Abstract

Thyroid hormone (TH) is essential for brain development both before and after birth. We have used gene expression microarrays to identify TH-regulated genes in the fetal cerebral cortex prior to the onset of fetal thyroid function to better understand the role of TH in early cortical development. TH levels were transiently manipulated in pregnant mice by treatment with goitrogens from gestational day (GD) 13–16 and/or by injection of TH 12 h before sacrifice on GD 16. The transcriptional response to exogenous TH in the GD 16 fetal cortex was potentiated by transient goitrogen treatment, suggesting that the hypothyroxinemic brain is a different substrate upon which TH can act, or that robust compensatory mechanisms are induced by transient hypothyroxinemia. Several known TH-responsive genes were identified including Klf9, and several novel TH-responsive genes such as Appbp2, Ppap2b, and Fgfr1op2 were identified in which TH response elements were confirmed. We also identified specific microRNAs whose expression in the fetal cortex was affected by TH treatment, and determined that Ppap2b and Klf9 are the target genes of miR-16 and miR-106, respectively. Thus, a complex redundant functional network appears to coordinate TH-mediated gene expression in the developing brain.

Introduction

Thyroid hormone (TH) is essential for normal brain development (Bernal and Nunez 1995; Thompson and Potter 2000; Zoeller and Rovet 2004). This is perhaps best known in the postnatal time period (Rastogi and LaFranchi 2010; Lafranchi 2011), where a great deal of research has focused on optimizing TH replacement therapy for children with congenital hypothyroidism (Chan and Rovet 2003; Rovet 2005; Donaldson and Jones 2013). But TH is also important for brain development in utero (Zoeller and Rovet 2004; De Groot et al. 2012), and it is now recognized that TH of maternal origin is the only source of TH in the fetus during the first trimester (e.g., Milanesi and Brent 2011). Despite the recognition of the importance of TH in brain ontogeny, the developmental events affected by TH and the mechanisms underlying TH effects are incompletely understood. This is important because knowledge of the neurodevelopmental events under TH regulation and the mechanisms and developmental timing underlying this regulation are key to identifying potential impacts of TH insufficiency or excess on brain development as well as to inform clinical care of pregnant women and neonates.

TH is thought to affect brain development predominantly by acting on one of several isoforms of the TH receptor (TR) (Bernal 2007). These receptors are classified as “type II” nuclear receptors (Flamant et al. 2006), which bind to DNA regulatory regions in the absence of ligand and either suppress the transcription of positively regulated genes, or enhance the transcription of negatively regulated genes. Upon binding cognate ligand, these receptors undergo a conformational change resulting in activation (or suppression) of these genes. Thus, low TH during development causes brain damage because the unliganded TR is inappropriately regulating gene transcription (Hashimoto et al. 2001); consistent with this concept is the observation that the genetic deletion of the TR in a mouse can rescue many aspects of development from the deleterious effects of low TH (Morte et al. 2002).

Because TH regulates the transcription of genes that code for proteins involved in brain development, an important strategy for understanding the developmental events regulated by TH is to identify TH-responsive genes in the developing brain. Genomic approaches have provided important insights into the neurodevelopmental processes affected by TH, but also the impact of differences in developmental timing and severity of TH insufficiency on brain development. For example, Royland et al. (2008) reported that genes driving important developmental processes after birth are sensitive to relatively modest perturbations of circulating levels of T4, and that the magnitude of the effect on gene expression is dependent upon the degree of TH insufficiency. Likewise, our previous work (Dong et al. 2005) characterized the impact of hypothyroidism on the transcriptional profile in the cerebellum of developing mice showing that some of these changes had permanent consequences.

The predominant model to study TH regulation of brain development is that of low T4 (i.e., hypothyroidism). However, this approach may be misleading because if TH controls cell fate specification in the brain (Dugas et al. 2012; Pascual and Aranda 2013; Picou et al. 2012; Sirakov et al. 2013), then the hypothyroid brain has a different cellular composition than a normal brain and, as a result, TH administration to a hypothyroid animal may be acting on a different neural substrate. This may be of clinical relevance to thyroid status in pregnant women.

To gain insight into these issues, we performed a 2 × 2 experiment in which pregnant mice were treated briefly with a mixture of goitrogens (perchlorate plus methimazole [MMI]), and the effect of acute exposure to TH was then evaluated. The timing of goitrogen treatment was based on the work of Auso et al. (2004) who characterized the effects of a modest and transient TH insufficiency during fetal development on neuronal migration in the cortex. The TH-injection paradigm was based on the work of Dowling et al. (2000) who identified changes in gene expression in the fetal cortex using an acute TH injection model. We now report that a short period of low maternal TH dramatically alters the transcriptional response of the fetal cerebral cortex to exogenous TH treatment, an observation consistent with the hypothesis that the hypothyroid brain is a different substrate upon which TH can act. In addition, we identified important transcription factors that may mediate the actions of TH on cell fate determination in the fetal cerebral cortex and also identify novel TH response elements in these genes. Finally, we have identified specific microRNAs (miRNAs) whose expression in the fetal cortex is affected by TH treatment and propose that, in addition to regulating mRNA levels directly by interacting with cis regulatory elements, TH may also regulate mRNA levels indirectly through the actions of specific miRNAs.

Materials and Methods

Animal Models and Experimental Design

All animal procedures followed the NIH Guidelines for the Care and Use of Experimental Animals and were approved by the University of Massachusetts Amherst Institutional Animal Care and Use Committee (IACUC).

Timed-pregnant C57BL/6 mice (n = 20; Harlan, Indianapolis, IN, USA) arrived on gestational day (GD) 11 and were housed individually in plastic cages under a 12:12 light cycle (6000–1800 hours) with food available ad libitum. Mice were randomly assigned to 1 of 4 groups (control, hypo, hypo+, and hyper; 5 per group) and treated as described in Table 1. Mice in the control and hyper groups were provided with fresh drinking water containing 2% sucrose from GD 13 to GD 16 (time of sacrifice). Mice in hypo and hypo+ groups were provided daily with fresh drinking water containing 2% sucrose (to mask bitterness) and 1% perchlorate and 0.025% MMI from GD 13 to sacrifice on GD 16. Twelve hours prior to sacrifice, all mice received a single subcutaneous injection. Control and hypo mice received vehicle (0.9% saline in 100 µL volume); the hypo+ group received 25.0 µg/100 g body weight (bw) T4 and 2.5 µg/100 g bw T3 to restore physiological levels of TH; the hyper group received 100.0 µg/100 g bw T4 and 10.0 µg/100 g bw T3 to model hyperthyroidism.

Table 1

Animal treatment groups for genomic analysis of TH-regulated genes in the fetal cortex

Group From GD 13 to GD 16 12 h prior to GD 16 i.p. GD 16 
Control 2% sucrose water 0.9% saline Sacrifice 
Hypo 2% sucrose water+ 1%Per/0.025MMI 0.9% saline 
Hypo+ 2% sucrose water+ 1%Per/0.025MMI 25 μg T4/2.5 μg T3/100 g bw 
Hyper 2% sucrose water 100 μg T4/10 μg T3/100 g bw 
Group From GD 13 to GD 16 12 h prior to GD 16 i.p. GD 16 
Control 2% sucrose water 0.9% saline Sacrifice 
Hypo 2% sucrose water+ 1%Per/0.025MMI 0.9% saline 
Hypo+ 2% sucrose water+ 1%Per/0.025MMI 25 μg T4/2.5 μg T3/100 g bw 
Hyper 2% sucrose water 100 μg T4/10 μg T3/100 g bw 

Tissue Collection and Processing

Dams were killed by exposure to CO2. Fetuses were dissected from the uterine horn and embryonic membranes, and then frozen on crushed dry ice. Tissue samples were taken from the fetuses to determine sex using PCR for SRY (Yang and Zoeller 2002). Fetal cerebral cortices were removed from each mouse embryo prior to freezing; to be consistent, the left half was used to determine tissue T4 levels, and the right half was used for microarray and quantitative reverse-transcriptase PCR (qRT-PCR) analysis.

T4 Levels in Fetal Cortex

T4 was extracted from each individual half cortex using a methanol-based extraction procedure. Samples were homogenized in 1 mL methanol containing 1 mM PTU and 25 µM iodopanoic acid to block endogenous deiodinases (which convert T4 to T3 and could alter values of total T4 in the tissue during the extraction). This extract was then added to a tube containing a tracer amount of 125I labeled T4. Tubes were then centrifuged for 20 min at 4 °C, and the pellet was resuspended in 500 µL of a cell lysis buffer (5 mM PIPES, 85 mM KCl, 0.05% NP-40), and protein content was measured in a small aliquot using BCA protein assay reagent (Pierce Biotechnology, Rockford, IL, USA) according to the manufacturer's instructions.

The supernatant from the methanol extract was dried overnight in a speed-vac, and the residue was resuspended in 50 µL methanol, 200 µL chloroform, and 50 µL barbital buffer. Tubes were centrifuged for 5 min to separate aqueous and organic phase. The aqueous phase was transferred into a new 0.5-mL tube and vacuum dried. The residue was counted using a gamma counter (CobraII, Packard, Meriden, CT, USA) to determine the T4 extraction efficiency. The extraction efficiency was 65% or higher. After counting, 10 µL of barbital buffer was added to each tube and vortexed. These samples were then assayed for T4 using an in-house developed RIA described previously (Sharlin et al. 2006). Estimated T4 levels were normalized to the protein concentration. Serum T4 levels in dams were measured using commercially available solid phase RIA kits from MP Biomedicals (Orangeburg, NY, USA).

Gene Expression Microarray Analysis

Total RNA was extracted from half cortices of individual female fetuses using RNeasy Lipid Tissue Mini kits (Qiagen, Germantown, MD, USA) according to the manufacturer's instructions. The quality of total RNA was evaluated by A260/A280 ratio (found to be at least 1.8 for each sample) and by electrophoresis on an Agilent Bioanalyzer (RIN number was 8.5 or above). The Yale Center for Neuroscience Microarray Consortium carried out the Affymetrix microarray analysis using the GeneChip Rat Genome 230 2.0 Array. Preparation of labeled cRNA for hybridization onto Affymetrix GeneChips followed the recommended Affymetrix protocol. Briefly, double-stranded cDNA was synthesized from 1 to 5 μg of total RNA by using a Superscript Choice System (Invitrogen), with an HPLC-purified oligo (dT)24 primer containing a T7 RNA polymerase promoter sequence at the 5′ end (Proligo LLC, Boulder, CO, USA). The second cDNA strand was synthesized by using Escherichia coli DNA polymerase I, RNase H, and DNA ligase. Labeled cRNA was generated from cDNA by in vitro transcription using a GeneChip IVT labeling kit (Affymetrix, Inc., Cleveland, OH, USA) following the manufacturer's instructions and incorporating biotinylated synthetic analog as pseudouridine reagent, a reaction mediated by MEGAscript T7 RNA Polymerase. Biotin-labeled cRNA was purified using the GeneChip cleanup module prior to fragmenting to a size of 35–200 bases by incubating at 94 °C for 35 min in fragmentation buffer. Fragmented cRNA was hybridized to the arrays for 16 h at 45 °C.

After hybridization, arrays were washed using an Affymetrix fluidics station and stained with streptavidin-phycoerythrin (10 mg/mL, Molecular Probes). Washed arrays were scanned on an Affymetrix GeneChip scanner 3000. Scanned output files were visually inspected for hybridization artifacts. All preprocessing of the data was conducted using the R package. The data were normalized by robust multiarray average using the ReadAffy function in the Affy R library (Irizarry et al. 2003). For the Fs statistic (Cui et al. 2005), a shrinkage estimator was used for the gene-specific variance components, and the associated P-values (unadjusted P) for all the statistical tests were estimated using the permutation method (30 000 permutations with residual shuffling). Clustering was conducted on genes with unadjusted P < 0.05 using GeneSpring GX (Agilent Technologies, Missisauga, ON, Canada). These P-values were then adjusted for multiple comparisons using the false discovery rate (FDR) approach (Benjamini et al. 2001). All data are available through the Gene Expression Omnibus (GEO) website, accession number GSE43927. The least squares mean (Goodnight and Harvey 1997), a function of the model parameters, was used to estimate the fold change for each pairwise comparison of interest (control vs. each of the 3 treatment groups: hypo, hypo+, and hyper).

Genes exhibiting a significant difference compared with controls, with adjusted FDR P < 0.05 and fold change no <1.5, were input into ingenuity pathway analysis (Ingenuity System, Inc., Redwood City, CA, USA) to perform network and pathway analysis.

Cell Culture

Neuroblastoma cells (N2aβ at 1 × 105; a gift from Dr J. Puymirat, University of Laval, QC, Canada) were seeded in 6-well plates in DMEM/F12 (1:1) with 10% FBS (charcoal treated, Medicorp, Montreal, QC, Canada) and 1% antibiotics (Life Technologies, Burlington, ON, Canada). After 24 h, 10 nM T3 was added to the cells. RNA was extracted using mirVana™ miRNA isolation kits (Life Technologies) after 24 or 48 h of incubation in the presence of T3.

Real-Time qRT-PCR Analysis of Gene Expression

Total RNA was extracted from male cerebral cortex or cultured cells using mirVana™ miRNA isolation kits (Life Technologies), and reverse transcribed into cDNA using SuperScript III (Life Technologies). qRT-PCR was performed in duplicate for each sample. Primer sequences are available in Supplementary Table 1. A CFX96 real-time qPCR detection system (Bio-Rad Laboratories, Canada) was employed to detect SYBR green 1 in amplified product. Gene expression levels were normalized to Hprt expression. PCR efficiency was examined using the standard curve for each gene. Primer specificity was verified using a melting curve. Statistical testing was performed using Student's t-test.

TRE Prediction and Electrophoretic Mobility Shift Assay

Potential TREs in selected genes were predicted in silico as previously described (Paquette et al. 2011). Briefly, we searched the 10 kb of DNA sequence from −8 to +2 kb (relative to the transcription start site (TSS) of the selected genes), recognizing that not all genes will be regulated by TREs within the promoter region (Jones et al. 2007). We scanned these sequences for putative TREs with a position weight matrix (PWM) that was developed based on, and cross-validated by, TREs gathered from the literature. The scan searched for 3 different types of TREs: direct repeats with a 4-nucleotide spacer (DR4); inverted repeats with no spacer (IR0); and everted repeats with a spacer of 6 nucleotides (ER6). Sequences that conformed to the PWM model with high likelihood were tested to investigate TR binding to the predicted TREs using electrophoretic mobility shift assay (EMSA). EMSA was conducted using recombinant TRβ and RXRα produced in vitro by incubating the expression plasmids, TRβ and RXRα (OriGene Technologies, Rockville, MD, USA) with the TnT Quick Coupled Transcription/translation System (Promega, Madison, WI, USA) for 90 min at 30 °C, along with no template control (NTC). The recombinant TRβ was confirmed by western blot analysis using an anti-TRβ antibody (J52, Santa Cruz Biotechnology, Inc., Dallas, TX, USA) as described previously (Dong et al. 2009). Twenty-six basepairs oligonucleotides (Eurofins MWG Operon, Huntsville, AL, USA) containing the selected putative TREs were modified by adding fluorescent Cy5 at the 5′ end. Complementary oligonucleotides were denatured in a buffer containing 50 mM NaCl, 1 mM EDTA, and 10 mM Tris (pH 8.0) at 95 °C for 5 min, and then annealed by ramping down to 25 °C over 45 min.

Recombinant TRβ and RXRα or NTC were incubated with 0.5-pmol Cy5-labeled probes in binding buffer (5 mM MgCl2, 0.05% NP40, 2.5% glycerol, and 50 ng/μL poly dI-dC, Pierce Biotechnology) for 20 min in the presence or absence of anti-TRβ antibody with or without preincubation with 100-pmol unlabeled probes for 10 min. Reaction solutions were electrophoresed on 5% native polyacrylamide gels in 0.5 × Tris/borate/EDTA buffer at 100 V for ∼50 min at 4 °C in the dark. The gels were scanned using a Typhoon Trio+ imager (GE Healthcare, Mississauga, ON, Canada).

MiRNA Microarray Analysis

In addition to 4 samples each from GD 16 control and hyper groups, we also included 4 male cortex samples each from postnatal day (PND) 15 control and hyper groups for miRNA microarray analysis. The PND 15 samples were produced in a similar manner to the fetal tissue and both the method and effects on gene expression are described elsewhere (Paquette et al. 2011). The major difference in method, other than the age of pups, was that TH was injected 4 h before sacrifice to derive the PND 15 hyper mice, rather than 12 h, as was used for GD 16 samples. RNA extraction and quality control were performed using the same approaches for both sets of samples. Total RNA was labeled with Cy3 using the miRNA Complete Labeling and Hybridization kit (Agilent). Labeled RNA was hybridized on 8 × 15K mouse miRNA microarray slides (Agilent). Arrays were scanned using an Agilent G2505B scanner (5 μm resolution). Feature extraction (version 10.7.3.1, Agilent) was used to acquire the fluorescence intensity of each probe.

The quality of the microarray data was evaluated using Agilent Feature extraction quality control metrics. Data were normalized in R using cyclic-lowess (Bolstad et al. 2003). Ratio-intensity plots, boxplots, and cluster analyses were used to identify potential outliers. All but one sample from the PND 15 hyper group passed the quality control tests and was used for subsequent analyses (i.e., n = 3 for PND 15 hyper, and n = 4 for all other 3 groups). miRNAs were identified as being significantly changed in abundance as described above for the mRNA arrays.

Real-Time qRT-PCR Analysis of miRNAs

The miScript PCR system (Qiagen) was used to examine the expression of specific miRNAs and their precursors. Briefly, 1 μg of total RNA was used for polyadenylation of mature miRNAs and reverse transcription into cDNA using oligodT primers with a universal tag. qRT-PCR was performed in duplicate for each sample using a primer complementary to the universal tag and a miScript primer or miScript precursor assay specific to each miRNA. RNU6B was used to normalize the expression of the target miRNAs and precursors. Fold change was calculated according to the 2−ΔΔCt method (Livak and Schmittgen 2001), and Student's t-test was applied to test for significant differences.

Identification of miRNA Target Genes Using 3′ UTR Reporter Assays

The web-based tools microRNA.org-Targets and Expression (http://www.microrna.org/microrna/getGeneForm.do) were used to predict the target genes of miRNAs and identify candidate target miRNAs of TH-regulated genes.

Reporter vectors were constructed by inserting portions of the 3′ UTR of Klf9, Ppap2b, and Fgfr1op2 into a pLUC-3 UTR vector downstream of the firefly luciferase expression cassette (custommade by Signosis, Inc., Sunnyvale, CA, USA). The inserts were ∼200 bp long, and included the sequence aligned with the 8-mer or 7-mer seed of the indicated miRNAs (refer to Fig. 5A). NIH3T3 (ATCC, Rockville, MD, USA) cells were cultured with DMEM, containing 10% bovine calf serum (BCS) and seeded in 6-well plates at 1 × 105 cells/well. Cells were co-transfected with 0.5-μg pLUC-3 UTR reporter vector (containing firefly luciferase), 0.5-μg pRL-TK (containing renila luciferase), and 37.5 nM mirVana™ miRNA mimics or inhibitors (Life Technologies). Each transfection was performed in duplicate. Luciferase activity was quantified using a Dual Luciferase kit in a luminometer (Promega). Briefly, cells were lysed by adding 200 μL of lysis buffer, and shaking for 15 min at room temperature. Cell lysis sample (20 μL) was added to 96-well opaque plates, to which 50 μL LARII was added to measure firefly luciferase activity. After this, 50 μL of Stop & Glow was added and renila luciferase was measured to normalize for transfection efficiency. The ratio of the normalized luciferase activity in the presence of miRNA mimics to the normalized luciferase activity in the presence of miRNA inhibitors was used to calculate the repression of miRNA on the expression of the genes investigated. Experiments were repeated at least 3 times. The averages of these experiments are presented.

Results

Mouse Model of Transient Gestational Hypothyroxinemia or Hyperthyroxinemia

T4 levels in serum of dams or fetal cortices were measured in 4 experimental groups including control, hypo, hypo+, and hyper as described in the Materials and Methods (n = 5 each). Serum T4 levels in dams in hypo group were not significantly different from those of control dams, although the concentration of T4 in the fetal cortices of hypo dams was slightly, but significantly decreased (Fig. 1). Serum T4 levels in hypo+ were significantly elevated as was the concentration of T4 in the cortices of their fetuses. Likewise, serum T4 levels in hyper group were even more elevated, as were T4 levels in the fetal cortex.

Figure 1.

T4 levels. (A) Serum total T4 in dams at the time of sacrifice on GD 16. One-way ANOVA revealed significant differences among the means (F3,27 = 258; P < 0.0001). Serum total T4 was significantly elevated in goitrogen-treated dams injected with the combination of T4 and T3 (adjusted P = 0.0001). Likewise, dams in the hyper group exhibited the highest level of serum total T4 (adjusted P = 0.0001). (B) T4 levels in fetal cortex on GD 16. One-way ANOVA revealed significant differences among the means of T4 in cerebral cortex (F3,16 = 18.63; P < 0.0001). T4 was significantly decreased in hypo relative to control (P < 0.05) and increased in hypo+ and hyper relative to control (P < 0.01 and P < 0.001, respectively; n = 5 or 6 for each group).

Figure 1.

T4 levels. (A) Serum total T4 in dams at the time of sacrifice on GD 16. One-way ANOVA revealed significant differences among the means (F3,27 = 258; P < 0.0001). Serum total T4 was significantly elevated in goitrogen-treated dams injected with the combination of T4 and T3 (adjusted P = 0.0001). Likewise, dams in the hyper group exhibited the highest level of serum total T4 (adjusted P = 0.0001). (B) T4 levels in fetal cortex on GD 16. One-way ANOVA revealed significant differences among the means of T4 in cerebral cortex (F3,16 = 18.63; P < 0.0001). T4 was significantly decreased in hypo relative to control (P < 0.05) and increased in hypo+ and hyper relative to control (P < 0.01 and P < 0.001, respectively; n = 5 or 6 for each group).

Gene Expression Analysis in Fetal Cortex

Affymetrix microarrays were used to analyze the effect of manipulating maternal thyroid status on the transcriptional profile in the fetal cortex before the onset of fetal thyroid function. Across all groups, the expression of 18 000 genes significantly changed in response to some treatment with an unadjusted P-value < 0.05. Hierarchical cluster analysis on these genes revealed 2 main branches (Supplementary Fig. 1), with control and hypo mice on one branch and hypo+ and hyper mice on the other. These data were filtered to retain genes that had a fold change ≥ 1.5 and a FDR P ≤ 0.05. The filtered gene list revealed 11 (hypo), 875 (hypo+), and 161 (hyper) transcripts that were significantly up- or downregulated relative to controls (Supplementary Table 2). None of the 11 genes identified in the hypo group were differentially expressed in the hypo+ group. Because the model used to induce hypothyroidism produced little effect at the gene expression level, our subsequent analyses focused on the mRNA changes induced in the hyperthyroid model.

Of the 161 genes (125 up and 36 down) that were differentially expressed in hyperthyroid cortices compared with control, 115 genes were identified in the Ingenuity Knowledge Database and were used to construct functional networks. The biological functions assigned to the most significant networks correspond to assembly and bundling of microtubules, patterning of cerebral cortex, polarization of hippocampal neurons, as well as mental retardation. The complete list of processes identified as significantly altered in hyperthyroid animals is provided in Supplementary Table 3.

Detailed Characterization of TH Regulation of Specific Genes

The microarray analysis was performed on cerebral cortex from female fetuses; to validate the integrity of these data, we performed qRT-PCR on cortices from male littermates on 10 selected genes. These genes were chosen because they were significantly upregulated in the hyper group analyzed using microarray (FDR P ≤ 0.05), and there was prior evidence in the literature that these genes might be influenced by TH. Nine of these 10 genes were confirmed to be more abundant in the male cerebral cortex from the hyper group using qPCR (Table 2). Because we employed only 3 samples for each of the control and hyper groups, we considered a P-value of <0.1 to be significant. A known direct target of TH action, Klf9 (Denver and Williamson 2009), was among these, but we also identified several genes that were not previously known to be controlled by TH. To test whether these genes might be direct targets of TH action, we tested the effect of T3 treatment on the expression of 6 of these genes in N2aβ cells. We selected these genes on the basis that they were upregulated by at least 2-fold in the hyper group compared with controls. N2aβ cells were treated with 10 nM T3 for 24 or 48 h, and mRNA levels for 6 transcripts were analyzed (Fig. 2). With the exception Appbp2, all genes were significantly increased by T3 treatment confirming their positive regulation by TH.

Table 2

qRT-PCR confirmation of microarray results in hyperthyroid fetuses

Gene symbol Gene name Fold change microarray Fold change qRT-PCR P-value (qRT-PCR) 
Klf9 Kruppel-like factor 9 6.4 3.0 <0.05 
Appbp2 Amyloid protein-binding protein 2 2.3 2.0 <0.05 
Ppap2b Phosphatidic acid phosphatase type 2B 2.1 8.4 <0.1 
Fgfr1op2 FGFR1 oncogene partner 2 2.1 2.6 <0.1 
Srrm2 Serine/arginine repetitive matrix 2 2.0 2.4 <0.05 
Myt1 Mylein transcription factor 2.0 3.2 <0.1 
Ndfip2 Nedd4 family interacting protein 2 1.8 1.1 >0.1 
Mbd5 Methyl-CpG binding domain protein 5 1.7 2.4 <0.05 
Sfpg Splicing factor proline/glutamine rich 1.6 1.7 <0.05 
Ube3c Ubiquitin protein ligase E3C 1.5 1.8 <0.05 
Gene symbol Gene name Fold change microarray Fold change qRT-PCR P-value (qRT-PCR) 
Klf9 Kruppel-like factor 9 6.4 3.0 <0.05 
Appbp2 Amyloid protein-binding protein 2 2.3 2.0 <0.05 
Ppap2b Phosphatidic acid phosphatase type 2B 2.1 8.4 <0.1 
Fgfr1op2 FGFR1 oncogene partner 2 2.1 2.6 <0.1 
Srrm2 Serine/arginine repetitive matrix 2 2.0 2.4 <0.05 
Myt1 Mylein transcription factor 2.0 3.2 <0.1 
Ndfip2 Nedd4 family interacting protein 2 1.8 1.1 >0.1 
Mbd5 Methyl-CpG binding domain protein 5 1.7 2.4 <0.05 
Sfpg Splicing factor proline/glutamine rich 1.6 1.7 <0.05 
Ube3c Ubiquitin protein ligase E3C 1.5 1.8 <0.05 
Figure 2.

Confirmation, in cultured cells, of TH regulation of genes identified by microarray analysis of fetal cortices. The effects of T3 treatment on the expression of selected genes was evaluated in the neuroblastoma cell line N2aβ stably transfected with the human TRβ1 gene. Cells were cultured in medium with charcoal-treated FBS for 24 or 48 h in the presence or absence 10 nM T3. qRT-PCR was used to examine gene expression. Experiments were repeated 3 times, and the results of one representative experiment are shown here. ***P < 0.01, **P < 0.05, *P < 0.1.

Figure 2.

Confirmation, in cultured cells, of TH regulation of genes identified by microarray analysis of fetal cortices. The effects of T3 treatment on the expression of selected genes was evaluated in the neuroblastoma cell line N2aβ stably transfected with the human TRβ1 gene. Cells were cultured in medium with charcoal-treated FBS for 24 or 48 h in the presence or absence 10 nM T3. qRT-PCR was used to examine gene expression. Experiments were repeated 3 times, and the results of one representative experiment are shown here. ***P < 0.01, **P < 0.05, *P < 0.1.

To determine if these genes are direct targets of TH action, we first performed an in silico search for putative TREs in the 10-kb DNA sequence (−8 to +2 kb relative to TSS) of Appbp2, Fgfr1op2 and Ppap2b using a program based on known TREs as the training set (containing the TRE in Klf9; see Materials and Methods). Predicted TREs for these genes are shown in Table 3. Priority for TRE follow-up validation was given to probes where the TRE is in the sense orientation. Two probes of Appbp2 in the sense orientation have exactly the same sequences as the 2 half-sites of probes B and C; therefore, the probes of this gene in the antisense orientation were chosen for further validation.

Table 3

Sequence and location of predicted TREs for indicated genes

Gene Chromosome Dist from TSS Predicted TREs TRE orientationa EMSA probe 
Ppap2b −6650 aggaca nnnn gggcta Sense 
−5536 aggtaa nnnn gggaca Sense 
−129 aggcca nnnn aggata Antisense  
Fgfr1op2 −6067 aggaca nnnn gggcta Sense  
−2046 aggaca nnnn gggcta Sense 
−485 aggctt nnnn aggtaa Antisense  
Appbp2 11 −4313 aggaca nnnn gggcta Sense  
−6512 aggaca nnnn gggcta Sense  
−7244 aggcta nnnn aggaca Antisense 
−3369 cggtca nnnn aggaca Antisense 
−1674 acgtca nnnn aggtaa Antisense 
Gene Chromosome Dist from TSS Predicted TREs TRE orientationa EMSA probe 
Ppap2b −6650 aggaca nnnn gggcta Sense 
−5536 aggtaa nnnn gggaca Sense 
−129 aggcca nnnn aggata Antisense  
Fgfr1op2 −6067 aggaca nnnn gggcta Sense  
−2046 aggaca nnnn gggcta Sense 
−485 aggctt nnnn aggtaa Antisense  
Appbp2 11 −4313 aggaca nnnn gggcta Sense  
−6512 aggaca nnnn gggcta Sense  
−7244 aggcta nnnn aggaca Antisense 
−3369 cggtca nnnn aggaca Antisense 
−1674 acgtca nnnn aggtaa Antisense 

aSense indicates that the TRE is located on the same strand as the gene; antisense indicates that TRE is located on the complementary strand as the gene.

Six probes (A–F, as shown in the last column in Table 3) were developed based on the putative TREs to investigate binding affinity to TR using EMSA. DR4 was used as the positive control. The specific binding of DR4 with TRβ and RXRα is shown in Figure 3A (black arrow). Binding to DR4 was diminished by adding competing unlabeled DR4 or anti-TRβ antibody as expected. DR4 did not bind with TRβ in the absence of RXRα (data not shown) in the binding reaction. Binding was also found for probes B, C, and D, which are located in the promoters of Ppap2b, Fgfr1op2, and Appbp2, respectively (Fig. 3B). We noted that the binding of Appbp2 with TRβ and RXRα was relatively weak. The other 3 probes selected for these genes did not bind with TRβ and RXRα (Fig. 3C). All 3 of the validated TREs are located in the upstream region of the promoters, further than 2000 bp from TSS. The results are consistent with the interpretation that TH regulates the expression of these genes through TR–TRE binding and transcriptional activation.

Figure 3.

Confirmation of TRβ-binding to putative TREs in targeted genes using EMSA. (A) DR4 was used as a positive control to optimize the EMSA system. A specific shift was observed, as indicated by the black arrow shown in the reaction of DR4 with TRβ and RXRα, but not with the no-template control. Unlabeled DR4 or anti-TRβ antibody competed for binding, while IgG did not. (B) Probe B, C, and D (last column of Table 3) were shown to cause a specific shift (indicate with a black arrow). (C) There was no specific shift observed for probes A, E, or F (last column of Table 3). In B and C: Lane 1, Cy5-labeled probe (Cy5); Lane 2, Cy5 + TRβ + RXRα; Lane 3, Cy5 + NTC; Lane 4, Cy5 + TRβ + RXRα + 200 × cold probe; Lane 5, Cy5 + TRβ + RXRα + antibody; Lane 6, Cy5 + TRβ + RXRα + IgG.

Figure 3.

Confirmation of TRβ-binding to putative TREs in targeted genes using EMSA. (A) DR4 was used as a positive control to optimize the EMSA system. A specific shift was observed, as indicated by the black arrow shown in the reaction of DR4 with TRβ and RXRα, but not with the no-template control. Unlabeled DR4 or anti-TRβ antibody competed for binding, while IgG did not. (B) Probe B, C, and D (last column of Table 3) were shown to cause a specific shift (indicate with a black arrow). (C) There was no specific shift observed for probes A, E, or F (last column of Table 3). In B and C: Lane 1, Cy5-labeled probe (Cy5); Lane 2, Cy5 + TRβ + RXRα; Lane 3, Cy5 + NTC; Lane 4, Cy5 + TRβ + RXRα + 200 × cold probe; Lane 5, Cy5 + TRβ + RXRα + antibody; Lane 6, Cy5 + TRβ + RXRα + IgG.

Effects of TH on miRNA Expression During Cerebral Cortex Development

To investigate the effects of TH on the regulation of miRNAs potentially involved in neurodevelopment, we compared miRNA expression profiles of hyperthyroid and euthyroid fetal cortices on GD 16 using miRNA microarrays. No miRNAs in this analysis passed our stringent filtering criteria: FDR P ≤ 0.05 and fold change≥1.5. The lack of changes in miRNAs may be the result of a low dynamic range and sensitivity of the DNA microarray platform for detecting small miRNA changes (Git et al. 2010), combined with the relatively modest impact expected for the effects of TH on miRNA expression in the brain. However, because we have previously identified TH-regulated miRNAs in livers of pups on PND 15 (Dong et al. 2010), we wanted to pursue this issue further. Therefore, we also produced miRNA profiles using microarrays for the cerebral cortices of hyperthyroid and control pups sampled on PND 15. PND 15 mice were rendered hyperthyroid by injection of TH 4 h prior sacrifice as part of a previously published study (Paquette et al. 2011). This treatment resulted in a 4.5-fold increase in serum TH levels, but did not significantly change miRNA expression in the cerebral cortex.

We surmised that the microarray analysis of miRNAs would not be fruitful for the developing brain at this time, but we had no a priori reason to choose specific miRNAs for analysis in the GD 16 brain using a targeted qPCR approach. Therefore, we sought to identify candidates for qRT-PCR by comparing the miRNA expression profiles from GD 16 and PND 15 mice. Our reasoning is that there is a sharp increase in circulating TH levels on PND 15 relative to GD 16, and it is a critical time window for brain development in rodents (Anderson 2008). Comparison of miRNA profiles at GD 16 and PND 15 revealed significant differences in the expression of 99 miRNAs (FDR P ≤ 0.05 and fold change≥2, Supplementary Table 4). We then sought to identify miRNAs that may enhance or facilitate the expression of TH-target genes. Because miRNAs predominantly suppress gene expression by binding to 3′ UTRs, we focused on miRNAs that had been downregulated (which would lead to upregulation of TH-regulated genes) at PND 15 relative to GD 16 mice. Among the 99 differentially expressed miRNAs, 27 were decreased on PND 15 relative to GD 16 (Table 4). We chose 7 of these 27 to pursue using qRT-PCR based on the apparent abundance in the microarray analysis. The expression of all of the miRNAs except miR-135b was confirmed to be suppressed on PND 15 relative to GD 16 (Fig. 4A). The expression of 3 miRNA precursors of 6 confirmed miRNAs were also suppressed on PND 15 (Fig. 4B).

Table 4

Downregulated miRNAs in the cerebral cortices of PND 15 versus GD 16 euthyroid mice

Probe Ratio PND15/G16* 
mmu-miR-130b 0.09 
mmu-miR-20a 0.09 
mmu-miR-25 0.09 
mmu-miR-20b 0.10 
mmu-miR-92a 0.11 
mmu-miR-18a 0.12 
mmu-miR-19a 0.12 
mmu-miR-93 0.14 
mmu-miR-106b 0.15 
mmu-miR-15b 0.15 
mmu-miR-19b 0.18 
mmu-miR-449a 0.22 
mmu-miR-130a 0.22 
mmu-miR-16 0.23 
mmu-miR-301a 0.26 
mmu-miR-17 0.26 
mmu-miR-135b 0.28 
mmu-miR-17* 0.30 
mmu-miR-15a 0.30 
mmu-miR-135a 0.31 
mmu-miR-181d 0.35 
mmu-miR-181b 0.38 
mmu-miR-99a 0.39 
mmu-miR-9 0.40 
mmu-miR-99b 0.46 
mmu-miR-125b-5p 0.48 
Probe Ratio PND15/G16* 
mmu-miR-130b 0.09 
mmu-miR-20a 0.09 
mmu-miR-25 0.09 
mmu-miR-20b 0.10 
mmu-miR-92a 0.11 
mmu-miR-18a 0.12 
mmu-miR-19a 0.12 
mmu-miR-93 0.14 
mmu-miR-106b 0.15 
mmu-miR-15b 0.15 
mmu-miR-19b 0.18 
mmu-miR-449a 0.22 
mmu-miR-130a 0.22 
mmu-miR-16 0.23 
mmu-miR-301a 0.26 
mmu-miR-17 0.26 
mmu-miR-135b 0.28 
mmu-miR-17* 0.30 
mmu-miR-15a 0.30 
mmu-miR-135a 0.31 
mmu-miR-181d 0.35 
mmu-miR-181b 0.38 
mmu-miR-99a 0.39 
mmu-miR-9 0.40 
mmu-miR-99b 0.46 
mmu-miR-125b-5p 0.48 

Bold indicates the miRNAs that were chosen to perform qRT-PCR.

*FDR P < 0.05 for all ratios.

Figure 4.

Expression of miRNAs and precursors in the cerebral cortex of euthyroid or hyperthyroid mice on GD 16 or PND 15. (A) Confirmation of differences in the expression of selected miRNAs in euthyroid mice from GD 16 to PND 15. With the exception of miR-135, the expression of these miRNAs declined from GD 16 to PND 15. (B) Change of expression of miRNA precursors from GD 16 to PND 15. (C) Effects of TH on the expression of selected miRNAs on GD 16 or PND 15 mice. TH downregulated all of the miRNAs on GD 16. (D) Effects of TH on the expression of selected miRNA precursors at GD 16 or PND 15. All of the precursors except for miR-301 were downregulated by TH on GD 16; no significant changes were found for PND 15. *P < 0.05.

Figure 4.

Expression of miRNAs and precursors in the cerebral cortex of euthyroid or hyperthyroid mice on GD 16 or PND 15. (A) Confirmation of differences in the expression of selected miRNAs in euthyroid mice from GD 16 to PND 15. With the exception of miR-135, the expression of these miRNAs declined from GD 16 to PND 15. (B) Change of expression of miRNA precursors from GD 16 to PND 15. (C) Effects of TH on the expression of selected miRNAs on GD 16 or PND 15 mice. TH downregulated all of the miRNAs on GD 16. (D) Effects of TH on the expression of selected miRNA precursors at GD 16 or PND 15. All of the precursors except for miR-301 were downregulated by TH on GD 16; no significant changes were found for PND 15. *P < 0.05.

These 6 miRNAs were used as candidates for qRT-PCR comparison of control and hyperthyroid cortex for both developmental time points. qRT-PCR analysis revealed that the expression of these miRNAs decreased significantly in hyperthyroid cortices relative to control mice on GD 16, but no significant changes were found for PND 15 (Fig. 4C). With the exception of miR-301a, the precursors of these miRNAs were also decreased in hyperthyroid mice on GD 16, but not on PND 15, compared with control mice (Fig. 4D).

Identification of miRNA Target Genes

In silico analysis revealed that 19 upregulated mRNAs in the cerebral cortex of the hyper group on GD 16 are predicted targets of miR-16 (including Ppap2b), while 22 are predicted targets of miR-106 (including Klf9 and Fgfr1op2, Fig. 5A). We used 3′ UTR luciferase reporter assays to experimentally validate the regulatory roles of these miRNAs on the expression of Ppap2b, Klf9, and Fgfr1op2. We found that miR-16 effectively suppressed the luciferase activity of the vector containing the 3′ UTR of Ppap2b (Fig. 5B), while miR-106 suppressed the luciferase activity of Klf9 (Fig. 5C). The results indicate that TH regulates the expression of Ppap2b and Klf9 through direct control via TR–TRE binding and activation, but that these genes are also regulated independently by miR-16 and miR-106.

Figure 5.

Identification of miRNA target genes. (A) MicroRNA.org-Targets and Expression was used to predict the miRNAs that target the genes PaPpa2b, Klf9, and Fgfr1op2. (B) Validation of the targets of miR-16 by 3′ UTR luciferase assays. The luciferase vector containing the 3′ UTR of the indicated genes and miR-16 mimics or miR-16 inhibitors, as well as pRL-TK were co-transfected into NIH3T3 cells. Luciferase activities were measured 24 h later. Renila luciferase activity was used to normalize the firefly luciferase activity derived from 3′ UTR reporter vectors. The ratio of the normalized luciferase activity in the presence of miR-16 mimics to the normalized luciferase activity measured in the presence of miR-16 inhibitors (y-axis) was used to calculate the repressive activity of miR-16 on the expression of the indicated genes. Ppap2b was confirmed to be a target of miR-16. (C) Validation of the targets of miR-106 by 3′ UTR luciferase assays. Similar methods to those described in B were used, except co-transfection was with miR-106 mimics or inhibitors. Klf9 was confirmed to be a target of miR-106. Experiments were repeated at least 3 times, and the average of these experiments is presented. *P < 0.05.

Figure 5.

Identification of miRNA target genes. (A) MicroRNA.org-Targets and Expression was used to predict the miRNAs that target the genes PaPpa2b, Klf9, and Fgfr1op2. (B) Validation of the targets of miR-16 by 3′ UTR luciferase assays. The luciferase vector containing the 3′ UTR of the indicated genes and miR-16 mimics or miR-16 inhibitors, as well as pRL-TK were co-transfected into NIH3T3 cells. Luciferase activities were measured 24 h later. Renila luciferase activity was used to normalize the firefly luciferase activity derived from 3′ UTR reporter vectors. The ratio of the normalized luciferase activity in the presence of miR-16 mimics to the normalized luciferase activity measured in the presence of miR-16 inhibitors (y-axis) was used to calculate the repressive activity of miR-16 on the expression of the indicated genes. Ppap2b was confirmed to be a target of miR-16. (C) Validation of the targets of miR-106 by 3′ UTR luciferase assays. Similar methods to those described in B were used, except co-transfection was with miR-106 mimics or inhibitors. Klf9 was confirmed to be a target of miR-106. Experiments were repeated at least 3 times, and the average of these experiments is presented. *P < 0.05.

Discussion

We now report that mild and transient reduction in T4 availability to the mouse fetal cerebral cortex, prior to the onset of fetal thyroid gland function, produces relatively mild effects on gene expression in the fetal cortex. Supplemental TH administration to the dam produced more dramatic effects, and these effects appeared to be even more robust if the dam had been previously treated briefly with goitrogenic agents. This latter finding may reflect differences in development produced by goitrogen treatment and/or in the activation of adaptive responses to goitrogenic agents. We also report that an important target of TH action in the fetal cortex is Klf9, which was previously shown to mediate differentiative effects of TH in the postnatal brain (Dugas et al. 2012). In addition, we identified a number of new direct gene targets of TH action in the fetal cortex and their TREs. The identity of some of these genes may provide new insight into our understanding of the mechanism by which TH regulates early cortical development. Finally, we show that TH may control gene expression in the developing brain both by direct actions on mRNA transcription and through the regulation of specific miRNAs.

A critical feature of our experimental design is that we manipulated maternal thyroid status before the onset of fetal thyroid function on GD 17.5 (Fisher et al. 1977). Therefore, our finding of significant differences in cortical T4 among the treatment groups is consistent with findings that T4 of maternal origin can cross the placenta and be taken up in the fetus of both animals (Porterfield and Hendrich 1992; Porterfield and Stein 1994) and humans (Vulsma et al. 1989; Kester et al. 2004). Because we did not perfuse the fetus to remove blood from tissues, our T4 measurements include both T4 in the blood and T4 in the parenchyma of the cortex. However, our main goal with these measures was to ensure that maternal treatments affected fetal exposure to T4, and this was confirmed. We measured T4 and not T3 because tissue levels of T3 are not a good index of TH action in the brain (Galton et al. 2007) and therefore would not be particularly informative for our purpose.

Cluster analysis of gene expression profiles demonstrated that fetuses of dams not provided with supplemental TH (hypo and control groups) exhibited highly correlated expression profiles, while those receiving supplemental TH (hypo+ and hyper groups) clustered together on a separate branch. This analysis closely reflects thyroid status given that TH levels in the circulation of euthyroid fetuses is already very low compared with adult levels (Escobar-Morreale et al. 1995), and Per/MMI exposure may only slightly diminish TH effects compared with the major spike in TH caused by the injections in the hypo+ and hyper groups.

The minimal transcriptional response to goitrogen treatment—only 11 genes were identified as having altered expression relative to controls—stood in stark contrast to the robust transcriptional response to TH treatment. It was unexpected that goitrogen-treated animals given TH (hypo+) exhibited 875 genes whose expression was altered relative to controls, compared with 161 genes in euthyroid animals given TH (hyper). This potentiation may reflect an efficient adaptive response in the fetal brain that would increase transport of T4 to the brain and/or increase the conversion of T4 to T3 in the brain. Both T4 transport into the brain and conversion to the genomically active T3 are important steps in controlling TH action (e.g., Morte, Ceballos, et al. 2010). However, we did not observe an increase in 5′-deiodinase (type 1 or type 2) mRNA levels in the Hypo group on the microarrays, nor did we observe a decrease in 5′-deiodinase mRNA levels in either of the TH-treated groups.

It is also possible that goitrogen treatment altered the trajectory of development such that the cellular phenotype upon which TH acts is sufficiently different from euthyroid animals that TH treatment produces a dramatically different transcriptional response between hypothyroid and euthyroid animals. We know that TH affects fate specification of at least some cell types (Kapoor et al. 2012; Picou et al. 2012); thus, it is possible that the goitrogen-treated fetal brain is different from that of the untreated brain and this is reflected in the difference in the profile of TH-regulated genes. In addition, Klf9 was very significantly upregulated by TH treatment and this gene product is known to mediate TH action on cell differentiation at other times during development (e.g., Dugas et al. 2012). These are not mutually exclusive interpretations of our current data and it is likely to be important to determine.

Our findings may appear at first glance to contradict previous work, in that others have reported more robust effects of maternal hypothyroidism—and the lack of effect of exogenous TH—on the transcriptional profile in the fetal brain (e.g., Morte, Diez, et al. 2010; Grijota-Martinez et al. 2011). In addition, others have reported effects of maternal hypothyroidism on the expression of specific genes in the fetal cortex such as Reelin (Pathak et al. 2011), Nsp (Nrgn), and Oct1 (Pou2f1) (Dowling et al. 2000) that we did not observe in the current microarray data. These differences are likely to be a manifestation of the differences in timing of the observations, severity of hypothyroidism, and the dose or route of delivery (e.g., injection versus drinking water or osmotic pump) of TH, as well as the technical approach to identifying changes in gene expression. However, this variability is likely relevant to the underlying biology in that human studies demonstrate that the dose and timing of T4 supplementation in congenital hypothyroidism have different impacts on neurobehavioral outcome (Selva et al. 2005). Moreover, TH insufficiency that occurs at different times during development has different impacts on brain development (Rovet and Daneman 2003; Zoeller and Rovet 2004).

Considering this variability, it is difficult to make predictions at this time about the impact of different experimental approaches on brain development or on gene expression changes underlying these developmental events. For example, long-term hyperthyroidism may have different effects on cortical gene expression than the short-term (i.e., 12 h) TH exposure employed in the current paradigm. Our current strategy of acute exposure to TH was employed to increase the likelihood that we would capture “early” response genes that would be enriched in direct targets of TH action, and this appears to be the case.

In the current work, we validated the microarray findings in 2 ways. First, we used qPCR to confirm the microarray results using 10 targeted genes. These genes were among the top induced genes in the hyper group, and in some cases were previously reported to be regulated by TH. We used cortical tissue from male siblings of those animals employed for the microarray study. We recognize that there may be sex differences; however, 9 of 10 of these genes exhibited an increase in expression. These findings support the interpretation that there are no sex differences in the response of these genes to TH, as well as to support the reliability of the microarray data. In a second approach, we evaluated the effect of 10 nM T3 in a neuroblastoma cell line (Neuro2A) that had been stably transfected with the TRβ1 gene (Lebel et al. 1994) on the expression of 6 of these same mRNAs. Five of these genes exhibited increased expression at either 24 or 48 h or both, which also provides substantial confidence in the microarray data.

The identity of the genes whose expression was affected by TH in the fetal cortex may provide important insight into the role of TH during cortical development. The ingenuity pathway analysis indicated that patterns of genes were altered that are involved in the development of microtubules, patterning of the cerebral cortex and neuronal polarization. In addition, the gene Klf9 exhibited the greatest response to TH both in vivo and in vitro. Because Klf9 mediates the effect of T3 on oligodendrocyte differentiation (Dugas et al. 2012) later in development, it highlights the possibility that TH may be affecting fate specification of cells earlier in development as well. In addition, we provide evidence for TREs in 3 other genes: Ppap2b, Fgfr1op2, and Appbp2. Using EMSA, these data suggest that the 4 nucleotide spacer and the flanking sequence are essential to determine the potential binding of TREs by TR. However, because EMSA is a fully in vitro technique, these data may not always be applicable to the in vivo state. Chromatin immunoprecipitation (ChIP) would provide more direct evidence for this, but was outside the scope of the current work. Clearly, further work is required to refine the in silico approach to move this field forward.

We also evaluated the potential role of miRNAs in mediating gene regulatory effects of TH in the fetal cortex. Our initial strategy was simply to identify miRNA species that were affected by maternal TH status; however, we had to abandon that approach when the miRNA microarrays did not reveal candidate miRNAs that were affected. We made a similar observation in the cerebral cortex of PND 15 pups that had been treated with TH, suggesting that the microarrays are not sufficiently sensitive to pick up small changes in miRNA levels. However, we had no a priori reason to evaluate a battery of targeted miRNAs for analysis by qRT-PCR, so we identified differences in miRNA abundance between GD 16 and PND 15. These differences may or may not have been related to TH status, though TH levels change dramatically between these 2 developmental times. Comparison of cortical miRNA expression on GD 16 and PND 15 revealed a more than 10-fold decrease in the expression of miR-130, miR-20, and miR-25 from GD 16 to PND 15.

We found that, at GD 16, acute TH treatment of the dam resulted in decreased expression of select miRNAs and their precursors, while TH treatment of PND 15 pups had little effect on these miRNAs. The results suggest that TH-mediated regulation of these miRNAs, as with many other features of brain development, is dependent on the developmental stage. MiRNAs mediate their effects through binding to the 3′ UTR of target mRNA transcripts, which leads to mRNA degradation or inhibition of translation. It is estimated that over one-third of the transcriptome is regulated by miRNAs (Farh et al. 2005; Fahlgren et al. 2007). Thus, the observed differences in cortex miRNA transcriptome between GD 16 and PND 15 in the present study may play important roles in regulating tissue differentiation and development. Given that declining expression of miRNAs may lead to increased expression of their targets, genes identified as upregulated in the hyperthyroid group provide an excellent candidate list to mine for potentially affected targets. In silico analysis suggests that the most upregulated genes (Klf9, Fgfr1op2, and Ppap2b) in hyperthyroid fetal cortex are the targets of miR-16 or miR-106. Further searching showed that ∼30% of the upregulated genes in the hyperthyroid GD 16 fetuses are predicted targets of miR-16 or miR-106. Our analysis with 3′ UTR reporter assays confirms that Klf9 and Ppap2b are indeed targets of miR-106 and miR-16, respectively. These data suggest that upregulation of Klf9 and Ppap2b expression by TH can occur both by TR–TRE transcriptional activation and by the repression of miR-106 and miR-16 reducing their post-transcriptional inhibition.

In conclusion, we found that transient maternal hypothyroxinemia potentiates the transcriptional response to exogenous TH in the fetal cerebral cortex before the onset of fetal thyroid function. This was an unexpected finding that strongly suggests that even transient and mild hypothyroxinemia during development affects the fetal brain in a way that we still do not fully understand. The major strengths of this study include a unique experimental design and the breadth of technologies we employed to understand the effects. A limitation of the study is that we do not understand the full implications of this study, or the degree to which this observation is generalizable to other development times both because there are few studies of a comparable time, or of a comparable manipulation of thyroid status. However, these observations may be important when considering treating mild hypothyroidism during pregnancy with supplemental TH. We have identified several direct gene targets of TH action, as well as demonstrated TH regulation on the function of some miRNAs during brain development in mouse fetus. Thus, there appears to be a complex and redundant (direct gene regulation combined with gene regulation through miRNAs) functional network that coordinates TH-mediated effects on the developing pathway. These findings provide new insight into our understanding of the mechanism by which TH regulates early cortical development.

Supplementary Material

Supplementary material can be found at: http://www.cercor.oxfordjournals.org/

Funding

This work was funded in part by the Health Canada Genomics Research and Development Initiative to H.D, A.W., M.G.W., and C.L.Y., by grant ES010026 to R.T.Z., and by a generous gift to R.T.Z. by Mrs Audrey McMahon.

Notes

We thank Dr Jack Puymirat for the N2aβ cells. Mr Rémi Gagné developed the program that was used to identify potential TREs. We thank Dr Bhaja K. Padhi and Dr Guillaume Pelletier for helpful comments on the manuscript. Conflict of Interest: None declared. 

References

Anderson
GW
.
2008
.
Thyroid hormone and cerebellar development
.
Cerebellum
 .
7
:
60
74
.
Auso
E
Lavado-Autric
R
Cuevas
E
Del Rey
FE
Morreale De Escobar
G
Berbel
P
2004
.
A moderate and transient deficiency of maternal thyroid function at the beginning of fetal neocorticogenesis alters neuronal migration
.
Endocrinology
 .
145
:
4037
4047
.
Benjamini
Y
Drai
D
Elmer
G
Kafkafi
N
Golani
I
.
2001
.
Controlling the false discovery rate in behavior genetics research
.
Behav Brain Res
 .
125
:
279
284
.
Bernal
J
.
2007
.
Thyroid hormone receptors in brain development and function
.
Nat Clin Pract Endocrinol Metab
 .
3
:
249
259
.
Bernal
J
Nunez
J
.
1995
.
Thyroid hormones and brain development
.
Eur J Endocrinol
 .
133
:
390
398
.
Bolstad
BM
Irizarry
RA
Astrand
M
Speed
TP
.
2003
.
A comparison of normalization methods for high density oligonucleotide array data based on variance and bias
.
Bioinformatics
 .
19
:
185
193
.
Chan
S
Rovet
J
.
2003
.
Thyroid Hormones in the fetal central nervous system development
.
Fet Mat Med Rev
 .
14
:
177
208
.
Cui
X
Hwang
JT
Qiu
J
Blades
NJ
Churchill
GA
.
2005
.
Improved statistical tests for differential gene expression by shrinking variance components estimates
.
Biostatistics
 .
6
:
59
75
.
De Groot
L
Abalovich
M
Alexander
EK
Amino
N
Barbour
L
Cobin
RH
Eastman
CJ
Lazarus
JH
Luton
D
Mandel
SJ
et al
2012
.
Management of thyroid dysfunction during pregnancy and postpartum: an Endocrine Society clinical practice guideline
.
J Clin Endocrinol Metab
 .
97
:
2543
2565
.
Denver
RJ
Williamson
KE
.
2009
.
Identification of a thyroid hormone response element in the mouse Kruppel-like factor 9 gene to explain its postnatal expression in the brain
.
Endocrinology
 .
150
:
3935
3943
.
Donaldson
M
Jones
J
.
2013
.
Optimising outcome in congenital hypothyroidism; current opinions on best practice in initial assessment and subsequent management
.
J Clin Res Pediatr Endocrinol
 .
5
(Suppl. 1)
:
13
22
.
Dong
H
Paquette
M
Williams
A
Zoeller
RT
Wade
M
Yauk
C
.
2010
.
Thyroid hormone may regulate mRNA abundance in liver by acting on microRNAs
.
PLoS One
 .
5
:
e12136
.
Dong
H
Wade
M
Williams
A
Lee
A
Douglas
GR
Yauk
C
.
2005
.
Molecular insight into the effects of hypothyroidism on the developing cerebellum
.
Biochem Biophys Res Commun
 .
330
:
1182
1193
.
Dong
H
Yauk
CL
Rowan-Carroll
A
You
SH
Zoeller
RT
Lambert
I
Wade
MG
.
2009
.
Identification of thyroid hormone receptor binding sites and target genes using ChIP-on-chip in developing mouse cerebellum
.
PLoS One
 .
4
:
e4610
.
Dowling
AL
Martz
GU
Leonard
JL
Zoeller
RT
.
2000
.
Acute changes in maternal thyroid hormone induce rapid and transient changes in gene expression in fetal rat brain
.
J Neurosci
 .
20
:
2255
2265
.
Dugas
JC
Ibrahim
A
Barres
BA
.
2012
.
The T3-induced gene KLF9 regulates oligodendrocyte differentiation and myelin regeneration
.
Mol Cell Neurosci
 .
50
:
45
57
.
Escobar-Morreale
HF
Obregon
MJ
Escobar del Rey
F
Morreale de Escobar
G
.
1995
.
Replacement therapy for hypothyroidism with thyroxine alone does not ensure euthyroidism in all tissues, as studied in thyroidectomized rats
.
J Clin Invest
 .
96
:
2828
2838
.
Fahlgren
N
Howell
MD
Kasschau
KD
Chapman
EJ
Sullivan
CM
Cumbie
JS
Givan
SA
Law
TF
Grant
SR
Dangl
JL
et al
2007
.
High-throughput sequencing of Arabidopsis microRNAs: evidence for frequent birth and death of MIRNA genes
.
PLoS One
 .
2
:
e219
.
Farh
KK
Grimson
A
Jan
C
Lewis
BP
Johnston
WK
Lim
LP
Burge
CB
Bartel
DP
.
2005
.
The widespread impact of mammalian MicroRNAs on mRNA repression and evolution
.
Science
 .
310
:
1817
1821
.
Fisher
D
Dussault
J
Sack
J
Chopra
I
.
1977
.
Ontogenesis of hypothalamic-pituitary-thyroid function and metabolism in man, sheep, rat
.
Rec Progr Horm Res
 .
33
:
59
107
.
Flamant
F
Baxter
JD
Forrest
D
Refetoff
S
Samuels
H
Scanlan
TS
Vennstrom
B
Samarut
J
.
2006
.
International Union of Pharmacology. LIX. The pharmacology and classification of the nuclear receptor superfamily: thyroid hormone receptors
.
Pharmacol Rev
 .
58
:
705
711
.
Galton
VA
Wood
ET
St Germain
EA
Withrow
CA
Aldrich
G
St Germain
GM
Clark
AS
St Germain
DL
.
2007
.
Thyroid hormone homeostasis and action in the type 2 deiodinase-deficient rodent brain during development
.
Endocrinology
 .
148
:
3080
3088
.
Git
A
Dvinge
H
Salmon-Divon
M
Osborne
M
Kutter
C
Hadfield
J
Bertone
P
Caldas
C
.
2010
.
Systematic comparison of microarray profiling, real-time PCR, and next-generation sequencing technologies for measuring differential microRNA expression
.
RNA
 .
16
:
991
1006
.
Goodnight
CJ
Harvey
W
.
1997
.
Least-Squares Means in the Fixed-Effects General Linear Models
.
SAS Technical Report. Cary, NC, USA: SAS Institute
.
Grijota-Martinez
C
Diez
D
Morreale de Escobar
G
Bernal
J
Morte
B
.
2011
.
Lack of action of exogenously administered T3 on the fetal rat brain despite expression of the monocarboxylate transporter 8
.
Endocrinology
 .
152
:
1713
1721
.
Hashimoto
K
Curty
FH
Borges
PP
Lee
CE
Abel
ED
Elmquist
JK
Cohen
RN
Wondisford
FE
.
2001
.
An unliganded thyroid hormone receptor causes severe neurological dysfunction
.
Proc Natl Acad Sci USA
 .
98
:
3998
4003
.
Irizarry
RA
Bolstad
BM
Collin
F
Cope
LM
Hobbs
B
Speed
TP
.
2003
.
Summaries of Affymetrix GeneChip probe level data
.
Nucleic Acids Res
 .
31
:
e15
.
Jones
I
Ng
L
Liu
H
Forrest
D
.
2007
.
An intron control region differentially regulates expression of thyroid hormone receptor beta2 in the cochlea, pituitary, and cone photoreceptors
.
Mol Endocrinol
 .
21
:
1108
1119
.
Kapoor
R
Desouza
LA
Nanavaty
IN
Kernie
SG
Vaidya
VA
.
2012
.
Thyroid hormone accelerates the differentiation of adult hippocampal progenitors
.
J Neuroendocrinol
 .
24
:
1259
1271
.
Kester
MHA
Martinez de Mena
R
Obregon
MJ
Marinkovic
D
Visser
TJ
Hume
R
Morreale de Escobar
G
.
2004
.
Iodothyronine levels in the human developing brain: major regulatory roles of iodothyronine deiodinases in different areas
.
J Clin Endocrinol Metab
 .
89
:
3117
3128
.
Lafranchi
SH
.
2011
.
Congenital hypothyroidism: delayed detection after birth and monitoring treatment in the first year of life
.
J Pediatr
 .
158
:
525
527
.
Lebel
JM
Dussault
JH
Puymirat
J
.
1994
.
Overexpression of the beta 1 thyroid receptor induces differentiation in neuro-2a cells
.
Proc Natl Acad Sci USA
 .
91
:
2644
2648
.
Livak
KJ
Schmittgen
TD
.
2001
.
Analysis of relative gene expression data using real-time quantitative PCR and the 2(-Delta Delta C(T)) method
.
Methods
 .
25
:
402
408
.
Milanesi
A
Brent
GA
.
2011
.
Management of hypothyroidism in pregnancy
.
Curr Opin Endocrinol Diabetes Obes
 .
18
:
304
309
.
Morte
B
Ceballos
A
Diez
D
Grijota-Martinez
C
Dumitrescu
AM
Di Cosmo
C
Galton
VA
Refetoff
S
Bernal
J
.
2010
.
Thyroid hormone-regulated mouse cerebral cortex genes are differentially dependent on the source of the hormone: a study in monocarboxylate transporter-8- and deiodinase-2-deficient mice
.
Endocrinology
 .
151
:
2381
2387
.
Morte
B
Diez
D
Auso
E
Belinchon
MM
Gil-Ibanez
P
Grijota-Martinez
C
Navarro
D
de Escobar
GM
Berbel
P
Bernal
J
.
2010
.
Thyroid hormone regulation of gene expression in the developing rat fetal cerebral cortex: prominent role of the Ca2+/calmodulin-dependent protein kinase IV pathway
.
Endocrinology
 .
151
:
810
820
.
Morte
B
Manzano
J
Scanlan
T
Vennstrom
B
Bernal
J
.
2002
.
Deletion of the thyroid hormone receptor alpha 1 prevents the structural alterations of the cerebellum induced by hypothyroidism
.
Proc Natl Acad Sci USA
 .
99
:
3985
3989
.
Paquette
MA
Dong
H
Gagne
R
Williams
A
Malowany
M
Wade
MG
Yauk
CL
.
2011
.
Thyroid hormone-regulated gene expression in juvenile mouse liver: identification of thyroid response elements using microarray profiling and in silico analyses
.
BMC Genomics
 .
12
:
634
.
Pascual
A
Aranda
A
.
2013
.
Thyroid hormone receptors, cell growth and differentiation
.
Biochim Biophys Acta
 .
1830
:
3908
3916
.
Pathak
A
Sinha
RA
Mohan
V
Mitra
K
Godbole
MM
.
2011
.
Maternal thyroid hormone before the onset of fetal thyroid function regulates Reelin and downstream signaling cascade affecting neocortical neuronal migration
.
Cereb Cortex
 .
21
:
11
21
.
Picou
F
Fauquier
T
Chatonnet
F
Flamant
F
.
2012
.
A bimodal influence of thyroid hormone on cerebellum oligodendrocyte differentiation
.
Mol Endocrinol
 .
26
:
608
618
.
Porterfield
SP
Hendrich
CE
.
1992
.
Tissue iodothyronine levels in fetuses of control and hypothyroid rats at 13 and 16 days gestation
.
Endocrinology
 .
131
:
195
200
.
Porterfield
SP
Stein
SA
.
1994
.
Thyroid hormones and neurological development: update 1994
.
Endocrine Rev
 .
3
:
357
363
.
Rastogi
MV
LaFranchi
SH
.
2010
.
Congenital hypothyroidism
.
Orphanet J Rare Dis
 .
5
:
17
.
Rovet
J
Daneman
D
.
2003
.
Congenital hypothyroidism: a review of current diagnostic and treatment practices in relation to neuropsychologic outcome
.
Paediatr Drugs
 .
5
:
141
149
.
Rovet
JF
.
2005
.
Children with congenital hypothyroidism and their siblings: do they really differ?
Pediatrics
 .
115
:
e52
e57
.
Royland
JE
Parker
JS
Gilbert
ME
.
2008
.
A genomic analysis of subclinical hypothyroidism in hippocampus and neocortex of the developing rat brain
.
J Neuroendocrinol
 .
20
:
1319
1338
.
Selva
KA
Harper
A
Downs
A
Blasco
PA
Lafranchi
SH
.
2005
.
Neurodevelopmental outcomes in congenital hypothyroidism: comparison of initial T4 dose and time to reach target T4 and TSH
.
J Pediatr
 .
147
:
775
780
.
Sharlin
DS
Bansal
R
Zoeller
RT
.
2006
.
Polychlorinated biphenyls exert selective effects on cellular composition of white matter in a manner inconsistent with thyroid hormone insufficiency
.
Endocrinology
 .
147
:
846
858
.
Sirakov
M
Skah
S
Nadjar
J
Plateroti
M
.
2013
.
Thyroid hormone's action on progenitor/stem cell biology: new challenge for a classic hormone?
Biochim Biophys Acta
 .
1830
:
3917
3927
.
Thompson
CC
Potter
GB
.
2000
.
Thyroid hormone action in neural development
.
Cereb Cortex
 .
10
:
939
945
.
Vulsma
T
Gons
MH
de Vijlder
JJ
.
1989
.
Maternal-fetal transfer of thyroxine in congenital hypothyroidism due to a total organification defect or thyroid agenesis
.
N Engl J Med
 .
321
:
13
16
.
Yang
J
Zoeller
RT
.
2002
.
Differential display identifies neuroendocrine-specific protein-A (NSP-A) and interferon-inducible protein 10 (IP-10) as ethanol-responsive genes in the fetal rat brain
.
Brain Res Dev Brain Res
 .
138
:
117
133
.
Zoeller
RT
Rovet
J
.
2004
.
Timing of thyroid hormone action in the developing brain: clinical observations and experimental findings
.
J Neuroendocrinol
 .
16
:
809
818
.

Author notes

Hongyan Dong and Seo-Hee You have contributed equally to the work and should be considered Co-first authors.