LoBex was significant compared with T3, which is in good agreement with the Tag-Seq data. However, in the time course, it took 4 days for inhibition of mafb.L by UVI to become significant compared with the T3 induction.

Figure 7.

RXR agonist potentiation of T3 induction occurred rapidly, whereas inhibition by UVI displayed delayed kinetics. Tails from tadpoles 1wk-PF were treated with the indicated compounds for the specified times before extraction of RNA and qRT-PCR. Data are presented as the log2 transform of normalized relative expression. (a) mxd1.L, (b) sox4.L, (c) fosl2.L, and (d) mafb.L. Error bars represent the SEM. For all graphs, symbol labels are in accordance with Table 1. Significance was determined using two-way ANOVA and Tukey MCT. Histograms show the log2FC for the 12-h time point. *Represents significance from vehicle control: *P < 0.05; **P < 0.01; ***P < 0.001. ^#Represents significance from T3-alone: ^#P < 0.05; ^##P < 0.01; ^###P < 0.001.

TR target genes affected by RXR ligands in a gene-specific manner

Tag-Seq results showed that T3 strongly induced the TRβ gene (thrb;Fig. 8a), RXR agonists significantly potentiated the T3 induction of both alleles, and the l-allele was inhibited by UVI. As expected for a direct target gene, thrb was already significantly induced at 12 hours when assayed in the time course (Fig. 8b); however, only LoBex cotreatment was significantly different than T3 at 12 hours. By 2 days of treatment, the RXR ligands had no significant effect on thrb expression when assayed by qRT-PCR, and this result was maintained at 4 days of treatment. The gene for TRα (thra; Supplemental Fig. 1) was modestly induced by T3, but that induction was not significantly affected by any of the RXR ligands. Neither T3 nor the RXR ligands had much effect on the expression of the RXR genes (Supplemental Fig. 1), except that rxrg.S was downregulated by T3 (rxrg.L was not detected) as described in mammals (63). The only significant effect the RXR ligands had on the expression of the RXR genes was inhibition of the rxra.S allele by TBT (the rxra.L allele was not differentially expressed under any condition).

Figure 8.

Some known direct TR-target genes displayed gene-specific regulation with both Tag-Seq and time course analysis. Tag-seq data are represented as in Figs. 4 and 6, and time course data are represented as in Figs. 5 and 7. (a and b) thrb alleles were modestly but significantly potentiated by RXR agonists with Tag-Seq but not significantly with qRT-PCR over a time course. (c and d) T3 induction of thibz was strongly inhibited by cotreatment with UVI but was not affected by RXR agonist cotreatment. (e and f) T3 induction of dio3.L was significantly inhibited by UVI cotreatment, and HiBex as the sole treatment was a strong inhibitor of dio3.L expression compared with vehicle, which was not completely recovered by T3 cotreatment. *P < 0.05; ***P < 0.001. ^#Represents significance from T3-alone: ^#P < 0.05; ^###P < 0.001.

The thibz (TH/bZIP) is a well-characterized TR-direct target gene (50) that is also a transcription factor. The S allele for thibz was not assigned in the X. laevis version 9.1 genome and so could not be assayed by Tag-Seq, but the l-allele was both significantly induced by T3 and strongly inhibited by UVI (Fig. 8c); however, the RXR agonists did not potentiate T3 induction. Over the time course, at 2 days UVI significantly inhibited T3 induction (Fig. 8d), such that there was no significant increase in thibz compared with 12 hours. At 4 days, UVI still inhibited T3 induction. In agreement with the Tag-Seq data, the RXR agonists did not potentiate T3 action at any time point.

The gene encoding iodothyronine deiodinase 3, dio3, is a known TR target gene (64). With Tag-Seq, the S allele was not detected, but dio3.L was activated by T3 (Fig. 8e), and UVI significantly, but modestly, inhibited T3 induction. TBT and LoBex cotreatment did not potentiate T3 action, and over the time course, UVI, TBT, or LoBex cotreatment had no significant effect on dio3 expression (Fig. 8f). Surprisingly, HiBex in the absence of T3 strongly downregulated dio3 expression by Tag-Seq (Fig. 8e). Although cotreatment of HiBex with T3 activated dio3 significantly compared with vehicle, it did not recover to or potentiate T3-only induction. Fifty-one genes were upregulated by T3 and downregulated by HiBex, but only dio3 and slc16a2.S (Mct8 S-allele) were activated more than twofold by T3 and repressed by >50% by HiBex, and T3-HiBex activation was significantly less than that of T3-alone. Interestingly, slc16a2.L was activated by T3, HiBex, and T3-HiBex all to the same extent (approximately twofold) (Supplemental Fig. 1).

None of the p160 coactivators that can bind TR were differentially expressed under any conditions, and the genes for both NCoR and SMRT corepressors (ncor1 and ncor2, respectively) were modestly activated by T3 with mixed significance by the RXR ligands. Importantly, the direction of corepressor expression was the opposite of what one would expect for potentiating TR activation (T3 activated the corepressors, and some RXR agonists significantly potentiated that) (Supplemental Fig. 1).

UVI inhibited T3 activation of the TRE reporter in TH-competent NF-54 tadpoles

The potentiation of T3 signaling by RXR agonists suggests that RXR agonism jump-starts TH competence in the semicompetent NF-48 tadpoles. We found it interesting that UVI-mediated RXR antagonism abrogated TH competence in NF-48 tadpoles. We hypothesized that UVI treatment of competent tadpoles (NF-54) should inhibit T3 activation of the TRE-driven Luc reporter. In NF-48 animals, we needed to assay the entire head to acquire a measurable signal from a single animal because of the animal’s small size. However, in NF-54 animals we were able to reproducibly extract active Luc activity from the brain, tail (the posterior third), and hind limbs. Figure 9 shows the results of treating NF-54 TRE-Luc‒reporter tadpoles for 2 days with 10 nM T3 or 10 nM T3 plus 600 nM UVI. In all tissues, UVI significantly inhibited T3 induction of the reporter to the extent that the induction was not significantly different from that of vehicle treatment, showing that UVI is effective in tissues undergoing different cell fates (e.g., resorbing a tail vs growing a limb). Extracting active Luc from the different tissues cannot be equally efficient; therefore, we do not consider differences in the level of T3 induction between tissues to be comparable.

Figure 9.

UVI cotreatment with T3 abrogated T3 induction of the TRE-Luc reporter in NF-54 tadpole tissues. TRE-Luc reporter activity in NF-54 tadpole tissues treated for 2 days with the indicated treatments. Bars represent fold activation over vehicle. Black bars, T3-only; slashed bars, T3 + UVI. Error bars represent SEM. *Represents significance from vehicle control (one-way ANOVA with Sidak MCT: ***P < 0.001). ^#Represents significance from T3-alone: ^##P < 0.01; ^###P < 0.001.

Discussion

The poor biological outcomes that result from inappropriate TH signaling during development (1–5) illustrate the importance of identifying compounds that can disrupt TH signaling as well as their mode of action. On the basis of in vitro screening of a mammalian pituitary cell line bearing an integrated TRE-driven Luc reporter, we discovered that RXR-specific [but not retinoic acid receptor (RAR)–specific] agonists could activate the reporter (21). This was unexpected, as the RXR-TR heterodimer in most in vitro cell lines is refractive to the action of RXR ligands, putting it into the category of nonpermissive RXR-NR heterodimers. However, others have shown that pituitary cells show apparent permissiveness (i.e., response to RXR ligands on T3-driven reporters and target genes), although molecular proof remains elusive (20). Furthermore, RXR agonists in vivo repress the HPT axis to the point of overt hypothyroidism (22, 23), which has limited their use pharmacologically as activators for permissive heterodimers in a variety of potential clinical applications. This raises the question of why this might be necessary if RXR agonism does not affect signaling through TRs. In general, evidence for RXR ligands affecting TH signaling outside the HPT axis is lacking.

One of the RXR agonists that activated the TRE-Luc reporter in the mammalian cells was TBT, a prevalent environmental pollutant that has been shown to covalently bind to a conserved Cys residue at the opening to the ligand-binding pocket of RXRs (only RXRs have a Cys in that position) and activate the receptor (65). In an earlier report, we showed that the ability of TBT to affect TH signaling was not limited to in vitro reporter assays in a pituitary cell line but rather that TBT radically potentiated resorption phenotypes in a precocious metamorphosis assay using 1wk-PF X. laevis tadpoles (33). In addition to activating RXRs, TBT has been shown to activate PPARγ (43). Therefore, we decided to use a pharmacological RXR-specific agonist, Bex, in our precocious metamorphosis assay, as we report here. Results with 30 nM Bex were almost identical to those with 1 nM TBT in terms of both morphology and a transgenic frog TRE-Luc reporter response. These results prompted us to test an RXR antagonist, and surprisingly, we found that UVI abrogated the action of T3 in our precocious metamorphosis assays.

To determine how the RXR ligands were affecting TH signaling, we undertook a transcriptomic analysis. We chose to use Tag-Seq in lieu of traditional RNA-seq because the outbred nature of the X. laevis tadpoles suggested that a larger number of bioreplicates rather than greater sequencing depth would be most informative. By sequencing only at the 3′-end of genes, Tag-Seq requires less sequencing depth to gain good significance on DEGs (50). A final consideration was that the allo-alleles that are tetraploid in X. laevis are most distinct in the untranslated regions of the genes, and this biological feature might make Tag-Seq a better choice than traditional RNA-seq to definitively align sequencing data. Although side-by-side analysis from split RNA samples was required to test this hypothesis, the fact that only 4% of reads mapped to multiple genes suggests that Tag-Seq worked well for this purpose. In addition, we looked for different patterns of expression by our different treatment groups across allo-alleles, which could indicate differential regulation. In most cases, the two alleles appeared to be regulated similarly; however, although the S-allele of mmp9 was expressed, unlike the l-allele, it was not regulated by T3.

Although the “classic” model for gene regulation by TRs is that TRs repress genes in the absence of T3 and activate genes in the presence of T3, transcriptomic studies have shown large numbers of genes downregulated by T3 addition (66, 67). We found near-equal percentages of genes regulated in each direction by T3. Interestingly, our heat map analysis showed that the RXR agonists potentiated and UVI inhibited T3 action, even when T3 was repressing rather than activating gene expression. This suggests that the mechanisms through which the RXR ligands are working are not specific to the direction of the T3 response but rather to TH signaling as a whole.

Very few genes were differentially expressed in the tails of tadpoles treated with UVI in the absence of T3, suggesting that liganded RXR signaling is not prevalent during this stage of development. UVI has been implicated as a teratogen of Xenopus tropicalis embryogenesis (68); however, it was not clear whether the results were due to general toxicity or to RXR antagonism. UVI has also been shown to activate a PPARγ reporter from X. tropicalis (69); however, the EC₅₀ was 12.6 μM, which is >20-fold higher than the concentration we used. Furthermore, neither allele of PPARγ showed expression in the NF-48 tails with our Tag-Seq experiments. TBT and LoBex alone also did not affect the expression of many genes; however, HiBex did. Of note, HiBex alone did not result in any noticeable changes in morphology nor in any lethality or general toxicity-associated behavior (e.g., swimming in small circles or ventral side up). This is interesting because HiBex did highly activate dhrs3 and the cyp26 genes, which are the primary means of degrading retinoic acids (55–57); this suggests that repression of signaling through RARs does not result in overt defects at this stage. In addition, Bex has a >1 μM K_d for binding to RARs (44), suggesting that RXR signaling can induce the degradation of retinoic acids. HiBex cotreatment with T3 did not potentiate T3 action more than LoBex, suggesting that the HiBex-alone activation of T3-induced genes was independent of potentiation of the T3 signal.

Going into the transcriptomic analysis, we found that, on the basis of the morphological results, the simplest model for RXR agonist potentiation and RXR antagonist inhibition of TH signaling was the potentiation and inhibition, respectively, of transcript levels for the proteases that carry out tail resorption. Although transcription of all the proteases was potentiated by the RXR agonists, UVI cotreatment did not significantly repress transcription of any of them. This surprised us because resorption was essentially prevented during the course of our assay. Previous research has shown that mmp11 and mmp13l genes are expressed in fibroblasts that surround the notochord and in subepidermal fibroblasts, which then invade the neighboring tissue to promote the resorption process (47). Because UVI did not significantly inhibit their transcription but resorption was severely inhibited morphologically, it may be that UVI was acting to inhibit transcription factors that drive the fibroblast invasion program. In addition, the proteins responsible for the posttranslational activation of the MMPs were not differentially regulated. Independent of the mechanism by which UVI cotreatment prevents resorption, the protease expression data also indicate that RXR agonist potentiation and RXR antagonist action may not entirely operate through the same molecular mechanisms, or we would expect a higher correlation of reciprocal phenotypes, particularly inhibition.

The heat map in Fig. 3f illustrates the global nature of the effects of RXR ligands on T3-regulated gene expression. Overwhelmingly, HiBex, LoBex, and TBT potentiated the effect of T3, whether that effect was activation or repression, and UVI inhibited T3 action, again independently of whether T3 activated or repressed a gene. This global result suggests that the effects of RXR ligands occur early in the TH signaling pathway. When we turned our focus to the genes that were activated by T3, inhibited by UVI, and potentiated by Bex and/or TBT, we found ∼20% of the genes were transcription factors, which is in keeping with expected early expression genes. Many of these transcription factors have been implicated in gene programs of natural metamorphosis, including sox4, nfib, fosl2, mxd1, and nr3c2 (49, 70). For example, the Shi Laboratory recently showed that mxd1 (Mad1), the transcriptional repressor that competes with Myc for binding to Max, is a T3-induced gene of intestinal maturation during metamorphosis, promoting apoptosis of larval intestinal epithelium (71). Intriguingly, although absolute fold activation cannot be directly compared in different tissues from different laboratories, the fold activation of mxd1 by T3 alone in the NF-48 tadpoles we used is very low compared with the natural metamorphic levels measured in the Shi Laboratory (71); however, the TBT or Bex potentiation of T3 we saw is very much in line with the levels measured during natural metamorphosis. Taken together, our results indicate that RXR agonists potentiate T3 signaling by increasing the tail TH competence, and it raises the question of whether RXR agonism is a bona fide competence factor during natural metamorphosis. The inhibition of dio3 by HiBex, whose enzymatic activity is strongly linked to its level of transcription (64, 72), fits with this hypothesis. However, LoBex and TBT potentiated tail resorption without repressing dio3, indicating that repression of dio3 is not the only mechanism by which RXR agonists potentiated T3 action.

We measured gene expression after 2 days of treatment with ligands and measured morphology after 5 days of treatment. The induction of MMP genes known to be active during metamorphic resorption at 2 days followed by significantly measurable resorption at 5 days suggests that this is an informative choice in time points. However, changes in transcription do not necessarily correlate with changes in protein expression. Loss-of-function and gain-of-function studies are required to prove the role of a particular gene in a pathway such as metamorphosis. Our Tag-Seq data provide several genes to test with these methods. For example, for the transcription factors in Figs. 6 and 7 that were significantly inhibited by cotreatment with UVI and significantly potentiated by cotreatment with TBT or Bex, we can hypothesize that uncoupling their expression from TH signaling (i.e., making their expression constitutive) should result in the morphological effects of RXR agonist treatment in the absence of an RXR agonist. Conversely, using genome editing to make a null allele of a gene that is essential to mediating the signal from RXR agonists should, at a minimum, make cotreatment with an RXR agonist ineffective. Furthermore, such studies can help define the epistatic relationships between the different proteins, which are not understood for TH signaling during development, when a gain-of-function of one gene is combined with a loss-of-function of another.

Our data suggest that the RXR agonists were driving TH competence in the semicompetent NF-48 tadpoles to a more complete competence. We took tadpoles from our TRE-Luc reporter line at the premetamorphic NF-54 stage, when they are fully competent to respond to TH, and treated them with T3 in the presence and absence of UVI for 2 days to see whether the RXR antagonist could inhibit competence. At this later stage, the animals had begun to develop their hind limbs and were large enough to harvest individual tails, brains, and hind limbs to assess Luc activation in tissues undergoing different fates. Our results showing that UVI abrogated T3 induction of the Luc reporter in all three tissues indicate that the effect of RXR antagonism is global in terms of tissue and that it is functional in higher-competence tadpoles.

Our other major finding is that TBT is acting essentially identically to a designed RXR agonist in our biological system. Recently, Shoucri et al. (73) demonstrated that TBT acting through RXR inhibited expression of enhancer zeste homolog 2 (EZH2) in mouse multipotent mesenchymal stems cells to change the chromatin landscape to favor commitment to adipocytes at the expense of osteoblasts. In our tadpole tails and in a microarray analysis of tail resorption genes during natural metamorphosis (49), ezh2 was activated by T3, and we found that induction was significantly potentiated by TBT. However, cells in the apoptotic tadpole tail were undergoing a very different fate than mesenchymal stem cells committing to a particular differentiation pathway. Nevertheless, both studies highlight the underappreciated role of RXR agonists and signaling in developmental decisions and the danger of poor biological outcome from disruption of those pathways by environmental pollutants.

In summary, cotreatment of RXR ligands with T3 resulted in global changes in transcription in the tadpole tail, where RXR agonists potentiated the action of T3 and RXR antagonism inhibited T3 signaling, independent of the direction of T3 regulation. Importantly, TBT and Bex behaved identically, indicating that TBT was functioning as an RXR agonist. The molecular mechanisms through which these actions occur remain to be determined but may include RXR ligands regulating the chromatin landscape rather than functioning only as a permissive partner to TR. Our use of a precocious metamorphosis assay may have been fortuitous, functioning as a sort of “time auxotroph” by uncovering a potential role for RXR liganding in TH competence and setting a start point for determining how TH competence is generated in all metamorphic tissues during natural metamorphosis.

Glossary

Abbreviations:

1wk-PF

1 week postfertilization

Bex

bexarotene

DEG

differentially expressed gene

dio3

iodothyronine deiodinase 3

DMSO

dimethyl sulfoxide

HiBex

375 nM bexarotene

HPT

hypothalamic-pituitary-thyroid

LoBex

30 nM bexarotene

log2FC

log2 fold change

Luc

luciferase

LXR

liver X receptor

MCT

multiple comparisons test

MDS

multidimensional scaling

MMP

extracellular matrix metalloprotease

MMR

Marc Modified Ringers

MS-222

tricaine methanesulfonate

NR

nuclear receptor

PPARγ

peroxisome proliferator–activated receptor γ

qRT-PCR

quantitative RT-PCR

RAR

retinoic acid receptor

RNA-seq

RNA sequencing

RXR

retinoid X receptor

Tag-Seq

poly-A-primed RNA sequencing

TBT

tributyltin

TH

thyroid hormone

TR

thyroid hormone receptor

TRE

thyroid hormone response element

UC

University of California

UVI

UVI 3003

VA

vitamin A