Bacterial Lipocalins: Origin, Structure, and Function

Bacterial Lipocalins Are Codistributed with Outer Membrane Proteins

In the year 2000, 20 lipocalins were identified in Gram-negative bacteria, but that number now exceeds 90. Among the newly identified sequences are three that belong to Gram-positive bacteria, namely, Nocardia and two species of Corynebacteria. These organisms belong to the so called CMN branch (for Corynebacteria, Mycobacteria, Nocardia) of the actinomycetes. The CMN branch represents the most highly derived of the Gram-positive bacteria and do not actually display Gram-positive staining. Instead, the presence of mycolic acids in these organisms generates a permeability barrier extrinsic to the inner membrane that requires staining by the so-called “acid-fast” procedure. The CMN organisms are regarded by some as being intermediate between Gram-negative and Gram-positive bacteria, because the mycolic acid layer requires porin channels to facilitate nutrient uptake and these adopt a transmembrane β-barrel architecture¹³ - a decidedly Gram-negative characteristic.¹⁴ Despite this similarity, the 16s rRNA phylogeny places the CMN branch firmly within the Gram-positive actinomycetes. Therefore, the bacterial lipocalins are associated with organisms that produce either LPS or mycolic acids, the two classes of lipids associated with bacterial permeability barriers that harbor β-barrel membrane proteins. Although lipocalins do not appear among the Mycobacteria, the fatty acid binding-protein fold that resembles lipocalins has been identified in these organisms (Pedro Alzari, personal communication). The mycolic acids of Mycobacteria are unusually long and create a thicker permeability barrier compared to Nocardia and Corynebacteria. To date, no lipocalins have been identified among the Gram-positive or archaebacterial organisms that lack the defining characteristics of outer membranes.

Not All Bacterial Lipocalins Are Bacterial

BLAST searches using E. coli Blc as a probe can usually discern between bacterial and eukaryotic lipocalins, because the latter sequences display a number of obvious insertions/ deletions. For example, apolipoprotein D (ApoD) and lazarillo, members of a group of eukaryotic lipocalins closely related to Blc, possess an insertion between β-strands G and H that is absent from Blc.¹² However, several eukaryotic lipocalins cannot easily be distinguished from the bacterial homologues (Table 1). Most striking among these are a group of six bacterial-like lipocalins from plants. Chromosome 5 of Arabidopsis thaliana encodes a lipocalin that appears to contain a single 84 nucleotide intron with the typical 5'-donor and 3'-acceptor splice sites. Removal of the intron yields a 186 amino acid polypeptide, which exhibits greatest similarity with the bacterial lipocalins. Plant genomes are unique among eukaryotes in that some encode most of the enzymes needed to make the hydrophobic anchor of Gram-negative LPS known as lipid A or endotoxin.¹⁵ Green plants are thought to have acquired plastids by a recent endosymbiosis with Gram-negative cyanobacteria, and might consequently have retained some outer membrane components.¹⁶ These same components were apparently lost from the mitochondrion after it was derived by an earlier endosymbiosis with a Gram-negative α-proteobacterium.¹⁷ Indeed, at least one lipocalin is encoded in the genome of the cyanobacterium G. violaceus. The six bacterial-like lipocalins of plants lack any obvious signal sequence, suggesting they are soluble cytosolic proteins. The eukaryotic acquisition of bacterial genes that code for cytosolic functions is now thought to be an inevitable consequence of endosymbiosis.¹⁸ Another bacterial-like lipocalin is derived from the marine yeast D. hansenii, and this protein exhibits a typical eukaryotic signal peptide that likely targets it to the endoplasmic reticulum.

Two Sub-Clades of Bacterial Lipocalins

The bacterial lipocalins are defined as belonging to Clade I,¹⁹ but we can divide them further into two groups distinguished by the absence (Clade IA) or presence (Clade IB) of a lipocalin subgroup that displays a pair of cysteine residues flanking β-strands B and C in an apparent disulfide-bonding configuration (Fig. 1). The bacterial-like lipocalins from plants belong to a subgroup within Clade IB that appears to have lost this cysteine pair, whereas the D. hansenii lipocalin displays the apparent disulfide. These observations can be rationalized with the reducing cytosolic and oxidizing extracellular environments to which these lipocalins are targeted. Inexplicably, some cytosolic lipocalins within Clade IB still retain this pair of cysteine residues. The two nonconsecutive disulfides found in the closely related group of eukaryotic lipocalins that include ApoD and lazarillo each appear to have their own antecedents within the B. fungorum2 and G. violaceus lipocalins from Clade IB (Fig. 2). Thus, it appears that Clade IB lipocalins are more closely related to the eukaryotic lipocalins than are those from Clade IA. An interesting feature within Clade IA is the N-terminal fusion with a short-chain dehydrogenase/reductase domain in two Bacteroides lipocalins. Similar domains are employed for carbohydrate epimerization during the synthesis of sugars needed for LPS biosynthesis.²⁰ Intriguingly, the Chlorobium genome reveals examples of both Clade IA and Clade IB bacterial lipocalins, suggesting that an early gene duplication event produced these two extant groups.

Vertical versus Horizontal Descent

In a previous analysis, we interpreted the absence of a homologue from an archaebacterium as indicating that lipocalins could not have existed in the last common ancestor of the three domains of life, and that they were probably acquired in eukaryotes horizontally by endosymbiosis.¹² Horizontal transfer is probably partly correct, especially in light of the bacterial-like lipocalins of plants and yeast described here. However, a recent study from Rivera and Lake suggests that a genome fusion event occurred between an archaebacterium and a proteobacterium to create the first eukaryotic cell; this framework provides a more parsimonious explanation for the acquisition of most eukaryotic lipocalins by vertical descent from the proteobacterial component of the ancestral chimera.²¹ This model builds on ideas from Gupta,²² who proposes that monoderm Gram-positive bacteria are ancestral to both the monoderm archaebacteria and the diderm Gram-negative bacteria/CMN actinomycetes.²³ Our observations clearly fit with a model where the lipocalins represent a late development in bacterial evolution, having arisen in the diderm bacteria. However, this is not so late a development as to preclude ancestry to the first eukaryotic cell according to the archaebacterial-proteobacterial fusion model.²¹ The earlier three domain model predicts that proteins acquired vertically by eukaryotes (Eukarya) will likely have ancestors in the archaebacteria (Archaea) if ancestors already exist in the eubacteria (Bacteria).²⁴

Structural Similarities between PagP and Retinol Binding Protein

A striking similarity is seen between the structures of the β-barrel outer membrane protein PagP and the prototypic lipocalin serum retinol-binding protein (RBP).^25,26 Both PagP and RBP are 8-stranded antiparallel β-barrels with a deep lipid-binding pocket located inside one end of the barrel and a flanking α-helix at the other end (Fig. 3A and B). The structures superpose with the proper alignment of the β-strands (i.e., A-H of PagP superposes with A-H of RBP) with an r.m.s.d. of 2.9 Å over 93 Cα atoms. A structural similarity between the outer membrane protein OmpA and the lipocalins has also been noted, and several of the key differences between these proteins was pointed out, including the length of the barrels and the distinctly different protein interiors.²⁷ However, our case for similarity is greater because the PagP and RBP barrels have a higher percentage of alignable positions, and have similar inner lipid-binding pockets. Furthermore, in the lipocalins, a long loop spans strands A and B and typically forms a lid that partially closes the ligand-binding site.¹¹ It is intriguing that the long L1 loop from PagP is at an equivalent position and is strongly implicated in the function of this enzyme,^28,29 however, additional structural data will be required to test whether this loop interacts directly with ligand and/or the upper region of the barrel.

We cannot rule out the possibility that this structural similarity represents a case of convergent evolution. However, based on the absence of lipocalins from those bacteria that also lack β-barrel membrane proteins, we cannot currently exclude the hypothesis that PagP and lipocalins share a common ancestor. This hypothesis is supported by the observation that many bacterial lipocalins are anchored in the outer membrane as lipoproteins.² The lipocalin-lipoprotein might represent an intermediate state in the adaptation between membrane-bound and soluble globular domains. Outer membrane β-barrel domains should be suitably adapted to occupy both soluble and membrane-bound conformations because both environments are encountered during the outer membrane assembly process.⁶

Structure and Function of the Bacterial Lipocalin Blc

E. coli Blc was recently cocrystallized in the presence of the 18-carbon unsaturated fatty acid known as vaccenic acid, and data collected at 1.8 Å resolution (V. Campanacci, R.E. Bishop, L. Reese, S. Blangy, M. Tegoni, and C. Cambillau; submitted). The complex is isostructural with that of native Blc and two monomers are contained in the asymmetric unit. The Blc monomer has a typical lipocalin fold consisting of a β-barrel with eight anti-parallel strands and an a-helix at the C-terminus. Inside the cavity of monomer B, a well defined elongated electron density could easily accommodate a vaccenic acid molecule (Fig. 4A). The molecule is bent at the position of the cis-double bond (Z11-12). Careful examination of the relationship between monomers A and B revealed that they interact tightly together: the interaction covers 786 Ų and 825 Ų of water accessible surface (WAS) area, on monomer A and B, respectively. Given that the total WAS area of each monomer is 7800 Ų, the buried WAS area represents ˜10% of the total, a value indicating that dimerization is not a crystallization artefact.³⁰ For comparison, these values of interacting surface area are comparable to those observed in immunoglobulin fragments/protein complexes.³¹ The interaction between the two monomers was present and essentially identical in the original native structure, but it escaped earlier observation and was not described.³

The comparison of subunits A and B indicate significant conformational differences that might be necessary for dimerization. In order to fit within the other monomer, strands F and G, and particularly the loops between them, move up to 5 Šin one monomer relative to the other. The dimer interface involves in large part these loops, and other residues among which the aromatics Tyr113, Tyr137, and several surface-exposed phenylalanines including Phe53 A & B, Phe108 A & B, Phe109 A & B and Phe112 A & B together form an inter-dimer hydrophobic core. The presence of many exposed phenylalanines is an uncommon feature at the surface of a monomeric globular protein, and has to be regarded as a hallmark of protein-protein or protein-lipid interactions. Both subunits are involved in the binding site of Blc, accounting for the stoichiometry of one vaccenic acid molecule per dimer. The vaccenic acid interacts with both subunits and covers 89 Ų and 171 Ų of the WAS area of monomers A and B, respectively.

It should be stressed that Blc is anchored to membranes by a covalently attached lipid modification at the amino terminus of the protein. The manner in which the lipid-anchor of each monomer is arranged relative to the other is a critical issue. The Blc construct employed has the signal peptide and the first four residues (CSSP) removed and replaced by the Gateway ATTB1 sequence.³ We have modelled into the dimeric structure the four original residues of Blc, as well as the major part of the anchoring lipid, S-sn-1,2-diacylglycerylcysteine, that we included at each N-terminal cysteine. This model reveals that the lipid-anchors of each of the 2 monomers are close to each other in the dimer, and are located on the same face in a geometry compatible with the insertion and stabilization of the dimer into the membrane (Fig. 4B). The binding site for vaccenic acid is located opposite to the membrane insertion site where it is expected to face the periplasmic space.

Tryptophan fluorescence quenching studies indicate that dimeric Blc binds fatty acids and phospholipids in a micromolar Kd range. An exposed and unfilled pocket seemingly suited to bind a polar group extending from the fatty acid prompted investigation of lysophospholipids (LPLs), which were found to bind in a nanomolar Kd range. Favorable steric and electrostatic interactions are observed when a glycerophosphoethanolamine moiety is modeled into this region of the structure. Given the high affinity of Blc for LPLs, it seems likely that Blc might fulfil a role in cell envelope LPL transport. Although LPLs are key inner membrane intermediates of phospholipid metabolism, we do not know of any evidence to indicate that they are exported to the outer membrane. However, exogenously supplied LPLs can be taken up by deep-rough LPS mutants and converted by reacylation into glycerophospholipids using inner membrane-associated enzymes.^32-34 To date, no accessory factors needed for LPL uptake have been identified. In wild-type cells, at least two enzymes can generate LPLs in the outer membrane, namely, the phospholipase OMPLA³⁵ and the lipid A palmitoyltransferase PagP.³⁶ Both enzymes preferentially generate the sn-1 LPL regioisomers, but these are known to spontaneously rearrange into the more stable sn-2 LPLs.³⁷ Only sn-2 LPLs were tested for binding to Blc.

The LPL products sn-2-lysophosphatidic acid and sphingosine-1-phosphate, together with the structurally related platelet activating factors, have been shown to function as potent biological mediators.^38,39 In eukaryotic cells, sn-2 LPLs can be generated by phospholipases that mobilize arachidonic acid for eicosanoid-mediated signal transduction. Additionally, sn-2 LPLs are enzymatic products of the lecithin:cholesterol acyltransferase LCAT, which is, together with ApoD, a key component of the plasma high density lipoprotein particle. Mammalian ApoD and Drosophila lazarillo are the closest eukaryotic homologues of Blc and the only eukaryotic lipocalins that, like Blc, are anchored to lipid membranes. The ligand for lazarillo is unknown, but ApoD binds a variety of ligands including arachidonic acid.⁴⁰ Perhaps the high affinity of Blc for LPLs might help shed some light on the enigmatic functions of lazarillo and ApoD.