Structure, function and flexibility of human lactoferrin Edward N. Baker*, Bryan F. Anderson, Heather M. Baker, M. Haridas, Geoffrey B. Jameson, Gillian E. Norris, Sylvia V. Rumball and Clyde A. Smith Department o f Chemistry and Biochemistry, Masse)' University, Palmerston North, New Zealand X-ray structure analyses of/bur different jbrms of human lactoferrin ( diferric, dieupric, an oxalate-substituted dicupric, and apo-lactoJerrin ), and o[" bovine di/erric lactoferrin, have revealed various ways in which the protein structure adapts to different structural and Junctional states. Comparison o f diferric and dicupric lactoferrins has shown that different metals can, through slight variations in the metal position, have difJerent stereochemistries and anion coordination without an)' significant change in the protein structure. Substitution of oxalate Jor carbonate, as seen in the structure of a hybrid dicupric complex with oxalate in one site and carbonate in the other, shows that larger anions can be accommodated by small side-chain movements in the binding site. The multidomain nature of lactoJerrin also allows rigid body movements. Comparison o f human and bovine lactoJerrins, and o f these with rabbit serum transJerrin, shows that the relative orientations 0[ the two lobes in each molecule can var.v ; these variations may contribute to differences in their binding properties. The structure of apo-lactojerrin demonstrates the importance o f large-scale domain movements .for metal binding and release and suggests that in solution an equilibrium exists between open and closed forms, with the open.[brm being the active binding species. These structural jorms are shown to be similar to those seen for bacterial periplasmie binding proteins, and lead to a common model Jbr the various steps in the binding process.

Ke),words: Lactoferrin;X-ray crystallography;protein structure; stereochemistry

Introduction Lactoferrin is an iron-binding protein which, together with the homologous serum transferrin, serves to control iron levels in body fluids by sequestering and solubilizing ferric iron 1'2. Its presence in leucocytes 3, and in many exocrine secretions 4 (e.g. milk, saliva, tears, mucosal and genital secretions), together with its ability to bind to a wide variety of cells 5, has further been associated with other postulated functions 2. These include roles in the immune and inflammatory responses s, as an antibacterial agent 6, and as a growth factor 7. The molecular properties of lactoferrin are well established, through a variety of chemical and physical studies 1'2, amino acid sequence determinations 8'9 and high resolution X-ray crystallographic analyses of several different structural forms 1°-~3. Our aim is to determine the molecular basis for the remarkable binding properties of iactoferrin, including ( 1) the extremely tight ( K ~ 102°) but reversible binding of two Fe 3+ ions per molecule, (2) the absolute requirement for a bound anion (normally CO 2-) with each Fe 3÷ ion, (3) the acceptance by lactoferrin of a wide variety of other metal ions and anions in place of Fe 3+ and CO32-, (4) the characteristic differences between different lactoferrins and between lactoferrin and transferrin 1'2, and (5) the unexpected

*To whom correspondenceshould be addressed. Presented at '1990 IUPAB Satellite Congress', 7 l0 August 1990, Palmerston North, New Zealand 0141 8130/91/030122-08 ~C) 1991 Butterworth-HeinemannLimited 122 Int. J. Biol. Macromol., 1991, Vol. 13, June

structural and functional similarities L~ relating lactoferrin (and other transferrins) to a group of bacterial periplasmic binding proteins involved in active transport and chemotaxis 15,16. A key feature of the structure and function oflactoferrin is the importance of flexibility, at various levels, ranging from small variations in the positions occupied by different metals and anions, to large-scale rigid body movements of entire domains. Here we describe the structure of human lactoferrin, the differences found in the various structural forms and the relevance of these to mechanisms of binding and release.

Experimental Protein purification and crystallization

Human lactoferrin was isolated from fresh colostrum as described previously aT. Under these conditions, with metal ion contamination rigorously excluded during the purification, the resulting protein typically had a residual iron content of some 8 10% saturation, as estimated from the spectral ratio A 2 8 o / A 4 6 6 . This material was used for crystallization of apo-lactoferrin in both its native and deglycosylated forms ~. For the latter an endoglycosidase preparation from Flavobacterium meningosepticum was used for the carbohydrate removal. Diferric and dicupric lactoferrins were prepared by addition of 2 equivalents of ferric nitrilotriacetate and cupric chloride respectively. For the preparation of the hybrid complex of dicupric

Human lactoferrin: E. N. Baker et al. Table 1 Crystal data a for different forms of lactoferrin Form of lactoferrin

Human apo Human apo deglycosylated Human diferric b Human dicupric b Human dicupric-oxalate c Bovine diferric b

a

b

(A)

c (•)

Space group

Reference

(A) 222.0 152.1 156.3 155.9 155.9 138.4

115.6 94.6 97.4 97.0 97.1 87.1

77.8 55.8 55.85 56.0 56.2 73.6

P2t2x21 P212121 P212x21 P212121 P212121 P212121

17 17 18 13 13 19

aAll crystal forms orthorhombicwith ct = fl = 7 = 90° bWith carbonate as the anion in both sites cWith oxalate in one anion site and carbonate in the other

lactoferrin with oxalate in one site and carbonate in the other, a 50-molar excess of sodium oxalate was added to previously prepared dicupric lactoferrin, displacing carbonate from one site 13. All these forms of human lactoferrin were crystallized by dialysis under similar conditions of low ionic strength, at pH 7.8, with small amounts (typically 10% v/v) of added alcohoP 3'18. Bovine lactoferrin was prepared from bovine colostrum and crystallized as before19, except that further purification by isoelectric focusing of the protein sample was necessary for reproducible crystallization. Crystal data for these various forms are summarized in Table 1.

Structure analyses The three-dimensional structure of diferric lactoferrin, Fe2Lf, was determined by multiple isomorphous replacement at 3.2 A resolution, from diffractometer data t°. The structure of apo-lactoferrin, apoLf, was solved by molecular replacement at 2.8 A resolution 12, using the Fe2Lf structure as the search model. The crystals of dicupric lactoferrin, Cu2Lf, and the hybrid oxalatecarbonate complex of dicupric lactoferrin, Cu2oxLf, are isomorphous with those of Fe2Lf; thus the Fe2Lf structure could be used as the starting model for the refinement of these structures. The 2 Fe 3 + and 2 CO 2ions, the amino acid side-chains involved in metal ligation, and all solvent molecules were, however, removed from this model prior to refinement, to avoid bias; the corresponding groups in Cu2Lf and Cu2oxLf were then included as they appeared in subsequent

electron density maps. All of these structures have been refined by restrained least squares methods, using synchrotron data to 2.2 A, 2.0 A and 2.1 /~, for Fe 2 Lf, apoLf and Cu2Lf respectively and diffractometer data to 2.5 A resolution for Cu2oxLf. The current state of each of these refinements is summarized in Table 2. In the case of bovine lactoferrin, diffractometer data to 4.5/~ resolution have been collected and the structure solved by molecular replacement, again using the human Fe2Lf structure as search model. The current crystallographic R factor, 0.18, after rigid body and restrained least squares refinement, and the appearance of density for the iron atoms (which had been omitted from the search model) indicate the correctness of the solution. The bovine lactoferrin sequence 2° has been used to include side-chains which are similar or identical to those in human lactoferrin, but at this resolution no detail can be drawn from the model.

Results and discussion

Structure of lactoferrin The structure is most easily described in terms of that of diferric lactoferrin 1°'11. The molecule consists of a single polypeptide chain of 691 residues, which is folded into two globular lobes, representing the N-terminal and C-terminal halves (the N-lobe and C-lobe). The two lobes have similar sequences (approx. 40 % sequence identity) and three-dimensional structures 'r.m.s. deviation 1.0

Table 2 Refinement details for lactoferrin Structure

Protein atoms

Ions

Water molecules

Ra

Resolution (/~)

r.m.s, dev.b (~)

apoLf Fe2Lf

5342 5322

1 Cl 2 Fe 3÷

406 431

0.181 0.179

2.0 2.2

0.020 0.018

Cu2Lf

5331

2 Cu 2÷

172

0.208

2.1

0.015

-

0.196

2.5

0.023

-

0.180

4.5

0.021

2 co~ 2 co3~ Cu2oxLf

5323

2 Cu 2÷

1 co~1 C20~Bovine Lfc

4780

2

Fe 3+

Crystallographic R-factor, R = Y IIFoI - IF¢ II/E IFol br.m.s. = root-mean square deviation from standard bond lengths cDiferric bovine lactoferrin

lnt I Riol Macxornnl. 1991_ Vol. 13. June

123

Human lactoferrin." E. N. Baker et al.

~

2

Figure 1 Stereo ribbon diagram for diferric-lactoferrin, showing the organization of the molecule. The N-lobe is above, the C-lobe below. Also indicated are the four domains (N1, N2, C1, C2), the connecting helix (H) and the region of hydrophobic contacts between the lobes (C). The iron atoms are shown by filled circles and the glycosylation sites by stars. The interdomain 'backbone' strands in each lobe can be seen behind the iron atoms

for 85 % of Co atoms after superposition), indicative of a gene duplication. Each lobe contains one specific metal binding site in a deep cleft between two dissimilar domains. The molecule as a whole thus comprises four domains, N ! and N2 in the N-lobe, and C1 and C2 in the C-lobe (see Figure 1). The two iron binding sites are extremely similar. The metal is coordinated by four protein side-chains, i.e. 2 Tyr, 1 Asp and 1 His (Tyr 92 and 192, Asp 60 and His 253 in the N-lobe and the corresponding Tyr 435 and 528, Asp 395 and His 597 in the C-lobe). Two oxygens from the bidentate carbonate ion complete a distorted octahedral geometry round the metal. As well as providing two ligands to the iron atom, the carbonate ion occupies a pocket formed by two positively charged groups, an arginine side-chain (Arg 121 in the N-lobe, and Arg 465 in the C-lobe) and the N-terminus of helix 5, making good hydrogen bonds with both ~1. This arrangement is shown schematically in Figure 2. Other important features of the binding region are (1) a substantial water-filled cavity, beyond the Arg side-chain

but within the interdomain cleft, and (2) several other basic side-chains in the vicinity of the iron site.

Effects of metal and anion substitution Comparison of the Fe2Lf and Cu2Lf structures shows the effect of substituting Fe 3÷ by Cu 2+. The protein structure is essentially undisturbed (r.m.s. deviation between C o positions for Cu2Lf and Fe2Lf being 0.25 A), the only significant differences being in the positions of the metal ions. In the N-lobe the Cu atom is displaced 0.7 A from the iron position in a direction away from Tyr 92; this also causes the anion to become monodentate, and the copper geometry to be square pyramidal. In the C-lobe the Cu atom is displaced only 0.2 A from the iron position and remains 6-coordinate, though with a slightly different geometry. Thus different metal ion stereochemistries and variations in anion coordination can be accommodated without any change in the protein structure, and in particular without any change in the closure of the two domains over the metal. This may not be true of ions as large as lanthanides, however, which

AsP

HIS 253 (

/

597~ 0' ~ ~

....

. . H4"N'~ H

TYR~ 92

(435)

Figure 2 Schematic diagram of the metal and anion binding site in lactoferrin. Residue numbers are for the N-lobe (with corresponding residues in the C-lobe given in parentheses)

124

Int. J. Biol. Macromol., 1991, Vol. 13, June

Human lactoferrin." E. N. Baker et al.

Figure 3 Stereo diagram showing the copper and carbonate binding in the C-lobe of dicupric-lactoferrin (solid lines). The positions taken by the oxalate ion and the two displaced side-chains (Arg 465 and Tyr 398), when oxalate is substituted for carbonate in Cu2oxLf, are shown with broken lines

have radii ~ 0.3 A larger than Fe 3 +, and this presumably explains their weaker binding. Adjustment of the structure to accept anions larger than carbonate is illustrated by the Cu2oxLf complex, which has a carbonate ion in the N-lobe site and oxalate in the C-lobe. The oxalate ion binds in 1,2-bidentate fashion, as predicted from model-building 21, with one carboxylate group hydrogen bonded, like carbonate, to the helix N-terminus (Figure 3). The extra bulk of the oxalate ion requires that the Arg 465 side-chain move some 2 A away from the metal site. This, in turn, displaces the side-chain of the adjacent Tyr 398, but these movements can easily be accommodated within the large, solvent-filled cavity in the interdomain cleft, without any other change in the protein structure. Thus, this large internal cavity apparently allows sufficient flexibility, in the side-chains around the binding site, for the accommodation of larger anions than carbonate.

Domain flexibility in lactoferrin Relative movement of domains is a common feature of multi-domain proteins and is frequently crucial for biological function 22. Examples include the flexible hinges between immunoglobulin domains 22, which have been shown to allow relative domain movements of up to 55 °, and the helix-mediated domain closures seen in a number of enzymes 23, whereby a bound substrate is enclosed. The two-lobe, four-domain structure of lactoferrin allows considerable scope for flexibility, which could include relative movement of the N- and C-lobes, or of the two domains which comprise each lobe. Variations in the orientations of the two lobes have already been noted in comparing diferric human lactoferrin, Fe2Lf, with diferric rabbit transferrin, Fe2Tf24; superposition of the C-lobe on to the N-lobe requires a rotation of 180 ° coupled with a translation of 25 ,~, for Fe2Lf, but a rotation of 167 ° and translation of 25 A for Fe2Tf, implying a difference of 13 ° in the lobe orientations between the two proteins. Comparison of the human and bovine diferric lactoferrin structures has now also shown a difference in the relative orientations of the two lobes, this time approximately 10 °. These relative movements are possible because there is a 'cushion' of mainly hydrophobic side-chains between the two lobes, TM, and because the single covalent connection between them lies on the outside of this contact area (see Figure 1). The

functional significance, if any, of these lobe movements is unknown, but they clearly could be important for receptor binding if both lobes are simultaneously involved. They could also influence metal binding properties since the hinge region (see later), of the N-lobe at least, is in the vicinity of the inter-lobe contact region. Changes in lobe contacts could thus alter the relative stabilities of the open and closed apo- and holo- forms (see below). Relative movements of the two domains of each lobe have a much more obvious functional relevance because of the fact that the metal and its associated anion are bound deep in the interdomain cleft. The polypeptide chain folding in Fe2 Lf suggested that this cleft could be opened via a hinge in two 'backbone strands' of extended chain which run behind the metal site, connecting the two domains 14. The Fe2Lf structure also offers indirect evidence for such movements 1~. Superposition of the N- and C-lobes shows that the corresponding individual domains superimpose better than do the whole lobes (r.m.s. deviations 1.06 A for domain C1 on domain N1, 1.06 A for C2 on N2, and 1.32 A for the entire C-lobe on the N-lobe). This results from the fact that the C-lobe is more closed than the N-lobe, the difference being a relative domain rotation of 6 °. The most dramatic evidence of domain movements, however, is seen in the N-lobe of apo-lactoferrin ~2. The relationships between the domains when apoLf is compared with Fe2Lf are summarized in Table 3, and the two structures are shown in Figure 4. There are small differences ( ~ 8 °) between the orientations of the N1 domain relative to the C-lobe domains, presumably as a Table 3 Domain relationships between Fe2Lf and apoLf~

apoLf

N1 N2 C1 C2

N1

Fe2Lf N2

0.45/~ -

54.2° 0.51 /~ -

C!

C2

8.6° 55.3° 0.35 A

7.9 ° 54.1 ° 1.3° 0.40 A

' The diagonal elements(N1 vs N1 etc.) give the r.m.s, deviation in C~ positions when the equivalent domains of Fe2Lf and apoLf are superimposed. The off-diagonal elements show the relative angular rotations between pairs of domains in the two structures.

Int. J. Biol. Macromol., 1991, Vol. 13, June

125

Human lactoferrin." E. N. Baker et al. The nature of the conformational change in the N-lobe has elements both of the simple hinge seen in some immunoglobulins and proposed from the Fe2 Lf structure, and of the helix-mediated domain closures of enzymes mentioned above. The two extended 'backbone' strands do show a quite abrupt hinge at about residues His 91 and Val 250 (Figure 5). This is achieved with generally small changes in the main chain torsion angles, the major differences being 45 ° in ~(90), 26 ° in q5(91 ), 20 ° in q5(250) and 24 ° in ~9(250). No Gly residues are involved, and no change in the hydrogen bonding pattern occurs; it is a simple flexing of an antiparallel /%ribbon. The second feature of the domain movement, illustrated in Figure 6, is that it involves two helices, helices 5 and 11, which lie across each other at an angle of ~ 75 °. Helix 11 (residues 321-332), which provides a third crossover between the domains, does not move significantly during the conformational change, remaining associated with domain NI. Helix 5 (residues 121 136), which is centrally involved in the organization of domain N2, and which has the anion-binding site at its N-terminus, pivots on helix 11 and carries the rest of domain N2 with it; the rotation axis lies approximately along helix 11. Some repacking of the helix contacts appears to occur and there are several specific changes in hydrogen bonding. In Fe 2Lf hydrogen bonds link the side-chains of Thr 122 and Asn 126, near the N-terminus of helix 5, with Tyr 324, at the N-terminus of helix 11. In apoLf these hydrogen bonds are broken but a new salt bridge links Arg 133, at the C-terminus of helix 5, to Glu 335, just following the C-terminus of helix 11. These hydrogen bonds may help to stabilize the two structures. Figure 4 C, plots of diferric-lactoferrin (above) and apolactoferrin (below). Note the open N-lobe and closed C-lobe in apo-Lf result of some readjustment in the interlobe contact area, but the most striking change is the 54 ° rotation of the N2 domain relative to the N1 domain. This causes the N-lobe binding cleft to be opened wide. That this is a rigid body movement is shown by the very close correspondence when the individual domains of apoLf and Fe2Lf are superimposed (Table 3); the r.m.s. deviation is only 0.45 A for N1 and 0.51 A for N2, for all C~ atoms.

2~::.

Functional implications of the apo-lactoferrin structure The apo-lactoferrin structure shows two striking features when compared with Fe2Lf (Figure 4), the large scale conformational change in the N-lobe, described above, and the fact that the C-lobe binding cleft is closed even though no metal is bound. These different metal-free configurations seen in the two lobes offer important clues about the processes of binding and release. The open N-lobe cleft suggests a sequence of steps in the binding process. A consequence of the ~/fl folding pattern is that a number of helix N-termini line the walls of the binding cleft. These, together with several basic side-chains (Arg 121, Arg 210, Lys 296, Lys 301) should provide a positive potential which attracts the CO3z- ion 257

":"

c~

' ~,~

" 75

251

6

Figure 5 Stereo diagram showing the divergence of the two 'backbone' polypeptide strands in apo-lactoferrin (solid lines) compared with metal-substituted lactoferrins (open lines). The hinge point is at about residues 91 and 250. The hydrogen bonding pattern (dotted lines) is unchanged by the movement

126

Int. J. Biol. Macromol., 1991, Vol. 13, June

Human lactoferrin: E. N. Baker et al. N2

5

,

hi1

N1

Figure 6 Schematic diagram of the N-lobe of human lactoferrin showing the change from the 'open' form of apo-lactoferrin (left) to the 'closed' form of iron-lactoferrin (right). Helices are shown as cylinders, fl-strands as arrows. In the conformational change helix 5 in domain N2 appears to pivot on helix 11. The hinge point in the 'backbone' fl-strands is shown with an arrow

into the open cleft, to bind to the specific pocket provided by Arg 121 and the N-terminus of helix 5. Chemical and kinetic arguments 25'26 do in fact suggest that the anion binds first. With the CO ] - ion bound, four of the six iron ligands would then be in place on domain N2, i.e. the two CO 2- oxygens together with Tyr 92 and Tyr 192, which in the open structure remain with domain N2. Thus, we anticipate that metal binding initially occurs on domain N2, which then rotates to enable the metal to complete its coordination, to Asp 60 and His 253, some 8 - 9 A away on domain N1. Support for this sequence of events also comes from the spectroscopic detection of quaternary protein-anion-metal-chelate complexes when the metal ion is added as a chelate complex 26. Such intermediates should be located on domain N2, and the chelate ligand should then be displaced as the conformational change occurs and the two remaining protein ligands bind. The closed C-lobe presents a most intriguing contrast to the open N-lobe. Its structure is hardly changed from that in Fe2Lf (Figure 7) despite the absence of a metal ion. There is no significant difference in the extent of

domain closure (difference only 1.3°; see Table 3) and the r.m.s, deviation in C~ positions between the two structures is essentially the same whether the whole lobe is superimposed (r.m.s. deviation 0.44 A) or the domains are superimposed individually (r.m.s. deviation 0.35 A for C1, 0.40 A for C2). Several possible explanations for the closure of the C-lobe can be ruled out. Firstly, the electron density gives no indication of a metal ion or anion which might hold the two domains together. A C1- ion is found in the specific anion site but there it interacts only with one domain (C2). Secondly, there appears to be nothing intrinsic to the structure which would prevent both domains being open simultaneously. The C2 domain does not contact any part of the N-lobe even when a rotation of 54 ° analogous to that of the N2 domain is applied. This modelled open C-lobe structure does, however, highlight one important difference between the N-lobe and the C-lobe. In the latter a disulphide bridge 483-677 (disulphide 7 in the nomenclature of Williams 2v) links the C-terminal ends of helices 465-481 and 664-678 (Figure 7). These two helices are equivalent to those

§

Figure 7 Superposition of the C-lobes of diferric lactoferrin ~open lines) and apolactoferrin (solid lines), demonstrating that there is essentially no change, except in one very flexible loop, seen below the binding cleft. The helices 5 and 11 in each lobe are indicated, and the disulphide 7, which is believed to inhibit opening of the C-lobe, is also shown

Int. J. Biol. Macromol., 1991, Vol. 13, June

127

Human lactoferrin." E. N. Baker et al. involved in the N-lobe domain rotation (5 and 11). Although disulphide bridges are conformationally flexible, they have a restricted number of preferred geometries, and their C, ... C, distance falls into a relatively narrow range, 5.5 to 7 A. If a domain rotation exactly equivalent to that in the N-lobe is applied, the C~ ... C, distance of disulphide 7 is stretched to 8.5 A. This implies that this disulphide, which has no equivalent in the N-lobe, must restrict domain movements in the C-lobe, and probably does not allow a conformational change as large or as facile as that in the N-lobe. This is consistent with the known tighter iron binding and slower iron release by the C-lobe 2'28, and with solution studies which suggest it is less flexible 28. Despite the restriction imposed by disulphide 7, it seems very unlikely that it would completely prevent opening of the C-lobe, though the domain movement is likely to be of lesser magnitude. Our conclusion is that for a flexible, multidomain protein such as this, a number of structures are possible in solution, and in particular that an equilibrium exists between the open and closed forms of apo-lactoferrin, with the open form being the active binding species. The 'one open, one closed' form seen in the crystal structure may then be selected by crystal packing requirements, which preferentially 'freeze-out' one of the several structures. If this is so, the energy difference between open and closed forms must be very slight since the crystal packing contacts are weak and few in number (from the N-lobe only 9 atoms out of more than 2500 make direct contacts < 3.4 A with neighbouring molecules, while for the C-lobe only 11 such contacts exist). Furthermore, it implies that if a closed form is relatively stable even in the absence of iron, the additional energy from metal binding must largely account for the great stability of the iron-lactoferrin complex and of other metal-substituted transferrins.

Comparison with other binding proteins The various structural forms identified for lactoferrin have close parallels in the bacterial periplasmic binding proteins whose structures have been characterized by Quiocho and co-workers 15. The latter proteins, which bind small molecules and ions, are important in active transport and chemotaxis 16. Structurally they resemble a single lobe of lactoferrin with respect to size, domain structure, polypeptide chain folding and the location of the binding site in the interdomain cleft x4. For these proteins, too, binding and release has been shown to involve a hinge-bending conformational change 29. Taking the two groups of proteins together, the structural forms determined crystallographically, and their significance for binding and release, are illustrated diagrammatically in Figure 8. They fit the following scenario. In solution the ligand-free protein exists in an equilibrium between open and closed forms. The open unliganded form is exemplified by the N-lobe of apo-lactoferrin and by the unliganded Leu/Ile/Valbinding protein 3°, while the closed unliganded form is represented by the C-lobe of apo-lactoferrin. Binding occurs to the open form to give an intermediate in which the ligand is only bound to one domain. This has been seen crystallographically in a complex made by diffusing leucine into crystals of the Leu/Ile/Val-binding protein 3°. Finally closure of the domains occurs to allow the ligand to bind to both domains, giving the closed, liganded

128

Int. J. Biol. Macromol., 1991, Vol. 13, June

IL

IL

Figure 8 Schematic representation of the structural changes associated with binding and release by transferrins and by bacterial periplasmic binding proteins. For the unliganded form (left) an equilibrium is presumed to exist between open and closed states. Binding is to the open form, giving an intermediate in which the ligand is bound to one domain only; domain closure then gives the fully-closed liganded form (lower right) structure seen for both lobes of metal-substituted lactoferrins and for the arabinose-, galactose- and sulphate-binding proteins ~5.

Conclusions The structural results described here demonstrate several levels of flexibility in the lactoferrin molecule, which can be presumed to apply also to other members of the transferrin family, given their close structural similarity. The complexes in which Cu 2÷ and oxalate are substituted for Fe 3÷ and carbonate demonstrate that there is sufficient flexibility in and around the binding site for other metal ions and anions to be accommodated without any disturbance of the stable closed structure, which is seen unchanged in all three complexes. The stability of this structure is further emphasized, for the C-lobe at least, by its presence, essentially unchanged, in the apo-lactoferrin structure, despite the absence of a bound metal. Thus metals of similar charge and size to Fe 3 + should be bound almost equally well, with just small variations in their positions, within the same structure. Transferrins should be good carriers for these metal ions. Likewise, the structure of the oxalate complex shows that the large internal cavity between the domains, by allowing side-chain movements, allows the substitution of other, larger, anions for carbonate. Clearly, however, the metal ion and anion are synergistic in their requirements, and the preferred binding mode for a given anion may be more or less compatible with a given cation. Also, very large cations (e.g. lanthanides) and anions may no longer be compatible with the same stable closed structure; this would explain their weaker binding, or lack of binding, to one or both sites in some cases. The rigid body domain and lobe movements seen in lactoferrin itself (from comparison of the apo- and

Human

holo-structures), as well as in the comparison of lactoferrin with other transferrins, are not unexpected for a multidomain protein. The domain movements lead to a structural model for the binding process which fits chemical and kinetic observations and matches, in a very striking way, conclusions drawn for bacterial binding proteins. The question of release remains a puzzle. Iron release is triggered by reduced pH (below pH 6.0 for transferrin, below 4.0 for lactoferrin 31) and is influenced also by the binding of non-synergistic anions in secondary sites 32, and by ionic strength 33. We speculate that these factors act to destabilize the 'closed' state, perhaps through environmental changes around the hinge region, and thus favour opening of the binding cleft, and subsequent metal release. In fact the apo-lactoferrin structure described here may give an illustration of this. Metal release by transferrin extrapolates to zero at low ionic strength 33, implying that the open structure is progressively less favoured as the ionic strength is reduced. Assuming that this applies also to lactoferrin, the closed C-lobe seen in the apo-lactoferrin crystal may arise because under the low ionic strength crystallization conditions, a closed structure is favoured, even when no metal is present. The flexibility of domain and lobe orientations is also likely to influence receptor interactions. The discrimination by transferrin receptors between the liganded and unliganded forms of the protein 34 presumably implies that the receptor contacts at least two different domains, whose relative orientations are altered by metal binding. Likewise conformational changes in the receptor, as are found to occur at the lower intraceUular pH 35, may help to promote iron release by perturbing the bound transferrin domains. Discrimination by transferrin receptors between transferrin and lactoferrin 36 and between different transferrins may also result from a 'two-domain' recognition, bearing in mind the differences in lobe orientations seen for lactoferrin and transferrin and for human and bovine lactoferrins. There is no doubt that flexibility plays a key functional role for members of this family of proteins, and more structural variations are to be expected as more crystal structures become available.

References 1 2 3 4 5 6 7 8 9 l0 ll 12 13

We gratefully acknowledge financial support from the U S National Institutes of Health (grant no. HD-20859), the Medical Research Council of New Zealand, and the New Zealand Dairy Research Institute. We are also grateful for the support of colleagues at Massey University, and for the many stimulating discussions we have had with the transferrin research group at Birkbeck College, London.

Aisen, P. and Listowsky, I. Annu. Rev. Biochem. 1980, 49, 357 Brock, J. H. in 'Metalloproteins' (ed. P. Harrison), MacMillan Press, London, Part 2, 1985, p 183 Baggiolini, M., De Duve, C., Masson, P. L. and Heremans, J. F. J. Exp. Med. 1970, 131, 559 Masson, P. L., Heremans, J. F. and Dive, C. Clin. Chim. Acta 1966, 14, 735 Birgens, H. S., Hansen, N. E., Karle, H. and Ostergaard Kristensen, L. Br. J. Haematol. 1983, 54, 383 Arnold, R. R., Russell, J. E., Champion, W. J., Brewer, M. and Gauthier, J. J. Infect. lmmunol. 1982, 35, 792 Hashizume, S., Kuroda, K. and Murakami, M. Biochim. Biophys. Acta 1983, 763, 377 Metz-Boutigue, M-H., Jolles, J., Mazurier, J., Schoentgen, F., Legrand, D., Spik, G. and Jolles, P. Eur. J. Biochem. 1984, 145, 659 Pentecost, B. T. and Teng, C. T. J. Biol. Chem. 1987, 262, 10134 Anderson, B. F., Baker, H. M., Dodson, E. J., Norris, G. E., Rumball, S. V., Waters, J. M. and Baker, E. N. Proc. Natl. Acad. Sci. USA 84, 1769 Anderson, B. F., Baker, H. M., Norris, G. E., Rice, D. W. and Baker, E. N. J. Mol. Biol. 1989, 209, 711 Anderson, B. F., Baker, H. M., Norris, G. E., Rumball, S. V. and Baker, E. N. Nature (London) 1990, 344, 784 Smith, C. A., Baker, H. M. and Baker, E. N. J. Mol. Biol. (in

press) 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30

Acknowledgements

l a c t o f e r r i n . " E . N . B a k e r et al.

31 32 33 34 35 36

Baker, E. N., Rumball, S. V. and Anderson, B. F. Trends Biochem. Sci. 1987, 12, 350 Quiocho, F. A. Phil. Trans. Roy. Soc. Lond. 1990, B326, 341 Adams, M. D. and Oxender, D. L. J. Biol. Chem. 1989, 264, 15739 Norris, G. E., Baker, H. M. and Baker, E. N. J. Mol. Biol. 1989, 209, 329 Baker, E. N. and Rumball, S. V. J. Mol. Biol. 1977, 111, 207 Norris, G. E., Anderson, B. F., Baker, E. N., Baker, H. M., G~irtner, A. L., Ward, J. and Rumball, S. V. J. Mol. Biol. 1986, 191, 143 Mead, P. E. and Tweedie, J. W. Nucleic Acids Res. 1990, 18_ 7167 Baker, E. N., Anderson, B. F., Baker, H. M., Haridas, M., Norris, G. E., Rumball, S. V. and Smith, C. A. Pure Appl. Chem. 1990, 62, 1067 Bennett, W. S. and Huber, R. CRC Crit. Rev. Biochem. 1984, 15, 291 Lesk, A. M. and Chothia, C. J. Mol. Biol. 1984, 174, 175 Bailey, S., Evans, R. W., Garratt, R. C., Gorinsky, B., Hasnain, S., Horsburgh, C., Jhoti, H., Lindley, P. F., Mydin, A., Sarra, R. and Watson, J. L. Biochemistry 1988, 27, 5804 Kojima, N. and Bates, G. W. J. Biol. Chem. 1981, 256, 12034 Cowart, R. E., Kojima, N. and Bates, G. W. J. Biol. Chem. 1982, 257, 7560 Williams, J. Trends Biochem. Sci. 1982, 7, 394 Kretchmar, S. A. and Raymond, K. N. J. Am. Chem. Soc. 1986, 108, 6212 Mao, B., Pear, M. R., McCammon, J. A. and Quiocho, F. A. J. Biol. Chem. 1982, 257, 1131 Sack, J. S., Saper, M. A. and Quiocho, F. A. J. Mol. Biol. 1989, 206, 171 Mazurier, J. and Spik, G. Biochim. Biophys. Acta 1980, 629, 399 Williams, J., Chasteen, N. D. and Moreton, K. Biochem. J. 1982, 201,527 Kretchmar, S. A. and Raymond, K. N. Inorg. Chem. 1988, 27, 1436 Jandl, J. H. and Katz, J. H. J. Clin. Invest. 1963, 42, 314 Turkewitz, A. P., Schwartz, A. L. and Harrison, S. C. J. Biol. Chem. 1988, 263, 16309 van Bockxmeer, F. M. and Morgan, E. H. Comp. Biochem. Physiol. 1982, 71A, 211

Int. J. Biol. M a c r o m o l . , 1991, Vol. 13, J u n e

129