If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
β-Lactoglobulin (β-LG) is the major whey protein of ruminant species and is also present in the milks of many, but not all, other species. Its amino-acid sequence and 3-dimensional structure show that it is a lipocalin, a widely diverse family, most of which bind small hydrophobic ligands and thus may act as specific transporters, as does serum retinol binding protein. Bovine β-LG binds a wide range of ligands, but this may not be the reason for its presence in milk. In reviewing the structure and physicochemical properties of the protein, we present the structures of the ligands cholesterol (at a resolution of 2.0Å, R = 0.221; Rfree = 0.295) and vitamin D2 (at a resolution of 2.4Å, R = 0.212; Rfree = 0.297) each bound to the central binding cavity of bovine β-LG at pH 7.3. Neither ligand is fully visible in the electron density maps, and the less well-ordered regions are the polar end groups at the mouth of the binding site. In a separate experiment, a mercury ion was bound to the free Cys121 (at a resolution of 2.2Å, R = 0.218; Rfree = 0.288) in a way that transmitted a small structural change through Asp137 via Arg148 to the dimer interface. It is not clear if the known dissociation that arises from the reaction of β-LG with HgCl2 results from this perturbation.
In reviewing the structural studies that reveal the ligand binding sites for long-chain fatty acids, retinoids, and steroids, only the central location, common to all lipocalins so far examined, is occupied under the conditions used. We find that there is no crystallographic evidence of another ligand binding site in our crystals grown in approximately 1.3 M citrate, although low ionic strength studies in solution indicate the possible presence of at least one other low affinity site. The apparent ability of the binding site to accommodate a wide range of ligands may point to a possible physiological function. However, by considering the lipocalin family in general, and the species distribution of β-LG in particular, some speculation as to the physiological function can be made. β-Lactoglobulin has been reported as being implicated, inter alia, in hydrophobic ligand transport and uptake, enzyme regulation, and the neonatal acquisition of passive immunity. However, these functions do not appear to be consistent between species. Sequence comparisons among members of the lipocalin family reveal that glycodelin, found in the human endometrium during early pregnancy, is the most closely related to β-LG. Although the function of glycodelin is also unknown, it appears to have effects on the immune system and/or to be involved in differentiation. It is proposed that β-LG, over-expressed in the lactating mammary gland of many, but not all, species, is primarily an important source of amino acids for the offspring of those animals that produce it, but that this function arose by gene duplication from the physiologically essential glycodelin. The other functions that have been associated with β-LG in the neonate are, therefore, fortuitous.
). It has been reported as being present in many (perhaps most) but not all, mammalian species with rodent, human, and lagomorph milks being notable by its absence. Whereas genetic variants of a single gene have been widely reported, particularly in the cow, several nonruminant species, such as the horse and the cat, have at least two distinct genes. Unlike the other main whey protein, α-LA, which has a definite enzyme-modulation function in lactose synthase, β-LG does not apparently have any such identifiable function. The protein family to which β-LG belongs, the lipocalins (
), presents a widely diverse series of functions, most of which involve some ligand-binding function, and it has been speculated that a similar function must be the physiological reason for the significant quantities of β-LG found in milk (the domestic cow produces ∼2 to 3 g L−1, dependent to varying extents on the breed, diet, and stage of lactation).
have dealt in some detail with the aggregation and denaturation phenomena associated with milk processing. It is the purpose of the present paper to examine some of the molecular properties of the protein under more or less physiological conditions as they relate to the ligand-binding properties. In most of what will be described, it is the bovine protein to which reference is made. Finally, we shall consider the physiological function of the protein in light of these ligand-binding properties.
β-Lactoglobulin is a small protein, soluble in dilute salt solution as befits a globulin, with 162 amino acid residues (Mr ∼18,400) that fold up into an 8-stranded, antiparallel β-barrel with a 3-turn α-helix on the outer surface and a ninth β-strand flanking the first strand (see Figure 1). It is this strand that forms a significant part of the dimer interface in the bovine and ovine proteins but, while still present in porcine β-LG, is not involved in the formation of the dimer that forms at low pH. The genetic variants referred to above in the ruminant species result in relatively minor amino acid differences, but the processing properties appear to be significantly affected even by these small changes, such that consideration has been given to removing the less favorable variants by selective breeding (
). The so-called calyx, or β-barrel, is conical and is made of β-strands A-D forming one sheet, and strands E-H forming a second. Strand A bends through a right angle such that the C-terminal end forms an antiparallel strand with H; strands D and E also form a less significant interaction completely closing the calyx. It is this central cavity, the calyx, that provides the ligand-binding site. On the outer surface of the β-barrel, between strands G and H, is the 3-turn α-helix. The loops that connect the β-strands at the closed end of the calyx, BC, DE, and FG are generally quite short, whereas those at the open end are significantly longer and more flexible.
In particular, the EF loop acts as a gate over the binding site. At low pH, it is in the “closed” position, and binding is inhibited or impossible, whereas at high pH it is open, allowing ligands to penetrate into the hydrophobic binding site. The “latch” for this gate is Glu89, the residue implicated in the Tanford transition observed, although not identified, as having an abnormally high pKa (
The free Cys121, with its reactive thiol group, has a pH-dependent activity that parallels that of the Tanford transition, and its involvement in the denaturation and aggregation behavior is now firmly established (
). The residue is situated on the outer surface of the β-barrel on strand H, under the α-helix, and quite some distance from the EF loop. The effect of this is to mask its accessibility to solvent, particularly at low pH. Reaction with heavy metals like Hg2+ or Cd2+ does occur at most pH values and leads to a dissociation of the dimer. Indeed, reaction of the Cys with almost any thiol reagent appears to enhance dissociation into monomers (
The CD and GH loops in many of the crystallographic analyses are indistinct, and the similarly indistinct nature of the extended AB loop led to the original misthreading of the amino acid sequence into the first electron density map (
). β-Strand I is on the opposite side of strand A from strand H and forms an antiparallel interaction with the same strand of the second monomer. The other dimer-stabilizing interaction comes from the AB loop, where there is an ion pair from Asp33 of one subunit to Arg40 of the other. It is interesting to note that although strand I also exists in the porcine and equine β-LG structures, albeit with a slightly modified sequence, as do the Asp and Arg residues, no similar dimerization occurs at physiological pH. Indeed, site-directed mutagenesis designed to disrupt one or the other of these interactions, or better, both, produces a monomeric bovine protein (
; Wilson and Holt, to be published). There is no direct interaction between the α-helixes, which are antiparallel on the outer surface of the molecule with the polypeptide chains, some 12 Å apart. The closest approach of any atoms of the helix in the A subunit from the triclinic lattice X structure to the helix in the B subunit is 5.8Å, and this involves the flexible side chains of Asp137A and Arg148B and vice versa.
The structure of β-LG has also been determined by NMR techniques by several groups (
). The bovine protein is monomeric at pH values below 3, and the structure has been determined at this pH. Although there is no matching crystal structure at this pH where flat hexagonal plates of poor diffraction quality can be grown, the structure of the core of the protein is well conserved. However, significant differences do exist, and these have been critically reviewed (
) and even before the fortuitous finding of the retinol-β-LG complex during a study of retinol-binding protein (RBP), to which it later emerged that β-LG is related, the binding of hydrophobic ligands to the protein was well established (
). Since then, many other ligands have been observed to bind to the protein by a variety of methods, including equilibrium dialysis, affinity chromatography, electron spin resonance spectroscopy, spectrophotometry, and perhaps the most popular, perturbation of the intrinsic Trp fluorescence (
The central, or main, ligand-binding site seems to accommodate linear molecules like fatty acids and also retinol with the cyclohexenyl ring system inside. It is clear that the lining of the pocket is very hydrophobic, and although we do not see any well-defined water or endogenous ligand molecules in the native protein as isolated from milk, it seems improbable that the calyx is actually empty; rather, there is only poorly ordered solvent, and/or a low occupancy of fatty acids (
), that cannot be detected above the noise in our experimental electron density. If all the molecules that can bind are bound within the cavity, then there is scope for binding nutritionally or medically useful molecules, either directly or after some modification to the protein's binding site (
). However, such speculation is outside the scope of this article, although it is to the nature of the ligand binding site(s) of β-LG that this paper now turns, relying mainly on direct, crystallographic evidence.
Materials and Methods
Preparation of the Complexes
Protein, ligands, and other chemicals were purchased from Sigma (St. Louis, MO). The general chemicals were of Analar or greater purity. A 20 mg/mL solution in 20 mM Tris pH 8.0 of β-LG A was made to contain sufficient ligand stock solution (cholesterol, vitamin D2: 50 mM ligand in ethanol) to give a molar ratio of one part protein to two parts ligand. The mixture was incubated for 2 h at 40°C.
Crystallization and Data Collection
A solution of the β-LG-ligand complex was crystallized at 17°C by the hanging drop method using 1.25 M sodium citrate, 0.1 M HEPES, pH 7.3, as the precipitant. Each hanging drop contained 8 μL of β-LG-ligand complex solution added to 8 μL of the reservoir solution. After approximately 7 d at 17°C, rhombohedral crystals of lattice Z (P3221) had grown. For the mercury derivative, several crystals of the native protein of 0.2 to 0.3 mm in length were immersed in mother liquor containing 4 mM HgCl2 and left for 1/2 h before mounting.
A crystal of 0.2 to 0.3 mm in length was collected in a 0.4-mm Cryoloop (Hampton Research Inc., Alisa Viejo, CA), dipped briefly in immersion oil (Type B, Cargille), and frozen by plunging into liquid N2. The frozen crystal was then transferred to a magnetic goniometer head in a stream of N2 at 100 K (Cryostream, Oxford Cryosystems, Oxford, UK). Diffraction data for the β-LG-vitamin D2 complex were collected in 1.5° oscillations each of 30 min duration on a 300-mm MarResearch imaging plate system mounted on an ENRAF-Nonius FR571 rotating anode generator operating at 40 kV and 80 mA and producing Cu-Kα radiation from a graphite crystal monochromator. The β-LG-Hg complex data were collected in a similar manner on the Mar-FR571 system in Edinburgh. The β-LG-cholesterol complex data were collected at the CCLRC Daresbury Laboratory Synchrotron Radiation Source using station 9.5 with λ = 1.3 Å. All data were processed by DENZO and reduced with SCALEPACK (
). After some 20 cycles of positional and individual B-factor refinement interspersed with interactive modeling, the R-factor for the β-LG-complex had dropped to around 28% with the R-free around 36%. The first difference electron density map (|Fo| − |Fc|, αc) calculated at this stage showed (Figures 1a, 2a, 3a) that the central calyx binding site was occupied by a molecule the size and shape consistent with its being cholesterol (Figure 1a) or vitamin D2 (Figure 2a). Refinement proceeded with the addition of water molecules, then the ligand was introduced, and positional and B-factor refinement continued. From the final electron density map calculated with coefficients (2|Fo| − |Fc|, αc) and shown in Figures 1b and 2b, it was clear that much of the ligand fitted the density well, and further, there was no other significant electron density unaccounted for. The final statistics, including average temperature factors, are presented in Table 1.
for retinol. There is always some uncertainty in dealing with ligands that are only sparingly soluble in water, particularly in cases where the association is not particularly tight. We have adopted a method that works moderately well for crystallographic analysis but which, of course, requires a fairly high salt medium (typically ∼1.3 M citrate) to grow and stabilize the crystals, a pH above 7, to ensure the binding site is open and a significant protein concentration (20 mg/ml, ∼1 mM).
Figure 1a shows the initial electron density map of the cholesterol-β-LG complex showing the ribbon of density in the same binding pocket that is occupied by fatty acids (
). The electron density is not long enough to encompass the whole cholesterol molecule, and hence there was some ambiguity in building it into the available density. We assumed that the 3-hydroxy group would not be buried in the center of the protein, and consequently modeled it such that a long hydrogen bond (∼3.7Å) was formed with the carbonyl oxygen of residue Pro38 at the mouth of the calyx (Figure 1c). Refinement was carried out to improve this initial fitting, and the final statistics are shown in Table 1. The model is adequate and is certainly a plausible interpretation of the density. What was immediately obvious from the electron density map, however, was that part of the exposed fused-ring system of cholesterol was ill-defined in a manner reminiscent of the polar groups of the retinoids, retinal, and retinoic acid. A superposition of the cholesterol moiety with palmitic acid (
) shows that the independently refined cholesterol occupies a broadly similar space and, although the degree of detail that is achievable in the structure reported here is less than one would hope for, there is little doubt that the binding that we observed is consistent with the association constant (
) and also with the other ligand-bound structures.
A similar situation arose with vitamin D2. The bulk of the electron density was not enough to cover the entire ligand, and so there was some difficulty in knowing which end should be inserted into the binding calyx (Figure 2a). The refinement statistics (Table 1) did not discriminate between either of the two principal binding modes (that with the hydroxyl group innermost, and that with it in the more chemically reasonable external arrangement), and we have, therefore, chosen that with the hydroxyl group bound externally. In this case, however, the 3-hydroxyl group appears to form an H-bond with the carbonyl of Lys 69, and that of Pro38 interacts with Lys 60 (Figure 2b). Once again, the electron density is poor for the outer extremity of the ligand but, in this case, the innermost aliphatic chain is also less well defined than in cholesterol, retinal, or palmitate.
A mercury derivative made by reaction with the free cysteine of β-LG was used to solve the original crystal structure by multiple isomorphous replacement (
). Data collected from a soaked crystal produced the expected large peak close to the Sγ of Cys121 (Figure 3a, b) and, further, Figure 3a shows that there is no ligand discernible in the central binding pocket, although, following refinement, there is evidence of a second, weaker binding site near the extremity of the EF loop. When the structure of the native protein in the same space group was superimposed using the core β-strands F, G, and H that support Cys121, it was found that there was very little obvious movement within the dimer interface; the helix did not move significantly relative strands F, G, and H, nor did strand I (Figure 3c). Several sidechains had moved, although, in some cases, this does not affect the other subunit. Glu131, Glu134, and Lys138 on the upper surface of the helix had adopted different conformations—the ring of Phe136 had rotated by some 25° and that of His146 on strand I by about 90°. Whereas the guanidino group of Arg40 remained largely unmoved, the sidechain of Asp33 had rotated by approximately 90° around χ1. This would affect the ion pair interaction between Asp33 and Arg40, which is part of the interface that holds the dimer together (
). Perhaps most significantly, the side chain of Arg 148 in the center of strand I at the dimer interface is hydrogen bonded to Asp137 on the helix. Thus, any small movement of the helix will be transmitted to the central side chain of the crucial interfacial β-strand. However, it is unclear why this should result in dissociation, since Arg in particular has a long and flexible side chain, capable of absorbing small movements caused by a slight roll in the helix toward the interface, produced when the Hg atom binds. The I strand hardly moves at all (Figure 3c). It is even less clear at this stage why this binding should have any effect on Asp33 or Arg40, some 20 and 15 Å from the Asp and Arg residues, respectively.
The evidence presented here is consistent with all of the structure-based binding studies on the lipocalin family in that the observed binding site for hydrophobic ligands with moderately low dissociation constants is within the central calyx of the structure. No evidence for binding at an alternative site was obtained, although the work from other laboratories, most notably those of Swaisgood and de Wolf (
), provides a possible explanation for the diverse reports of more than a single binding site. It appears that discrepancies in the reported ligand affinities may be method-dependent. Using the intrinsic fluorescence of Trp can give results that indicate significantly tighter ligand binding than other methods, especially equilibrium dialysis. This effect is particularly noticeable in the case in which the Trp fluorescence decreases on ligand addition (
), a surface binding site with a low affinity together with the main, central site would appear to satisfy most of the reported experimental observations. A crystallographic analysis, particularly of a sparingly soluble ligand, is unlikely to show a distinguishable binding site when the Kd > 10 μM, and this may well be why we have been unable to locate any other binding site on the protein.
The location of this putative second binding site is also uncertain. The original report of retinol binding by β-LG would appear to be in error (
), and although it is tempting to assign the site to the region around the free thiol group “under” the helix on the outer surface of the protein, there is no structural evidence for this being the case, with the exception of those ions and molecules that are known to react with cysteine. However, the distinct structural variability of binding of the polar ends of the various ligands, where the protein also has a relatively open “vestibule” at the mouth of the calyx, could indicate that this may be a subsidiary site. It is certainly large enough to accommodate a second molecule when the ligand is small (
It is known that reaction of the free thiol at Cys121 increases the monomer concentration over that of the dimer, and it has also been shown that it takes only a relatively small change to the interface to dissociate the molecule (
). By soaking mercury ions into crystals of β-LG, it has been possible to trap an otherwise unstable state of the protein, and the small changes that occur involving the helix residue Asp137 and Arg148 on the I-strand at the dimer interface, show that a modification, remote from the interface, can be transmitted to it and thereby may enhance dissociation. Dissociation is believed to occur before unfolding and disulphide interchange during the thermal denaturation of β-LG (
), and so this structure may represent an early stage of the unfolding pathway.
No definite physiological function has been ascribed to β-LG, although several suggestions have been made, the more compelling of which favor a role in molecular transport or, possibly, as some form of modulator (
). Most suggestions concerning the function, understandably, have concentrated on either the lactating cell or more usually the neonate, and a transporter role seems reasonable, since many lipocalins are transporters. β-Lactoglobulin binds both fatty acids and retinol, and the structure is similar to the known transporter, plasma RBP. The nature of the ligand transported, apart from being generally hydrophobic, is not clear, however. Fatty acids, rather than retinal, are found as endogenous ligands in milk, but not all species have a β-LG that binds fatty acids (
), and it seems improbable that the true function will vary from species to species. Similarly, retinol is significantly more soluble in the fat phase of milk and thus will probably be transferred from mother to offspring by that route. A signaling or activity-modulator role appears to be less likely, not just because of the paucity of similar roles reported for other lipocalins (
), but also because the data supporting these various activities appear rather circumstantial. Further, for such an important role, one might expect the presence of β-LG in the milks of all species, not just some. What does not appear to have been considered in detail until recently (
) is that the function is directly related to maternal physiology.
When the species variation of β-LG sequences is examined along with that of the other lipocalins, what emerges is a family tree like that shown in Figure 4. The RBP are clearly distinct from the lactoglobulins, but there is one protein present in the endometrium, now called glycodelin (formerly PP14, inter alia), that is the closest relative to the β-LG-II gene product (
). Notice also that there is ruminant pseudo-gene identified in both cow and goat that is also most nearly related to the sequence of β-LG-II. Might it be that the protein glycodelin reflects the true, original function of β-LG as a protein involved in some aspect of fetal development in all mammals? In many species, the gene has undergone a gene duplication event and is expressed during lactation for nutritional purposes. More recently, some species, like rodents, lagomorphs, and man, have lost the function of one of these genes through formation of a pseudo-gene, resulting in the gene ceasing to be expressed. The biological properties of the milk protein, such as ligand binding and inhibiting harmful bacterial adhesion in the intestine (
) that have been identified, are clearly useful to each species but are subsidiary to those of neonatal nutrition. On the other hand, it is known that low levels of β-LG are expressed in the cow throughout the dry period in a manner that is distinct from α-LA and the caseins (
). Whether this is indicative of another more physiological role for β-LG in the changes that occur to the mammary gland during this period remains unclear at present.
Baboon (Papio hyandras) appears to have both β-LG and glycodelin (A.T. Fazleabas, personal communication) and two obvious questions arise. First, is there a pseudo-gene for β-LG in the human genome? Second, does the ruminant endometrial tissue contain glycodelin? In a preliminary experiment, searching the annotated human genome (
) with the human glycodelin protein sequence produced a large number of lipocalin sequences, most of which are not lactoglobulins. However, several fragments that may be pseudogenes have been identified and are undergoing further analysis. It also appears that the pseudogene identified by Pena et al. (2001; NCBI: AH011480, AF403023/24/25 http://www.ncbi.nih.gov/entrez/viewer.fcgi?18765895) may be just isolated exons of glycodelin; at least the alignment of the DNA shows a perfect match to part of that coding for glycodelin. Attempts at detecting glycodelin, as distinct from RBP, in the fetuses of sheep or cattle are yet to be attempted.
Intriguingly, the molecular function of glycodelin during fetal development is not clearly understood either (
), but the similarity with β-LG described above clearly implies involvement of some essential small molecule. A retinoid of some sort is an obvious possibility because of the similarity to RBP, but neither retinol nor retinoic acid appear to bind to glycodelin (
). Some involvement with long-chain or polyunsaturated fatty acids such as arachidonic acid, a precursor of prostaglandins, would provide a possible alternative physiological ligand, although there are no reports of this interaction. There is clearly considerable scope for further work in the area.
We report here the medium resolution structures of complexes of bovine β-LG with cholesterol, vitamin D2, and with mercury ions. The electron density for the two organic ligands is not sufficient to provide unambiguous evidence for the exact location of all atoms in the structure, but the likely arrangement is that with the hydrophobic ends being innermost. From the structure of the mercury-Cys121 complex, we are unable to identify with certainty why modification of Cys121 destabilizes the dimer.
Based on the species distribution and lipocalin family relationships, a function is proposed that accounts for much of the interspecies variation in properties.
Financial support from the BBSRC and a European Union Concerted Action is gratefully acknowledged. We are grateful to many friends and colleagues but particularly to Mike Boland, Stella Bury, Andrew Coulson, Lawrie Creamer, Geoff Jameson, Mahesh Narayan, and Su-Ying Wu for stimulating discussions and to the MADGELAS partners for their continued interest.
Gapped BLAST and PSI-BLAST: A new generation of protein database search programs.