Thyroxine Targets Different Pathways of Internalization of Type I1 Iodothyronine 5”Deiodinase in Astrocytes*

In the brain, thyroid hormone dynamically regulates levels of the short-lived plasma membrane protein, type 11 iodothyronine 5”deiodinase. In cultured astrocytes, thyroxine modulates deiodinase levels by acti- vating cytoskeletal-plasma membrane interactions that increase the rate of inactivation of the enzyme. Here we characterized the effects of these thyroxine- dependent cytoskeletal interactions upon the route of internalization of the deiodinase by following the in- tracellular transit of the affinity-labeled substrate- binding subunit of the deiodinase (p29). Thyroxine rapidly induced the inactivation of the deiodinase and initiated the binding of p29 to F-actin. By 40 min, >76% of the p29 had been transported to an endosomal pool, which was followed by dissociation of the F-actin- p29 complex. There was no significant accumulation of p29 in the dense lysosomes seen in the presence of thyroxine. In the absence of thyroxine, p29 was inter- nalized and transported to the dense lysosomes at a rate parallel to the inactivation rate of the deiodinase (tn 0.76 and 0.64 h, respectively) without involvement with the microfilaments. These data demonstrate that thyroxine targets type I1 iodothyronine 5”deiodinase to an endosomal pool by activating specific protein-F-actin interactions involved in microfilament-mediated intracellular protein trafficking. Type

. These actions of thyroxine occur independently of transcription and/or translation, and thyroxine is >loofold more potent than the more traditionally bioactive T3 in eliciting these actions (3-9). Using cyclic nucleotide-stimulated cultured astrocytes that express high levels of 5'D-I1 and exhibit all the regulatory aspects seen in the brain in vivo (4)(5)(6), we have shown that thyroxine dynamically regulates 5'D-I1 activity by promoting interactions between the enzyme and the F-actin stress fibers that lead to inactivation and internalization of the enzyme ( 5 ) .
Eukaryotic cells selectively transport membrane proteins through various distinct membrane pools by mechanism(s) that are poorly understood. One example of membrane protein internalization is the endocytotic pathway (10-12). While the distinct membrane compartments involved in this pathway have been extensively characterized, the mechanism behind the movement and targeting of internalized vesicles to their respective subcompartments remains obscure. Increasing evidence points toward a major role for the cytoskeleton in the regulation of such intracellular protein trafficking (13)(14)(15)(16)(17)(18). While the microtubules are most often associated with the intracellular movement of organelles, microfilaments have recently been shown to have an important role in these processes (17)(18)(19)(20)(21)(22)(23)(24)(25)(26).
Thyroxine-mediated turnover of 5'D-I1 provides an excellent model for the study of the role of the microfilaments in intracellular protein trafficking. For example, cyclic nucleotide-stimulated, thyroxine-deficient astrocytes have elevated 5'D-I1 levels, lack the radial F-actin stress fibers, and the turnover of 5'D-I1 is relatively slow (5-7). Addition of thyroxine rapidly restores F-actin stress fibers (<5 min) (7), and 5'D-I1 levels rapidly fall due to thyroxine-initiated binding of the enzyme to the F-actin stress fibers ( 5 ) . Using immunocytochemistry and confocal microscopy, we have shown that the plasma membrane-bound 5'D-I1 is quickly translocated an intracellular pool and that microtubules do not play a role in this cytoskeletal-mediated 5'D-I1 translocation ( 5 ) . However, the destination of the internalized 5'D-II, the identity of the intracellular membrane compartment(s) used during its transit, and the molecular events that mediate the exchange of this membrane protein between the different membrane compartments are unknown.
In this study, we have utilized the thyroxine-stimulated inactivation of 5'D-I1 in astrocytes to examine the role of the F-actin stress fibers in the internalization pathway(s) of short-lived membrane proteins. Using density gradient centrifugation to follow the intracellular transit of the affinitylabeled substrate-binding subunit of 5'D-II, we show that, after thyroxine stimulates the binding of 5'D-I1 to F-actin, the microfilaments transport this plasma membrane-bound enzyme to the endosomes. In the absence of thyroxine, 5'D-I1 is transported to the dense lysosomes without the use of the F-actin cytoskeleton. These data demonstrate that thy-Pathways of 5'0-11 roxine targets the internalization of type I1 5"deiodinase to an endosomal pool by activating specific protein-F-actin interactions involved in microfilament-mediated intracellular protein trafficking.
Confluent cells from passages 2-4, containing >95% astrocytes (22), were utilized for experiments. Cultures were grown in supplemented DMEM without serum for at least 8 h to remove serum iodothyronines, followed by a 16-h stimulation period with 1 mM BtSAMP and 100 nM hydrocortisone to induce steady-state levels of 5'D-I1 (14) prior to initiation of experimental procedures.

Affinity Labeling
Cells were affinity-labeled with 10 nM B~A c [ '~~I ] T , for 20 min as described previously (5). Labeling media was then removed, and cells were washed once in Hanks' solution. Affinity-labeled cells were then treated with either 10 nM thyroxine, T3, or no hormone in the presence of 1 mg/ml bovine serum albumin. Cells were harvested by scraping, suspended in ice-cold 150 mM sodium chloride, 20 mM sodium phosphate buffer, pH 7.4 (PBS), and collected by centrifugation.
Transferrin Receptor Labeling (28) Confluent astrocytes grown in medium containing 10% serum were washed free of growth medium with excess Hanks' solution. Cells were then incubated in supplemented DMEM without serum for 30 min at 4 "C followed by labeling with 5 pCi of '261-labeled transferrin for 2 h at 4 "C. The transferrin-labeled cells were then washed free of unbound radiolabel, and the medium was replaced with warmed (37 "C) supplemented DMEM without serum. At the times indicated, cells were harvested by scraping in phosphate-buffered saline, resuspended in 500 pl of sucrose buffer, homogenized with one freeze-thaw cycle, and fractionated through 16% and then 8% Percoll gradients as described above. Fractions (0.5 ml) were collected and counted.

Western Analysis
Actin and the @-subunit of the Na+/K+ ATPase were detected by Western analysis (29). Actin polymers were depolymerized at low ionic strength in 0.05 mM calcium chloride, 1 mM dithiothreitol, 3 mM sodium azide, 2 mM Tris-HCI, pH 8.0 (depolymerizing buffer), denatured in a boiling water bath for 5 min, and applied directly to nitrocellulose using the Minifold I1 slot blot apparatus (Schleicher and Schuell) (30). Blots were blocked for 1 h with 50 mM Tris-HC1, 0.05% (v/v) Tween-20, 50 mg/ml powdered milk, 2 mM calcium chloride, pH 8.0, then probed for 16 h at room temperature with an affinity-purified polyclonal rabbit antisera for either actin (1:500 dilution) or the @-subunit of the Na+/K+ ATPase (1:lOOO dilution). Immune complexes were visualized by radioautography following incubation of the blots with 1261-labeled protein G (-300,000 cpm/ ml). Quantification of actin and the @-subunit of the Na+/K+ ATPase was determined by scanning densitometry.
Isolated Fractions-Aliquots (125 pl) from fractions obtained from Percoll gradients were diluted 5-fold with Triton buffer, triturated, and incubated for 10-15 min at 25 "C. Aliquots were then spun for 130,000 g-min in a microcentrifuge, and the supernatants were discarded. The pellets were resuspended in 100 r l of depolymerizing buffer, triturated, and incubated for 16 h at 4 "C. Proteins were then resolved by SDS-PAGE, and radioautograms were analyzed as described above.

Vesicle Isolation
Aliquots (75 pl) from fractions obtained from Percoll gradients were diluted to a total volume of 375 pl in sucrose buffer in the presence and absence of 0.5% taurodeoxycholate. Aliquots were then centrifuged for 6 X lo6 g-min, and the pellets were resuspended in 100 ~1 of depolymerizing buffer, triturated, and incubated for 16 h at 4 "C. Proportionate aliquots from the pellets (50 pl) and supernatants (187 pl) were then analyzed by SDS-PAGE and radioautography as described above.

Miscellaneous Methds
Protein was determined using the method of Bradford (32) with human immunoglobulin as the protein standard. The Student's t test was used to compare the differences between experimental points. When indicated, results are presented as the mean & S.D.

Subcellullar
Fractionation of Cultured Astrocytes-We have previously used 16% Percoll gradients and affinity labeling to examine the subcellular localization of 5'D-II (5). After shortterm (20 min) treatment of thyroid hormone-deficient astrocytes with 10 nM thyroxine, there is a redistribution of the affinity-labeled 29-kDa substrate-binding subunit of 5'D-I1 (p29) within fractions that contain markers for the plasma membrane, endosomes, endoplasmic reticulum, and buoyant lysosomes. Because of the overlap of organelles in these fractions, we could not determine if this shift of 5'D-11 represents the transport of the affinity-labeled enzyme from the plasma membrane to another membrane pool.
In order to gain better resolution of organelles of similar density, the cytosolic fractions were discarded and fractions, from the 16% Percoll gradients with a buoyant density range of 1.034-1.050 g/ml that contain >90% of the membranebound organelles (5), were refractionated on 8% Percoll gradients. The distribution of marker enzymes after refractionation on 8% Percoll gradients is shown in Fig. 1. Na'/K+ ATPase, a marker for the plasma membrane, is found predominantly in fractions 9 and 10 (density -1.041 g/dl), along with the major protein peak. The distribution of the @-subunit of the enzyme is broader than that of the catalytically active  enzyme (found in a single peak in fraction lo), demonstrating the differential sensitivity between the assay for enzyme catalytic activity and the analysis of enzyme immunoreactivity. Actin is found predominantly in fractions 9 and 10 in the form of F-actin since all of the G-actin is found in the cytosolic fractions (5) and discarded prior to refractionation.
As seen in previous studies (5, 331, the lysosomal marker P-glucuronidase identifies two populations of lysosomes. The less dense population (fractions 8-11) consists of the buoyant lysosomes and contains -38% of the enzymatic activity in the gradient, with the major peak in fraction 9. The dense lysosomes contain -35% of the enzymatic activity and are found in fractions 13-18 (density 1.043-1.046 g/dl). The endoplasmic reticulum marker, glucose-6-phosphatase, also exhibits the bimodal distribution seen previously (5), with the major peak (fraction 11) of the less dense population migrating at a slightly higher density than the plasma membrane markers.
The endosomes were identified by following the internalized transferrin receptor (Fig. 1). Surface labeling of the transferrin receptor with ['251]transferrin at 4 "C resulted in -70% of the radioactivity comigrating with the plasma membrane markers in fraction 9 and 10. After warming the cells to 37 "C for 10 min, -20% of the ['251]transferrin is found in fractions 12 and 13 (density 1.042 g/dl). This represents internalization of the receptor-ligand complex to the endosomal pool, with -55% of the ['251]transferrin remaining in the plasma membrane pool. A minor fraction (15-20%) of the radioactivity appears in fraction 22 and is unchanged by warming. It is unlikely that this minor population represents any specific intracellular organelle and, thus, represents dense aggregates of transferrin formed during refractionation.
Effect of Thyroxine on the Intracellular Transit of 5'D-11- The greater separation of organelles achieved by successive gradient isolation allowed the effects of thyroid hormone on intracellular transit of 5'D-I1 to be determined. Shown in Fig.  2A is the internalization pathway of 5'D-I1 in cells treated with 10 nM thyroxine. As shown previously (5), 80-90% of the affinity-labeled substrate-binding subunit of 5'D-I1 (p29) is initially found in fractions comigrating with the plasma membrane (fractions 9, lo), with 510% of the p29 found in higher density fractions. Little movement of p29 between gradient fractions was initially detected after treatment with thyroxine, despite the marked decrease in 5'D-I1 catalytic activity (Fig. 2 A ) . Since -60% of p29 becomes associated with F-actin within 20 min after the addition of thyroxine (5) and F-actin comigrates with the plasma membrane (Fig. l), the lack of a dramatic shift in density of p29 coinciding with the loss of catalytic activity is anticipated. T o establish that the formation of a p29-F-actin complex accounts for the lack of movement of p29 from the plasma membrane fractions after thyroxine treatment, organelles migrating in fraction 10 after 20-min treatment with thyroxine were diluted &fold and the cytoskeletal components separated from the membrane vesicles by low speed centrifugation at 130,000 g-min. Most of the p29 (444 k 230 arbitrary units/fraction) was recovered in the cytoskeletal pellet along with the F-actin stress fibers. Depolymerization of the microfilaments present in fraction 10 prior to separating the membrane vesicles from the cytoskeleton resulted in the complete lost of p29 ( 4 arbitrary units/ fraction) from the cytoskeletal pellet.
In cells treated with thyroxine for 60 min, p29 shifts to a single fraction of higher density (fraction 12, density 1.042 g/ dl (Fig. 2, A and B). Fraction 12 contains the peak of the marker for the endosomal pool (transferrin, Fig. 2B). This endosomal-p29 pool is distinct from the plasma membrane, the two populations of lysosomes, and the major peak of the endoplasmic reticulum. There is no apparent transfer of p29 to the dense lysosomes after 60 min of thyroxine treatment. The small fraction of the total p29 (10-15%) spread over higher density fractions (fractions 15-21, Fig. 2 A ) remains stable in the presence of thyroxine. Incubation of thyroxinetreated astrocytes with either the lysosomotropic agents NH4C1 or chloroquine (34) or the endocytosis inhibitor monodancylcadaverine (35) blocked the thyroxine-stimulated internalization of p29 (Table I).
Intracellular Transit of p29 in Thyroid Hormone-deficient or T3-treated Astrocytes-Since dynamic regulation 5'D-I1 inactivation is an extranuclear action of thyroid hormone, the traditionally bioactive T3 is >100-fold less potent than thyroxine (3-9), and 10 nM T3 has no effect on 5'D-I1 inactivation over that seen in thyroid hormone-deficient conditions. Shown in Fig. 3 is the internalization pathway of 5'D-I1 in cells treated with 10 nM TB. In contrast to the translocation of p29 solely to an endosomal pool in thyroxine-treated astrocytes, treatment of astrocytes with T3 results in the translocation of p29 to the dense lysosomes. By 20 min, -20% of the p29 is found in fractions 13-18, comigrating with the dense lysosomes. After 60 min, only 30% of the p29 remained in the plasma membrane fractions, while -50% of the p29 is found in to the dense lysosomes.
As shown in Table 11, the loss of p29 from the plasma membrane fraction and subsequent accumulation in the lysosomes parallels the decrease in enzyme catalytic activity in cycloheximide-treated cells. By 60 min, the total cellular content of p29 decreased by 30-40% consistent with the degradation of this polypeptide in the lysosomes. The intracellular pathway of p29 under thyroid hormone-deficient conditions is identical to that seen in the Ts-treated cells (data not shown).
Treatment of thyroid hormone-deficient astrocytes with NH4C1, chloroquine, and monodancylcadaverine completely blocked accumulation of p29 in the dense fractions (Table I). Translocation of p29 from the plasma membrane fractions was blocked completely by NH&1 and partially by monodancylcadaverine and chloroquine.
Role of the Actin Cytoskeleton i n the Intracellular Transport of p29"To examine the cytoskeletal-protein interactions involved in the intraorganelle movement of 5'D-11, we combined density gradient centrifugation with Triton X-100 extraction. In thyroid hormone-deficient astrocytes, >90% of the p29 is are expressed as a percent of the total p29 in the gradient as determined by scanning densitometry; right panels, thyroid hormonedeficient, BbcAMP-stimulated cultured astrocytes were treated with 10 PM cycloheximide plus 10 nM thyroxine for varying times. Sonicates were assayed for 5'D-I1 activity as described under "Experimental Procedures." Results are expressed as the percent of initial activity remaining. All points represent the mean of closely agreeing values in two or more experiments. B, subcellular localization of $9 after thyroxine-stimulated internalization. Fractions 8-16 were obtained after treatment with thyroxine for 60 min and were analyzed for p29 by autoradiography, the @-subunit of the Na+/K+ ATPase by Western analysis and glucose-6-phosphatase and @-glucuronidase by enzyme activity. The quantity of p29 and the @-subunit of the Na+/ K+ ATPase are expressed as integrated optical density, in arbitrary units (au), per fraction. Glucose-6-phosphatase and @-glucuronidase enzyme activity is expressed the optical density at ASO and Aw, respectively, in arbitrary units. The distribution of transferrin was determined in a separate experiment by labeling cultured astrocytes with ['251]transferrin for 2 h at 4 "C, warming the cells to 37 "c for 5 min, then fractionating the cells as described above. Fractions 8-16 were then counted, and the data were expressed as cpm/fraction. Results represent the mean of duplicate values obtained in two experiments.

TABLE I Effect of metabolic inhibitors on the intracellular transport of 5'D-II
Serum-free, BtzcAMP-stimulated astrocytes were affinity labeled then treated for 60 min with 10 nM thyroxine or no hormone in the presence of endocytosis inhibitors and lysosomotropic agents. Homogenate5 were fractionated through Percoll gradients as described under "Experimental Procedures," and the distribution of p29 was determined in 0.5-ml fractions. The results are presented as percent of the total p29 in fractions corresponding to the plasma membrane, endosomes, and dense lysosomes. Each value represents the mean of duplicate values in two experiments.  Leftpanels, thyroid hormone-deficient, BtzcAMP-stimulated cultured astrocytes were affinity-labeled then treated with 10 nM T3 for varying times prior to fractionation through 16 then 8% Percoll gradients as described under "Experimental Procedures." Results are expressed as a percent of the total p29 in the gradient as determined by scanning densitometry. Right panels, thyroid hormone-deficient, BtzcAMPstimulated cultured astrocytes were treated with 10 PM cycloheximide plus 10 nM T, for varying times. Sonicates were assayed for 5'D-I1 activity as described under "Experimental Procedures." Results are expressed as the percent of initial activity remaining. All points represent the mean of closely agreeing values in two or more experiments.
Triton-soluble ( 5 ) . Consistent with our previous observations ( 5 ) , -60% of the p29 is lost from the Triton-soluble pool during the first 20 min of thyroxine treatment (Fig. 4). Since there is no significant change in the total p29 content 20 min after treatment with thyroxine ( 5 ) , this represents a shift of the p29 to the Triton-insoluble F-actin pellet. By 60 min of thyroxine treatment, -90% of the p29 is once again found in the Triton-soluble pool, indicating dissociation of the enzyme-F-actin complex. The reappearance of p29 in the Triton-soluble pool coincides with the shift of the p29 to the endosomal pool. No significant shift of p29 from the Triton-soluble pool at any time point was observed under

TABLE I1
Effect of T3 on 5'0-11 activity and the internalization of p29 Half-life of p29, thyroid hormone-deficient, BtzcAMP-stimulated cultured astrocytes were affinity-labeled then treated with 10 nM TO for varying times prior to fractionation through 16 then 8% Percoll gradients as described under "Experimental Procedures." The halflife of the loss of p29 from the plasma membrane and the accumulation of p29 into the dense lysosomes was determined by analyzing the integrated optical density of p29 in the plasma membrane (fractions 9 and 10) and the dense lysosomes (fractions 14-16), respectively, at 0, 20, 40, and 60 min after the addition of TO. Half-life of 5'0-11, the half-life of 5'D-I1 was determined by treating thyroid hormone-deficient, BtzcAMP-stimulated cultured astrocytes with 10 nM T3 in the presence of 100 PM cycloheximide for 0-60 min and assaying cell sonicates for 5'D-I1 catalytic activity as described under "Experimental Procedures." Results represent the mean of values obtained in two experiments. The binding of p29 to F-actin was examined directly by isolating the p29-containing fractions on Percoll gradients, extracting the isolated fractions with Triton X-100 and then examining the amount of p29 in the insoluble F-actin pellets after centrifugation. As shown in Fig. 5, the quantity of p29 found in the F-actin pellet increases -%fold after 20 min of thyroxine treatment. With longer times, the p29 content in the F-actin pellet decreases back to initial levels, again coinciding with the shift of p29 to an endosomal pool.
Demonstration That p29 Is Internalized as P a r t of a Vesicle in Thyroxine-treated Astrocytes-Whether the p29 subunit of 5'D-I1 is translocated from the plasma membrane to the endosomes as an isolated subunit or as part of a vesicleassociated holoenzyme was then examined. The ability of lysosomotropic agents and endocytosis inhibitors to block the internalization of p29 is consistent with the idea that p29 is internalized in association with a membrane vesicle. As shown in Fig. 6, all of the p29 transferred to the endosomal pool by 60 min after treatment with thyroxine is found in the high speed membrane pellet after centrifugation in the absence of detergent. Addition of 0.5% taurodeoxycholate to the endo-soma1 fraction prior to centrifugation solubilized 65.6% of the p29, indicating that p29 is internalized as part of a membranebound vesicle.

DlSCUSSlON
Transport of proteins, lipids, and organelles within cells is a fundamental cellular process that is poorly understood. The individual intracellular components of the endocytotic pathway have been extensively characterized, but the locomotive force that translocates and targets the internalized vesicles to their eventual location within the cells remains obscure. Recent studies suggest that the cytoskeleton may be a key component in the intracellular transport of a variety of organelles and vesicles. In particular, disruption of the cytoskeletal elements has been shown to disrupt the endocytotic pathway roid hormone-deficient, Ht.$AMI'-stimulated cultured astroc.ytes were affinity-labeled, treated with 10 nM thyroxine for 60 min. collected, and fractionated through 16 then 8 5 I'ercoll gradient3 as described under "Experimental I'rocedures." Fraction 12 was ohtained and aliquots were diluted with sucrnse buffer in the presence and absence of 0.5% taurodeoxycholate and centrifuged for 6 X 10" g-min as described under "Experimental Procedures." Shown is representative radioautogram of radiolabeled proteins in the supernatant in macrophages (13.14). rat Kupffercells (181,rat hepatocyt,es (17),and 3T3 cells (19,20). Most studies on the mechanisms of cytoskeletal-mediated vesicle transport have focused on the role of the microtubules and their associated motor proteins, as the microtubules have been shown to be major determinants of tubular lysosome morphology (13, 14) and are responsible for the internalization of the acetocholine recept.or (36) and the asialo-glycoprotein receptor (17).
Our laboratory has focused on the microfilaments as the key component in the intracellular transport of the shortlived integral membrane protein, type I1 iodothyronine 5'deiodinase. This enzyme is highly regulated by thyroxine, which acts to markedly increase the turnover of 5'D-I1 (3-9).
The actions of thyroxine on the turnover of 5'D-I1 are blocked by chemical disruption of the microfilaments with cytochalasins (5, 6). Subsequent studies have shown that t,hyroxine stimulates actin polymerization (7) and promotes the binding of 5'D-I1 to F-actin, which leads to internalization of an Factin-5'D-I1 complex (5). Disruption of the microtubules with colchicine has no effect on the inactivation/internalization of 5'D-11 in the presence or absence of thyroid hormone. Thus, the hormonally regulated turnover of type I1 iodothyronine 5"deiodinase is an excellent model in which to study the role of the cytoskeleton, specifically the microfilaments, in intracellular protein trafficking. The present study was designed to examine where in the cell 5'D-I1 is transported to and the effect of thyroxine and, hence, the role of the microfilaments, on the route of internalization of 5'D-11. We have shown that 5'D-I1 is internalized via both microfilament-dependent and microfilamentindependent mechanisms and that the targeting of 5'D-I1 to a particular endocytotic pathway is regulated by thyroxine (Fig. 7). In the presence of thyroxine, 5'D-I1 binds to F-actin and the microfilaments translocate the enzyme from the plasma membrane to the endosomes. Following transport to the endosomes, 5'D-I1 is uncoupled from the microfilaments. Interestingly, the enzyme remains in the endosomal pool for extended periods, as further transport of 5'D-I1 to the dense lysosomes was not observed in the time frames examined. 5'D-I1 is also internalized into an endosomal pool via a microfilament-independent mechanism in the absence of thyroxine. However, in contrast to the thyroxine-treated pathway, the enzyme is rapidly transferred from the endosomes to the dense lysosomes, preventing any accumulation of 5'D-I1 in the endosomes. Whether the endosomal population 5'D-I1 that is cycled through in the absence of thyroxine is the same endosomal pool that 5'D-I1 is transported to by the microfilaments in the presence of thyroxine is uncertain.
Previously, functions attributed to the actin cytoskeleton have been limited to the regional distribution of proteins and surface receptors within the plasma membrane (36-40) and global changes in plasma membrane organization following acid stimulation in gastric parietal cells (41). However, disruption of the microfilaments has also been shown to block transport of asialo-glycoproteins (17) and labeled phagosomes (18) from the endosomes to the lysosomes. The present study now demonstrates a primary role of the microfilaments in the endocytosis of the short-lived membrane protein, 5'D-11. In addition, since thyroxine regulates the presence of the F-actin stress fibers by modulating actin polymerization (7), these data demonstrate that the stress fibers are responsible for the hormonally regulated endocytosis of 5'D-11.
While we have shown that thyroxine stimulates actin polymerization and targets 5'D-I1 to a microfilament-dependent endocytotic pathway, the identity of the targeting signal regulated by thyroxine remains unclear. Because the actions of thyroxine on the microfilaments that result in differential pathways of endocytosis of 5'D-I1 occur independently of interactions with the nuclear thyroid hormone receptor (6, 7), other sites of action must be considered. The ability of thyroxine to promote actin polymerization alone is insufficient to explain the activation of the microfilament-mediated pathway of endocytosis since restoration of the stress fibers in thyroid hormone-deficient astrocytes with retinoic acid does not increase inactivation/internalization of 5'D-I1 (42). Likewise, deiodination of thyroxine is not the targeting signal, as covalent modification of the active site of the substratebinding subunit of 5'D-I1 with the affinity label, BrAcT4, fails to alter the thyroxine-dependent internalization of the enzyme.
Although high affinity binding sites for thyroid hormone have been shown to be present in the erythrocyte plasma membrane (43), cell surface receptors for thyroxine have not been described in astrocytes. However, the demonstration that 5'D-I1 is internalized as part of a vesicle raises the possibility that thyroxine may be interacting with other membrane proteins that act as chaperons to bind the enzyme to F-actin during the microfilament-mediated internalization.
These as yet unidentified proteins may play a key role in intracellular targeting. Implicit in this hypothesis is that potential chaperon proteins are constitutively expressed, as the effects of thyroxine on endocytosis of 5'D-I1 occurs in the absence of new proteins synthesis.
The ultimate fate of the 5'D-I1 transported to the endosomes by the microfilaments in the presence of thyroxine is unknown. We were unable to detect further transport of the endosomal5'D-I1 to other cellular compartments in the time frames examined. Preliminary studies have failed to show evidence of recycling of 5'D-I1 back to the plasma membrane (44). Further studies are necessary to determine the metabolic fate of this enzyme.
In summary, we have shown that several endocytotic pathways exist within astrocytes and that targeting to a specific pathway is hormonally regulated. Thyroxine targets 5'D-I1 to the endosomes by activating enzyme-F-actin interactions important in microfilament-dependent intracellular protein trafficking.