Analysis of Sequence and Structure Relationships in the Calycin Protein Superfamily

The Calycin protein superfamily [1,2], an example of a so-called "structural superfamily" where obvious similarities of three-dimensional structure are not easily discernible at the sequence level, is composed of three distinct families of ligand binding proteins. The Lipocalins form a diverse and rapidly burgeoning family of predominantly small secreted proteins; members of the family exhibit high affinity and selectivity for hydrophobic molecules and have been generally characterized as extracellular transport proteins; in this respect, the family is typified by long-established members such as retinol binding protein (RBP), which transports retinol through the blood to target cells. The Fatty Acid-Binding Proteins, or FABPs, another recently identified family of small, almost exclusively intracellular proteins, which again bind hydrophobic molecules, are thought to be involved in lipid metabolism. These proteins are typified by cellular retinol binding protein (cRBP), which is believed to function in the intracellular trafficking of retinol. The Avidins, such as streptavidin, a small soluble protein from the bacteria Streptomyces avidinii, and its homologue hen egg-white avidin, are proteins with a remarkable affinity for the vitamin biotin, as well as other hydrophobic ligands, and have found important applications in biotechnology.

Structural Relationships

Crystallographic studies of several lipocalins have shown them to possess a conserved folding pattern dominated by an 8 stranded calyx- or cup-shaped antiparallel b-barrel which encloses an internal ligand binding site [1]. The 8 strands of the barrel (labelled A to H) are linked by a succession of +1 connections, which may be formally regarded as b-hairpin loops (labelled L1 to L7), giving the simplest possible topology for a closed b-sheet. Loops L2 to L7 are typical of short b-hairpins, while loop L1 is a large W loop which forms a lid folded back to cap the internal ligand binding site. This loop varies in conformation between different lipocalins but maintains its overall shape, size, and position. A short 310 helix, directly before strand A, helps to close off the other end of the barrel. Beyond the last strand of the b-barrel, is an a-helix which folds back to pack parallel with strands G and H.

The crystal structures of several FABPs comprise a well conserved 10 stranded antiparallel b-barrel, again with a repeated +1 topology, although it is not continuously hydrogen bonded, with a wide discontinuity between strands D & E1. A short 310 helix, leading into strand A, partly closes off one end of the barrel. In the FABPs the region between strands A and B form two short but well defined a-helices which pack onto the top of the lipid binding site. If considered as a loop connecting two strands then this feature is very similar to loop L1 of the lipocalin fold (also a large loop folded back to cap the internal cavity).

The avidin fold is dominated by 8 b-strands which form an antiparallel b-barrel enclosing the internal biotin binding site [2]. These 8 strands are linked by a succession of seven +1 connections (L1 to L7). These loops, although they vary in size and conformation, are all typical of short b-hairpins. In the lipocalins and FABPs, loop L1 is a large, atypical W-loop, while in the avidins loop L1 is, by contrast, a short, typical b-hairpin. A short 310 helix, directly before strand A, helps to close off one end of the barrel.

The structures of a lipocalin, human retinol binding protein, and a FABP, rat intestinal fatty acid-binding protein (I-FABP), can be superimposed onto that of streptavidin [2]. The topology of all three structures is very similar: each form antiparallel b-barrels with repeated +1 connections, although the FABP barrel is more flattened or elliptical than that of the lipocalins and the avidin barrel is both more circular and compact, in cross-section, than that of the lipocalins. The three structures can be equivalenced so that the initial four and final two strands of each barrel match to form contiguous structurally conserved regions (SCRs), with little or no correspondence between strands from the centre of their closed b-sheets or between any intervening loop regions. Such structural alignments show that the first SCR of the calycin common core has the same unusual conformation (a 310 helix leading into a b-strand) and location within the fold of the lipocalins, FABPs, and avidins. It contains within it a highly conserved sequence motif common to all three families.

Sequence relationships

In recent years, the apparent size and diversity of both the lipocalin and FABP families has grown significantly, as new members are proposed on the basis of sequence similarity, to encompass a large number of proteins. In contrast, the number of extant avidin sequences remains small.

The common core of the lipocalin fold is dominated by three large SCRs, which bear characteristic, conserved sequence motifs. Using possession of these three key motifs as a definition of family membership allows identification of the largest self-consistent subset within the whole set of related sequences [1,3]. This analysis shows the lipocalin protein family to be composed of a main core group of quite closely related proteins: the kernal lipocalins, and a smaller number of more structurally, and possibly functionally, divergent sequences (such as the a-1-acid glycoproteins, odour binding proteins, and Von Ebner's gland proteins): the outlier lipocalins. It may be misleading to say a protein is a lipocalin, with all that that implies for shared structure and function (such as cell surface receptor binding), because it has some global sequence similarity to an established member of the family; rather, for certain inferences, particularly structural ones, to be drawn with confidence from such sequence similarity, all three key characteristic motifs must also be present.

Moreover, within the lipocalin fold, residues from both SCR1 and SCR2 make several specific contacts with a conserved feature near the end of SCR3: the terminal residue of strand G. This is an arginine in most lipocalins and a lysine in some others. The end of this side chain makes hydrogen bonds to a ring of acceptors, main chain carbonyls derived from either the 310-helix leading into strand A or from the apex of loop L6, and its side chain packs, in an extended conformation, against the totally conserved tryptophan at the start of strand A. These highly conserved residues form a characteristic patch on the surface of the proteins [1].

A similar arrangement of side chains interactions - a tryptophan, from the first strand of the b-barrel, packing against an arginine from near the end of the last - is also observed in the FABP structures, although this interaction is not as extensive as that seen in the lipocalins, for example there is no equivalent to the SCR2 interaction.

Examination of the available structures of the chicken avidin and streptavidin reveals that a very similar arrangement of interacting residues is also, seemingly, a conserved feature of the third member family of the calycins.

It is clear then that in all three families comprising the calycin superfamily, a similar pattern of packing between characteristically conserved residues is apparent. An arginine or lysine, able to form a number of potential hydrogen bonds with the main chain carbonyls of a short 310 helix, packs across a conserved tryptophan [4]. This conservation of particular residues and their interactions across the three member families suggests that this structural signature, seemingly characteristic of the calycin superfamily, might also correspond to a characteristic sequence signature. To test this contention, two short sequence motifs, including the residues participating in these key interactions, were identified within the first and last SCRs of the overall calycin common core and used as the basis for a regular expression summarising the different residue types observed at each alignment position. Combining these with motif separations gives a search pattern whose performance compares favourably with those of much more sophisticated search methods [5,6].

The implications of this work are clear, a structural signature characteristic of almost the whole calycin family corresponds to a sequence pattern able to identify a significant subset of their sequences. These conserved side chain-side chain and side chain-main chain interactions act to pin together the two ends of the calycin b-barrel. But what is the role of this key structural motif ? It may stabilize the overall protein structure and help maintain the b-barrel fold, it may guide its formation as a key interaction in the folding pathway or it may have some functional role, perhaps as a protein-protein recognition site. Alternatively, in these present day structures, the sequences that give rise to the individual folding patterns of each family may have evolved stability independent of this residue packing and all we see, in this interaction, is an echo of an important past interaction; still favourable and / or important but no longer essential: an evolutionary relic or leftover of the now distant common calycin ancestor protein.

Discussion

Beyond an obvious similarity of function, the triumvirate of protein families composing the calycin superfamily is characterized by a similar overall folding pattern (an antiparallel b- barrel, with a repeated +1 topology, possessed of an internal ligand binding site), within which large amounts of the lipocalin, FABP, and streptavidin structures can be meaningfully equivalenced to form a number of significant discrete structurally conserved regions [1,2]. This is indicative of a close relationship between the characteristic folds of these families. The first SCR, within the common core characteristic of the three families, corresponds to an unusual structural feature (a short 310 helix leading into a b-strand, the first of the barrel), which is conserved in its conformation and location within the fold of the lipocalins, FABPs, and avidins and displays a characteristic pattern of sequence conservation. Notwithstanding the lack of significant global sequence similarity between any of these three groups, this motif is present in virtually all of the calycins.

Protein structures are rather better conserved than their sequences and in extreme cases close relationships are not apparent from the analysis of sequence alone: two views of this phenomenon have emerged. One maintains that although structural propinquity has been retained, evolutionary divergence from a common ancestor has occurred beyond which any significant sequence similarity remains. The other proposes a convergent mechanism whereby structures resemble each by chance because of the limited number of thermodynamically stable folds. In the calycins, the retention of characteristic and conserved sequence signatures corresponding to a common pattern of side chain packing argues for the former explanation and supports the conjecture that the lipocalins, FABPs, and avidins share an evolutionary relationship [4].

References

1 Flower, D.R., North, A.C.T., and Attwood, T.K. (1993), Structural and sequence relationships in the lipocalins and related proteins, Prot. Sci. 2, 753-761.
2 Flower, D.R., (1993), Structural relationship of streptavidin to the calycin protein superfamily, FEBS Letters, 333, 99-102.
3 Flower, D.R., North, A.C.T., and Attwood, T.K. (1991), Mouse oncogene protein 24p3 is a member of the lipocalin protein family, Bioch.Biophys.Res.Comm. 180, 65-74.
4 Flower, D. R. (1995), A Structural Signature characteristic of the Lipocalin Protein Family, Prot. Pep. Lett. 2, 341-346.
5 Sansom, C.E., North, A.C.T., and Sawyer, L. (1994), Bioch. Biophys. Acta 1208, 247-255.
6 Johnson, M.S., Overington, J.P, and Blundell, T.L. (1993), J. Mol. Biol. 231, 735-752.

Database Searching for Inhibitors of Secretory Phospholipase A2 using CLIX

As part of an ongoing program to devise specific PLA2 inhibitors as anti-inflammatory agents we have used CLIX [1] to find potential new ligands for secretory PLA2 (sPLA2) from the Cambridge Structural Database (CSD) [2]. CLIX represents a novel approach to database searching. Rather than using a simple pharmacophore to define a search, CLIX uses the protein structure to define an array of target sites that are used to find complementary structures in the database. The hits were scored using GRID version 10.0 [3].

In order to be confident that the software was working in our hands we used it to reproduce the crystal structure of human sPLA2 complexed with a transition state analogue, L-1- O-octyl-2-heptylphosphonyl-sn-glycero-3-phosphoethanolamine [4]. The ligand was reduced in size and complexity in order to provide a more rigorous test of the method. To include the full structure would almost certainly preclude alternative orientations of the ligand within the active site and therefore provide an unrealistic estimate of the predictive power of the software. A CSD format database containing this one structure was formed and searched using CLIX.

Having satisfied ourselves that the program is producing sensible results we then searched the CSD. Steric complementarity of the first hit is good, and the oxygen of one of the carboxylate groups is placed adjacent to the calcium ion. The interaction energy used for scoring will be dominated by the presence of the calcium ion. This is why all the hits are carboxylic acids.

CLIX has successfully reproduced the inhibitor-ligand crystal structure and retrieved convincing hits that provide new starting points for synthesis. We intend to convert other commercial and in-house databases into CLIX format and also to convert target points into a pharmacophore for conventional database searching. An obvious refinement of this whole approach would be to construct a conformational database first.

References

1 Lawrence,M.C. and Davis, P.C., PROTEINS:Structure,Function & Genetics (1992), 12, 31-41.
2 Cambridge Structural Database System, Cambridge Crystallographic Data Centre, 12 Union Road, Cambridge, CB2 1EZ.
3 Goodford, P.J., J.Med.Chem. (1985), 28, 849-857.
4 Scott, D.L., White, S.P., Browning, J.L., Rosa, J.J., Gelb, M.H. and Sigler, P.B., Science (1991), 254, 1007- 1010.