Corresponding Author
Jack L. Leonard, PhD
Dept of Cellular and Molecular Physiology
UMASS Medical School
55 Lake Avenue North
Worcester, MA 01655
Fax 5088565997
jack.leonard@umassmed.edu

Introduction
Thyroid hormone is required for the normal development of the mammalian brain where it regulates a diverse set of developmental programs that include i) cell proliferation and migration, ii) apoptosis, iii) neuronal integration, and iv) dendritic arborization (see [1-4] for reviews). The window of time for thyroid hormone-dependent regulation of these processes is limited to pre- and perinatal life in rodents. Despite such obvious biology, the molecular events mediating the morphogenic actions of thyroid hormone have proven elusive. While it is clear that many of the actions of thyroid hormone are mediated by T3-dependent regulation of gene expression managed by specific chromatin-bound thyroid receptors (TR) (see the following reviews [5-7]), recently, nongenomic actions of thyroid hormone have also been validated [8-10]. These latter actions of thyroid hormone have become more important with the discovery that the brain of the TR-null mouse shows few, if any, of the developmental defects observed in the hypothyroid animal. In this review, we will examine the evidence for a nongenomic mechanism of action of thyroid hormone in brain by considering: 1) what role the cytoskeleton and the extracellular matrix play in mediating the actions of thyroid hormone; 2) which iodothyronine(s) are responsible for regulated cell trafficking; and 3) the nature of the thyroid receptor(s) that modulates neuronal migration.

The cytoskeleton as a target for thyroid hormone in the brain
The influence of altered thyroid status on brain development was first documented in the late 1960s by the demonstration that altered thyroid status disturbed the maturation of the cerebellum and led to defects in granule cell migration, Purkinje cell arborization, the timing of apoptosis, and neuronal integration [11, 12]. The cell’s cytoskeleton was one of the first choices as a target for thyroid hormone action [13-16] because of its central role in defining the architecture and motility of the cell. Analysis of microtubule polymerization in the developing cerebellum revealed that the expression of the tau family of microtubule-associated protein and at least 5 isoforms of tubulin were regulated by thyroid hormone. It is still not clear whether these changes in transcript abundance are due to a direct action on gene expression, are mediated by other genes, or are regulated by post-transcriptional mechanism(s) [17, 18].
Microfilaments are the other major component of the cell’s cytoskeleton and is composed of fibrils and fiber bundles of polymers of the mechano-chemical protein actin. Both T4 and reverse T3 (rT3) dynamically regulate actin polymerization in astrocytes by a nongenomic process [19, 20]. It is this ability of thyroid hormone to rapidly reorganize the actin cytoskeleton in the developing cerebellum that provides a key component of a nongenomic process capable of regulating neuronal migration. In vivo, the actin fibers used for pathfinding and guidance of the migrating neurite/growth cone during neuronal maturation are disassembled in the cerebellum of neonatal hypothyroid rats [21-23], and this defect can be repaired by single injection of thyroxine [24]. Total cellular actin levels do not change with thyroid status; only the relative proportions of polymers (fibrous) actin to monomeric actin are influenced by T4 [25].
A second component of a nongenomic regulatory process that can directly impact neuronal migration is derived from the fact that neuronal migration/neurite guidance is directed by external cues from the extracellular matrix (ECM) protein laminin. Laminin is a product of astrocytes that is held in polymer arrays on the astrocyte cell surface by specific transmembrane receptors composed of integrin subunits. These integrin receptors cluster to form focal contacts that are anchored in place by actin filaments that bind to their cytoplasmic tails [26, 27], and integrin clustering is necessary to anchor laminin arrays on the cell surface. T4 regulates the organization of microfilaments in both astrocytes and neurons, especially those in neuronal processes, in vitro and in vivo [28]. One important consequence of loss of the microfilaments in the developing cerebellum of hypothyroid neonates is the temporal disruption of laminin deposition ~7-10 days [29, 30]. T4 replacement, given at least one day prior to the critical time period for granule cell migration (~7-10 days after birth), normalizes the timing and topological complexity of laminin deposition and, thereby, facilitates the orderly migration of granule neurons. Interestingly, a laminin receptor has recently been identified as a specific T4 binding protein on the cell membrane [31], although the contribution of such binding to the organization of the ECM is unknown. Thus, by modulating the organization of the actin cytoskeleton, thyroid hormone directly regulates the deposition and organization of a key guidance cue used by migrating granular neurons.

Which iodothyronine is responsible for regulating cell trafficking?
While T3 is widely thought to be the bioactive form of thyroid hormone, the T4-dependent regulation of actin polymerization is an exception to this rule. Comparison of the ability of individual iodothyronines to initiate actin polymerization in astrocytes revealed that T4, and its metabolically inert metabolite, reverse T3 (rT3)—two iodothyronines that do not regulate transcription—are at least 100-fold more potent that the transfactor activator, T3. Similarly, acute hormone replacement with either T4 or rT3 (6 hr treatment) completely restores microfilament organization to normal in the cerebellum of 14 day old, hypothyroid neonates, while acute T3 replacement fails to correct this defect [32]. These findings show that the effector protein(s) mediating the non-genomic regulation of actin polymerization have a ligand preference very different from that of the transcription regulating TRs. Using microfilament remodeling in astrocytes to evaluate iodothyronine potency, the thyroid hormone dependent effector(s) were shown to prefer iodothyronines with a fully substituted phenolic ring, i.e. two iodines (such as T₄ and rT₃) and an alanine side chain with a neutral or net positive charge [33]. Removal of one phenolic ring substituent or the presence of a net negative charge on the alanine side chain destroys the ability of the iodothyronine to regulate actin polymerization. Importantly, inactive analogs with negatively charged alanine chains were “reactivated” by blocking the free carboxyl group of the alanine side chain [33]. Thus, the effector molecule mediating thyroid hormone-dependent actin polymerization possesses a unique set of thyroid hormone binding properties that distinguish it from all the known TH binding proteins, including the TR and the cell surface receptor [31].

The role of thyroid receptors: neuronal migration
While the T4-dependent regulation of actin polymerization is clearly a nongenomic event, in mammals, most of the actions of thyroid hormone are mediated by chromatin bound T3 receptors. These T3 receptors are encoded by two genes—TRα (NR1A1) and TRβ (NR1A2). The TRα encodes at least four gene products, the T3-binding TRα 1 and three that do not bind T3; TRα2, TRΔα 1 and TRΔα2. In the cerebellum, T3 binding, TRβ gene products are found in the nuclei of oligodendrocytes and Purkinje neurons [34, 35]. On the other hand, granular neurons express TRα gene products suggesting that they may be responsible for thyroid hormone dependent granule cell migration [36].
The lack of overt developmental defects in the brains of most TR knockout mice was unforeseen and two possible explanations for this anomaly have emerged. The first is that gene repression by the unliganded TRα mediates the developmental defects observed in the developing brain of neonates; the unliganded TRα1 appears to suppress hormone-induced tissue development in frogs [37]. A second is that a biology other than T3 regulated gene expression—a nongenomic action—mediates thyroid hormone’s influence on the developmental program of the brain.
The first possibility is supported by the finding that selective deletion of the T3-binding gene product of the TRα1 gene (TRα1^-/-) eliminates the delay in granule cell migration observed after chemically induced hypothyroidism [36]. While it is generally presumed that the persistence of the external granule layer of the neonatal hypothyroid cerebellum is due to a delay in granule cell migration, a delay in the timing of the proliferation of the granule cell precursors also contributes to the preservation of a thickened external granule layer. Importantly, neonatal hyperthyroidism is associated with the premature termination of granule cell precursor proliferation [16, 38], while neonatal hypothyroidism is associated with a shift in the timing of granule cell proliferation by about 7-10 days. The role of the T3-binding TRα1 in the regulation of granule cell precursor proliferation remains to be determined.
The unliganded TR is also thought to play a role in the survival of the developing neonate. Deletion of the entire TRα gene locus (TRα⁰⁰) [39] greatly improved the survival of congenitally hypothyroid (Pax8^-/-) neonatal mice [40], although the growth retardation observed in these athyroidic neonates. It now appears that poor survival of the Pax8^-/- neonate is a consequence of impaired gut maturation [40]—a developmental defect shared by the TRα ^-/- neonate that lack the full-length TRα1 and TRα2, but express truncated TRΔα1 or TRΔα2 [41]. The TRα ^-/- shows a progressive fall in circulating T4 and T3 beginning 12-14 days after birth that leads to death between 3 and 5 weeks of life and like the hypothyroid Pax8^-/- mouse, T3 treatment beginning on day 21 normalizes gut maturation and improves survival. More recently, the role of the unliganded TRα1 in the neonatal survival has been questioned. Selective inactivation of the full-length TRα1 and the truncated TRΔα1 (TRα1^-/-) gene products failed to improve the survival of the congenitally hypothyroid Pax8^-/- neonatal mouse [42]. This raised the possibility that TRα2 and/or TRΔα2 may be responsible for some of the developmental defects that lead to the poor survival of congenitally hypothyroid neonates.

Direct analysis of the role of the four TRα gene products TRα1, TRΔα1 TRα2, TRΔα2 on thyroid hormone dependent microfilament remodeling and laminin deposition by astrocytes was done with TRα1^-/-, TRα2^-/-, and TRα⁰⁰ mice and the data are summarized in Table I. None of these targeted gene deletions altered total cellular actin in astrocytes, but the actin cytoskeleton was disassembled in cells lacking TRα1, TRΔα1 (TRα1^-/-) and in mice lacking all for TRα gene products, TRα1, TRΔα1 TRα2, TRΔα2 (TRα^0/0). The TRα2^-/- mouse lacking TRα2, TRΔα2 [43] showed a normal actin cytoskeleton Similarly, astrocytes from the TRα1^-/-, and TRα^0/0 mice did not assemble laminin arrays on their cell surface, while astrocytes from TRα2^-/- or wild-type mice showed abundant arrays of extracellular laminin. Direct examination of the programmed deposition of laminin in the molecular layer of the neonatal cerebellum in TRα1^-/-, TRα2^-/-, TRα^0/0 mice revealed that the 7-10 day delay in laminin deposition observed in the hypothyroid cerebellum was also observed in the TRα1^-/- and TRα⁰⁰, but not the TRα2^-/- animal [43]. These data suggest that the full-length TRα2 and the truncated TRΔα2 do not contribute to the organization of laminin on the astrocyte cell surface. While, both the full-length TRα1 and its truncated partner TRΔα1 appear to be candidate(s) for the effector mediating laminin deposition, the failure of T3 to modulate actin polymerization or to facilitate laminin deposition in astrocytes or cerebellar extracts [10, 33, 44] suggests that the full length TRα1 does not participate in these biological events. Thus, these findings raise the possibility that non T3-binding TRα gene products, such as TRΔα1 and/or TRΔα2, are likely candidates for the effector molecule(s) that mediate T4-dependent regulation of cerebellar development and launch an important new avenue of research for the future.


Summarizing table:
Cytoskeletal actions of iodothyronines

A. The cytoskeleton as a target for thyroid hormone in the brain
1. Early development of the cerebellum

	a. Components of the microtubules, tubulin and MAPs, show hormone dependent changes in transcript and protein abundance
	b. The organization of actin fibers are regulated by thyroid hormone
	c. Cell migration requires an intact cytoskeleton

2. Astrocyte function

	a. The actin polymerization is regulated by thyroid hormone
	b. Integrin receptor clustering is requires an intact actin cytoskeleton
	c. Laminin deposition on the astrocyte cell surface is regulated by thyroid hormone

B. Iodothyronines responsible for regulating cell trafficking in the cerebellum
1. Astrocytes in culture

	a. T4, and rT3, but not T3, initiate hormone-dependent actin polymerization
	b. Iodothyronine specificity is determined by the charge on the alanine side chain and by the presence of two iodines on the phenolic ring.

2. In organ culture, and in vivo

	a. T4, and rT3 initiate hormone-dependent actin polymerization, and neurite outgrowth
	b. T4, and rT3 initiate laminin deposition in the molecular layer of the cerebellum
	c. pharmacological doses of T3 do not influence actin polymerization, neurite outgrowth or laminin deposition

C. The role of thyroid receptors neuronal migration
1. Astrocyte function

	a. TRalpha gene products are required for hormone dependent actin polymerization and laminin deposition
	b. deletion of three of the four potential TRalpha gene products, TRalpha1, TRalpha2 and delta alpha2 have no effect on hormone dependent actin polymerization and laminin deposition
	c. deletion of all TRalpha gene products, including delta alpha1, leads to the disruption of the action cytoskeleton and eliminates laminin deposition.

2. Developing cerebellum

	a. the T3-binding TRalpha1, and the non-T3 binding TRalpha2 or delta alpha2 are not required for actin polymerization or laminin deposition
	b. the absence of delta alpha1 disrupts actin polymerization and laminin deposition