Nongenomic Actions of Thyroid Hormone in Brain Development

Thyroid hormone (TH) is essential for normal brain development (1, 2). The absence of TH during the critical period of neurogenesis leads to multiple irreversible morphological abnormalities due, in large part, to defects in neuronal migration and neuronal process outgrowth. Up until the late 1970s, work had focused on the influence of TH on the cytoskeleton of the developing neuron (3–6), but this avenue was dropped with the discovery that the thyroxine metabolite, 3′,3,5-triiodothyronine (T₃), interacted with a transcription factor that regulated specific target genes (7). These TH dependent transfactors belong to the c-erbA family of gene products that encode receptors for the steroid and thyroid hormones (8). More recently, the discovery that the brain developments normally in animals lacking these chromatin bound thyroid receptors, once again has raised the possibility that non-genomic actions of TH participate in the morphogenic actions of this hormone, especially in the brain. In this brief review, I will focus on the recent experimental support resurrecting the concept that TH dependent regulation of the cytoskeleton plays a key role in the developmental program of the brain.

The rat cerebellum completes its developmental program during the first 2–3 weeks of life and was one of the first identified targets for TH action. It is widely used to study the developmental effects of TH (9–11). In the euthyroid rat cerebellum, granule cell precursor neurons located in the External Granular Layer (EGL) undergo intense proliferation during the first week of life. Following a 24 hour period of lateral migration along the interface between the EGL and the Molecular Layer (ML) (Fig. 1, Stage 1), the granule cell sends a single radial projection through molecular layer into the Internal Granular Layer (IGL) (Fig. 1, Stage 2), and the cell body then migrates down this projection into the IGL (Fig. 1, Stage 3), where it synapses with axons in the climbing fibers from higher brain centers (12). Radial migration of granule cells reaches a maximum between postnatal days 10–14 and by 21 days of life the EGL has all but disappeared. During granule cell migration, there is extensive process outgrowth and dendritic arborization by the Purkinje cells, the target for granule neuron axonal projections. In the hypothyroid rat cerebellum, the EGL is retained for up to 24 days and there is increased granule cell death in the IGL presumably due to delayed granule cell migration from the EGL and a failure of the maturing granule cell to make connections (13, 14). As a result, the dendritic arbor of the Purkinje cell is markedly blunted in the hypothyroid rat (15) due, in large part, to decreased synaptogenesis (16). Importantly, TH replacement, specifically T₄ or rT₃, during the first two weeks of life rescues granule cell migration and normalizes Purkinje cell arborization and the cerebellum to complete its maturation according to the euthyroid developmental program. While the consequences of hypothyroidism on cerebellar development are well recognized, the biochemical and molecular bases for the effects of this morphogenic hormone on neuronal integration remain elusive.

Figure 1. Developmental Fate of the maturing Granule Neuron in the Cerebellum.

EGL, external granule layer; ML, molecular layer, IGL, internal granule layer.

Transcriptional actions of TH on the brain developmental program

The major TH produced by the thyroid gland is thyroxine (T₄) which is deiodinated in peripheral tissues to the transcriptionally active iodothyronine, 3,5,3′-triiodothyronine (T₃) (17) or to the transcriptionally inert metabolite, 3,3′,5′-trioodothyronine (reverse T₃, rT₃), which is the predominant iodothyronine produced during fetal life (18–20). Many of the actions of TH are initiated by the binding of T₃ to specific chromatin-bound, ligand-activated receptors (thyroid receptors, TRs) (21–23). There are two TR genes that encode at least 10 gene products—only three are chromatin bound, T₃ binding receptors (24). While the predominant T₃-binding TRs, TRβ1 and TRα1, are developmentally expressed in the rat brain, the most abundant TR isoform expressed throughout development is the non-T₃-binding TRα2 (25, 26). TRα1 does not show developmental changes in expression and is the predominant isoform in cerebellar granule cells (27). On the other hand, TRβ1 is the predominant isoform in the Purkinje cells (28) and shows a marked increase in abundance just prior to birth and remains elevated until postnatal day 10–12; a pattern that coincides with, rather than precedes, TH-dependent neurite integration (25, 29). While differential display and microchip analysis have identified a number of cerebellar gene products that are altered by thyroid status, few genes have been shown to be directly regulated by T₃ (ie contain an identified thyroid hormone-response element) (for review see (26, 30). Thus, the identification of developmentally important, T₃-responsive genes in the brain has been difficult at best.

Recent work done in transgenic mice suggests that T₃-induced transcriptional regulation is not the sole contributor to the brain developmental program. Mice lacking one or more of the TR isoforms show abnormalities in the auditory system, the thyroid-pituitary axis, the heart, the skeletal system and the small intestine, but few, if any, abnormalities in brain development (31, 32). This is true for both single receptor knockout mice (33, 34) and for knockout mice devoid of all known TRs (35, 36).

Since the developmental deficits observed when nuclear TRs are missing are markedly less than those observed in hypothyroid neonates, it has been proposed that the unliganded TRs participate in this developmental program (32, 37). Because TRs bind to the thyroid response element of target genes in the unliganded state and in vitro studies have shown that the unliganded receptor represses transcription (38), some have proposed that during brain development T₃ acts on target genes to release gene repression, rather than to activate gene expression (32). However, this hypothesis cannot account for the developmental defects observed in hyperthyroid neonates, and ignores the presence of the most abundant TR transcriptional repressor that can not bind T₃, TRα2 (25, 39). Importantly, there is no direct in vivo demonstration of regulated gene repression by either the unliganded T₃-binding TRs or by TRα2; instead all in vivo data has been inferential and based on the presence of a more benign phenotype than predicted. Thus, it seems likely that TH-dependent processes other than gene regulation contribute to the developmental program of the brain.

Non-genomic actions of TH in the brain: Regulation of actin polymerization

One well-recognized non-genomic action of TH is the T₄-dependent regulation of actin polymerization. In astrocytes and the neuronal processes of granular neurons grown in the absence of TH only 40–60% of the cellular actin is polymerized (F-actin) and the ctin cytoskeleton is disorganized (Figure 2). Both T₄ and rT₃, but not T₃, rapidly (~10–20 minutes) increase the F-actin content to 90–95% of the cellular actin and restore the microfilaments without altering total actin content or gene expression (40–42).

Figure 2. Effects of rT₃ on microfilament organization in astrocytes.

Astrocytes were grown on p-lysine coated coverslips for 48 h in DMEM containing 10% calf serum and antibiotics. Twenty-four hours before staining, growth medium was replaced with serum-free DMEM containing 1 mg/ml BSA ± 10 nM rT₃. After 24 hrs, coverslips were then washed, fixed with 4% paraformylaldehyde and the actin cytoskeleton visualized with alexafluor⁴⁸⁸-phallacidin as described in ref (41).

These cell culture findings mimic those observed during cerebellar development in vivo, where the F-actin content of the hypothyroid rat cerebellum remains significantly lower than that of the euthyroid cerebellum until postnatal day 15–16 (43). A single injection of either T₄ or rT₃ to the hypothyroid neonate on day 14 will normalize the F-actin content in the cerebellum within 3 h without altering the total cerebellar actin content (42). As observed in cell culture, T₃ has no effect on either F-actin or total actin content in the hypothyroid rat cerebellum. These data indicate that TH initiated changes in actin polymerization during brain development is one action of TH that does not require the TRs and is an especially attractive mechanism of action because the two effector hormones, T₄ and rT₃, are the two predominant thyroid hormones produced during fetal life (17–20).

Actin polymerization, microfilament organization and neuronal migration

During normal brain development developing neurons migrate over long distances and project axons along specific pathways towards target cells using a cohort of transmembrane sensors tethered in place by the actin cytoskeleton (44, 45). Thus, the ability of the developing neuron to regulate actin polymerization in neurites is essential for the interpretation of extracellular guidance cues (45–48). Chemical disruption of neuronal actin cytoskeleton markedly impairs neuronal growth cone motility and pathfinding ability in vitro (48, 49). In light of the effects of TH on actin polymerization, we found that the lack of TH led to a marked reduction in granule cell migration from cerebellar explants and a dramatic decrease in neurite outgrowth and process formation. Replacement with T₃ did not correct these defects; however, physiological replacement with either the T₄ or rT₃ normalized granule cell migration and neurite formation (50). Direct analysis of the actin cytoskeleton in the cells migrating from cerebellar explants revealed a large decrease in F-actin content in the cellular projections when grown in TH-deficient or T₃-treated medium and that both T₄ or rT₃ restored the F-actin content of granule neurons to normal (50). These data illustrate the ability of transcriptionally inert iodothyronines such as T₄ and rT₃ to regulate neuronal cell migration by modulating the organization of the neuronal actin cytoskeleton.

Regulation of the actin cytoskeleton in the supporting astrocytes is also essential for the organization of extracellular neuronal guidance molecules during brain development. One such guidance molecule is laminin, an astrocyte-derived extracellular matrix protein that appears on the astrocyte surface during neuronal migration (51–53). Secreted laminin binds to integrins and is organized in specific patterns on the astrocyte surface by macromolecular focal contacts (54, 55). Focal contacts then transduce mechanical forces through their associated microfilaments. Microfilament disassembly in the TH-deficient or T₃-treated astrocyte prevents focal contact formation on laminin coated surfaces, while abundant focal contacts are found on T₄-treated cells (56). The loss of focal contacts prevents newly secreted laminin from binding to the astrocyte surface and leads to the loss of this guidance molecule from the cell surface. In the presence of T₄ or rT₃, focal contacts form normally leading to the accumulation of secreted laminin in macromolecular arrays of the astrocyte’s cell surface (50, 57, 58). These TH-dependent alterations in laminin organization on the astrocyte surface occur in the absence of any changes in laminin γ chain mRNA expression, laminin protein synthesis, or the rate of laminin secretion (57).

These in vitro findings also have an in vivo counterpart in the developing cerebellum of hypothyroid neonates where the appearance of laminin derived migration pathways are both delayed and diminished in quantity (58). These data provide the essentials to construct a physiological pathway where TH-dependent regulation of the polymerization state of actin in the astrocyte and the neuron modulates the production and recognition of guidance cues—cues that if disrupted lead to abnormal neuronal migration and neuronal process formation—and lead to the morphological deficits observed in the cretinous brain.