Molecular and Structural Characterization of the Heat-resistant Thyroxine-binding Globulin-Chicago*

Thyroxine-binding globulin (TBG) is the main trans- port protein for thyroxine (T 4 ) in blood. It shares con-siderable sequence homology with (cid:97) 1 -antitrypsin (AT) and other members of the serine proteinase inhibitor (serpin) superfamily of proteins. The crystallographic structure of AT has been determined and was found to represent the archetype of the serpins. This model has been used for structure-function correlations of TBG. Sequence analysis of the heat-resistant variant TBG- Chicago (TBG-CH) revealed a substitution of the normal tyrosine 309 with phenylalanine. For further analysis, vectors containing the coding regions of normal TBG (TBG-N) and TBG-CH were constructed, transcribed in vitro , and expressed in Xenopus oocytes. Both TBGs were secreted into the culture medium and could not be distinguished by gel electrophoresis. Scatchard analysis of T 4 binding to TBG-N and -CH revealed no significant differences in binding affinity. The rate of heat denaturation of TBGs was determined by measurement of resid- ual T 4 binding capacity after incubation at 60 °C for various periods of time. The half-life values of denaturation of TBG-N and -CH were 7 and 132 min, respec- tively. The tyrosine 309 to phenylalanine substitution of TBG-CH involves a highly conserved phenylalanine

Heat stability of proteins has been a subject of intensive research for several decades. The introduction of genetic engineering techniques (1,2) and the need for heat-resistant proteins for research (3), food biotechnology (4), and other industrial processes (5) has stimulated further research in this area (6 -8). Early claims of universally applicable mechanisms to predict heat stability of proteins have been followed by reports that deny the existence of general strategies to improve protein stability (9,10). However, the stabilizing effects of several alterations of protein structures have been well established. These include an increase in the hydrophobicity of the protein core (11)(12)(13)(14), improved packing density (15), interaction of ␣-helix dipoles with charged residues (8), disulfide bond formation (16 -18), N-linked glycosylation (19), and the assembly into quaternary structures (10). It has also been shown that the combination of stabilizing mutations leads to a cumulative effect on protein stability (1, 9, 20 -22).
Current knowledge about stabilizing factors stems mostly from the analysis of wild-type proteins and their less stable variants. Only a few proteins with significant increases in thermal stability have been described. These include the well characterized barnase (23,24), -repressor protein (15,25), subtilisin (26), kanamycin nucleotidyltransferase (20), and bacteriophage T4 lysozyme (20,27,28). However, the gain of thermal stability in most of these variants is offset by a loss of function, i.e. reduced enzyme activity.
Thyroxine-binding globulin (TBG) 1 is the main transport protein for thyroxine (T 4 ) and triiodothyronine in human serum (29). It is synthesized in the liver and secreted into the blood stream as a 54-kDa glycoprotein (30). The primary structure of human TBG and the organization of the TBG gene have been described (31,32). The mature protein contains 395 amino acids in a single polypeptide chain and oligosaccharides attached to four of the five potential N-linked glycosylation sites (33). Familial TBG defects follow a X-linked inheritance pattern, consistent with the presence of a single TBG gene on the long arm of the X chromosome (34,35).
So far, six partial and three complete TBG deficiency variants have been analyzed at the gene level (35,36). All partial deficiency variants have different degrees of heat lability associated with one or two nucleotide substitutions resulting in amino acid replacements. A causal relationship of the mutation and impaired function has been established for three of these TBG variants (37,38). Screening of serum samples has revealed several more heat-labile, but only one heat-resistant, thyroxine-binding globulin variant (39). This unique serum sample belonged to an African-American man from a Chicago family that had no stigmata of thyroid dysfunction. Except for resistance to heat and acid denaturation, the TBG of the propositus (TBG-CH) had normal thyroxine-binding kinetics, a microheterogeneous isoelectric focusing pattern and immunological properties identical to the normal TBG (TBG-N) and was present in serum in normal concentration (39).
TBG belongs to the serine protease inhibitor (serpin) superfamily, a heterogeneous group of more than 60 proteins, including ␣ 1 -antitrypsin (AT), ␣ 1 -antichymotrypsin, and corticosteroid-binding globulin, among others (40 -42). The crystallographic structure of AT has been determined (43) and was found to represent the archetype of the serpins (40). Since attempts to crystallize TBG have failed (44), the AT model was used for structure-function correlations of the TBG molecule.
We now present the amino acid sequence of TBG-CH deduced by gene sequencing; confirm by in vitro expression that the single amino acid substitution is sufficient to impart heat resistance to the molecule and provide an explanation for the increase in thermal stability of TBG-CH by structure modeling.

MATERIALS AND METHODS
Reagents-Restriction endonucleases, DNA modifying enzymes, and size markers were from New England Biolabs, Life Technologies, Inc., and Boehringer Mannheim. Oligonucleotide primers used for amplification and sequencing have been described previously (45). The expression vector pSELECT, the altered sites mutagenesis kit, the Gemini-II in vitro transcription kit, recombinant RNasin, rabbit reticulocyte lysate, and canine microsomal membranes were obtained from Promega Biotech. L-Methionine used for the cell-free synthesis of unlabeled TBG was from Sigma. The cap analog m 7 G(5Ј)ppp(5Ј)G was from Pharmacia Biotech Inc. Taq DNA polymerase was from Perkin-Elmer. ␣-35 S-dATP (specific activity Ͼ 800 Ci/mmol) for DNA sequencing and 14 C-methylated protein size markers were from Amersham Corp. L-[ 35 S]Methionine (specific activity Ͼ 1200 Ci/mmol) for metabolic labeling of proteins and [ 125 I]T 4 (specific activity, 1500 Ci/g) for binding analysis were from DuPont NEN. All other reagents were of analytical grade.
Construction of a Genomic DNA Library Containing the TBG-CH Gene-Leukocyte DNA was prepared from blood (47) of a subject known to expess both TBG-CH and TBG-Slow (TBG-S) (39). 20 g of this DNA were partially digested with EcoRI and run on a 0.8% SeaPlaque (FMC, San Diego, CA) agarose gel in TAE buffer (46). DNA from the 9 -23 kilobase pair region was excised, isolated by gel extraction, and ligated into EMBL4-EcoRI-arms (Stratagene, La Jolla, CA). The resulting -vector library was packaged with the Gigapack Gold in vitro packaging extract (Stratagene) and transfected into E. coli P2392 and grown on LB plates. Plaques were lifted with BA 85 nitrocellulose filters (Schleicher & Schuell, Dassel, Germany). Plaque hybridization, further propagation, rehybridization, and large scale preparation of positive clones was performed essentially as described previously (46), using a 32 P-labeled TBG-cDNA probe synthesized with a nick translation kit from Amersham labeled with [␣-32 P]dCTP (specific activity Ͼ 3000 Ci/mmol, DuPont NEN) and purified with G-50 quick spin columns (Boehringer Mannheim, Mannheim, Germany). Clones were screened by allele-specific amplification (48), and those that did not have the TBG-S genotype were assumed to have the TBG-CH genotype.
Sequence Analysis-A -vector containing the TBG-CH gene was digested with BamHI, and the resulting 5.5-kilobase pair insert was subcloned into a pGEM-7Zfϩ vector. The coding regions and exonintron junctions of the TBG gene in this vector were sequenced by the dideoxynucleotide termination method (49) using the Sequenase-2 kit from U. S. Biochemical Corp.
Construction of Vectors-A vector containing the full-length cDNA of TBG-N had been constructed before (37). The TBG insert of this vector was excised with BamHI, separated by agarose gel electrophoresis, isolated with miniprep filters (Millipore, Bedford, MA), and subcloned into the pSELECT expression vector by standard techniques (46). This construct was designated pSpT-N. Site-directed mutagenesis with the oligonucleotide 5Ј-TTCAGCATGCCTTTTCTGAAAATGC-3Ј (the underlined T is substituted for the normal A) was performed to obtain a vector coding for TBG-CH. Briefly, single-stranded DNA was generated from the vector pSpT-N and incubated with the mutagenesis oligonucleotide and with a repair oligonucleotide to reinstitute ampicillin resistance of the pSELECT vector. The complete second strand was synthesized with DNA-polymerase and -ligase, and the resulting reaction products were used to transform the DNA repair-deficient BMH7118 strain of E. coli. Mutated vectors were selected by supplementing the medium with ampicillin. After preparation of vector DNA, a second round of transformation in JM109 E. coli was performed to obtain pure vector preparations. Individual clones for both the TBG-N and TBG-CH constructs were tested by restriction endonuclease analysis for proper orientation of the coding region relative to the T7 RNA polymerase promoter and verified to contain the complete and proper nucleotide sequence by sequencing alkali denatured double-stranded vector DNA as template.
In vitro transcription and cell-free translation were performed as described previously (37).
Preparation of Oocytes and RNA Injection-Ovaries were removed from mature Xenopus laevis (Nasco) (50) and suspended in OR-IIa medium (83 mM NaCl, 2.5 mM KCl, 1 mM MgCl 2 , 1 mM Na 2 HPO 4 , and 5 mM HEPES, pH 7.6) (51). After manual dissection of the follicles, oocytes were dissociated from the surrounding connective tissue by incubation in OR-IIa containing 0.2% collagenase type IA (Sigma) for 2 h with shaking at room temperature. The liberated oocytes were then rinsed extensively in OR-IIb medium (OR-IIa with 1 mM CaCl 2 and 100 g/ml gentamycin), and stage VI oocytes (52) were separated and kept up to 3 days in OR-IIb with daily medium changes. After injection with 100 nl of sRNA (0.5 g/l), oocytes were kept on ice for 1 h and then for 2-6 h at 19°C in OR-IIb. Intact oocytes (Ͼ95%) were transferred to 24-well plates (Costar) and kept in OR-IIc (OR-IIb with 1 mM sodium pyruvate), 5 l/oocyte, at 19°C for up to 4 days, with daily exchange of medium. Typically, 100 oocytes were injected with each sRNA preparation. Control oocytes were either injected with water or noninjected, with identical results. In some experiments, the oocytes were metabolically labeled by addition of 250 Ci of [ 35 S]methionine/500 l of medium.
Extraction and Analysis of TBG Synthesized in Oocytes-The medium of microinjected oocytes was removed, supplemented with 1 mM phenylmethylsulfonyl fluoride, and stored at Ϫ20°C until further use.
SDS-PAGE-Products of cell-free translation or synthesized in Xenopus oocytes were analyzed by the method of Laemmli (53), using polyacrylamide gels at 10% T and 2.7% C. Stacking gels were 3.75% T, 2.7% C. Gels were dried and autoradiographed at Ϫ90°C on X-AR5 film (Eastman Kodak Co.) with an intensifying screen.
Measurement of T 4 Binding to TBG-Parameters of T 4 binding to TBG were measured by a method previously described in detail (54). Briefly, TBG preparations were incubated with [ 125 I]T 4 in the presence of increasing amounts of unlabeled T 4 . TBG-bound T 4 was separated from free T 4 with anion exchange resin beads (Mallinckrodt), and the protein-bound 125 I activity was determined. The affinity constants (K a ) of TBG preparations were determined by the method of Scatchard (55).
Heat Denaturation-The rate of heat denaturation of TBG was determined by measurement of the residual T 4 binding capacity of TBG in samples incubated at 60 Ϯ 0.1°C for various periods of time. Data were expressed as a percentage of the T 4 binding capacity of TBG-N before heat exposure. Half-lives (T1 ⁄2 ) of denaturation were calculated from the slope of the linear regression semi logarithmic plots.

Sequence Analysis of the TBG-CH Gene-A genomic
-EMBL4 library was prepared from EcoRI-digested leukocyte DNA obtained from an heterozygous member of the TBG-CH family expressing both TBG-CH and TBG-S. The latter can be identified by its slower electrophoretic mobility, its cathodal shift on isoelectric focusing, and the substitution of the normal aspartate with asparagine due to a mutation in codon 171 (48). Screening of the EMBL4 library with a TBG-cDNA probe and subsequent analysis by allele-specific amplification specific for TBG-S identified clones that contained the TBG-CH allele, which is contained in a single 14-kilobase pair EcoRI DNA fragment. For further analysis, a 5.5-kilobase pair BamHI fragment of the TBG-CH gene containing the entire coding region was subcloned and sequenced. Fig. 1 shows the structure of the TBG gene and the strategy of sequencing. Analysis of the TBG-CH gene revealed a single base substitution replacing the normal adenine 2767 with thymine, resulting in the substitution of the normal tyrosine 309 with phenylalanine (Fig. 2).
Synthesis of Normal and Mutant TBG in Reticulocyte Lysate-An expression vector for TBG-N was constructed by subcloning a TBG-cDNA into the pSELECT phagemid. A vector containing the mutation specific for TBG-CH was then constructed by site-directed mutagenesis. The linearized vectors were transcribed in vitro to obtain synthetic mRNAs (sRNAs). The products of cell-free translation of TBG-N and TBG-CH sRNAs in reticulocyte lysate were indistinguishable on SDS-PAGE. Both variants had a major band at 44 kDa, and several new bands of higher molecular weight after signal peptide processing and core glycosylation (Fig. 3).

Expression of Normal and Mutant TBG sRNAs in Xenopus
Oocytes-Oocytes were removed from X. laevis, culled, injected with TBG sRNAs, and incubated in medium with [ 35 S]methionine. The TBGs synthesized and secreted into the medium were then submitted to SDS-PAGE. No significant differences were observed between TBG-N and TBG-CH (Fig. 4). Both showed a microheterogeneous 60-kDa product that was secreted in similar amounts. The TBGs from microinjected Xenopus oocytes were also submitted to isoelectric focusing. Again, no differences were found between TBG-N and TBG-CH (data not shown), as shown previously for the respective serum TBGs (39).
Analysis of T 4 Binding to TBG-The amount of biologically active TBG synthesized in Xenopus oocytes was examined by measurement of the T 4 binding characteristics. TBGs were synthesized as described above, but without the addition of [ 35 S]methionine to the medium. Scatchard analysis of the secreted TBGs revealed no significant differences in T 4 binding affinity and T 4 binding capacity/oocyte for TBG-N and TBG-CH (Fig. 5). The amounts of secreted TBG-N and TBG-CH as quantified directly by SDS-PAGE (Fig. 4) and by radioimmunoassay (data not shown) were also not significantly different. These results are in agreement with data obtained from serum TBG of a subject expressing only the TBG-CH gene (39).
Heat Resistance of TBG-CH-For heat denaturation, unlabeled TBG-N and TBG-CH were expressed in Xenopus oocytes as described above. The rate of heat denaturation of the TBGs was determined by measurement of residual T 4 binding capacity with a standard resin T 4 binding assay after various incubation periods. At 60°C, the half-life of denaturation of TBG-N was 7 min and that of TBG-CH was 132 min, almost 20 times as long (Fig. 6). DISCUSSION The enhancement of protein stability by rational design is one of the great goals of protein engineering. General principles are still not available because of the complex interactions and the strong positional and context dependence of the effect of a particular amino acid substitution on the stability of a specific protein (8). The properties of natural proteins often depend on just one amino acid at a specific site, as shown by the many deleterious point mutations in proteins such as ␣ 1 -antitrypsin, myoglobin, and hemoglobin. In most cases, the loss of stability is accompanied by a loss of biological activity. This has been confirmed by systematic analysis of proteins by site-directed mutagenesis.
Heat resistance has been described in a few natural variants and some engineered proteins. However, the gain of thermal stability in most of these variants is offset by a loss of function, i.e. reduced enzyme activity. Most TBG variants identified to date have unaltered or decreased heat stability. The altered properties of some of these variants can be explained by the changes in their primary structure, i.e. loss of a negative charge (TBG-S (48)) or creation of a new site for N-linked glycosylation (TBG-Gary (56)). In other variants (TBG-Montreal (45), TBG-CD5 (57), TBG-Quebec (58), TBG-San Diego (59), and TBG-PDJ (60)), alterations of the primary structure are more subtle, and functional studies are required to understand the effect of the amino acid replacement on the structure of the protein (37, 38, 61). All heat-sensitive TBG variants have also defects in their T 4 binding affinity and show increased concentrations of denatured TBG in serum, compatible with a general defect of The variant TBG-CH is unique in its pronounced heat resistance with preservation of normal T 4 binding affinity, electrophoretic mobility, and serum concentration. The isolated increase in stability of TBG-CH can thus be thought of as a specific effect of the mutant amino acid on the molecule (Fig. 2).
The choice of the Xenopus oocyte system to analyze the properties of TBG-CH was based on the previous use of this translation system to characterize transport and secretion abnormalities of mutant forms of AT (62,63) and TBG (TBG-Montreal, TBG-CD5, TBG-Gary, and TBG-CDJ). Microinjected Xenopus oocytes have been shown to be a legitimate surrogate system for the study of inherited TBG variants, since the biological, physical, and immunological properties of these variant TBGs from serum of affected individuals were faithfully reproduced by them (37,38,61).
In vitro translated TBG-CH had the same properties as the respective serum TBG, confirming that the substitution of the normal tyrosine at position 309 by phenylalanine is the cause for the increased heat stability of this variant TBG. The residue corresponding to tyrosine 309 of TBG-N (tyrosine ϭ Y, Fig. 7) corresponds to a highly conserved phenylalanine (phenylalanine ϭ F, at position 312 in AT, Fig. 7) in the serpin superfam-ily of proteins, of which TBG is a member. Comparison with the crystallographic structure of the archetypical serpin, AT, shows that this amino acid resides in a deep intramolecular pocket and ties the ␣-helix hI1 to the molecule, 2 thus stabilizing the tertiary structure. A substitution with tyrosine with its larger and hydrophilic side chain would disrupt this interaction. Accordingly, stabilization of the TBG-CH molecule is most likely due to hydrophobic interactions of the better fitting side chain of phenylalanine. TBG-CH with its phenylalanine for tyrosine substitution is thus more closely related to the serpins than TBG-N.