To identify potential novel versions of the β-GF that bind soluble ligands, we initiated comprehensive structural comparisons with various previously characterized members of the fold (see β-grasp fold in SCOP database) using the DALI program. Several of these searches retrieved the C-terminal domain of the transcobalamin protein with significant Z-scores. For example, searches initiated with MoaD proteins (PDB: 1v8c, 1vjk) retrieved the C-terminal domain of transcobalamin (PDB: 2bbc) with Z-scores of ~7. Transcobalamin is an animal-specific protein that binds cobalamin (vitamin B12), and is involved in its uptake by animal cells 9. Transcobalamin contains an N-terminal α/α toroid domain, and a C-terminal α/β domain 10 that corresponded to the β-GF domain recovered in the above searches. Further, DALI searches initiated with the C-terminal domain of transcobalamin recovered a diverse set of previously known β-GF domains such as MoaD (PDB:1vjk), YukD (PDB: 2bps) 2Fe-2S ferredoxin (PDB:1feh) and the middle domain of the Nqo1 subunit of the bacterial and mitochondrial NADPH-quinone oxidoreductase I complex (PDB: 2fug. S chain) with Z-scores in the range of 5–7. The structural alignments generated by these searches showed that the transcobalamin C-terminal domain aligns completely with all core structural elements of the β-GF, including a β-sheet of 5 strands and a helix between strands 2 and 3. However, the transcobalamin C-terminal domains are distinguished by the presence of a unique β-hairpin after the conserved helix of the β-GF (Figure 1A). The N-terminal α/α toroid domain and the C-terminal β-GF domain cooperate in ligand-binding by sandwiching a single B12 molecule between them 10. Systematic searches for contacts between the B12 ligand and the C-terminal β-GF domain in transcobalamin showed that the unique insert plays a prominent role in binding the ligand by contributing several direct or solvent-mediated interactions 10. Additional contacts with the ligand are also made by residues from the core β-GF such as those from strand 3, the end of strand 4 and the "ascending connector" between strand 4 and 5 (Figure 1A). These observations suggest that the C-terminal domain of transcobalamin represents a novel adaptation of the β-GF for small-molecule ligand interactions.

To better understand the diversity of this class of ligand-binding β-GF domains and their phyletic spread we initiated sequence profile and hidden Markov model (HMM) searches for homologs using PSI-BLAST and the HMMER package respectively. In addition to orthologs of transcobalamin, intrinsic factor and solo C-terminal domains from fishes, these searches retrieved numerous prokaryotic proteins, which were either present as stand-alone β-GF domains or in large multidomain proteins. For example, a search initiated with the β-GF domain of puffer fish transcobalamin (Tetraodon nigroviridus, gi: 47226456, region: 325–425) recovered closely related eukaryotic orthologs and paralogs fused to N-terminal α/α toroid modules (iteration 1), solo transcobalamin C-terminal domains with predicted signal peptides (e.g. XP_689937, Danio rerio, iteration 2, e-value: 3 × 10-12), and several prokaryotic proteins (e.g. BAC13773, Oceanobacterium iheyensis, iteration 3, e-value: 2 × 10-3). In order to exhaustively recover all divergent homologs, we conducted transitive searches with all above-detected members and also evaluated all hits below the threshold of PSI-BLAST searches for the presence of potentially homologous domains. We also prepared HMMs and PSSMs from the alignment of this region of all proteins recovered with significant expect-values (e < .01 with statistical correction for compositional bias) and used these to search all completely sequenced genomes. These searches consistently retrieved hits to multiple sequence repeats in a group of bacterial cell-surface/secreted sugar-binding proteins involved in polysaccharide export with significant e-values (e.g. Hahella periplasmic protein, HCH_02380 residues 852–990). Inclusion of polysaccharide export proteins in profiles for further searches additionally recovered the N-terminal region of the ComEA family of DNA uptake receptors of Gram positive bacteria (e.g. Clostridium ComEA, gi: 67874543, iteration 2, e = 10-6), PduS-like cobalamin reductases (e.g. E. coli cobalamin reductase iteration 7, e = 10-3), the middle domain of the 51 kD subunit (F chain) of the NADPH-quinone oxidoreductase complex I (Nqo1, E: 10-4, iteration 11) and the RnfC subunit of the oxidoreductases encoded by the bacterial Rnf operons 11(Rhodobacter RnfC, E: 10-6; iteration 14). This latter set of proteins was more similar to the homologous region recovered in the polysaccharide export proteins than to the transcobalamin C-terminal domain (Figure 1B). However, recovery of the middle domain of Nqo1 in sequence searches clearly confirms their relationship with transcobalamin C-terminal domains, because the former are also known, via structural analysis, to contain a similar β-GF domain 12 (See above and Additional file 1). This was additionally supported by separate secondary structure prediction for individual sub-groups with potentially homologous regions such as the ComEA N-terminal regions and the polysaccharide export proteins (Figure 1B).

We hereinafter refer to the homologous β-GF domains found in all these proteins as the

S

L

B

β

A comprehensive multiple alignment for the SLBB superfamily (Figure 1B) was prepared by combining alignments for individual groups constructed using the T-Coffee program, based on the structural superposition of transcobalamin C-terminal domain (2bbc) and Nqo1 middle domain (2fug; chain S). Much of the conservation seen across the entire superfamily is in the form of hydrophobic residues forming the stabilizing core of the fold. However, there was a notable sequence feature in the form of two strongly conserved glycine residues, one in the turn leading into the horizontal flange preceding the third β-strand (Figure 1A) of the β-GF and the other immediately downstream of the second conserved β-strand (Figure 1). This conservation pattern is a unique feature of the SLBB superfamily that distinguishes them from all other previously characterized β-GF domains, supporting a common ancestry for this set of domains within the β-GF.

The alignment also helped us to classify the SLBB superfamily into several distinct families. The Transcobalamin C-terminal domain clade is unified by the presence of the above-described β-hairpin insert within the β-GF that plays an important role in contacting the ligand (Fig 1A, see Additional file 1). This β-hairpin contains a conserved hydrophobic position that makes a stacking interaction with the aromatic ring of the base in cobalamin. However, the rest of the sequence in this region is poorly conserved as most other interactions occur through backbone oxygen or nitrogen atoms 10 (Figure 1B). Within animals, insects and most vertebrates have a single ortholog of the B12 binding protein, whereas the mammals have three distinct versions, transcobalamin I, transcobalamin II and the intrinsic factor. Besides animals, members of this clade are widely represented in Low GC Gram-positive bacteria and planctomycetes and less frequently in the euryarchaea.

The Nqo1-like clade includes at least two distinct families: 1) the first includes the NADPH-quinone oxidoreductase complex I subunit Nqo1 (51 kD/F chain), the RnfC oxidoreductase subunit, and the PduS-like cobalamin reductases. 2) The second family contains polysaccharide export proteins and the DNA receptor ComEA. This clade is unified by the presence of a small, often α-helical insert, in the "connector arm" between the fourth and fifth strands of the domain (Figure 1, see Additional file 1). In some cases, such as the ComEA protein, the helical segment is followed by a low complexity region; suggesting the presence of a disordered, extended loop. These proteins are also characterized by an sGG motif (where 's' is any small residue) around the second conserved glycine of the superfamily (Figure 1B). The Nqo1 subunit of the classical NADPH-quinone oxidoreductase complex I is present in all major bacterial lineages with well-developed electron-transport chains, in most mitochondriate eukaryotic lineages, and very rarely in euryarchaea. The RnfC proteins are strongly conserved in γ-proteobacteria, but are also found in some representatives of Low GC Gram positive bacteria and the Bacteroidetes/Chlorobi assemblage. The PduS protein is restricted to the Low GC Gram-positive bacteria and certain γ and δ proteobacteria. The ComEA proteins and polysaccharide export proteins show a nearly mutually exclusive complementary distribution. The ComEA family is chiefly present in Low GC Gram-positive bacteria and actinobacteria, whereas the polysaccharide export family is more widespread and widely present in proteobacteria, cyanobacteria, acidobacteria, planctomycetes, bacteroidetes/chlorobi, and more sporadically in a few other groups.

While the interaction between B12 and the transcobalamin-like SLBB domain involves the unique β-hairpin insert, key contacts are also contributed by the core fold (See above, Figure 1A), and in general the position of the bound ligand is comparable to that of the bound metal-sulfur cluster in the β-GF ferredoxins. The Nqo1-like clade shows its distinctive innovation in the region between strands 4 and 5, which also corresponds to the same general spatial location where the ligands are bound in the transcobalamin-like clade and β-GF ferredoxins (Figure 1B). This indicates that the structural innovation specific to the Nqo1-like clade might also be involved in binding a ligand at a similar position. This spatial location might thus represent a common site for soluble ligand interactions in the β-GF that is distinct from the C-terminal tail and the opposite protein surface that is key to the functional interaction of sulfur carriers like ThiS and MoaD and the ubiquitin-like proteins 13.

To investigate the functional diversification of the SLBB fold we used contextual analysis, which often provides insights into biochemical functions of poorly characterized protein domains or genes. Contextual analysis utilizes the information gleaned from the association of uncharacterized domains with other domains of known function and the tendency of genes whose products functionally interact to associate in conserved gene neighborhoods or predicted operons. 141516.

Most members of the transcobalamin C-terminal domain clade of the SLBB superfamily contain signal peptides, and several also contain the C-terminal Gram-positive anchor motif 17, suggesting that they are secreted or cell-surface proteins. A common domain architecture in this clade encountered in both eukaryotes and bacteria is the fusion of the SLBB to an α/α toroid domain. In bacteria the toroid may be present either N-terminal (e.g. Desulfotomaculum, gi: 88945170) or C-terminal (e.g. Bacillus, gi: 42782379) to the SLBB (Figure 2A). As the central cavity formed by the α/α toroid in transcobalamin plays a major role in binding B12 10, it is likely that the two domains cooperate in binding B12 in all these proteins. Additional architectures include fusions to domains typically found in extracellular proteins, such as one or more immunoglobulin-fold domains (e.g. Archaeoglobus; gi: 11498993 and Moorella; gi: 83590303), the FIVAR (Pfam entry: PF07554) sugar-binding domain (Clostridium, gi: 28210467), the fasciclin domain (Methanosarcina; gi: 21228740) and a β-propeller domain (Clostridium, gi: 28210494). Given that many of these domains are often involved in interactions with polysaccharides, they might play a role in tethering these proteins to the cell surface by binding peptidoglycan or capsular polysaccharides 181920212223. Often these multi-domain SLBB proteins occur in conserved operons that might additionally code a second paralogous extracellular SLBB protein (Figure 2). This might imply that different extracellular SLBB proteins interact together to form protein complexes on the cell surface. Interestingly, an analysis of the B12 biosynthesis pathways of all the bacteria that possess proteins with such SLBB domains showed they usually lacked key biosynthetic enzymes for B12. Furthermore, these SLBB proteins are generally encoded by predicted operons that also contain genes for CbiO-like ABC ATPase and the CbiQ-like integral membrane protein implicated in cobalt transport 24. These observations suggest that the primary role of this clade of SLBB proteins might be to scavenge B12 or its precursors from the environment. As the archaea which contain these SLBB proteins often possess an anaerobic B12 synthesis pathway, it is possible that these might instead be involved in scavenging a distinct metabolite. In Syntrophomonas the SLBB domain is found in putative extracellular enzymes fused to sulfite oxidase-like molybdopterin cofactor binding domain (e.g. gi: 71491441) 25. It is likely that in these proteins the SLBB provides a B12 cofactor that might be required by these enzymes.

Intracellular versions of the transcobalamin-like clade show fusions of the SLBB domain with two distinct winged HTH domains, namely those of the ArsR-like (e.g. gi:72395507, Methanosarcina) and AraC-like families (E.g. gi: 86604362, Lactobacillus) (Figure 2A). These proteins probably function as one-component transcription factors that respond to concentration of B12, its precursors or some other as yet unknown soluble ligands.

In the Nqo1-like clade, polysaccharide export proteins are predicted to be secreted or periplasmic proteins and contain an absolutely conserved N-terminal β-strand-rich domain followed by 1–8 repeats of the SLBB domain (Figure 2A). They appear to be part of a larger complex that is involved in transport of polysaccharides to the cell surface and are believe to associate with the outer membrane and periplasmic space in proteobacteria 26. Conserved gene-neighborhoods that encode these proteins are populated with proteins involved in the biosynthesis and modification of sugars or polysaccharides, which is consistent with their role in polysaccharide export (Figure 2). The related ComEA proteins of Gram-positive bacteria also contain a signal peptide followed by an N-terminal SLBB domain that is always fused to a pair of DNA-binding Helix-hairpin-Helix domains at their C-terminus. This is consistent with the role of the ComEA protein as a non-specific DNA receptor in the transformation competence mechanism of Gram-positive bacteria 2728. Prior studies suggest that this DNA receptor may be linked to the cell surface via the N-terminal region spanning the SLBB domain 27. Taken together these observations suggest that the SLBB domain in these proteins is likely to be critical for interaction with cell polysaccharides and/or sugars of the peptidoglycan. The complementary phyletic distribution of ComEA and polysaccharide export proteins is strongly correlated with the presence or the absence of the specialized Gram-positive cell wall (See above). This suggests that they probably diverged from a common ancestral polysaccharide/sugar-binding domain that was originally involved in uptake or extrusion of large molecules at the cell surface.

The remaining three groups of proteins, namely NqoI, RnfC and PduS, within the Nqo1-like clade of the SLBB superfamily share a common architectural core consisting of a fusion between an N-terminal Rossmannoid domain and an SLBB domain. Unlike classical Rossmann fold domains of oxidoreductases with 5–7 strands, the Rossmannoid domain of these proteins has a 4-stranded core in a 3214 order 12 coupled to a N-terminal two-stranded hairpin contributed by a module similar to the BBMs of RNA polymerases 29. The SLBB and this Rossmannoid domain are additionally combined to a variety of other domains in NqoI, RnfC and PduS proteins. The most common fusions seen in all three groups of proteins are those to 4Fe-S ferredoxin domains that flank the above two-domain core. The NqoI family also might contain a C-terminal tetrahelical bundle with an up-and-down topology that coordinates an Fe-S cluster via conserved cysteine residues, which in addition to the 4Fe-S ferredoxin is likely to provide an additional redox center for electron transport 12. A biotin/lipoate carrier-like Sandwich Barrel Hybrid Motif (SBHM) domain 29 is found respectively at the C- and N-termini of members of PduS and RnfC families.

The roles of these versions of the SLBB remain enigmatic; however there is evidence that the PduS protein might bind soluble ligands. The PduS gene typically belongs to a large mobile operon coding proteins required for the biogenesis of carboxysome/polyhedral bodies, which contains enzymes involved in propanediol degradation. The PduS has been shown to strongly bind cob(I)alamin and was characterized as a bifunctional cob(II)alamin and hydroxycobalamin (cob(III)alamin) reductase catalyzing the formation of cob(I)alamin. Cob(I)alamin is the immediate precursor of Ado-cobalamin, which serves as an essential coenzyme for the diol dehydratase in degradation of 1,2-propanediol 3031. It is likely that the SLBB domain in PduS, like that in transcobalamin, binds cob(I)alamin or HO-cobalamin, while the N-terminal Rossmanoid domain binds the flavin nucleotide cofactor for the redox reaction. Such a function is also supported by the observation that cob(I)alamin is highly reactive and needs to be shielded from the environment 30. The role of the fused SBHM domain seen in PduS proteins is less clear. However, given that the SBHM domain carries covalently associated ligands such as biotin/lipoate 3233, it might similarly carry cofactor ligands or intermediates in propanediol degradation such as 1,2-propanediol-1-yl radical 34. There is currently no evidence for a soluble ligand interacting with the related SLBB domain in the RnfC and NqoI. Nevertheless, crystal structures indicate an exposed location for the SLBB domains in these proteins, allowing the possibility that they might be allosterically regulated by hitherto unknown ligands interacting with this domain.