Type 2 iodothyronine deiodinase is the major source of plasma T₃ in euthyroid humans

Abstract

The relative roles of the types 1 and 2 iodothyronine deiodinases (D1 and D2) in extrathyroidal 3,5,3′-triiodothyronine (T₃) production in humans are unknown. We calculated the rate of thyroxine (T₄) to T₃ conversion by intact cells transiently expressing D1 or D2 at low (2 pM), normal (20 pM), and high (200 pM) free T₄ concentrations. Deiodinase activities were then assayed in cell sonicates. The ratio of T₃ production in cell sonicates (catalytic efficiency) was multiplied by the tissue activities reported in human liver (D1) and skeletal muscle (D2). From these calculations, we predict that in euthyroid humans, D2-generated T₃ is 29 nmol/d, while that of D1-generated T₃ is 15 nmol/d, from these major deiodinase-expressing tissues. The total estimated extrathyroidal T₃ production, 44 nmol/d, is in close agreement with the 40 nmol T₃/d based on previous kinetic studies. D2-generated T₃ production accounts for approximately 71% of the peripheral T₃ production in hypothyroidism, but D1 for approximately 67% in thyrotoxic patients. We also show that the intracellular D2-generated T₃ has a greater effect on T₃-dependent gene transcription than that from D1, which indicates that generation of nuclear T₃ is an intrinsic property of the D2 protein. We suggest that impairment of D2-generated T₃ is the major cause of the reduced T₃ production in the euthyroid sick syndrome.

Introduction

The monodeiodination of thyroxine (T₄) to 3,5,3′-triiodothyronine (T₃) activates the major secretory product of the iodine-sufficient human thyroid gland, producing approximately 80% of the circulating T₃ in humans. The types 1 and 2 iodothyronine deiodinases (D1 and D2) are members of a family of oxidoreductases that catalyze this reaction (1). These integral membrane proteins contain the rare amino acid selenocysteine (encoded by a UGA) in their active site. Both D1 and D2 require an as-yet-unidentified cofactor for the reaction. Several decades ago, D1 was identified in the liver and the kidney of rats and humans, and it is often assumed that this enzyme is the source of most of the plasma T₃ in humans. The more recent discovery of D2 mRNA and activity in human skeletal muscle suggests that D2 could also be a significant source for plasma T₃ production in humans, but this has not been quantitatively defined (1).

The biochemical and molecular properties of D2 seem ideal for extrathyroidal T₃ production. Its activity is tightly controlled by the concentration of its preferred substrate, T₄, since catalysis accelerates the ubiquitination of this enzyme, inactivating it and accelerating its degradation in proteasomes (2). The half-life of D2 in normal cells is 20–30 minutes in the presence of T₄. The transcription of the type 2 deiodinase (DIO2) gene is also negatively regulated by T₃ (3). In contrast, the D1 protein has a long half-life (>12 hours), and the transcription of the human type 1 deiodinase (DIO1) gene is markedly stimulated by T₃, just the opposite of what would be expected in a typical feedback loop (4, 5).

Hepatic T₃ production via D1 in the 1.5-kg human liver has been estimated to be only approximately 8 nmol/day, much less than the total daily extrathyroidal T₃ production of approximately 40 nmol (6). If accurate, this would predict the likelihood of a significant contribution by D2. It has not been possible to perform similar calculations for T₃ production by skeletal muscle D2 for 2 reasons. First, there was a paucity of data, which was only recently remedied with a report indicating that mean D2 activity is approximately 1 fmol T₄/mg protein/min in this tissue (7). Second, the catalytic efficiency of D2-catalyzed T₃ production is not known at physiological cofactor and free T₄ (FT₄) concentrations. The V_max/K_m ratio of the ping-pong reaction catalyzed by D1 is independent of the concentration of cofactor, but this is not the case for the sequential kinetics of D2-catalyzed T₄ 5′-deiodination. In fact, there also is no formal estimate of how efficient D1-catalyzed T₃ production is at the euthyroid FT₄ concentration of 20 pM.

A second important issue is to determine why in D2-expressing cells a significant portion of the T₃ produced enters the cell nucleus, where it binds to specific high-affinity thyroid hormone receptors (TRs), whereas most T₃ generated from the D1-catalyzed reaction remains in the cytoplasm or exits the cell (8). The first recognition of D2 activity was due to its catalysis of pituitary T₄ to T₃ conversion by a 6-n-propylthiouracil–resistant (PTU-resistant) process (9). This agent is a potent inhibitor of D1. The fact that the D2 is located in the ER, while D1 is in the plasma membrane, suggests an explanation, although the catalytic centers of both enzymes are in the cytosol (10).

To accurately define the relative roles of D1 versus D2 in extrathyroidal T₃ production in humans, the deiodination process must be analyzed at physiological cofactor and FT₄ concentrations. Since we do not know the identity of this cofactor, much less its concentration, we designed an in vitro model in which human D1 or D2 is transiently expressed in the same cell type and compared their catalytic efficiencies at FT₄ concentrations spanning the hypothyroid to thyrotoxic range. We also developed a method to quantify the transcriptional potency of the intracellular T₃ produced by D1 versus D2. Our results predict that D2-catalyzed T₃ production is the major source of plasma T₃ in euthyroid humans, while D1 is the major contributor when FT₄ is present at thyrotoxic concentrations. Furthermore, at equal rates of production, the T₃-generated by D2-catalyzed T₄ monodeiodination is 2- to 3-fold more effective in activating T₃-dependent gene transcription than that generated by D1, which indicates that this property is intrinsic to the D2 protein.

Results

Quantifying T₃ production in cells transiently expressing D1 or D2.

The production of ¹²⁵I^– from outer ring–labeled T₄ in intact cells can be analyzed by measuring the level of either [¹²⁵I]T₃ or ¹²⁵I^– in the medium, since equimolar amounts of both products are generated. The D1 or D2 activities were measured in cell sonicates harvested at the completion of the incubation. In human embryonic kidney epithelial (HEK 293) cells incubated in 0.1% BSA medium, approximately 50% of the labeled T₃ remains in the medium, and 50% is found in the cell pellet. The T₄ distribution differs, with a medium-to–cell pellet ratio of 9:1 (Figure 1A). Therefore, we based the calculation of T₄ to T₃ conversion on the rate of ¹²⁵I^– release. Cells were transfected with 100 ng of D1- or D2-expressing plasmid and incubated for 22 hours. The rate of T₄ deiodination by D1- or D2-catalyzed 5′-deiodination in intact cells was linear over the 22 hours, and less than 5% of the labeled T₄ appeared as iodide in untransfected cells (Figure 1B). Enzyme activities increased progressively, though not linearly, with increasing amounts of plasmid transfected in whole cells and cell sonicates (Figure 1, C–E). The rates of fractional T₄ deiodination by D1 and D2 in whole cells were quite similar (Figure 1D). The activities of both enzymes increased over the first 24 hours but were stable over the next 24 hours (Figure 1F).

Figure 1.

Model for quantification of T₃ production in cells transiently expressing D1 or D2. (A) Distribution of [¹²⁵I]T₃, [¹²⁵I]T₄, and I^– in HEK 293 cells incubated in 0.1% BSA medium for 22 hours. (B–H) Cells were transfected with 3–100 ng of D1- or D2-expressing plasmids and incubated for 22 hours in 1 ml 0.1% BSA DMEM, to which were added the indicated concentrations of T₄, including approximately 100,000 cpm/ml of [¹²⁵I]T₄. Whole-cell T₄ to T₃ conversion rates were determined from the fraction of [¹²⁵I]T₄ appearing as ¹²⁵I^–. The Vmax for each deiodinase was determined in cell sonicates, harvested after completion of the incubation. (B) Fractional T₄ deiodination over time. (C) Correlation between the quantity of plasmid transfected and fractional deiodination. (D) Correlation between Vmax and fractional deiodination. (E) D1 and D2 Vmax versus the quantity of plasmid transfected. (F) Time course of transient deiodinase expression over 72 hours. Effect of PTU (G) or T₄ (H) on fractional T₄ deiodination in D1- and D2-expressing cells. All data are expressed as the mean of duplicate samples in 3 independent experiments.

The addition of PTU (1–100 μM) caused a dose-dependent decrease in D1 activity but did not affect that of D2 (Figure 1G). On the other hand, an increase in FT₄ from 2 to 200 pM reduced the rate of fractional deiodination by D2 but did not affect that by D1 (Figure 1H). These differences are consistent with the PTU sensitivity of D1- but not D2-catalyzed T₄ deiodination and the posttranslational substrate-induced inactivation of D2 but not D1. We wished to express an amount of deiodinase that would not consume more than 30% of the T₄ over 24 hours but was as close as possible to the level in the most important tissues. This was achieved with 10–20 ng of D1- or D2-expressing plasmids.

Analyses of the deiodination products of cells expressing D1 and D2.

We estimated T₄ to T₃ conversion from the rate of appearance of enzymatically generated ¹²⁵I^–. Since T₄ is labeled only in the outer ring, the labeled T₃ and I^– produced have one half the specific activity of the parent compound (Figure 2A). Since D2 catalyzes only outer-ring deiodination of T₄, the T₃ production by D2 is equal to the labeled ¹²⁵I^– production multiplied by 2 to correct for the specific activity reduction. For D1, the calculation is more complex, since the V_max/K_m ratios for 5′- and 5-deiodination of T₄ are approximately equal (11). Therefore, 1 mol of T₃ and of 3,3′,5′-triiodothyronine (reverse T₃ [rT₃]) will be generated for every 2 molecules of T₄ deiodinated. Because of the much higher V_max/K_m ratios for D1-catalyzed 5′-monodeiodination of rT₃ and its products 3,3′-diiodothyronine (T₂) and 3′-monoiodothyronine (T1) than for T₄, the rT₃ is rapidly deiodinated, which leads to the release of all the ¹²⁵I^– (Figure 2A). This being the case, the [¹²⁵I]T₃ production should be equal to one-third of the ¹²⁵I^– generated multiplied by 2. We confirmed this experimentally by exposing D1-expressing cells to approximately 2 × 10⁶ cpm [¹²⁵I]T₄ for 24 hours. The ratio of ¹²⁵I^– (corrected for nonspecific ¹²⁵I^– released) to T₃ production was 3.3 ± 0.6 in 3 separate experiments (see D1 in Figure 2B). Note the absence of rT₃, T₂, or T1 in the chromatogram. A further consideration in D1-expressing cells is the slow rate of inner-ring deiodination of T₃, which would generate ¹²⁵I^– from the subsequent 5′-monodeiodination of the T₂ product (Figure 2A). Incubation of outer ring–labeled T₃ with D1-expressing cells for 24 hours caused the appearance of 15–20% of the T₃ label as ¹²⁵I^–, which indicates this is a detectable but inefficient deiodination pathway. This accounts for the slightly higher I^–/T₃ ratio observed after incubation of [¹²⁵I]/T₄ with D1 than the expected value of 3.0 (Figure 2B). Note that there is no deiodination of T₃ by D2.

Figure 2.

Chromatographic analysis of the products of D1- and D2-catalyzed deiodination. (A) Pathway of T₄ deiodination by D1. (B and C) HEK 293 cells transfected with 100 ng of D1 or D2 were homogenized in the medium after 24 hours incubation with labeled T₄ or T₃ and the deiodination products analyzed by HPLC. (B) Chromatographic patterns of the products of D1- or D2-catalyzed [¹²⁵I]T₄ deiodination. (C) Patterns of the products of outer ring–labeled [¹²⁵I]T₃ deiodination by D1- or D2-expressing cells.

Assessing the efficiencies of T₃ production by D1 and D2 at varying free T₄ concentrations.

To replicate physiological conditions precisely, it would be ideal to quantitate T₃ production when cellular D1 and D2 activities were comparable to those in deiodinase-expressing tissues. For liver, the D1 Vmax(T₄) is approximately 8 pmol/mg protein/min, and for skeletal muscle, the D2 Vmax is approximately 1 fmol/mg protein/min (6, 7). While we were able to achieve this level for D1, the T₃ production/24 hours in cells expressing D2 at Vmax values of 1–2 fmol/mg protein/min was too low to measure accurately. This is understandable, since the mass of D2-expressing skeletal muscle is in vast excess of that of the D1-expressing liver and kidneys. Accordingly, we expressed D2 at a Vmax of approximately 10 fmol/mg protein/min. To evaluate the effects of thyroid status on the efficiency of D1- and D2-catalyzed T₃ production, we incubated cells with FT₄ concentrations of 2 pM (typical of severe hypothyroidism), 20 pM (euthyroid state), and 200 pM (severe thyrotoxicosis). The fractional T₄ to T₃ conversion rates for D1 were unaffected by the FT₄ concentration, while those of D2-expressing cells decreased about 4-fold as FT₄ increased (Figure 3A). This can be explained by the observed decrease in D2 activity (in cell sonicates) produced by the higher ambient T₄ concentration due to catalysis-induced, posttranslational downregulation of D2. In contrast, D1 activity was unchanged during the incubation (Figure 3B). As a consequence, the quantity of T₃ produced by D2 decreased from about twice that produced by D1 at 2 pM FT₄ to about one-quarter that produced by D1 at 200 pM FT₄. It is remarkable that despite the approximately 700-fold higher Vmax estimates for D1 than for D2, the T₃ produced via these 2 pathways at 20 pM FT₄ was slightly higher with D2 (Figure 3C). The catalytic efficiency, the ratio of T₃ production in intact cells to the Vmax for T₄ to T₃ conversion in sonicates of the same cells, was higher for D2 than D1 at all FT₄ concentrations (Figure 3D).

Figure 3.

Human D1- and D2-catalyzed T₄ to T₃ conversion at different FT₄ concentrations. (A) Fractional T₄ to T₃ conversion in whole HEK 293 cells transiently expressing D1 or D2. (B) Vmax estimates for D1 and D2 in cell sonicates. (C) T₃ production in whole cells expressing D1 or D2. (D) Ratios of the whole-cell to cell sonicate T₃ production, or catalytic efficiency, of D1 and D2 enzymes.

Deiodination rates in cells expressing endogenous D1 and D2.

To confirm that endogenous D1 or D2 under control of homologous promoters would show the same characteristics as those found in HEK 293 cells expressing D1 or D2 driven by a heterologous promoter, we evaluated the patterns of deiodination in the porcine kidney tubule LLCPK1 cells for D1 and mesothelioma (MSTO) cells (12) for D2, respectively. Similar to results in HEK 293 cells, the T₃ production rate in intact LLCPK1 cells increased proportionately as FT₄ concentration increased (Figure 4, A and B), with no changes in D1 Vmax (Figure 4C). In other experiments (data not shown), we found no response to T₃ of the T₃-sensitive TRE3TKLUC reporter in LLCPK1 cells, which indicates the absence of functional TR in these cells. Therefore we cannot replicate in this system the increase in DIO1 gene transcription anticipated to result from T₄ to T₃ conversion. The ratio of T₃ production in intact cells to cell sonicate activity (catalytic efficiency) was 2- to 3-fold higher for porcine D1 in LLCPK1 cells than for the human D1 transfected in HEK 293 cells (Figure 4D).

Figure 4.

Whole-cell T₃ production in LLCPK1 cells expressing endogenous D1 at different FT₄ concentrations. (A) Fractional T₄ to T₃ conversion. (B) Whole-cell T₃ production. (C) Vmax estimates for T₃ production in cell sonicates. (D) Catalytic efficiency of T₄ to T₃ conversion by porcine D1 compared with transiently expressed human D1.

MSTO cells respond to changes in ambient FT₄ concentration in a manner similar to D2-expressing HEK 293 cells. The fractional conversion of T₄ to T₃ decreased progressively with increasing FT₄ concentrations, but T₃ production in whole cells increased until the FT₄ concentration reached 100 pM (Figure 5, A and B). We also examined the time course of T₄-induced loss of activity to see how rapidly this occurred. There were decreases in D2 activity over the first 13 hours, after which it reached a plateau (Figure 5D). Finally, since it is not known whether MSTO cells can respond to the T₃ produced by D2, we measured D2 mRNA expression after incubation to determine whether it had decreased. There was no effect of increasing FT₄ concentrations on the D2 mRNA, which indicated that the loss of D2 activity is completely posttranslational (Figure 5E). The catalytic efficiency of endogenous D2 in MSTO cells was similar to that in D2-expressing HEK 293 cells, except at the highest FT₄ concentration (Figure 5F).

Figure 5.

Whole-cell T₃ production in MSTO cells expressing endogenous D2 at different FT₄ concentrations. (A) The fractional T₄ to T₃ conversion. (B) Whole-cell T₃ production. (C) Vmax estimates for T₃ production in cell sonicates. (D) Time course of T₄-induced loss of D2 activity. (E) Effects of the FT₄ concentrations on D2 mRNA levels. (F) Catalytic efficiency of T₄ to T₃ conversion in MSTO cells compared with HEK 293 cells transiently expressing human D2.

D2-generated T₃ in human skeletal muscle.

β-Adrenergic–responsive D2 activity is present in human skeletal muscle cells (13). We used primary cultures of human skeletal muscle myoblasts (HSMMs) to examine whether these cells could convert medium [¹²⁵I]T₄ to [¹²⁵I]T₃. HSMMs were incubated for 22 hours with [¹²⁵I]T₄ in medium containing a rabbit anti-T₃ Ab (T₃-Ab) to block reuptake of [¹²⁵I]T₃ released (see Methods). This results in efficient binding of [¹²⁵I]T₃ plus weak binding of [¹²⁵I]T₄. HPLC was used to determine the T₃/T₄ ratios of the Ab-bound iodothyronines. In the absence of cells, after correction for recovery, this ratio was 0.016, and the ratio increased to 0.060 in the presence of HSMMs (Figure 6, A and B; see also Methods for calculations). A further increase to 0.090 was observed when cells were exposed to 0.5 mM dibutyryl cAMP (BT₂cAMP) to increase the level of D2 (Figure 6C) (13). With the 2-fold correction for loss of specific activity, this represents a 6- and 10-fold increase in the ratio of extracellular T₃ to T₄ produced by D2 in HSMMs over 22 hours, respectively.

Figure 6.

Chromatographic analysis of T₄ to T₃ conversion in HSMMs. Control medium incubated with no cells (A) or HSMMs (B and C) incubated in duplicate 60-mm dishes for 22 hours in 2 ml 0.1% BSA DMEM with approximately 1 μCi/ml [¹²⁵I]T₄ plus specific T₃-Ab (1:1,000) in the absence (B) or presence (C) of dibutyryl cAMP (0.5 mM).

To determine whether D2 is expressed at equal levels in different skeletal muscles, we obtained fresh samples of human sternocleidomastoid (n = 7), rectus abdominis (n = 20), and vastus lateralis (n = 3) muscle during surgery and assayed them for D2 activity at 0.15 nM T₄. No differences in D2 activity were found among these (0.021 ± 0.001 vs. 0.024 ± 0.001 vs. 0.023 ± 0.001 fmol/mg protein/min, respectively; P = 0.79).

The T₃ derived from D2-catalyzed T₄ to T₃ conversion is more potent than that from D1 in stimulating T₃-dependent gene expression.

D1 is expressed in plasma membrane and D2 in the ER. If this is the explanation for the greater contribution of D2- than D1-generated T₃ to TR-bound T₃, it should be possible to demonstrate a greater effect of the intracellular T₃ generated by D2 than by D1 on T₃-dependent transcription. To evaluate this, we transfected HEK 293 cells with D1- or D2-expressing plasmids, a TRα-expressing plasmid, and the T₃-responsive TRE3TKLUC construct. This was a highly T₃-responsive system (Figure 7A). Since we wished to monitor the effects of intracellular T₃ production, we prevented cellular reentry of the T₃ produced from T₄ without interfering with T₄ uptake by addition of T₃-Ab to the medium. In the presence of a 1;1,000 dilution of rabbit T₃-Ab, more then 98% of the labeled T₃ remained in the medium, whereas there was no effect on T₄ uptake (Figure 7B). The effectiveness of the T₃-Ab in preventing cellular T₃ uptake was confirmed by its blockade of the transcriptional response of TRE3TKLUC to medium T₃ in D1- or D2-expressing cells (Figure 7C).

Figure 7.

A T₃-responsive system for evaluating the contribution of D1- and D2-generated T₃ to nuclear T₃. (A–C) HEK 293 cells were transfected with D1- or D2-expressing plasmids, a TRα-expressing plasmid, and the T₃-responsive TRE3TKLUC construct (see Methods). (A) The TRE3TKLUC reporter responded to T₃ in a dose-dependent manner. (B) The addition of specific T₃-Ab prevents cellular uptake of T₃ but not T₄. (C) T₃-Ab blocks the transcriptional response to 20 pM T₃.

As a control for the potential T₄ induction of TRE3TKLUC, which could occur at high FT₄ concentrations, the same concentrations of FT₄ were added to cells transfected with a plasmid expressing an inactive D2 construct (see Methods). In the first experiment, performed after transfection of 100 ng D1 or D2 plasmids, luciferase (LUC) induction in cells expressing D1 was negligible at 2 pM, whereas a 3.6- and 7.3-fold increase occurred at 20 and 200 pM FT₄, respectively (Figure 8A). The mean LUC induction in D2-expressing cells was higher at all FT₄ concentrations — 2.7-, 7.5-, and 13.1-fold, respectively — but so was the T₃ production at comparable FT₄ concentrations (Figure 8, A and B). Nonetheless, note that 111 fmol of T₃ produced by D1 triggered a LUC induction similar to that observed with 37 fmol of T₃ generated by D2 (Figure 8A), which suggests that the D2-catalyzed T₃ production had an approximately 3-fold greater effect on gene transcription than that catalyzed by D1 (Figure 8C). We confirmed that this was indeed the case by increasing the level of transfected D1 4-fold, thereby increasing T₃ production by D1 to amounts comparable to those produced by D2 at the same FT₄ concentrations. Despite the higher T₃ production by D1, the TRE3TKLUC induction was the same with D2 (Figure 8B). The ratio of fold induction of TRE3TKLUC to T₃ production by D2 was again shifted to the left of that for D1 (Figure 8C). In a third experiment, we compared the effects of T₃ generated from 20 pM FT₄ by D1 or D2 with those of T₃ added directly to the medium, the latter in the absence of T₃-Ab (Figure 8D). While T₃ production by D2 was only twice that by D1, the induction of TRE3TKLUC was equivalent to that with a 5-fold higher quantity of T₃. Thus the transcriptional stimulation by D2-generated T₃ is consistently 2- to 3-fold greater than that from D1-generated T₃.

Figure 8.

The T₃ derived from D2-catalyzed T₄ to T₃ conversion is more potent than that from D1 in stimulating T₃-dependent gene expression. (A and B) TRE3TKLUC induction in D1- and D2-expressing HEK 293 cells. At any rate of T₃ production (see tables below graphs), D2-catalyzed T₃ production causes greater increases in T₃-dependent luciferase (LUC) expression than T₃ derived from the D1-catalyzed T₄ to T₃ reaction. (C) TRE3TKLUC induction versus T₃ production (log₁₀). D2-derived T₃ caused greater LUC induction than did D1-derived T₃. LUC expression was normalized to growth hormone as transfection efficiency control. (D) Effects of added T₃ and D1- or D2-derived T₃ on TRE3TKLUC induction. Cells transfected with inactive D2 were incubated with 1 ml 0.1% BSA DMEM, and the indicated amounts of T₃ were added. D1- and D2-expressing cells were incubated in the same medium containing T₃-Ab (1:1,000) and 20 pM FT₄.

Discussion

There are no prior studies addressing the relative roles of the 2 outer-ring human iodothyronine deiodinases in the peripheral conversion of T₄ to T₃. There are several reasons for this, including the fact that the putative thiol cofactor necessary for both D1- and D2-catalyzed reactions has not yet been identified. In addition, the assay conditions require the use of in vivo FT₄ of 20 pM, and the rate would be so low as to be difficult to quantitate. To address this issue, we modified a system used by us and others in which D1 or D2 are transiently expressed in HEK 293 cells at near physiological levels and exposed to a range of FT₄ concentrations that occur clinically (14, 15). This allows stable reactions over 18–24 hours at physiological substrate and cofactor concentrations (Figure 1).

The system developed in these experiments allowed us to compare the catalytic efficiencies of these enzymes under conditions that obtain in vivo. The characteristic properties of the 2 deiodinases, PTU sensitivity (D1) or resistance (D2) and posttranslational downregulation by T₄ (D2), were preserved (Figure 1). Another advantage to this system is that after the incubations, deiodinase activities can be quantitated in the same cells under conditions used to assay deiodinase activity in human tissues (Figure 3). This facilitates extrapolation of the results to the in vivo state. The system was further validated by the fact that it yielded results similar to those for cells expressing endogenous D1 or D2 (Figures 4 and 5).

The D1 activity in the cell sonicates in these transient expression studies was approximately 7 pmol/mg protein/min, virtually identical to that in human liver (6). To obtain reliable reaction rates, we studied D2-mediated T₄ to T₃ conversion at D2 activities approximately 10 fmol/mg protein/min (Figure 3B), about 10-fold that in human skeletal muscle (7). The striking results were that at a euthyroid FT₄ concentration of 20 pM, the D2-catalyzed T₃ production rate was slightly greater than that catalyzed by D1 despite the fact that the Vmax in cell sonicates was estimated to be 700-fold higher for D1 than for D2 (Figure 3, B and C). This is explained by the roughly 700-fold higher catalytic efficiency of D2 (Figure 3D).

What factors contribute to the remarkable difference in the catalytic efficiency of these 2 enzymes? First, each molecule of T₄ deiodinated via D2 produces 1 of T₃. Because D1 catalyzes the inner- and outer-ring deiodination of T₄ equally well, as predicted from Vmax/Km estimates, for each 2 moles of T₄ deiodinated by D1, only 1 mole of T₃ and a second of rT₃ is produced (Figure 2A) (6). However, this can only explain a catalytic efficiency difference of 2. The rest of the difference is primarily due to the much slower rate of D1-catalyzed T₄ to T₃ conversion under physiological conditions than under typical assay conditions. The assay conditions for D1 amplify the apparent Vmax to a much greater extent than those used for D2. This can be explained in part by differences in the kinetics of the enzyme: D1-catalyzed deiodination is a ping-pong reaction with the thiol cofactor acting as the second substrate, while the kinetics of D2 are sequential, with the cofactor thought to bind at the same time as the substrate. Relatively massive concentrations of the DTT cofactor, in our case 20 mM, are used for D1 and D2 assays. The catalytic activity of D1 decreases approximately 1,000-fold when the cofactor is reduced from 20 mM DTT to the more “physiological” 5 mM reduced glutathione (16). The Vmax is also reduced for D2 at lower cofactor concentrations, but by a much lower fraction (data not shown). Nonetheless, even considering these differences, D2 appears to be a more efficient enzyme than D1. In fact, D2 is so efficient that it is nearly impossible to quantitate the D2 protein by Western blotting, and only prolonged exposure of cell sonicates from ⁷⁵Se-labeled cells allows recognition of its presence (12). On the other hand, Western blotting or bromoacetyl [¹²⁵I]iodothyronine labeling allows ready quantitation of D1 (17, 18). Thus, we cannot establish as yet how much higher the turnover rate is for D2 than that reported for D1. The estimated Km(T₄) for D2, which is approximately 1,000-fold lower than that for D1, could also contribute to the higher catalytic efficiency of T₄ to T₃ conversion by D2 versus D1 at normal FT₄ concentrations.

These arguments pertain at euthyroid or hypothyroid FT₄ concentrations. As mentioned, D2 activity, but not that of D1, falls as FT₄ concentration increases due to substrate-induced ubiquitination (2). This is the explanation for the fact that at 200 pM FT₄, the D2 activity at the end of the incubation was about 10% of that found at 2 pM FT₄ (Figure 3B). For this reason, D1 would be predicted to be more important in T₄ to T₃ conversion in the thyrotoxic patient. This concept is supported by the comparisons of the effect of thyroid status on the sensitivity of T₄ to T₃ conversion in humans to inhibition by the D1-specific inhibitor PTU. We showed several decades ago that in hyperthyroid patients with Graves disease, there was a dose-dependent decrease of up to 50% in T₄ to T₃ conversion by administering up to 1,200 mg/d of PTU (19). On the other hand, 2 other studies showed that when even higher doses of PTU were used, a maximum decrease of 25% in serum T₃ occurred in athyreotic T₄-replaced, euthyroid individuals (20, 21). This is in excellent agreement with the predictions based on the current studies (Figure 3D). In addition to the posttranslational reduction of D2 by high FT₄ concentrations shown in this study, it has also been shown that transcription of the DIO2 gene is reduced by about 50% by a high level of T₃, while the human DIO1 gene is transcriptionally induced by T₃ (3, 5). In the present studies, the transcriptional effects of thyroid hormone on D1 and D2 expression were not seen either in the transiently expressing cells or in cells expressing endogenous deiodinase, since these do not express functional TR (Figures 4C and 5E.).

Another correlation between the predictions of these experiments and in vivo data with respect to the effective thyroid status comes from early work by Inada et al. (22). These studies showed that in hypothyroid patients, the fractional whole body conversion rate of T₄ to T₃ was 42%, while in the same patients made euthyroid with levothyroxine, this rate fell to 21%. We saw the same decrease in fractional T₄ to T₃ conversion by D2 as the medium FT₄ concentration was increased from 2 to 20 pM (Figure 3A). Lum et al. (23) also observed sustained elevations in serum T₃/T₄ ratios in athyreotic subjects partially withdrawn from T₄ therapy. Since T₄ concentrations are so low, this does not apply to total T₃ production, which was markedly reduced at 2 pM FT₄ (Figure 3C). The increased efficiency of T₄ to T₃ conversion in the hypothyroid state is expected from an increase in D2 activity due to the longer half-life of the D2 protein, increased DIO2 gene transcription, and reduced DIO1 gene transcription (2, 3, 5).

To test the applicability of these catalytic efficiency estimates to the in vivo state, we extrapolated these data to predict the quantities of T₃ from T₄ produced by D1 and D2 in humans in various thyroid states. This can be done if the ratios of intact cell to cell sonicate T₃ production are applied to deiodinase estimates in human tissues. D1 activity in human liver is approximately 8 pmol/mg protein/min, similar to that achieved in this study after correction for differences in assay conditions (5 vs. 20 mM DTT and microsomal vs. total cellular protein) (6, 24). The D2 activity estimate of approximately 1 fmol/mg protein/min is from a recent study in which D2 activity was measured in the sternocleidomastoid or the rectus abdominis muscle in euthyroid subjects (7). The present studies are the first to our knowledge to demonstrate that human skeletal muscle cells produce extracellular T₃ from T₄ (Figure 6). Other assumptions regarding the mass of human liver and skeletal muscle and the protein content/wet weight of these tissues are shown in Table 1. The conversion efficiencies for D1 and D2 are as reported in Figure 3.

Table 1.

Predicted rates of peripheral human T₃ production by D1 and D2 based on experimentally determined catalytic efficiency at different ambient FT₄ concentrations

In a 70-kg euthyroid human, the hepatic D1-catalyzed T₄ to T₃ conversion is predicted to produce approximately 15 nmol of T₃/d and that from skeletal muscle D2 approximately 29 nmol/d (Table 1). The total, 44 nmol/d, is quite close to the predicted extrathyroidal T₃ production rate of 40 nmol/d according to many in vivo kinetic studies (25). In the euthyroid state, two-thirds of this is predicted to be derived from D2-catalyzed T₄ monodeiodination. In the hypothyroid state, D2 is predicted to account for at least 70% of the extrathyroidal T₄ to T₃ conversion, but it would be the source of a much lower fraction, approximately 30%, in the thyrotoxic state (Table 1). It should be noted that the contribution of D1 is overestimated in the hypothyroid state and underestimated in the hyperthyroid state since the catalytic efficiency figures do not take into account the transcriptional effects of thyroid status, namely, reduction and enhancement of DIO2 and DIO1 gene expression, respectively.

The following assumptions were made. We ignored D1 in kidney and the D2 in brain and skin and assumed that cellular T₄ uptake in all tissues is the same as it is in HEK 293 cells. We also assumed that the catalytic efficiencies of D1 and D2 in vivo are the same as those in tissue culture and that the effect of the cofactor present in HEK 293 cells is not significantly different from that supporting D1- and D2-mediated catalysis in liver and skeletal muscle. This assumption is supported by the comparisons of the results in transfected cells with those in cells expressing endogenous D1 and D2 (Figures 4 and 5). The D2 activity at the start of the incubation was higher than that at the end of the experiment due to posttranslational downregulation (Figure 3B). It appears that this reaches equilibrium value by about 12 hours of incubation (Figures 3C and 4D). Thus, T₃ production would be somewhat higher in the first than in the second half of the incubation. Our model indicates that the estimated production by D2 was only about 10–15% higher than it would have been if the terminal levels of D2 had been present throughout the incubation period. Last, the in vivo concentration of FT₄ is maintained at a constant level by the hypothalamic pituitary thyroid axis. In our experiments, the fraction of total T₄ converted to T₃ in the system was less than 30% (except in the LLCPK1 experiments). Thus, depletion of substrate does not significantly reduce the T₃ production in this system (Figures 3A, 4A, and 5A).

The other major difference between D1- and D2-catalyzed T₃ production that we addressed in these studies is the explanation for the unique characteristic of D2-expressing tissues, namely, that D2-catalyzed T₄ to T₃ contributes a substantial fraction to the TR-bound T₃ (8). We have postulated that the location of D2 in the ER, as opposed to that of D1 in the inner surface of the plasma membrane, could improve the access of T₃ produced from T₄ to the nucleus (10). However, other explanations are also possible; for example, the rate of T₄ to T₃ conversion by D2 may be more rapid than that by D1. Yet, another possibility would be that in D2-expressing cells, there is a specific mechanism translocating D2-generated T₃ from the perinuclear ER into the nucleus. Specific cytosol-to-nuclear plasma T₃ transport mechanisms have been demonstrated (26).

We were able to modify the system to study transcriptional effects of the T₃ produced by the D1 or D2 pathway and at the same time estimate the rate of total T₃ production so that this could be taken into account (Figure 8). This required trapping the T₃ produced during the overnight incubation in the medium so that it could not induce transcription. These studies showed that when the induction of T₃-dependent gene expression is plotted as a function of T₃ production rates, D2-generated T₃ produces a 2- to 3-fold greater effect on the transcription of a T₃-dependent gene than does D1-generated T₃ (Figure 8C). Since these experiments are performed in the same cells, this difference cannot be explained by any cell type–specific property or by a faster rate of T₃ production by D2. They indicate that the effect of D2 in providing more nuclear T₃ than D1 very likely reflects its subcellular location in the ER as opposed to the plasma membrane.

In conclusion, our results argue that D2 is the major source of human extrathyroid-produced T₃ in the euthyroid state. It contributes a much lower fraction of the T₃ in the thyrotoxic patient, accounting for the fact that PTU inhibits T₄ to T₃ conversion to a greater extent in thyrotoxic than in euthyroid individuals. A major source of T₃ in the hyperthyroid patient, of course, is direct secretion by the thyroid gland, which may account for as much of the T₃ production as does D1-catalyzed deiodination (19). The important role of D2 in the euthyroid state and its half-life of less than 45 minutes would argue that the rapid decrease in T₃ in the sick patient is more likely to result from a decrease in D2-catalyzed T₄ to T₃ conversion than from the modest decrease in hepatic D1 activity. This is supported by the complete absence of D2 activity in skeletal muscle in patients dying in an intensive care unit (27). However, in chronic severe illness at least, this decrease is also supplemented by an increase in the inner-ring deiodination of T₄ and T₃ by type 3 iodothyronine deiodinase in the liver and skeletal muscle (27).

Methods

Reagents.

Reagents were from Calbiochem-Novabiochem or Sigma-Aldrich. Outer ring–labeled [¹²⁵I]T₃ or T₄ (specific activity, 4,400 Ci/mmol) was from PerkinElmer. Purification of [¹²⁵I]T₃ and [¹²⁵I]T₄ was performed on LH-20 columns just before use to reduce ¹²⁵I^– to less than 1%.

Transfection and expression studies.

MSTO, LLCPK1, and HEK 293 lines were obtained from ATCC. Cells were grown and maintained in DMEM supplemented with 10% FBS. HSMMs were obtained from Cambrex Corp. and cultured according to the manufacturer’s instructions using Cambrex-supplied SkGM2 medium. NH2-terminal FLAG-tagged wild-type human D1 and D2 plasmids were used (17). The Flag tag does not change kinetic properties.

For studies of catalytic efficiency, D1 or D2 expressing plasmids (3–100 ng) were transiently expressed in HEK 293 cells using CaPO4 precipitation (28). A TK-hGH plasmid was used to control for transfection efficiency (28). At 48 hours after transfection, cells in 6-well plates were washed twice with sterile PBS and then cultured for 18–24 hours in 1 ml serum-free 0.1% BSA (resulting in an FT₄ concentration of 2.7%; ref. 29) in DMEM plus variable T₄ concentrations ([total T₄], 74–7,400 nM; [FT₄], 2–200 pM) including approximately 100,000 cpm/ml [¹²⁵I]T₄. Experiments were performed in duplicate or triplicate for each condition.

Assay of 5 ′-deiodinase activity in intact transfected cells and in cell sonicates.

Deiodinase activity in intact cells was assayed as described previously with the following modifications (14, 15). At completion of the experiment, 300 μl of medium was removed and added to 200 μl horse serum and protein precipitated by the addition of 100 μl 50% trichloroacetic acid (TCA) followed by centrifugation at 12,000 g for 2 minutes. The ¹²⁵I^– generated was expressed as the fraction of the total T₄ counts minus the nonspecific deiodination in untransfected control cells (<5% of the total [¹²⁵I]T₄ counts) and corrected for the 50% reduction in the specific activity relative to T₄.

For measurements of activity in cell sonicates, the remaining medium was removed and the cells washed twice with PBS, harvested, and sonicated in 0.25 M sucrose in PE buffer (0.1 M potassium phosphate and 1 mM EDTA) with 10 mM DTT. For deiodinase assays, we used 100–150 μg cell sonicate; 10 μM [¹²⁵I]T₄ (for D1) or 10 nM T₄ (for D2); and 20 mM DTT in a final volume of 300 μl PE. These concentrations are approximately 5 times the estimated Km(T₄) for each enzyme (2 μM for D1 and 2 nM for D2). Incubation was for 60–120 minutes at 37°C, and ¹²⁵I^– was separated from labeled T₄ by TCA precipitation. Results are the mean of values derived from at least 2 separate experiments.

To assess whether T₄ to T₃ conversion in human skeletal muscle could increase the ratio of extracellular T₃ to T₄, we incubated HSMMs in 60-mm dishes for 22 hours in 2 ml 0.1% BSA DMEM with approximately 1 μCi/ml [¹²⁵I]T₄ without or with 0.5 mM BT₂cAMP. Rabbit polyclonal T₃-Ab (1:1,000) was added to the medium to trap released [¹²⁵I]T₃. A sample of the same medium was incubated with no cells as a control. At the end of the incubation, the 2-ml samples of medium were incubated with 50 μl of protein A/G agarose suspension at 4°C overnight and centrifuged at 15,000 g for 5 minutes and the pellet rinsed twice with cold PBS. Ice-cold methanol (150 μl) was added and the tubes vortexed vigorously for 2 minutes and allowed to stand for 2 hours at room temperature. The T₃ recovery with this procedure was 40%. T₄ displayed a weak cross-reactivity with T₃-Ab, resulting in the binding of 6% of the [¹²⁵I]T₄. The methanol extract containing more than 90% of the labeled iodothyronines was subjected to HPLC as described below so that the ratio of [¹²⁵I]T₃ to [¹²⁵I]T₄ in the medium could be determined.

To analyze the effect of D1- versus D2-catalyzed T₃ production on T₃-dependent gene transcription, HEK 293 cells in 6-well plates were transfected by CaPO4 precipitation with D1- or D2-expressing plasmid, 3 μg pTK-LUC containing 3 copies of the DR-4 5′ TRE from the human DIO1 gene (TRE3TKLUC) (5), 0.2 μg of a plasmid expressing the mouse thyroid hormone receptor α (T₃Rα) (30), and 1 μg of TK-hGH plasmid for correction of transfection efficiencies (28). Results are expressed as the mean induction of T₄-treated D1- or D2-transfected wells paired with cells transfected with an inactive D2, a COOH-terminal FLAG-tagged human D2 in which Ala replaced Glu163 (31). To prevent reentry of the T₃ produced from T₄ deiodination after it exited the cell, we added T₃-Ab to the medium (1:1,000) (32). Luciferase activity was measured by a commercial kit (Luciferase Assay System; Promega). Results are expressed as the mean of 3 experiments.

Reverse-phase HPLC.

Cells were homogenized in 1 ml medium after 24 hours incubation with appropriate amounts of [¹²⁵I]T₄ or T₃ and vortexed vigorously. Aliquots of 150 μl were mixed with ice-cold methanol and centrifuged. The resultant supernatant was mixed with an equal volume of 0.02 M ammonium acetate (pH 4) and the distribution of ¹²⁵I-labeled products determined by HPLC as described previously (33). The radioactivity of each iodothyronine peak was measured using a Radiomatic 500TR Flow Scintillation Analyzer (PerkinElmer).

Real-time PCR.

Total RNA was extracted using TRIzol reagent (Invitrogen Corp.) and used to synthesize cDNA using SuperScript First-Strand Synthesis System for RT-PCR (Invitrogen Corp.). The generated cDNAs were used in a real-time PCR using the QuantiTect SYBR Green PCR Kit (Bio-Rad Laboratories) in iCycler (Bio-Rad Laboratories). Standard curves representing 5-point serial dilution of mixed cDNA of the experimental and control groups were analyzed and used as calibrators of the relative quantification of product generated in the exponential phase of the amplification curve. The r² was greater than 0.99, and the amplification efficiency varied between 80% and 100%. Quantification was normalized to cyclophilin A mRNA expression. Oligonucleotides for human D2 (5′-ACTTCCTGCTGGTCTACATTGATG-3′ and 5′-CTTCCTGGTTCTGGTGCTTCTTC-3′) and cyclophilin A (5′-GTCAACCCCACCGTGTTCTTC-3′ and 5′-ACTTGCCACCAGTGCCATTATG-3) were designed using Beacon Designer 2.06 (PREMIER Biosoft International).

Skeletal muscle biopsies.

Biopsies of sternocleidomastoid, rectus abdominis, and vastus lateralis muscles were obtained during routine surgical procedures at Hospital de Clínicas de Porto Alegre. Exclusion criteria included a diagnosis of diabetes mellitus or impaired glucose intolerance, metastatic cancer, and hyper- or hypothyroidism. Samples were immediately snap-frozen in liquid nitrogen and stored at –70°C. The Ethics Committee of the Hospital de Clínicas de Porto Alegre approved the protocol, and patients provided informed consent.

Statistical analysis.

Data are mean ± SEM unless otherwise indicated. Statistical analysis was performed using 2-tailed Student’s t or ANOVA tests, and P < 0.05 was considered significant.