Structure of human lactoferrin (Lf) in its closed form showing important surface features. The chain is colored according to the flexibility of different parts of the molecule as seen in crystal structures, from blue (lowest B factors and lowest flexibility) through to red (highest B factors and highest flexibility). The most flexible features are the N-terminus (N), the PEST loop, loops at the entry to the iron-binding cleft of the N-lobe and the C-terminal part of helix H1. The 2 helices that give rise to antibacterial peptides, H1 and H9, are indicated, as are the glycosylation sites in human and bovine Lfs; glycosylation sites in human Lf are shown in magenta, those in bovine Lf in grey. Some coincide structurally but others do not. NLS, nuclear localization signal. 

Contexts in source publication

Context 1

... Fig. 3). Both proteins have very high affinities for Fe 3+ at physiological pH, with binding constants of ~10 20 (Aisen and Leibman 1972; Baker 1994), meaning that they have similar ability to sequester iron. A key difference, however, is that Lf is able to retain bound iron to a substantially lower pH; sTf loses iron at a pH of ~5.5, whereas Lf retains bound iron to a pH of ~3.5 (Mazurier and Spik 1980). Explanations for this difference have considered the role of second-shell residues around the N-lobe binding site of each protein, including a so-called dilysine trigger in sTf (Dewan et al. 1993; Peterson et al. 2002), but the primary reason stems from interactions be- tween the 2 lobes of Lf (Ward et al. 1996), which stabilize N-lobe binding. The greater pH stability of iron binding does mean that Lf is a somewhat more effective bacteriostatic agent and antioxidant than sTf, for its ability to scavenge and retain iron. It may not have much to do with iron absorption, however, since the uptake of iron from sTf is intimately tied to structural changes that occur when sTf binds to the transferrin receptor (Eckenroth et al. 2011). If Lf is to release iron in vivo, it seems that this can only occur if Lf is degraded or denatured or through structural changes concomitant with binding to receptors. No hard evidence of such mechanisms currently exists. As in other families of homologous proteins, the amino acid sequences of Lfs of different species are highly conserved in internal regions, reflecting the need to generate a stable, folded structure. In contrast, amino acid substitutions occur readily on the protein surface, where they do not affect protein folding but can generate unique, protein-specific functions. In the case of Lf, key differences from other transferrins include surface charge, glycosylation, and the genera- tion of adventitious binding sites for other proteins, nucleic acids, small molecules or ions, arising from local combina- tions of amino acid substitutions. Some of these sites are shown in Fig. 4 and are discussed later. All lactoferrins are highly positively charged, with an iso- electric point of 9 – 10. This cationic character is a major point of difference from other transferrin family members, which have pI values of 5 – 6. This cationic character must be a significant factor in the well-known ability of Lf to bind to a variety of cell types (Birgens 1984), nucleic acids (van Berkel et al. 1997), and a variety of proteins and other molecules (van Berkel et al. 1997). The distribution of surface charge is uneven, however, and differs in detail between the lactoferrins of different species. Likewise, whereas all lactoferrins are glycosylated, the number of potential N-glycosylation sites varies, from 1 (mouse Lf) to 5 (cow, goat, and sheep Lf), and the sites are widely distributed over the protein surface in different Lfs (Baker and Baker 2009). These variations highlight the likelihood that species-specific variations in Lf function (but not structure) may exist. It is also worth noting that there are regions of relative flexibility on the protein surface, even for iron-bound Lf with its highly stable overall conformation. These regions of increased flexibility can be recognized from their higher B factors (the crystallographic parameter that describes the extent to which an atom or group of atoms is able to oscillate about its mean position). Some of these flexible regions are external loops that are not tightly constrained, such as several loops at the entry to the N-lobe binding cleft and the loop on the C-lobe that carries the PEST degradation sequence (Hardivillé et al. 2010). Others are larger substructures; for example, most of the first helix H1 is relatively mobile, espe- cially at its C-terminal end where it abuts the flexible N- terminus. Still other flexible features simply extend out from the surface without restraint, such as the N-terminus and the glycan chains. For all Lfs that have been structurally characterized to date, one striking hot spot of positive charge stands out (Fig. 5). This is located on the N-lobe, encompassing the poly- peptide N-terminus (sequence GRRRRS in human Lf), the C- terminal end of the first helix Η 1 (residues 12 – 30 in human Lf), and a cluster of basic residues associated with another surface helix Η 9 (residues 263 – 279 in human Lf). The latter 2 helices are also the source of antibacterial peptides, discussed later. Another, much smaller, hot spot of positive charge is found in the interlobe region, associated with the helix that connects the 2 lobes. Intriguingly, several promi- nent negative patches can also be seen. The major basic region surrounding the N-terminus has been shown to be responsible for the binding of DNA, heparin, lipopolysaccharide, and glycosaminoglycans by Lf (Mann et al. 1994; van Berkel et al. 1997; van Berkel et al. 1997). This region also contains some of the most flexible elements in the protein structure (Fig. 4); the N-terminal residues proj- ect out into solution and are so mobile that they are rarely ob- served in Lf crystal structures. The C-terminal part of helix H1 is also somewhat mobile, and the relative flexibility of these 2 regions is probably an important factor in the binding of such a variety of molecules. Intriguingly, the Lfs from different biological species vary greatly in their N-terminal sequences, the 4 consecutive arginines of human Lf being replaced by 1 arginine and 1 lysine in bovine, goat, and horse Lf and just a sin- gle lysine in mouse Lf. This suggests that some properties seen for the human protein may not be replicated in the other Lfs and could be an adventitious property of the human Lf alone. Only comparative experiments with other Lfs can resolve this. The positive charge associated with the first helix 1 is a conserved feature of all Lfs, although the number of basic residues and their exact positions vary. This helix forms part of the lactoferricin (Lfcin) region, so called because proteolytic cleavage of Lf releases potent antibacterial peptides (Bellamy et al. 1992) called Lfcin H (human Lf) and Lfcin B (bovine Lf) (Gifford et al. 2005). The relationship between these peptides and the corresponding regions on intact Lf (Fig. 4) is intriguing because the conformations adopted by the peptides in their bioactive forms can be very different from those they have in the intact protein; Lfcin B, for example, forms an amphipathic b -ribbon structure rather than an a -helix (Vogel et al. 2002). Whether this region does indeed contribute to the antibacterial properties of intact Lfs is not clear; it is highly exposed on both iron-bound and apo forms of Lf, and the positive charge could enable binding to bacterial cell membranes. The detailed mechanism of action must be different, however, since the hydrophobic residues that are critical to Lfcin activity are arranged differently in the intact protein; some key residues are buried. A second antibacterial peptide derived from Lf, called lac- toferrampin, corresponds to another amphipathic helix on the protein surface, helix Η 9 (residues 263 – 279) and a few basic residues that follow it (van der Kraan et al. 2004). Again, some of the residues that contribute to the ability of the iso- lated peptide to disrupt cell membranes are buried in the intact protein, but the basic residues (4 Lys and 1 Arg in both human and bovine Lfs) contribute to the large positively charged patch surrounding the protein N-terminus. Their im- portance to the antibacterial properties of the intact protein is yet to be determined. In addition to its ability to bind to a variety of cell types, Lf has been implicated in binding to a number of physiologi- cally important proteins, including ceruloplasmin (Sabatucci et al. 2007), calmodulin (de Lillo et al. 1992), and osteopontin (Yamniuk et al. 2009). So far, there is little definitive knowledge concerning the binding sites for these proteins or their mode of interaction, although small-angle X-ray scatter- ing studies of the Lf – ceruloplasmin complex suggest that the positive patch surrounding the Lf N-terminus is involved (Sabatucci et al. 2007). The most unequivocal demonstration of protein binding to Lf concerns the protein PspA, a major virulence factor from Streptococcus pneumoniae. The crystal structure of a complex between the N-lobe of human Lf and the Lf-binding fragment of PspA shows that binding occurs through the docking of 2 helices from the pneumococcal protein against helix Η 1 (the Lfcin helix) of Lf, coupled with specific interactions with positively charged residues from helix H1 and the N-terminus (Senkovich et al. 2007). Given the highly exposed location of this region of Lf, and the fact that helices are often involved in protein – protein interactions, it is likely that other proteins bind to the same surface in similar ways. Much early research on Lf focused on the nature and role of its glycan chains, but this appears to have largely lapsed in recent years. This can in part be attributed to the fact that glycosylation appears at first sight to have little impact on Lf structure or function. Comparisons of glycosylated and non- glycosylated Lfs (van Berkel et al. 1995; H.M. Baker, unpublished work) show that glycosylation does not have a significant effect on properties such as iron binding and release, thermal stability, and the ability to bind other molecules such as glycosaminoglycans. Furthermore, crystallographic studies of Lfs show that only the first 1 or 2 sugar residues of the glycan chains can be seen in most cases and that their interactions with the protein structure are mini- mal, at most a few hydrogen bonds or nonbonded interactions. It is possible, however, that glycosylation may play a larger part in lactoferrin function than has been recognised to date. Complex carbohydrates form a major part of the surface of human epithelial cells and of the extracellular ...

Context 2

... into 2 lobes, N1 and N2 (N-lobe) or C1 and C2 (C-lobe), with the iron-binding site located between the 2 domains. The same structural organization is true of all transferrins. Depending on its metal ion status, Lf can adopt either of 2 conformational states. In its iron-bound state (Fe 2 Lf), 2 Fe 3+ ions are bound, each associated with a synergistically bound carbonate ion, and with the 2 domains of each lobe fully closed over the bound metal ion (Fig. 2A). This closed form is characteristic of all metal-bound Lfs, including complexes with other transition metal ions such as Cu 2+ and Mn 3+ and larger ions such as Ce 4+ (Smith et al. 1992, 1994; Kumar et al. 2000). Evidently, different metal ions can be bound, albeit with lower affinity than Fe 3+ , without disturbing the basic structure. This closed, metal-bound form is highly stable and relatively rigid as the metal ions lock the domains together. In contrast, in its apo (metal-free) state Lf adopts an open form in which the 2 domains of each lobe are opened wide (Fig. 2B). Metal binding and release is thus associated with large-scale conformational changes in which the domains close over the bound metal ion or open to release it (Anderson et al. 1990; Gerstein et al. 1993). There is strong evidence that the apo-protein is in some sort of dynamic equilibrium between open and closed states (Baker et al. 2002; Baker 1994). Although X-ray crystallography cannot usually moni- tor dynamic processes, when an equilibrium exists in solution, crystal formation can trap one out of multiple species present in solution. Thus crystallographic analyses have vari- ously visualized apo-Lf with both lobes open (human and camel Lfs (Khan et al. 2001; Baker et al. 2002)), both closed (horse Lf (Sharma et al. 1999)), and one open, one closed (human Lf (Anderson et al. 1990)), showing that the apo form is flexible, alternating between open and closed forms. Importantly, to interpret functional studies it is essential to know which conformational state applies, because the exposed surfaces and functional groups will be different and some properties will also be different. In terms of overall sequence and structure, lactoferrin is re- markably similar to other transferrins. Human Lf shares ~60% sequence identity with human sTf, which is very similar to the sequence similarity between different Lfs (~70% identity) (Baker 1994). The 2 proteins also share the same fold, and their metal binding sites are identical in terms of their metal-binding groups, differing only in some second- shell residues. Why, then, are they so different in their functional roles? We distinguish the following 3 factors: specific differences in iron release, differences in surface properties, and different biological locations Lf and sTf share essentially identical iron-binding sites, at least in terms of the iron ligands, which are completely conserved (2 Tyr, 1 His, 1 Asp, and the CO ion; Fig. 3). Both proteins have very high affinities for Fe 3+ at physiological pH, with binding constants of ~10 20 (Aisen and Leibman 1972; Baker 1994), meaning that they have similar ability to sequester iron. A key difference, however, is that Lf is able to retain bound iron to a substantially lower pH; sTf loses iron at a pH of ~5.5, whereas Lf retains bound iron to a pH of ~3.5 (Mazurier and Spik 1980). Explanations for this difference have considered the role of second-shell residues around the N-lobe binding site of each protein, including a so-called dilysine trigger in sTf (Dewan et al. 1993; Peterson et al. 2002), but the primary reason stems from interactions be- tween the 2 lobes of Lf (Ward et al. 1996), which stabilize N-lobe binding. The greater pH stability of iron binding does mean that Lf is a somewhat more effective bacteriostatic agent and antioxidant than sTf, for its ability to scavenge and retain iron. It may not have much to do with iron absorption, however, since the uptake of iron from sTf is intimately tied to structural changes that occur when sTf binds to the transferrin receptor (Eckenroth et al. 2011). If Lf is to release iron in vivo, it seems that this can only occur if Lf is degraded or denatured or through structural changes concomitant with binding to receptors. No hard evidence of such mechanisms currently exists. As in other families of homologous proteins, the amino acid sequences of Lfs of different species are highly conserved in internal regions, reflecting the need to generate a stable, folded structure. In contrast, amino acid substitutions occur readily on the protein surface, where they do not affect protein folding but can generate unique, protein-specific functions. In the case of Lf, key differences from other transferrins include surface charge, glycosylation, and the genera- tion of adventitious binding sites for other proteins, nucleic acids, small molecules or ions, arising from local combina- tions of amino acid substitutions. Some of these sites are shown in Fig. 4 and are discussed later. All lactoferrins are highly positively charged, with an iso- electric point of 9 – 10. This cationic character is a major point of difference from other transferrin family members, which have pI values of 5 – 6. This cationic character must be a significant factor in the well-known ability of Lf to bind to a variety of cell types (Birgens 1984), nucleic acids (van Berkel et al. 1997), and a variety of proteins and other molecules (van Berkel et al. 1997). The distribution of surface charge is uneven, however, and differs in detail between the lactoferrins of different species. Likewise, whereas all lactoferrins are glycosylated, the number of potential N-glycosylation sites varies, from 1 (mouse Lf) to 5 (cow, goat, and sheep Lf), and the sites are widely distributed over the protein surface in different Lfs (Baker and Baker 2009). These variations highlight the likelihood that species-specific variations in Lf function (but not structure) may exist. It is also worth noting that there are regions of relative flexibility on the protein surface, even for iron-bound Lf with its highly stable overall conformation. These regions of increased flexibility can be recognized from their higher B factors (the crystallographic parameter that describes the extent to which an atom or group of atoms is able to oscillate about its mean position). Some of these flexible regions are external loops that are not tightly constrained, such as several loops at the entry to the N-lobe binding cleft and the loop on the C-lobe that carries the PEST degradation sequence (Hardivillé et al. 2010). Others are larger substructures; for example, most of the first helix H1 is relatively mobile, espe- cially at its C-terminal end where it abuts the flexible N- terminus. Still other flexible features simply extend out from the surface without restraint, such as the N-terminus and the glycan chains. For all Lfs that have been structurally characterized to date, one striking hot spot of positive charge stands out (Fig. 5). This is located on the N-lobe, encompassing the poly- peptide N-terminus (sequence GRRRRS in human Lf), the C- terminal end of the first helix Η 1 (residues 12 – 30 in human Lf), and a cluster of basic residues associated with another surface helix Η 9 (residues 263 – 279 in human Lf). The latter 2 helices are also the source of antibacterial peptides, discussed later. Another, much smaller, hot spot of positive charge is found in the interlobe region, associated with the helix that connects the 2 lobes. Intriguingly, several promi- nent negative patches can also be seen. The major basic region surrounding the N-terminus has been shown to be responsible for the binding of DNA, heparin, lipopolysaccharide, and glycosaminoglycans by Lf (Mann et al. 1994; van Berkel et al. 1997; van Berkel et al. 1997). This region also contains some of the most flexible elements in the protein structure (Fig. 4); the N-terminal residues proj- ect out into solution and are so mobile that they are rarely ob- served in Lf crystal structures. The C-terminal part of helix H1 is also somewhat mobile, and the relative flexibility of these 2 regions is probably an important factor in the binding of such a variety of molecules. Intriguingly, the Lfs from different biological species vary greatly in their N-terminal sequences, the 4 consecutive arginines of human Lf being replaced by 1 arginine and 1 lysine in bovine, goat, and horse Lf and just a sin- gle lysine in mouse Lf. This suggests that some properties seen for the human protein may not be replicated in the other Lfs and could be an adventitious property of the human Lf alone. Only comparative experiments with other Lfs can resolve this. The positive charge associated with the first helix 1 is a conserved feature of all Lfs, although the number of basic residues and their exact positions vary. This helix forms part of the lactoferricin (Lfcin) region, so called because proteolytic cleavage of Lf releases potent antibacterial peptides (Bellamy et al. 1992) called Lfcin H (human Lf) and Lfcin B (bovine Lf) (Gifford et al. 2005). The relationship between these peptides and the corresponding regions on intact Lf (Fig. 4) is intriguing because the conformations adopted by the peptides in their bioactive forms can be very different from those they have in the intact protein; Lfcin B, for example, forms an amphipathic b -ribbon structure rather than an a -helix (Vogel et al. 2002). Whether this region does indeed contribute to the antibacterial properties of intact Lfs is not clear; it is highly exposed on both iron-bound and apo forms of Lf, and the positive charge could enable binding to bacterial cell membranes. The detailed mechanism of action must be different, however, since the hydrophobic residues that are critical to Lfcin activity are arranged differently in the intact protein; some key residues are buried. A second antibacterial peptide derived from Lf, ...

Context 3

... residues from helix H1 and the N-terminus (Senkovich et al. 2007). Given the highly exposed location of this region of Lf, and the fact that helices are often involved in protein – protein interactions, it is likely that other proteins bind to the same surface in similar ways. Much early research on Lf focused on the nature and role of its glycan chains, but this appears to have largely lapsed in recent years. This can in part be attributed to the fact that glycosylation appears at first sight to have little impact on Lf structure or function. Comparisons of glycosylated and non- glycosylated Lfs (van Berkel et al. 1995; H.M. Baker, unpublished work) show that glycosylation does not have a significant effect on properties such as iron binding and release, thermal stability, and the ability to bind other molecules such as glycosaminoglycans. Furthermore, crystallographic studies of Lfs show that only the first 1 or 2 sugar residues of the glycan chains can be seen in most cases and that their interactions with the protein structure are mini- mal, at most a few hydrogen bonds or nonbonded interactions. It is possible, however, that glycosylation may play a larger part in lactoferrin function than has been recognised to date. Complex carbohydrates form a major part of the surface of human epithelial cells and of the extracellular matrix and are the receptors for many viruses and pathogenic bacteria, ena- bling the viruses to gain entry into human cells (Varki 2007). The predominant receptors are sialylated glycans, typically with a terminal sialic acid residue linked to a penultimate galactose. Human viruses such as influenza virus (Viswanathan et al. 2010), rotavirus (Blanchard et al. 2007), norovirus (Chen et al. 2011), adenovirus, and many others target such receptors, binding to the glycans via sialic acid binding proteins. Likewise, adhesins on the surfaces of pathogenic bacteria frequently contain sialic acid binding domains that bind to sialylated glycans to initiate colonization (Pyburn et al. 2011). Other pathogens secrete toxins that disrupt immune responses by targeting key human gly- coproteins; examples include botulinum toxin B (Jin et al. 2006), tetanus toxin (Stein et al. 1994), and a family of superantigen-like toxins that are secreted by Staphylococcus aureus (Baker et al. 2007). Although Lf glycan structures vary, most of them termi- nate with a sialic acid residue joined by an a 2 – 6 linkage to galactose (Spik et al. 1988); the same disaccharide moiety that is targeted by many viruses and bacterial proteins. We have shown in pull-down assays that the staphylococcal toxins mentioned above bind to human Lf (M. Chung, H.M. Baker, E.N. Baker, J.D. Fraser, unpublished data), presum- ably through its sialylated glycans. Although its effects might only be local, given the large number of sialic acid residues presented by the glycans on epithelial cells, it is possible that bursts of Lf, which is released from neutrophils in response to attack by pathogens, may provide an initial level of protection by binding through its glycan chains to the sialic acid binding proteins of viruses and bacteria. Some of the reported antiviral and antibacterial activities of Lf may well be mediated through its glycans. Recent progress in several other aspects of Lf function can also be interpreted in terms of features of the molecular surface. The demonstration that Lf can exert proteolytic activity was traced to a Ser-Lys dyad (Ser259 and Lys73) on the N- lobe, close to the interlobe region (Hendrixson et al. 2003). This activity at first appeared to be limited to the cleavage of several bacterial-cell-surface proteins, the IgA protease and the Hap adhesin (Qiu et al. 1998). It now appears that its specificity is for Arg-rich sequences, with another example coming to light recently in the form of a heparin-binding protein from Neisseria meningitidis (Serruto et al. 2010). Early reports that human Lf could enter the cell nucleus and act as a transcription factor (He and Furmanski 1995) have been substantiated recently. It has also been established that delta-lactoferrin ( D Lf), which is expressed as a splice variant of human Lf lacking the first 45 residues of the full- length protein, similarly has the ability to act as a nuclear transcription factor (Mariller et al. 2007). These observations imply that Lf must possess at least one nuclear localization signal (NLS). Full-length human Lf has a recognizable NLS in the 4 consecutive Arg residues at its N-terminus (Penco et al. 2001), which are also implicated in DNA binding. This N-terminal Arg 4 sequence is flexible and highly exposed, which should facilitate nuclear import. In contrast, D Lf lacks the first 45 residues of Lf and must therefore possess a different NLS. This has recently been shown to be a bipartite NLS, formed by residues 442 – 457 (Lf numbering; RRSDTSLTWNSVKGKK) in the C-lobe (Mariller et al. 2007). This forms an exposed loop on the surface (Fig. 4), with the 2 clusters of basic residues separated by an irregular helix. Unlike the Arg 4 motif at the N-terminus, which is only present in human Lf, this C-lobe NLS is conserved in all Lfs. Finally, the group of T.P. Singh at the All India Institute of Medical Sciences in New Delhi has carried out a series of studies on bovine Lf and demonstrated that a variety of small molecules such as aspirin, diclofenac, niacin, trehalose, and other sugars bind at a surface site on the C-lobe. This work has not yet been published, and data on binding affinities or physiological significance are not available, but the structures of the complexes are available through the Protein Data Bank (). Low-affinity zinc binding sites have similarly been found on the surface of the C-lobe of bovine Lf (Jabeen et al. 2005). What these studies also show is that binding sites can be found on the protein surface that may be specific to a particular Lf, arising from adventitious sequence changes; differences at the equivalent positions on the human Lf surface suggest that it probably does not share the same binding activities. The 3-D structure of Lf provides a framework for more rigorous understanding of the many functions of this intriguing protein. Some activities of Lf clearly derive from its ability to scavenge iron and retain it with high affinity over a wide pH range. These are well understood in terms of the structural changes associated with metal binding and the differences in conformation and dynamics between the metal- bound and apo forms. Other properties, most of which relate to host defence and the role of Lf as part of the innate immune response, are much less well understood at the molecular level. These activities, which include the antibacterial, antiviral, antifungal, and antiparasitic activities of Lf, and its ability to bind to a variety of cell types are almost certainly mediated through surface features. Not only are binding sites on the molecular surface difficult to define, but there is a dis- tinct possibility that they may differ significantly among the Lfs of different species given that most sequence variation occurs on the surface. Polymorphism, for example, as seen at the N-terminus of human Lf (van Berkel et al. 1997), and heterogeneity of glycosylation (Spik et al. 1988; van Berkel et al. 1995) may also affect the properties of different prepa- rations of Lf. Some activities are almost certainly adventitious, likely to be present only on a subset of Lfs. If we are to come to a fuller understanding of Lf function, it is essential that carefully controlled experiments be under- taken to establish molecular mechanisms. This requires in- depth biochemical and biophysical analyses to extend the many reported phenomenological observations. As an example, an observation that Lf binds to a particular cell type or to a virus or other entity should ideally also ask questions such as the following: • Does it really bind, and in which biological locations, and to what? Labelled Lf or conjugates with green fluorescent protein would be very valuable tools. • Which form is bound? Both apo-Lf and Fe 2 Lf should be compared. • Do the Lfs of different species behave the same? Compar- isons should be carried out with the human, bovine, and mouse proteins. • Does binding occur through the protein or the carbohydrate? Comparisons should be made of both native and degly- cosylated or desialylated proteins. • Is it a charge effect or specific binding? Small molecules that disrupt charge interactions could be used. Finally, the greatest need if Lf function is to be analyzed more rigorously is for a good expression system that would allow the ready production of site-specific mutants. There are so many intriguing observations that the time is surely ripe for more in-depth analysis. We greatly acknowledge the many colleagues, research fel- lows, and students who have contributed to this work over the years and the wonderful lactoferrin research community internationally. We also thank the Health Research Council of New Zealand for financial ...

Context 4

... concomitant with binding to receptors. No hard evidence of such mechanisms currently exists. As in other families of homologous proteins, the amino acid sequences of Lfs of different species are highly conserved in internal regions, reflecting the need to generate a stable, folded structure. In contrast, amino acid substitutions occur readily on the protein surface, where they do not affect protein folding but can generate unique, protein-specific functions. In the case of Lf, key differences from other transferrins include surface charge, glycosylation, and the genera- tion of adventitious binding sites for other proteins, nucleic acids, small molecules or ions, arising from local combina- tions of amino acid substitutions. Some of these sites are shown in Fig. 4 and are discussed later. All lactoferrins are highly positively charged, with an iso- electric point of 9 – 10. This cationic character is a major point of difference from other transferrin family members, which have pI values of 5 – 6. This cationic character must be a significant factor in the well-known ability of Lf to bind to a variety of cell types (Birgens 1984), nucleic acids (van Berkel et al. 1997), and a variety of proteins and other molecules (van Berkel et al. 1997). The distribution of surface charge is uneven, however, and differs in detail between the lactoferrins of different species. Likewise, whereas all lactoferrins are glycosylated, the number of potential N-glycosylation sites varies, from 1 (mouse Lf) to 5 (cow, goat, and sheep Lf), and the sites are widely distributed over the protein surface in different Lfs (Baker and Baker 2009). These variations highlight the likelihood that species-specific variations in Lf function (but not structure) may exist. It is also worth noting that there are regions of relative flexibility on the protein surface, even for iron-bound Lf with its highly stable overall conformation. These regions of increased flexibility can be recognized from their higher B factors (the crystallographic parameter that describes the extent to which an atom or group of atoms is able to oscillate about its mean position). Some of these flexible regions are external loops that are not tightly constrained, such as several loops at the entry to the N-lobe binding cleft and the loop on the C-lobe that carries the PEST degradation sequence (Hardivillé et al. 2010). Others are larger substructures; for example, most of the first helix H1 is relatively mobile, espe- cially at its C-terminal end where it abuts the flexible N- terminus. Still other flexible features simply extend out from the surface without restraint, such as the N-terminus and the glycan chains. For all Lfs that have been structurally characterized to date, one striking hot spot of positive charge stands out (Fig. 5). This is located on the N-lobe, encompassing the poly- peptide N-terminus (sequence GRRRRS in human Lf), the C- terminal end of the first helix Η 1 (residues 12 – 30 in human Lf), and a cluster of basic residues associated with another surface helix Η 9 (residues 263 – 279 in human Lf). The latter 2 helices are also the source of antibacterial peptides, discussed later. Another, much smaller, hot spot of positive charge is found in the interlobe region, associated with the helix that connects the 2 lobes. Intriguingly, several promi- nent negative patches can also be seen. The major basic region surrounding the N-terminus has been shown to be responsible for the binding of DNA, heparin, lipopolysaccharide, and glycosaminoglycans by Lf (Mann et al. 1994; van Berkel et al. 1997; van Berkel et al. 1997). This region also contains some of the most flexible elements in the protein structure (Fig. 4); the N-terminal residues proj- ect out into solution and are so mobile that they are rarely ob- served in Lf crystal structures. The C-terminal part of helix H1 is also somewhat mobile, and the relative flexibility of these 2 regions is probably an important factor in the binding of such a variety of molecules. Intriguingly, the Lfs from different biological species vary greatly in their N-terminal sequences, the 4 consecutive arginines of human Lf being replaced by 1 arginine and 1 lysine in bovine, goat, and horse Lf and just a sin- gle lysine in mouse Lf. This suggests that some properties seen for the human protein may not be replicated in the other Lfs and could be an adventitious property of the human Lf alone. Only comparative experiments with other Lfs can resolve this. The positive charge associated with the first helix 1 is a conserved feature of all Lfs, although the number of basic residues and their exact positions vary. This helix forms part of the lactoferricin (Lfcin) region, so called because proteolytic cleavage of Lf releases potent antibacterial peptides (Bellamy et al. 1992) called Lfcin H (human Lf) and Lfcin B (bovine Lf) (Gifford et al. 2005). The relationship between these peptides and the corresponding regions on intact Lf (Fig. 4) is intriguing because the conformations adopted by the peptides in their bioactive forms can be very different from those they have in the intact protein; Lfcin B, for example, forms an amphipathic b -ribbon structure rather than an a -helix (Vogel et al. 2002). Whether this region does indeed contribute to the antibacterial properties of intact Lfs is not clear; it is highly exposed on both iron-bound and apo forms of Lf, and the positive charge could enable binding to bacterial cell membranes. The detailed mechanism of action must be different, however, since the hydrophobic residues that are critical to Lfcin activity are arranged differently in the intact protein; some key residues are buried. A second antibacterial peptide derived from Lf, called lac- toferrampin, corresponds to another amphipathic helix on the protein surface, helix Η 9 (residues 263 – 279) and a few basic residues that follow it (van der Kraan et al. 2004). Again, some of the residues that contribute to the ability of the iso- lated peptide to disrupt cell membranes are buried in the intact protein, but the basic residues (4 Lys and 1 Arg in both human and bovine Lfs) contribute to the large positively charged patch surrounding the protein N-terminus. Their im- portance to the antibacterial properties of the intact protein is yet to be determined. In addition to its ability to bind to a variety of cell types, Lf has been implicated in binding to a number of physiologi- cally important proteins, including ceruloplasmin (Sabatucci et al. 2007), calmodulin (de Lillo et al. 1992), and osteopontin (Yamniuk et al. 2009). So far, there is little definitive knowledge concerning the binding sites for these proteins or their mode of interaction, although small-angle X-ray scatter- ing studies of the Lf – ceruloplasmin complex suggest that the positive patch surrounding the Lf N-terminus is involved (Sabatucci et al. 2007). The most unequivocal demonstration of protein binding to Lf concerns the protein PspA, a major virulence factor from Streptococcus pneumoniae. The crystal structure of a complex between the N-lobe of human Lf and the Lf-binding fragment of PspA shows that binding occurs through the docking of 2 helices from the pneumococcal protein against helix Η 1 (the Lfcin helix) of Lf, coupled with specific interactions with positively charged residues from helix H1 and the N-terminus (Senkovich et al. 2007). Given the highly exposed location of this region of Lf, and the fact that helices are often involved in protein – protein interactions, it is likely that other proteins bind to the same surface in similar ways. Much early research on Lf focused on the nature and role of its glycan chains, but this appears to have largely lapsed in recent years. This can in part be attributed to the fact that glycosylation appears at first sight to have little impact on Lf structure or function. Comparisons of glycosylated and non- glycosylated Lfs (van Berkel et al. 1995; H.M. Baker, unpublished work) show that glycosylation does not have a significant effect on properties such as iron binding and release, thermal stability, and the ability to bind other molecules such as glycosaminoglycans. Furthermore, crystallographic studies of Lfs show that only the first 1 or 2 sugar residues of the glycan chains can be seen in most cases and that their interactions with the protein structure are mini- mal, at most a few hydrogen bonds or nonbonded interactions. It is possible, however, that glycosylation may play a larger part in lactoferrin function than has been recognised to date. Complex carbohydrates form a major part of the surface of human epithelial cells and of the extracellular matrix and are the receptors for many viruses and pathogenic bacteria, ena- bling the viruses to gain entry into human cells (Varki 2007). The predominant receptors are sialylated glycans, typically with a terminal sialic acid residue linked to a penultimate galactose. Human viruses such as influenza virus (Viswanathan et al. 2010), rotavirus (Blanchard et al. 2007), norovirus (Chen et al. 2011), adenovirus, and many others target such receptors, binding to the glycans via sialic acid binding proteins. Likewise, adhesins on the surfaces of pathogenic bacteria frequently contain sialic acid binding domains that bind to sialylated glycans to initiate colonization (Pyburn et al. 2011). Other pathogens secrete toxins that disrupt immune responses by targeting key human gly- coproteins; examples include botulinum toxin B (Jin et al. 2006), tetanus toxin (Stein et al. 1994), and a family of superantigen-like toxins that are secreted by Staphylococcus aureus (Baker et al. 2007). Although Lf glycan structures vary, most of them termi- nate with a sialic acid residue joined by an a 2 – 6 linkage to galactose (Spik et al. 1988); the same disaccharide moiety that is targeted by many viruses and bacterial proteins. We have shown in pull-down assays that the staphylococcal ...