335
biotopics they determine function. The obscrvntions on EPO suggest that the chain branching ofgLycans is at Least partly amino acid sequenceand structure-dependent although it is not possible to identify the signah responsible, at tbe level of protein structure. ??Second: sugan have a geat deal of influence upon both spec;fic (reccptor-mediated) and non-specific (physicochcmical) properties in viva. EPO and t-PA glycosylation variants tiavc widely differing plasma halflifes as a result of the masking or unmasking of terminal sugars which are recognized by various rxepton. EPO may undergo another type oi sprcitic recognition in uiut~which is a function of a i%nction of-the degree of glycan branching. Since the gLycan core itself is common to all .Wlinked glycans, it would not bc expected to play a specific role in the same sense. The implications for the production of therapeutic proteins are apparent. Terminal sugars and chain branching intluence both biological
activity and plasma half-life. Both are determined by the choice of host cciis but chain branching may also depend upon protein sequence. Careful selection and culrurc of host cells for proteins destined frrr &I r&o application is therefore crucial. Being able to control the glycan srructures present on plasm3 glycoproteirn~ although greatly desired, remains an intriguing, but unsolved multimillion-dollar problem.
Human serum albumin structure sdved
The three-dimensional molecular structure of human serum dbumin (HSA) has been solved and interpreted at atomic resolutioni. Scicntists’ reactions to this may well have differed widely. Some will have been relatively indiifiirent to information about a molecule that lacks ‘exciting’ bioiogirai function. Many were probably surprised that the atomic co-ordinates of such a well-known and abundant human protein did not already reside in the protein structure databanks. Most wcuid have been interested that a key drug-binding plasma protein has at last revealed its r~+clrlar xcrc6. rtiiir those who have worked with this serum protein ovfr the years, the article represents the end oTone era and marks the br~nning of a new period of technc-iogicai and therapeutic possibilities, now opened up by the availability of the high-resolution molxular model
Pieces ofthe puzzle
The background to studies ofaLbumin structure is fixioating. The first cluestothe stNcNrecamefionl the early work of J. Foster2. On the basis of physicochrmical and binding studies of albumin, there was a need to explain how so many different sorts ofmolecule could bind to albumin. Not only could albumin accommodate molecules as diverse as fatty acids and aspirin, but also they appeared to occupy the same sites and to a&ct the structure of the protein. At a time when there was no sequence information avarlable, Foster suggested that albumin ws. composed of four ‘domains’ which could rearrange to generate a range of binding sires or cavities (Fig. 1). Later, in 1975, the group5 of Meloun3 and Brown3 independend~ reported the complete amino acid sequence of HSA: Brown’s group Bo reported the sequence ofbovine
serum aIbur&r (USA). lc is a measure of how iast biochemical technolo;~ is moving tu recall that an amino acrd sequcxing project of this size (X5 at residuesj was a major achievement the time. the HSA sequence taking six years to complete. The amino acid sequences of both HSA and BSA contained evidence for Foster’s domain hypothesis, and revealed the prcxence of tandem repeats of three homologous regions, each cf\sbic!: contained further evidence of gene duplication. On rhc basis of the internal sequence horoology. Brown proposed 1 three-domain structure for serum albmnin and drscribcd a pbvsicat model in which ~:1ch domain consisted of two homobgous subdomains~. It was awumd that ligmds ryould not bind in the interfaces betwr0r domains. as suggested by Foster. but \vould bin-l in cavities within each domain. Va’ariations MI 0 &crrrc Subseqxntly. the albumin cDNA xx cloncds and patented by Gcnentech and other Ciotcchnofng)-
__I__--._-._ 0 1992. ElsevlgiSciencePubbshers Ltd(IN)
TBiEcti
Ixxl@ER
-_ 1992 f&43_ 10:
336
biotopics 17 disulphide bridges of HSA was originally predicted by &own on stereochemical grout&, and remained formally untested until the recent solution of the crystal strutture. &so! ing the ambiguity in linkage of disutphidc bonds, where two or mos: cyst&es are immediately adjacent in sequence, remains a diffictitt task, because there is no selective way of cleaving the pepcide bond betweeh thein. The disulphidc bridge assignment of Brown was confirmed by the crystal structure. The diverse nature of ligands bound by albumin and the presence of mulGplc sites is accounted for by just two binding
Figure 1 ’ Human serum albumin. The aminMenninus_of the molecule is on the right. Two of the homologous domiins form a compact structure on the right; the third domain is more sxtanded, from top left to the bottom of the structure. The bo multi-ligand binding sites are formed by partsof domains 2 and Z, and are shown here rjccupied by triiodobenzoic acid (white). The 17 disulohide bridges are shown in red. (Illuslration courtesy ot Dr Daniel Carter, Space Science Laboratory, Gabama, USA.) companies.
The companies
naturally
had their eyes upon the potential of recombinant albumin as a rheraprv tic product, probably as a rcplaccmcnt for human plasma albumin as a plasma cxpandcr in the management ofshock from fluid km. lnh$uin~y, the published patents disagreed upon the exact amino acid sequence. Academic groups later published both cl?NA and genomic scqucnccs (also with small variations): that of Dugaiczyk7 now being accepted as the most representative of HSA in the normal population! The diffcrenccs in the original cDNA sequences might conceivably have represented HSA polymorphisms because many natural variants are now k0i’WL I’\ :IU~~IIKXof rcse;lrch groupa, ~UictCd
hy
th
iWIltK~
established
l&d
Se~t!~IlW,
tha*, isolated
serum albumin fiagmetrrs (including domains 41-~dsubdumains) retlined tire :dzility tc bind phpsi&g&l ligat& of HSA such :LS htty-acids, bilirubin and tryptoph;riW, ‘rhis reinlorccd rhc domain theory of -~lhu*lrin,but despite the InosL:cxrensive studies, 111which a l;irgt umber tri I-ISA ligands hsd br:en studied in depth’, thi structural details of albuhave
‘fIRTECH OCTWFR 1992
(VOL101
min-binding sites remained &sive. For ncarty ‘20 years, protein crystallographic laboratories around the world have at one time or aunther attcmptrd to obtain crystals of HSA suitable for diffraction. but the protcin had resisted all such a&t.mpts cvrn a concerted recent effort by the doyen of blood proteins, Max Perutz (pers. commun.). Finally, the structure has yicldcd - not through novel tcchnolob~, but through ttz:: protein crystalloglphr+s two conventional strategies - patience and p.%stence. Predictions patterns
and emerging
The three-dimcnsionzl structurr of HSA verifies the dom.Gn hypothesis’. As predicted, rhc protein consists of six homologous subdomains which assemble to fom a heartshaped molecule in which the Nand C-terminal potiior; s arc nearer in space than originally l-elicveds. The main secondary structure present in WS.A is the a-h&, and HSA, tog&er with il3 hornoh~~~~~‘:relatives !ti-t&protein *md vitamin tlbinding protein), has be-n shown to represent a new mult;gcn:nr:family by virtue of the unique three-dimensional fold. The linkage pattcm cf’thc
cavjtics in the scc-
ond and third albumin domains. It finally turned out that ligands arc botmd within subdomains - a possibility not forcscen by the original modelers”. A longer-than-average polypeptidc chain in the first subdomain has apparently allowed a mclecular rearrangement which has closed a putative binding cavity in this regir,n. It was already known from studies ofisolarcd albumin fiagmeets that the N-terminal portion of the molecule did not bind HSA,ligand@. Differences in the side chains present in the second and third binding sites account for thr observed different ligand selectivities of these sites. A survey of a large number of physiological and non-physiological (drug) ligan&Findicates that there are either one or two strong binding sites on HSA, in accordance with the crystallographic evidence. The evotutionary solution, if that is the appropriate term, for binding a cariety ofdiffering organic ions (ranting from bjlii-tibin to palmitic acid) appears to be the provisiun of binding cavities with a variety of potential interactions, each accommodat-. ing a different stereochemistry. No doubt those with molecular graphics systems XL already designing or rcde&ming ligarlds to fit the cavities presented by this muiti-purpose macromolecule. Implications for biotechnology A number of opportunities sh. uld now arise km the publication ofthe
albumin structure. ‘The finr is implied in the ? L.r~rrt~ paper in that the structure
ol’ a roconrbinant
vcr-
sion of HSA was compared with the human plasma-derived ! rotein’. Since HSA is art establish& human therapeutic product, the crystal scrucmre rrprescnts an important step forward in the development of a
337
bio topics therapeutic equivalent derived from a non-human so Jrce. This is because it provides a sta: dard by which such products can ue assessed. Other opportunities TV utilize albumin are also mentioned in the paper. Albumin is well tolerated by humans: it is non-glycosylatcd and has a reasonably long half-life in plasma of about 17 days. Knowing the disposition of amino acids, both in the binding cavities and upon the surface, should allow albumin to be ‘m-engineered for novel drug-binding or macromolecule-canier properties. Such engineered proteins might be used, fat example, for detoxification or slow drug release.
tein scientists to elucidate the threedimensional structure invofved HSA joining a number of high-proiile, high-flying peptider - such as HIV reverse transcriptase - as a passenger on Space Lab. Although HSA samples on the First US Microgravity laboratory remained refractory to ctystallization, and the recent Nature paper’ represents the culmination of almost seven years earthbound research, &&action data from crystals of the tet-agonal form grown on the First Internationr‘ ’ iicrogravity Laboratory in January 1992 have proven the best so fat (D. Carter, perscomnum.)1 A small step for protein, but a giant leap forward foi biomedical science!
In summary, solution of the HSA ::tructure has satisfied the intellectual curiosity of some, but may benefit many more through development as a therapeutic molecule or a medical ‘device’. The determination of pro-
References 1 He, M. H. andCaner. D. C. (I 992) Nilur~
Meloon, B.. Mowck, 1.. and Kostb, V. (1975) l%!%srerr. sn, 21.~2137 Bchrms, P. 0.. Spickerman.A. M. and Ilrown,J. K. (1975) &f. PNc 34.591 Brown,J. K. (1977) in .#fwniw Sfwt~~. fhcliotl dud Uses (KOWKW~, V. M., Oratz, M. andRothschild,M. A., rdr), pp. 27-51,
Pcrgmnn IJrms Lawn. K. M.. Adelman,J., Bock. S. C.. Franke, A. E., Ho&., C. M., Najarian, R. C., Seeburg,Y. H. and Wion, K. I,. (1981) Nu&,i ui,Kes.O. 6103-6114 Dugaiczyk,A., Law. .%W. and Dennison,
3%. 209-219
Engineering kzymes for chemoenzymatic synthesis Part I: practical routes to aza-sugars and complex carbohydrates With various recombinant-DNA and Protein engineering techniques now available, enzyme-based technologies are emerging as a practical new route for the large-scale synthesis ofchiral intermediates and bioactive molecule:;. One class of molecules of particular interest in this respect is complex carbohydrates, oligosaccharides. their cot@gates and related molecules. This article describes the syntheses of novel monosaccharides 2nd aza-sugars using recombinant aldolases, and of complex oligosaccharides using recombinant glycosylttansferase~ coupled of sugar with is sibt regeneratioii nucleotidcs. Many enzymes are now available for the stereospecific synthesis of chiral synthons’. Recently, attention has been directed towards the development of more effective and stable enzymes for the synthesis of mol-
ecula ofincreasing complexityz. One class of such complex molecules is composed ofcarbohydrat
biotopics they determine function. The obscrvntions on EPO suggest that the chain branching ofgLycans is at Least partly amino acid sequenceand structure-dependent although it is not possible to identify the signah responsible, at tbe level of protein structure. ??Second: sugan have a geat deal of influence upon both spec;fic (reccptor-mediated) and non-specific (physicochcmical) properties in viva. EPO and t-PA glycosylation variants tiavc widely differing plasma halflifes as a result of the masking or unmasking of terminal sugars which are recognized by various rxepton. EPO may undergo another type oi sprcitic recognition in uiut~which is a function of a i%nction of-the degree of glycan branching. Since the gLycan core itself is common to all .Wlinked glycans, it would not bc expected to play a specific role in the same sense. The implications for the production of therapeutic proteins are apparent. Terminal sugars and chain branching intluence both biological
activity and plasma half-life. Both are determined by the choice of host cciis but chain branching may also depend upon protein sequence. Careful selection and culrurc of host cells for proteins destined frrr &I r&o application is therefore crucial. Being able to control the glycan srructures present on plasm3 glycoproteirn~ although greatly desired, remains an intriguing, but unsolved multimillion-dollar problem.
Human serum albumin structure sdved
The three-dimensional molecular structure of human serum dbumin (HSA) has been solved and interpreted at atomic resolutioni. Scicntists’ reactions to this may well have differed widely. Some will have been relatively indiifiirent to information about a molecule that lacks ‘exciting’ bioiogirai function. Many were probably surprised that the atomic co-ordinates of such a well-known and abundant human protein did not already reside in the protein structure databanks. Most wcuid have been interested that a key drug-binding plasma protein has at last revealed its r~+clrlar xcrc6. rtiiir those who have worked with this serum protein ovfr the years, the article represents the end oTone era and marks the br~nning of a new period of technc-iogicai and therapeutic possibilities, now opened up by the availability of the high-resolution molxular model
Pieces ofthe puzzle
The background to studies ofaLbumin structure is fixioating. The first cluestothe stNcNrecamefionl the early work of J. Foster2. On the basis of physicochrmical and binding studies of albumin, there was a need to explain how so many different sorts ofmolecule could bind to albumin. Not only could albumin accommodate molecules as diverse as fatty acids and aspirin, but also they appeared to occupy the same sites and to a&ct the structure of the protein. At a time when there was no sequence information avarlable, Foster suggested that albumin ws. composed of four ‘domains’ which could rearrange to generate a range of binding sires or cavities (Fig. 1). Later, in 1975, the group5 of Meloun3 and Brown3 independend~ reported the complete amino acid sequence of HSA: Brown’s group Bo reported the sequence ofbovine
serum aIbur&r (USA). lc is a measure of how iast biochemical technolo;~ is moving tu recall that an amino acrd sequcxing project of this size (X5 at residuesj was a major achievement the time. the HSA sequence taking six years to complete. The amino acid sequences of both HSA and BSA contained evidence for Foster’s domain hypothesis, and revealed the prcxence of tandem repeats of three homologous regions, each cf\sbic!: contained further evidence of gene duplication. On rhc basis of the internal sequence horoology. Brown proposed 1 three-domain structure for serum albmnin and drscribcd a pbvsicat model in which ~:1ch domain consisted of two homobgous subdomains~. It was awumd that ligmds ryould not bind in the interfaces betwr0r domains. as suggested by Foster. but \vould bin-l in cavities within each domain. Va’ariations MI 0 &crrrc Subseqxntly. the albumin cDNA xx cloncds and patented by Gcnentech and other Ciotcchnofng)-
__I__--._-._ 0 1992. ElsevlgiSciencePubbshers Ltd(IN)
TBiEcti
Ixxl@ER
-_ 1992 f&43_ 10:
336
biotopics 17 disulphide bridges of HSA was originally predicted by &own on stereochemical grout&, and remained formally untested until the recent solution of the crystal strutture. &so! ing the ambiguity in linkage of disutphidc bonds, where two or mos: cyst&es are immediately adjacent in sequence, remains a diffictitt task, because there is no selective way of cleaving the pepcide bond betweeh thein. The disulphidc bridge assignment of Brown was confirmed by the crystal structure. The diverse nature of ligands bound by albumin and the presence of mulGplc sites is accounted for by just two binding
Figure 1 ’ Human serum albumin. The aminMenninus_of the molecule is on the right. Two of the homologous domiins form a compact structure on the right; the third domain is more sxtanded, from top left to the bottom of the structure. The bo multi-ligand binding sites are formed by partsof domains 2 and Z, and are shown here rjccupied by triiodobenzoic acid (white). The 17 disulohide bridges are shown in red. (Illuslration courtesy ot Dr Daniel Carter, Space Science Laboratory, Gabama, USA.) companies.
The companies
naturally
had their eyes upon the potential of recombinant albumin as a rheraprv tic product, probably as a rcplaccmcnt for human plasma albumin as a plasma cxpandcr in the management ofshock from fluid km. lnh$uin~y, the published patents disagreed upon the exact amino acid sequence. Academic groups later published both cl?NA and genomic scqucnccs (also with small variations): that of Dugaiczyk7 now being accepted as the most representative of HSA in the normal population! The diffcrenccs in the original cDNA sequences might conceivably have represented HSA polymorphisms because many natural variants are now k0i’WL I’\ :IU~~IIKXof rcse;lrch groupa, ~UictCd
hy
th
iWIltK~
established
l&d
Se~t!~IlW,
tha*, isolated
serum albumin fiagmetrrs (including domains 41-~dsubdumains) retlined tire :dzility tc bind phpsi&g&l ligat& of HSA such :LS htty-acids, bilirubin and tryptoph;riW, ‘rhis reinlorccd rhc domain theory of -~lhu*lrin,but despite the InosL:cxrensive studies, 111which a l;irgt umber tri I-ISA ligands hsd br:en studied in depth’, thi structural details of albuhave
‘fIRTECH OCTWFR 1992
(VOL101
min-binding sites remained &sive. For ncarty ‘20 years, protein crystallographic laboratories around the world have at one time or aunther attcmptrd to obtain crystals of HSA suitable for diffraction. but the protcin had resisted all such a&t.mpts cvrn a concerted recent effort by the doyen of blood proteins, Max Perutz (pers. commun.). Finally, the structure has yicldcd - not through novel tcchnolob~, but through ttz:: protein crystalloglphr+s two conventional strategies - patience and p.%stence. Predictions patterns
and emerging
The three-dimcnsionzl structurr of HSA verifies the dom.Gn hypothesis’. As predicted, rhc protein consists of six homologous subdomains which assemble to fom a heartshaped molecule in which the Nand C-terminal potiior; s arc nearer in space than originally l-elicveds. The main secondary structure present in WS.A is the a-h&, and HSA, tog&er with il3 hornoh~~~~~‘:relatives !ti-t&protein *md vitamin tlbinding protein), has be-n shown to represent a new mult;gcn:nr:family by virtue of the unique three-dimensional fold. The linkage pattcm cf’thc
cavjtics in the scc-
ond and third albumin domains. It finally turned out that ligands arc botmd within subdomains - a possibility not forcscen by the original modelers”. A longer-than-average polypeptidc chain in the first subdomain has apparently allowed a mclecular rearrangement which has closed a putative binding cavity in this regir,n. It was already known from studies ofisolarcd albumin fiagmeets that the N-terminal portion of the molecule did not bind HSA,ligand@. Differences in the side chains present in the second and third binding sites account for thr observed different ligand selectivities of these sites. A survey of a large number of physiological and non-physiological (drug) ligan&Findicates that there are either one or two strong binding sites on HSA, in accordance with the crystallographic evidence. The evotutionary solution, if that is the appropriate term, for binding a cariety ofdiffering organic ions (ranting from bjlii-tibin to palmitic acid) appears to be the provisiun of binding cavities with a variety of potential interactions, each accommodat-. ing a different stereochemistry. No doubt those with molecular graphics systems XL already designing or rcde&ming ligarlds to fit the cavities presented by this muiti-purpose macromolecule. Implications for biotechnology A number of opportunities sh. uld now arise km the publication ofthe
albumin structure. ‘The finr is implied in the ? L.r~rrt~ paper in that the structure
ol’ a roconrbinant
vcr-
sion of HSA was compared with the human plasma-derived ! rotein’. Since HSA is art establish& human therapeutic product, the crystal scrucmre rrprescnts an important step forward in the development of a
337
bio topics therapeutic equivalent derived from a non-human so Jrce. This is because it provides a sta: dard by which such products can ue assessed. Other opportunities TV utilize albumin are also mentioned in the paper. Albumin is well tolerated by humans: it is non-glycosylatcd and has a reasonably long half-life in plasma of about 17 days. Knowing the disposition of amino acids, both in the binding cavities and upon the surface, should allow albumin to be ‘m-engineered for novel drug-binding or macromolecule-canier properties. Such engineered proteins might be used, fat example, for detoxification or slow drug release.
tein scientists to elucidate the threedimensional structure invofved HSA joining a number of high-proiile, high-flying peptider - such as HIV reverse transcriptase - as a passenger on Space Lab. Although HSA samples on the First US Microgravity laboratory remained refractory to ctystallization, and the recent Nature paper’ represents the culmination of almost seven years earthbound research, &&action data from crystals of the tet-agonal form grown on the First Internationr‘ ’ iicrogravity Laboratory in January 1992 have proven the best so fat (D. Carter, perscomnum.)1 A small step for protein, but a giant leap forward foi biomedical science!
In summary, solution of the HSA ::tructure has satisfied the intellectual curiosity of some, but may benefit many more through development as a therapeutic molecule or a medical ‘device’. The determination of pro-
References 1 He, M. H. andCaner. D. C. (I 992) Nilur~
Meloon, B.. Mowck, 1.. and Kostb, V. (1975) l%!%srerr. sn, 21.~2137 Bchrms, P. 0.. Spickerman.A. M. and Ilrown,J. K. (1975) &f. PNc 34.591 Brown,J. K. (1977) in .#fwniw Sfwt~~. fhcliotl dud Uses (KOWKW~, V. M., Oratz, M. andRothschild,M. A., rdr), pp. 27-51,
Pcrgmnn IJrms Lawn. K. M.. Adelman,J., Bock. S. C.. Franke, A. E., Ho&., C. M., Najarian, R. C., Seeburg,Y. H. and Wion, K. I,. (1981) Nu&,i ui,Kes.O. 6103-6114 Dugaiczyk,A., Law. .%W. and Dennison,
3%. 209-219
Engineering kzymes for chemoenzymatic synthesis Part I: practical routes to aza-sugars and complex carbohydrates With various recombinant-DNA and Protein engineering techniques now available, enzyme-based technologies are emerging as a practical new route for the large-scale synthesis ofchiral intermediates and bioactive molecule:;. One class of molecules of particular interest in this respect is complex carbohydrates, oligosaccharides. their cot@gates and related molecules. This article describes the syntheses of novel monosaccharides 2nd aza-sugars using recombinant aldolases, and of complex oligosaccharides using recombinant glycosylttansferase~ coupled of sugar with is sibt regeneratioii nucleotidcs. Many enzymes are now available for the stereospecific synthesis of chiral synthons’. Recently, attention has been directed towards the development of more effective and stable enzymes for the synthesis of mol-
ecula ofincreasing complexityz. One class of such complex molecules is composed ofcarbohydrat
