Structural characterization of three genetic ... - Wiley Online Library

Eur. J. Biochem. 251, 3292334 (1998)  FEBS 1998

Structural characterization of three genetic variants of human serum albumin modified in subdomains IIB and IIIA Monica GALLIANO 1, Scott WATKINS 2, Jeanne MADISON 2, Frank W. PUTNAM 2, Ulrich KRAGH-HANSEN 3, Roberto CESATI 4 and Lorenzo MINCHIOTTI 1 1 2 3 4

Dipartimento di Biochimica ‘A. Castellani’, Universita` di Pavia, Pavia, Italy Department of Biology, Indiana University, Bloomington, USA Department of Medical Biochemistry, University of Aarhus, Aarhus, Denmark Laboratorio di Analisi, Ospedale di Carate Brianza, Carate Brianza, Italy

(Received 4 September 1997) 2 EJB 97 1269/3

Three new genetic variants of human serum albumin have been detected in Italy by routine clinical electrophoresis. Albumin Milano Slow is common in Northern Italy, while albumins Liprizzi and Trieste, which are fast migrating, are rare and local variants. Isoelectric focusing analysis of the CNBr fragments obtained from the carboxymethylated alloalbumins in all cases localized the mutation to fragment CB5 (residues 3302446). The modified CNBr fragments were isolated on a preparative scale and subjected to tryptic digestion. Sequence determination of the abnormal tryptic peptides revealed that all the variants are caused by single point mutations : Trieste, Lys359→Asn, Milano Slow, Asp375→His, and Liprizzi, Arg410→Cys. These results were confirmed by sequence determination of a variant V8 peptide in the case of Trieste, and by DNA sequence analysis for the other two variants. The DNA analysis showed a G→C transversion at nucleotide position 11969 for albumin Milano Slow, and a C→T transition at position 13 251 for Liprizzi. The latter represents a mutation at a hypermutable CpG dinucleotide site. Albumins Trieste and Milano Slow, as most of the variants thus far described, have mutations involving residues on the surface of the molecule. In contrast, albumin Liprizzi represents the first example of a mutation in the most active binding pocket of the molecule, placed in subdomain IIIA. Keywords : human serum albumin; genetic variant; amino acid sequence; DNA sequence; point mutation.

Human serum albumin is a single-chain, unglycosylated protein of 585 amino acids which is folded into three homologous domains [1] like all other mammalian albumins. Each domain is composed of two subdomains, A and B, which are made up of tightly packed series of helices. Extended peptide strands connect the A and B subdomains, while helical stretches connect adjacent whole domains [2]. Although the functional role of this molecule is still under investigation, albumin is quantitatively the most important transport protein in plasma and can act as a major antioxidant in extracellular fluids [1, 2]. Inherited mutations produce circulating genetic variants (alloalbumins), which are detected during routine clinical electrophoresis. The number of known variants is greater for albumin than for any other protein except hemoglobin [1], but alloalbumins are not clearly associated with disease. However, the Arg218→His mutation causes the inherited condition of familial dysalbuminemic hyperthyroxinemia [3, 4]. The genetic variants of human serum albumin are under investigation in several laboratories in order to define their molecular defects and correlate them to the functional properties and stability of the molecule [1, 5, 6]. More than 100 electrophoretically different mutants have been detected with a frequency in the average population of 321031024 and 56 amino acid Correspondence to L. Minchiotti, Dipartimento di Biochimica ‘A. Castellani’, Universita` di Pavia, via Taramelli 3B, I-27100 Pavia, Italy E-mail: [email protected] Abbreviations. RP, reverse-phase ; HDX, heteroduplex; pI, isoelectric point.

substitutions have been so far characterized (see [5] and Table 4.8 in [1] for a complete review). Among them, 52 reflect singlebase changes in the structural gene and 4 are chain-termination mutants. Most of the point mutations involve charged amino acid residues on the surface of the molecule and are clustered in three regions: the propeptide and amino terminus, and the Cterminal regions of subdomains IIB and IIIB [5, 7, 8]. Mutations are rarely identified in the highly conserved stretches of the polypeptide chain and in the two hydrophobic cavities in subdomains IIA and IIIA that represent the principal regions of ligand-binding to serum albumin [1, 2]. In the present paper we report the structural characterization of two fast and one slow Italian alloalbumins, Trieste (Lys359→Asn), Liprizzi (Arg410→Cys), and Milano Slow (Asp375→His). Albumin Trieste has been characterized by protein sequencing, while the molecular defects of albumins Milano Slow and Liprizzi have been identified by both DNA and protein sequencing. All of them have previously unreported amino acid substitutions. Albumin Liprizzi represents the first instance of a mutation located in the binding pocket of subdomain IIIA, the most active in the molecule [2]. The position of this residue in the serpentine layout of the molecule [9] is identical to that of Arg218 in subdomain IIA, which is substituted in the familial dysalbuminemic hyperthyroxinemia syndrome [3, 4]. Albumins Milano Slow and Trieste have mutations located in a short segment, position 3542382, corresponding to the C-terminal portion of helix h9(II), the loop connecting h9(II) and h10(II) and the N-terminal region of the long helix h10(II)-h1(III) in the

330

Galliano et al. (Eur. J. Biochem. 251)

three-dimensional structure [2], that has the highest concentration of substitutions [10]. MATERIALS AND METHODS Electrophoretic survey. A fresh specimen of serum from an individual with the albumin Liprizzi trait was supplied by Roberto Cesati, while those of albumins Milano Slow and Trieste were from C. Petrini (Milan) and G. Ferrari (Tradate, Lombardy), respectively. Each variant was named according to its geographical origin. Normal (common) albumin is designated albumin A. All the donors were heterozygous and inheritance of the traits had been demonstated. EDTA-treated blood for DNA preparation was received in the case of Liprizzi and Milano Slow. A preliminary screening of the variants was performed by comparative cellulose acetate electrophoresis with the Chemetron apparatus at two different pH values : pH 5.0 (31 mM sodium acetate, 4 mM EDTA) and pH 8.6 (4 mM sodium barbital, 7.5 mM barbital, 1.2 mM calcium lactate). The electrophoretic mobility was faster than normal for variants Liprizzi and Trieste, and was slower for Milano Slow. Protein structural studies. The normal and variant albumins were separated on a column (1 cm3100 cm) of DEAESephadex eluted with 0.11 M sodium phosphate pH 5.75, or by HPLC on a DEAE-5PW column (21.5 mm3150 mm) (BioRad) using 20 mM sodium acetate with a 40-min pH gradient from pH 5.2 to pH 4.5 for the slow variant and from pH 5.2 to pH 4.0 for the fast variants at a flow rate of 5.0 ml/min [11]. SDS/PAGE was performed under reducing and non-reducing conditions in 8 % polyacrylamide gels. The reduced and carboxymethylated albumins were cleaved with CNBr to yield seven fragments (CB12CB7); the CNBr digests were compared by IEF in 8 M urea, pH range 2.528.0 [12]. For all three variants the site of the mutation was fragment CB5 (residues 3302 446). The modified CB5 fragments were isolated preparatively by gel filtration on a TSK G3000SW column (10 µm, 7.5 mm360 cm) (Toyo Soda) and isocratic elution with 35% acetonitrile, 0.1 % trifluoroacetic acid followed by reverse-phase (RP) HPLC on a Vydac C18 column (4.6 mm3250 mm) (Anspec). The purified CB5 fragments were digested with trypsin, and CB5 from Trieste was also digested with Staphylococcus aureus V8 protease (Boehringer). The V8 and tryptic digests were mapped by RP-HPLC on Vydac C18 or on a PepRPC HR 5/5 column (Pharmacia). Tryptic and V8 peptides are given the prefixes T and S, respectively, and are numbered consecutively in their order in the sequence [13]. The purified tryptic and V8 variant peptides were submitted to amino acid analysis and sequenced with a Hewlett-Packard model G 1000A sequencer (Centro Grandi Strumenti, Universita` di Pavia) or with an Applied Biosystem model 477B (Indiana). DNA structural studies. High-molecular-mass DNA was isolated from leukocytes by proteinase K treatment and phenol extraction [14]. Primer sets (designated A01A and A02A, etc.) that spanned specific exons of human serum albumin and their intron-exon junctions were used to PCR-amplify regions of the gene ranging over 2882464 bp in length [14]. PCR reactions were performed as described [14]. For heteroduplex (HDX) analysis the DNA samples were denatured and reannealed slowly to form heteroduplexes, and the heteroduplex products were resolved in mutation detection enhancement gels (AT Biochem, Malvern PA) [14]. For subcloning and DNA sequencing, the PCR products were digested with appropriate restriction enzymes and the resulting fragment was ligated into vector DNA (pBluescript, Stratagene) having compatible ends. Transformation, screening, and double-stranded DNA sequencing were performed as described [15].

Fig. 1. Isoelectric focusing of CNBr fragments from albumins Liprizzi, Milano Slow, and Trieste. Lane 1, normal albumin; lane 2, Liprizzi; lane 3, purified CB5 from Liprizzi ; lane 4, Milano Slow ; lane 5, purified CB5 from Milano Slow ; lane 6, Trieste; lane 7, normal albumin. Peptides were resolved in the pH range 2.528.0 in the presence of 8 M urea. CNBr fragments are numbered as CB127 according to their order in the known sequence of human serum albumin [13], and the positions given are for normal fragments. Each fragment, except the Cterminal CB7, may have two charged forms, owing to the homoserine/ homoserine lactone equilibrium. Fragment CB2 has not been identified, probably owing to its high solubility and low dye affinity. Microheterogeneity, mostly in the case of larger fragments, is due either to partial cleavage or oxidation of unreacted Cys residues or deamidation. The asterisks mark the abnormal CNBr fragments. The apparently higher concentration of the variant CB5 fragment in lane 4 might reflect its increased binding affinity for Coomassie blue.

RESULTS Trieste (Lys359→Asn). This variant was found in a single individual from Trieste (Friuli-Venezia Giulia region); on cellulose acetate electrophoresis it exhibited a fast 21 mobility at pH 5.0. The mutant and the normal albumins were isolated in a 1:1 ratio by DEAE-Sephadex chromatography, and by comparative IEF analysis of the CNBr fragments the mutation was localized in CB5 (Fig. 1, lane 6). The difference in isoelectric point (pI) of 20.420.5 suggested a 21 charge change. The modified CB5 fragment was isolated on a preparative scale by TSK gel filtration followed by RP-HPLC on a Vydac C18 column and digested with both S. aureus V8 protease and trypsin. The V8 profile on a PepRPC column revealed that fragment S38-39 eluted later than the normal (Fig. 2 B). The tryptic pattern on a Vydac C18 column showed the absence of peptides T48 and T49 and the presence of a new peptide (T48-49m) (Fig. 3A). Sequence analysis of both the modified V8 and tryptic peptides showed a Lys359→Asn mutation (Fig. 4). The protein change can be accounted for by a G→T/C transversion at nucleotide 10835 [13]. Milano Slow (Asp375→His). This variant is common in the Lombardy region and nearby Veneto, where up to 100 unrelated cases have been reported ; on cellulose acetate electrophoresis it exhibites a slow 11 mobility at pH 5.0. As the mutant did not separate from normal albumin by DEAE-Sephadex chromatography, the CNBr fragments were analyzed by IEF as a mixture of normal and variant. The isoelectric point of the variant CB5 was increased by 0.6520.75 (Fig. 1, lane 4). This finding, suggesting a 11.5 charge mutation, prompted us to examine the exons coding for CB5. HDX analysis of the exons encoding


331

Fig. 2. RP-HPLC elution profile of a V8 protease digest of fragment CB5 from normal (A) and Trieste (B) albumins. The lyophilized V8 protease digests were dissolved in 0.1 % aqueous trifluoroacetic acid (solvent A), and 100 µl (corresponding to about 5 nmol) were injected on a PepRPC HR 5/5 column equilibrated with solvent A. Peptides were eluted at a flow rate of 1 ml/min using the gradient indicated by the dashed line. Solvent B was 0.1 % trifluoroacetic acid in acetonitrile. Absorbance range, 1.0 full scale. Each peak was subjected to amino acid and N-terminal analysis. V8 peptides are given the prefix S and are numbered consecutively in their predicted order in the amino acid sequence [13]. S38-39m denotes the variant peptide resulting from the amino acid substitution in albumin Trieste.

CB5 indicated a mutation in exon 10. DNA sequencing of both strands of this exon showed a G→C transversion at nucleotide 11 969 [13] (Fig. 5A). In order to confirm the mutation by protein sequencing, the modified CB5 fragment was isolated on a preparative scale by TSK gel filtration followed by RP-HPLC on a Vydac C18 column. In the last step the variant CB5 eluted earlier than the normal one and was obtained in a pure form, as judged by IEF analysis (Fig. 1, lane 5). The elution profile of the tryptic digest on a Vydac C18 column showed a decreased retention time for the modified T50 (Fig. 3 B). Protein sequencing of this peptide identified the expected Asp375→His mutation (Fig. 4). After the structural work was finished, a separation of Milano Slow from normal albumin was obtained using a DEAE5PW column and the variant protein comprised half of the total albumin content. Liprizzi (Arg410→Cys). This variant, which showed a slightly fast electrophoretic mobility at pH 5.0, was found in a single individual from Liprizzi (Sicily region). Chromatography on DEAE-Sephadex and DEAE-5PW failed to isolate the variant. However, IEF analysis of the CNBr fragments, performed on a mixture of normal and variant albumin, indicated a mutation in CB5 (Fig. 1, lane 2) and suggested a 22 charge mutation (∆pI 5 20.6620.87). The modified CB5 fragment was obtained

Fig. 3. RP-HPLC elution profile of a tryptic digest of fragment CB5 from albumins Trieste (A), Milano Slow (B) and Liprizzi (C). The lyophilized tryptic digests were dissolved in 0.1 % aqueous trifluoroacetic acid (solvent A), and 100 µl (corresponding to about 5 nmol) were injected on a Vydac C 18 column equilibrated with solvent A. Peptides were eluted at a flow rate of 1 ml/min using the gradient indicated by the dashed line. Solvent B was 0.1 % trifluoroacetic acid in acetonitrile. Absorbance range, 1.0 full scale. Each peak was subjected to amino acid and N-terminal analysis. Tryptic peptides are given the prefix T and are numbered consecutively in their predicted order in the amino acid sequence [13]. T48-49m, T50m, and T52-53m denote the variant peptides resulting from the amino acid substitutions in albumins Trieste, Milano Slow and Liprizzi, respectively.

in a pure form (Fig. 1, lane 3) as described for the variant CB5 from Milano Slow, because it eluted later than the normal from the Vydac C18 column. The tryptic profile on a Vydac C18 column showed the absence of peptides T52 and T53 and the presence of an additional peptide (T52-53m) (Fig. 3C). Protein sequencing of the T52-T53m peptide indicated an Arg→Cys change at amino acid 410 (Fig. 4). The exons coding for CB5 were examined by HDX analysis, which indicated a mutation in exon 11. DNA sequencing of both strands of this exon showed a C→T transition at nucleotide 13251 [13], corresponding to the Arg410→Cys substitution (Fig. 5 B). SDS/PAGE electrophoresis with and without dithiothreitol (not shown) seemed to exclude a gross conformational

332


Fig. 4. Amino acid sequence analysis of variant peptides from albumins Trieste, Milano Slow, and Liprizzi. The tryptic (T) and V8 protease (S) peptide designation are defined in the text and the V8 protease and tryptic peptide profiles are given in Figs 2 and 3, respectively. The amino acid substitutions in the variant peptides are shown in boldface letters.

Fig. 5. Sequence ladder for albumins Milano Slow (A) and Liprizzi (B). The underlinings indicate the single nucleotide change G→C at position 11 969 in the genomic sequence [13] and the corresponding Asp375→His amino acid exchange for albumin Milano Slow (A), and the single nucleotide change C→T at position 13 251 [13] and the corresponding Arg410→Cys amino acid change for albumin Liprizzi (B).

change due to an alteration in the intrachain disulfide bonds caused by the presence of the additional Cys residue. DISCUSSION Distribution of mutations within the albumin molecule. The substitutions that have so far been identified are not randomly distributed within the albumin molecule; rather, most of them are clustered in three regions of the polypeptide chain: the propeptide and amino terminus, the C-terminal region of subdomain IIB (residues 3132382), and the C-terminal region of

subdomain IIIB (residues 4792580) [2, 5, 7, 8]. With the exception of the Ala320→Thr and the Arg218→His substitutions, all the known albumin mutations are located on the molecular surface and are exposed to the solvent [2]. In addition, eight among them coincide with proposed antigenic sites which represent about a third of the surface of the albumin molecule [1, 5]. The mutations of albumins Trieste and Milano Slow are in keeping with this finding. The mutation of albumin Trieste is in subdomain IIB, near the C-terminal of helix h9(II) [2], and that of Milano Slow is in the N-terminal region of helix h10(II)-h1(III) in the connecting segment from subdomain IIB to subdomain IIIA [2]. Both are located in the segment 3542382, which contains the highest concentration of known substitutions [10]. This region includes the C-terminal portion of helix h9(II), the loop connecting h9(II) and h10(II) and the N-terminal region of the long helix h10(II)-h1(III) and appears to be particularly exposed to the solvent. In addition, the mutation of albumin Trieste lies adjacent to one of the five sites proposed as an antigenic determinant (regions 3102316 1 3612364, two noncontigous segments in the short loop 6) [16]. Only a few mutations have been reported in subdomains IB, IIA, and IIIA. A possible reason for this rarity, as shown by the crystallographic data, is that subdomains IA, IB and IIA form a very compact structure, as do subdomains IIB, IIIA and IIIB [2]. In particular subdomains IB and IIIA, representing the core of each albumin moiety, appear to be deeply buried inside the molecule [2]. In addition, the low mutation frequency of subdomains IIA and IIIA is clearly related to the fact that the helical bundles of those two subdomains form the two hydrophobic binding cavities of the molecule [2]. Albumin Liprizzi represents the first case of a mutation in the binding pocket in subdomain IIIA, corresponding to Sudlow’s Site II. This region is the most active site in the molecule, preferentially binding many ligands, such as : tryptophan, digitoxin, ibuprofen, and azidothymidine [2]. Arg410 is located in one of the three major helices (h2) that line this hydrophobic cavity as is the corresponding Arg218 in the more specialized binding pocket in subdomain IIA, corresponding to Sudlow’s Site I [2]. Mutagenesis studies show that Arg218 exerts an unfavorable steric effect on the binding of thyroxine [17, 18]. Therefore, the Arg218→His mutation found in the inherited condition of familial dysalbuminemic hyperthyroxinemia creates a new high-affinity binding site for thyroxine in this cavity, in addition to the purported one in subdomain IIIA, and gives rise to a variant that binds thyroxine with abnormally high affinity [3, 4, 17, 18]. Arg410 also plays a crucial role for binding, as it primarily interacts with the carboxylate group of the test ligand 2,3,5-triiodobenzoic acid [2]. Therefore, the substitution of the positively charged and bulky guanidyl group of Arg410 with a small thiol group might alter the binding properties of albumin Liprizzi considerably. Moreover, the esterase activity of the variant might be impaired because the unusually high reactivity of the phenolic group of Tyr411 towards nucleophilic substitutions ˚ ) to the nitrogens of is probably due to its close proximity (2.7 A Arg410 [2]. Alloalbumins with mutations involving Cys residues. Three mutations involving Cys residues have been previously identified in the mature albumin molecule: albumin Hawkes Bay (Cys177→Phe) [19], albumin Asola (Tyr140→Cys) [8] and the C-terminal variant albumin Bazzano [7]. In all cases the mutation appears to affect disulfide bonding and reduces the amount of the circulating variant. The substitution of Cys177 in albumin Hawkes Bay abolishes the disulfide bridge with Cys168 which instead binds to nearby Cys124; the subsequent conformational change causes the molecule to be unstable, so that only 5% of


the variant form persists in plasma [19]. The new Cys140 in albumin Asola creates an 18th S-S bond in the molecule between this new Cys and the single thiol group of albumin, Cys34, and the conformational change reduces the variant to 25245% of the total albumin [8]. In albumin Bazzano Cys567 is absent so the final S-S bond cannot be formed and only 18% of the variant circulates [7]. The instability of the variants with altered disulfide bridges indicates that the integrity of the 17 S-S bonds is crucial for albumin’s folding and catabolism and supports the hypothesis that S-S interchanges make the molecule more effectively degraded by uptake into endocytotic vesicles [1]. The mutation of albumin Liprizzi introduces an additional Cys residue which is buried in the hydrophobic binding pocket in subdomain IIIA, far from a disulfide bridge and from Cys34. Therefore, the presence of the new Cys does not alter the disulfide pattern or cause gross conformational changes. Consequently, the stability of the molecule is unaffected and the variant represents about 50% of the total albumin content. Nucleotide mutations in the albumin gene. Of the 55 point mutations thus far reported, including the three described in the present paper, 37 are transitions (34 purines and 3 pyrimidines) and 18 transversions (15 purines and 3 pyrimidines). The total number of substituted bases are as follows: 34 G (61.82%), 15 A (27.27 %), 5 C (9.09 %), and 1 T (1.82 %); 49 purines (89.09%), and 6 pyrimidines (10.91 %). The high frequency of mutations involving purines, particularly guanine, may reflect either a hypermutability of these bases, or simply the fact that purine transitions often result in double charge substitutions in the mutant protein which can thus be easily detected electrophoretically. Albumin Liprizzi represents a new finding of a mutation at a hypermutable CpG dinucleotide site. The substitution of Milano Slow is the second mutation at residue 375, the other being Asp375→Asn in albumin Nagasaki-2, found in Japan [20]. Both are caused by a change of G at position 11969 which occurs in a CpG dinucleotide sequence [13], but only the latter reflects a C→T transition in the non-coding strand; 28 CpG dinucleotides are present in the region of the gene coding for the mature albumin molecule [13]. The theoretical point mutations caused by C→T transitions in both the sense and the antisense strands at these sites would produce 19 charged, 14 neutral, and 18 silent amino acid substitutions, 2 stop codons, and 3 splicing defects. Of these, 8 point mutants, and 1 stop codon in a case of analbuminemia, have been thus far identified [1, 5] (and present paper). Since alloalbuminemia is a benign trait and since only charged mutations can be identified by electrophoretic screening, these cases correspond to about half of those one can expect to detect. It has been previously reported that the hypermutability of CpG dinucleotides in the propeptide-encoding sequence results in a series of recurring proalbumin variants [21]. Our findings show that CpG dinucleotide hypermutability is also responsible for a substantial fraction of the defects in the human albumin gene coding for the mature protein. This work was supported by grants to M. G. and L. M. from the Ministero della Universita` e della Ricerca Scientifica e Tecnologica (Rome, Italy) and from Regione Lombardia (progetto 723, area 5.2.5), by grants to U. K.-H. from the Danish Medical Research Council and Fonden af 1870, and by a Collaborative Research grant of the North Atlantic Treaty Organization (CRG910029) to F. W. Putnam and by National Institutes of Health grant DK19221 to F. W. Putnam. The authors wish to thank Dr C. Petrini and G. Ferrari for kindly supplying the Milano Slow and Trieste sera, respectively. Appreciation is also expressed to Dr A. Cobianchi (Centro Grandi Strumenti, Universita` di Pavia) for sequence analysis and to A. Mortara and M. Bellaviti for expert technical assistance.

333

REFERENCES 1. Peters, T. Jr (1996) All about albumin : biochemistry, genetics and medical applications, Academic Press, San Diego CA. 2. Carter, D. C. & Ho, J. X. (1994) Structure of serum albumin, Adv. Protein Chem. 45, 1532203. 3. Sunthormthepvarakul, T., Angkeow, P., Weiss, R. E., Hayashi, Y. & Refetoff, S. (1994) An identical missense mutation in the albumin gene results in familial dysalbuminemic hyperthyroxinemia in 8 unrelated families, Biochem. Biophys. Res. Commun. 202, 7812 787. 4. Petersen, C. E., Scottolini, A. G., Cody, L. R., Mandel, M., Reimer, R. & Bhagavan, N. V. (1994) A point mutation in the human serum albumin gene results in familial dysalbuminaemic hyperthyroxinaemia, J. Med. Genet. 31, 3552359. 5. Minchiotti, L., Watkins, S., Madison, J., Putnam, F. W., KraghHansen, U., Amoresano, A., Pucci, P., Cesati, R. & Galliano, M. (1997) Structural characterization of four genetic variants of human serum albumin associated with alloalbuminemia in Italy, Eur. J. Biochem. 247, 4762482. 6. Kragh-Hansen, U., Pedersen, A. O., Galliano, M., Minchiotti, L., Brennan, S. O., Ta`rnoky, A. L., Franco, M. H. L. P. & Salzano, F. M. (1996) High-affinity binding of laurate to naturally occurring mutants of human serum albumin and proalbumin, Biochem. J. 320, 9112916. 7. Madison, J., Galliano, M., Watkins, S., Minchiotti, L., Porta, F., Rossi, A. & Putnam, F. W. (1994) Genetic variants of human serum albumin in Italy : point mutants and a carboxyl-terminal variant, Proc. Natl Acad. Sci. USA 91, 647626480. 8. Minchiotti, L., Galliano, M., Kragh-Hansen, U., Watkins, S., Madison, J. & Putnam, F. W. (1995) A genetic variant of albumin (albumin Asola ; Tyr140→Cys) with no free -SH group but with an additional disulfide bridge, Eur. J. Biochem. 228, 1552 159. 9. Brown, J. R. (1976) Structural origin of mammalian albumin, Fed. Proc. 35, 214122144. 10. Madison, J., Arai, K., Sakamoto, Y., Feld, R. D., Kyle, R. A., Watkins, S., Davis, E., Matsuda, Y.-i., Amaki, I. & Putnam F. W. (1991) Genetic variants of serum albumin in Americans and Japanese, Proc. Natl Acad. Sci. USA 88, 985329857. 11. Sakamoto, Y., Kitamura, K., Madison, J., Watkins, S., Laurell, C. B., Nomura, M., Higashiyama, T. & Putnam, F. W. (1995) Structural study of the glycosylated and unglycosylated forms of a genetic variant of human serum albumin (63 Asp→Asn), Biochim. Biophys. Acta 1252, 2092216. 12. Galliano, M., Minchiotti, L., Iadarola, P. & Porta, F. (1986) Screening of CNBr peptides from genetic variants of human serum albumin, Protides Biol. Fluids Proc. Colloq. 34, 8152 818. 13. Minghetti, P. P., Ruffner, D. E., Kuang, W. J., Dennison, O. E., Hawkins, J. W., Beattie, W. G. & Dugaiczyk, A. (1986) Molecular structure of the human albumin gene is revealed by nucleotide sequence within q11-22 of chromosome 4, J. Biol. Chem. 261, 674726757. 14. Watkins, S., Madison, J., Davis, E., Sakamoto, Y., Galliano, M., Minchiotti, L. & Putnam, F. W. (1991) A donor splice mutation and a single-base deletion produce two carboxyl-terminal variants of human serum albumin, Proc. Natl Acad. Sci. USA 88, 59592 5963. 15. Watkins, S., Madison, J., Galliano, M., Minchiotti, L. & Putnam, F. W. (1994) A nucleotide insertion and frameshift cause analbuminemia in an Italian family, Proc. Natl Acad. Sci. USA 91, 22752 2279. 16. Sakata, S. & Atassi, M. Z. (1980) Immunochemistry of serum albumin. X. Five major antigenic sites of human serum albumin are extrapolated from bovine albumin and confirmed by synthetic peptides, Mol. Immunol. 17, 1392142. 17. Petersen, C. E., Ha, C.-E., Jameson, D. M. & Bhagavan, N. V. (1996) Mutations in a specific human serum albumin thyroxine binding site define the structural basis of familial dysalbuminemic hyperthyroxinemia, J. Biol. Chem. 271, 19 110219 117. 18. Petersen, C. E., Ha, C.-E., Haroalli, K., Park, D. & Bhagavan, N. V. (1997) Mutagenesis studies of thyroxine binding to human serum

334


albumin define an important structural characteristic of subdomain 2A, Biochemistry 36, 701227017. 19. Brennan, S. O. & Fellowes, A. P. (1993) Albumin Hawkes Bay: a low level variant caused by loss of a sulphydryl group at position 177, Biochim. Biophys. Acta 1182, 46250. 20. Takahashi, N., Takahashi, Y., Isobe, T., Putnam, F. W., Fujita, M., Satoh, C. & Neel, J. V. (1987) Amino acid substitutions in inher-

ited albumin variants from Amerindian and Japanese populations, Proc. Natl Acad. Sci. USA 84, 800128005. 21. Brennan, S. O., Arai, K., Madison, J., Laurell, C. B., Galliano, M., Watkins, S., Peach, R., Myles, T., George, P. & Putnam, F. W. (1990) Hypermutability of CpG dinucleotides in the propeptideencoding sequence of the human albumin gene, Proc. Natl Acad. Sci. USA 87, 390923913.