ABC Transporters Involved in Export of Cell Surface Glycoconjugates

Abstract

Summary: Complex glycoconjugates play critical roles in the biology of microorganisms. Despite the remarkable diversity in glycan structures and the bacteria that produce them, conserved themes are evident in the biosynthesis-export pathways. One of the primary pathways involves representatives of the ATP-binding cassette (ABC) transporter superfamily. These proteins are responsible for the export of a wide variety of cell surface oligo- and polysaccharides in both Gram-positive and Gram-negative bacteria. Recent investigations of the structure and function of ABC transporters involved in the export of lipopolysaccharide O antigens have revealed two fundamentally different strategies for coupling glycan polymerization to export. These mechanisms are distinguished by the presence (or absence) of characteristic nonreducing terminal modifications on the export substrates, which serve as chain termination and/or export signals, and by the presence (or absence) of a discrete substrate-binding domain in the nucleotide-binding domain polypeptide of the ABC transporter. A bioinformatic survey examining ABC exporters from known oligo- and polysaccharide biosynthesis loci identifies conserved nucleotide-binding domain protein families that correlate well with themes in the structures and assembly of glycans. The familial relationships among the ABC exporters generate hypotheses concerning the biosynthesis of structurally diverse oligo- and polysaccharides, which play important roles in the biology of bacteria with different lifestyles.

INTRODUCTION

In prokaryotes, glycosylated macromolecules (glycoconjugates) often form the point of first contact between the organism and its immediate environment. Surface glycoconjugates play major roles in the maintenance of cell envelope structure, protection against host immune defenses, signaling events, and the formation of biofilms. The structures of bacterial glycans show tremendous diversity in terms of component sugars, noncarbohydrate modifications, and linkage configurations. They are also attached to a broad range of molecules, including lipids, peptidoglycan, and proteins.

Despite the remarkable diversity in oligo- and polysaccharide structures and the manners in which they are presented on the cell surface, the number of fundamentally different pathways used in polysaccharide synthesis and export across the inner membrane is relatively limited. For polysaccharides or oligosaccharides synthesized directly from activated precursors consisting of nucleotide mono- or diphosphoglycoses, there are currently just three known export strategies that have been described in any detail (Fig. 1); these have been termed “Wzx/Wzy dependent,” “ATP-binding cassette (ABC) transporter dependent,” and “synthase dependent,” based on characteristic components. These three general assembly-export strategies are not confined to the assembly of any one class of glycoconjugate. It is the steps that occur beyond synthesis and export (i.e., attachment of the glycan to a final acceptor molecule) that define the type of glycoconjugate that is formed. Most of our current understanding of these systems has been obtained in the context of lipopolysaccharide (LPS) O-antigen polysaccharide (O-PS) and capsular or exopolysaccharide (CPS or EPS) assembly, and these have been reviewed elsewhere (121, 173).

FIG. 1.

Models for processes involved in the synthesis and trans-cytoplasmic membrane export of bacterial glycans. In the Wzx/Wzy-dependent pathway, individual lipid-linked repeat units are synthesized by glycosyltransferase enzymes (illustrated in yellow in each model) located at the interface of the cytoplasm and the membrane. The lipid-linked repeat units are exported via the Wzx flippase (a putative transporter of unknown mechanism) and polymerized at the periplasmic face of the membrane. The polymer grows in a blockwise process, by addition of new repeat units to the reducing end of the glycan in a reaction requiring the Wzy polymerase. In the ABC transporter-dependent pathway, the chain is elongated by addition of monomers to the nonreducing terminus of a lipid-linked intermediate and is completed in the cytoplasm, prior to export via the ABC transporter. In the synthase-dependent system, a single protein is thought to serve as both a polymerase and an exporter, but the details of the export process are unknown. There is no unifying involvement of a lipid acceptor in the synthase pathway, and even the direction of chain growth may differ, depending on the system. Glycans (or glycoses) derived from these pathways can be attached to protein, lipid, LPS, or peptidoglycan acceptors. In the case of some glycoconjugates in Gram-negative bacteria, the final cellular location may depend on additional export pathways to transfer the molecule across the outer membrane.

In the Wzx/Wzy-dependent and ABC transporter-dependent pathways, the polysaccharide is built on a lipid acceptor. The activated precursors are cytosolic sugar nucleotides, and the glycosyltransferase reactions that transfer sugars to the lipid carrier occur at the cytoplasmic face of the membrane. Ultimately, both of these pathways lead to the formation of a lipid-linked polysaccharide located outside the cytoplasmic membrane. However, the intervening steps between chain initiation and completion are quite different. In the Wzx/Wzy-dependent pathway, individual undecaprenol diphosphate (Und-PP)-linked polysaccharide repeat units are assembled and exported across the membrane by a transport process requiring a Wzx protein homolog. Wzx is considered a “flippase,” but its precise mechanism of action is unknown. The newly exported lipid-linked repeat units then form the substrates for a polymerization reaction that requires a Wzy (putative polymerase) homolog, which extends the growing chain one repeat unit at a time at the periplasmic face of the cytoplasmic membrane. The polymerization process is controlled by a member of the PCP (polysaccharide copolymerase) family (32, 96). In contrast, polysaccharides assembled by ABC transporters are fully polymerized by sequential glycosyl transfer at the cytoplasmic face of the inner membrane. The glycan can be assembled as a Und-PP-linked intermediate, as is the case for most O-PSs (121). Alternatively, for some CPSs, it seems more likely that the acceptor is diacylglycerol phosphate (173). The completed molecule is then exported by the ABC transporter (see below), but it is not clear whether, under normal physiological conditions, chain extension must be completed before export can begin. Components of the Wzx/Wzy-dependent and ABC transporter-dependent pathways have been identified in both Gram-positive and Gram-negative organisms, and some aspects of the underlying biochemistry have been described. Significantly less is known about a third assembly process, the synthase-dependent pathway (Fig. 1). Synthases are processive glycosyltransferases and are involved in the formation of some important biological molecules, including bacterial cellulose (130), hyaluronan and chondroitin (in some producing organisms [171]), alginate (123), and poly-β-d-N-acetylglucosamine (GlcNAc) (60). In the synthase pathway, the nature of the acceptor on which the polymer grows is not always certain, nor is the process by which the polymer (or repeat units) is exported. It is possible that a single protein (the synthase) is sufficient for both polymerization and export activities in some cases (66, 171), but details of the process are unknown.

Other export-assembly mechanisms may emerge as more information becomes available. One candidate includes those processes where the direct donors for glycosylation are polyprenol-linked glycoses and one or more residues are transferred to an acceptor at the periplasmic face of the cytoplasmic membrane. Although limited in diversity, such systems have been identified in phage-mediated glucosylation of LPS O-PSs (50, 74) and in the modification of LPS lipid A by 4-amino-4-deoxy-l-arabinose (4-aminoarabinose) (reviewed in reference 116). In these pathways, a single glycose residue is exported, and in each case, a Und-P-linked glycose is the donor. This process is superficially similar to the activity of Wzx, but the mechanism(s) of export for the Und-P-linked glycoses is unknown. Although candidate flippases have been identified, there is no obvious similarity to other glycotransporters, including Wzx. Glycosylation by a Und-P-linked donor is also a hallmark of the biosynthesis of the long-chain arabinan part of the mycolyl-arabinogalactan-peptidoglycan complex in mycobacteria (reviewed in references 13 and 154). The donor is decaprenol-monophospho-arabinose (Dec-P-Ara), whose synthesis occurs without the involvement of sugar nucleotides, but the cellular compartment in which its activity occurs, and hence any requirement for a specific exporter, is currently unknown (see below).

Members of the ABC transporter superfamily participate in a wide range of glycosylation processes. The involvement of ABC transporters in glycan export was first identified in studies of CPS biosynthesis in Haemophilus influenzae (79) and Escherichia coli (109, 143). In the absence of the ABC transporter, polymer is synthesized by each of these systems, but it never leaves the cytoplasm. Later, a similar phenomenon was recognized in the biosynthesis of LPS O-PS in Yersinia enterocolitica O:3 (182). Subsequently, ABC transporters were identified in the assembly of teichoic acid in Bacillus subtilis 168 (83) and in the protein glycosylation system from Campylobacter jejuni (2). ABC transporters are now known to be involved in the export of representatives from all of the major classes of cell surface glycoconjugates. By mining genomic information to examine the distribution of “glyco-focused” ABC transporters, it is clear that they play pivotal roles in the glycobiology of many bacteria, even though there are many cases where the exact nature and structure of the export substrate have not been established. In each of these systems, the organism must overcome the same challenges. First, the export system must engage a biosynthesis pathway, often involving multiple proteins with different catalytic activities. Second, the exporter must handle a large hydrophilic (and sometimes charged) substrate, often attached to a hydrophobic lipid acceptor. Here we describe our current understanding of the structure and function of glycan ABC transporters and illustrate how bioinformatic analyses can extend this knowledge to other cell surface glycoconjugates.

STRUCTURE AND FUNCTION OF ABC TRANSPORTERS

ABC transporters represent a massive protein superfamily and play diverse roles in biological processes which span the biological kingdoms. In bacteria, ABC transporters are involved in the uptake of nutrients (e.g., some sugars, amino acids, and vitamins). They also participate in the export of molecules with remarkable structural diversity, including various drugs and other small-molecule inhibitors, proteins, lipids, and (in the context of this review) oligo- and polysaccharides.

ABC transporters contain four essential domains: two transmembrane domains (TMDs), which are integral membrane proteins, and two nucleotide-binding domains (NBDs) that associate with the TMDs on the cytoplasmic face of the membrane (Fig. 2 A and B). These domains can be organized as individual polypeptides or may be fused into multidomain proteins in a variety of formats. The TMDs of different transporters typically share low sequence similarity, and the number of α-helices varies, with 8 to 20 in importers and, typically, 12 in exporters (88, 97). The membrane-spanning α-helices of the TMDs form the transport channel, and the differences in the primary sequences of the TMDs may be a reflection of the wide variety of substrates used. In contrast to the sequence variability shared between the TMDs, the NBDs of ABC transporters possess the highly conserved sequences that define ABC proteins. The signature “LSGGQ motif” is used to identify members of the ABC transporter superfamily, and it resides between the Walker A and B motifs, which are found in a variety of ATP- and GTP-hydrolyzing proteins (167). ABC proteins also contain additional motifs (e.g., the Q and H loops) essential for ATP binding and hydrolysis (23, 136, 144, 179). Each NBD monomer forms two structural domains: a RecA-like domain, sharing structural similarities with other ATPases, and a smaller helical domain, which contains the signature motif. The ATP-binding sites are created at the interface of two NBDs, which align in the transporter in a head-to-tail arrangement (Fig. 2A and B). In some cases, the NBDs also contain accessory domains, which may aid in the transport of the substrate or be involved in regulatory mechanisms (14). A combination of structural information and biochemical evidence has led to proposals concerning the functions of the conserved ABC motifs in ATP hydrolysis and the mechanism by which ATP hydrolysis is coupled to substrate transport. In the working models, ATP binding induces conformational changes in the NBDs that force them into closer contact, creating the characteristic nucleotide sandwich (68). The ATP-binding sites are created by the juxtaposition of the Walker A motif of one NBD and the signature motif on its partner. Structures of various NBDs in nucleotide-bound and unbound states clearly reveal significant changes in the proximity of the NBDs (Fig. 2B).

FIG. 2.

Model for ABC exporters. (A) Structure of the complete Sav1866 multidrug exporter from S. aureus, illustrated from two angles rotated 90° around the vertical axis (35). Sav1866 is a half-transporter in which each subunit (identified by different shades of blue) contains one TMD fused to one NBD. The coupling helices that play a crucial role in opening the channel are highlighted in yellow. Although the Sav1866 crystal structure was obtained in the presence of ADP (magenta), its NBDs exhibit the sandwich dimer typically reserved for ATP-bound complexes (37, 144, 179). It has been suggested that during the crystallization process, the conformation of the transporter was shifted to the ATP-bound state (36). (B) Models of the open and closed conformations of ABC transporters. The model is based upon the information obtained from crystal structures of Sav1866 (open; outward facing) and ModABC (closed; inward facing), as described elsewhere (57). The model illustrates how ABC transporters may link ATP hydrolysis to distinct conformational changes in the transporter proteins. Conformational changes are transmitted from the NBDs with the hydrolysis of ATP to the TMDs through the interactions of the coupling helices (highlighted in yellow). ModBC is the molybdate/tungstate importer from Archaeoglobus fulgidus (57). The structural models were created from the structures under PDB accession numbers 2ONJ (Sav1866 with AMP-PNP) and 2ONK (ModBC), using PyMol (www.pymol.org/).

The bacterial multidrug exporter Sav1866 (35), the lipid A flippase MsbA (168), and the maltose transporter MalFGK₂E (105) have become influential models for ABC transporters in both prokaryotes and eukaryotes. The Sav1866 exporter from Staphylococcus aureus is a typical “half-transporter,” comprised of a homodimer in which each polypeptide contributes one TMD and one NBD (35) (Fig. 2A). The TMDs of the transporter consist of long helices that extend outward through the lipid bilayer. The TMDs interact with one another, forming two wings, with each one being composed of α-helices from each TMD subunit. The large pocket formed between the two TMDs in the crystal structure of Sav1866 is open to the periplasm and is thought to form the export cavity for the extruded antibiotic, as this part of the structure is not very hydrophobic and may therefore facilitate extrusion of hydrophobic substrates to the outer face of the membrane. The NBDs sit ∼25 Å away from the inner membrane due to long loops in the TMDs, and each TMD is in contact with both NBDs (35). TMD-NBD contacts are mediated by coupling helices in each of the TMDs and through the Q loop and the TEVGERV sequence of the X loop in the NBDs. The X loop precedes the ABC signature motif in the NBDs and is unique to exporters. It is important for the interactions of the NBDs with the TMDs (35, 103). In each TMD subunit, one coupling helix makes contacts with both NBD subunits, while the other makes contacts only with the NBD from the opposite subunit. The arrangement of ABC transporters involved in the export of substrates is thought to resemble that of the Sav1866 transporter (35). However, the mechanisms by which transporters work may vary, as recent evidence suggests that BtuCD (the vitamin B₁₂ importer) has a gate conformation distinct from those of other importers, such as those specific for molybdate (ModBC) (46).

In the maltose ABC transporter, two polypeptides (MalFG) provide the TMDs, while the NBD dimer consists of two copies of MalK. The activity of the transporter is aided by a periplasmic binding protein, MalE, and the structure of the entire complex has been solved (105, 106). TMD-NBD interactions in the MalFGK₂E complex are mediated by the EAA motifs in the MalF and MalG polypeptides (105). The EAA motif consists of two short helices. One of these is termed the coupling helix, but it is distinct from the coupling helices in Sav1866, and it interacts with the Q loop of a MalK monomer.

MsbA is another “half-transporter” and has been studied in detail (57, 136), with structures solved for multiple species and different conformational states (168). The varied conformational states are suggestive of a substantial structural change occurring during the catalytic cycle of the exporter, but this proposal remains to be confirmed in vivo. In the current working model, an ABC transporter cycles between the nucleotide-bound and unbound states. The separation of the NBDs is communicated to the TMDs via a coupling helix (Fig. 2B). The ensuing rigid-body motions in the TMDs result in a transition of the channel that alternately opens the lumen of the exporter to either the interior or exterior face of the cytoplasmic membrane. This is the basis of the alternate access model, in which the transporter switches between the “open” and “closed” states (122, 138). However, recent data also suggest that some ABC transporters do not fit within a single mechanistic model for generation of alternating access (23, 46, 69, 88).

Structures have now been solved for several ABC transporters, and a growing body of biochemical data is available. However, there remain a number of critical unresolved questions concerning their precise mode of action (62), particularly the molecular events that occur during the transition of the channel between open and closed states and during the energization of the transporter. ABC transporters are responsible for shuttling a vast variety of substrates across the membrane, from small peptides and drugs to long chains of polysaccharide or protein. It is still unclear how these molecules with diverse physical properties interact with the transporter. These questions will be answered only by further investigation using both biochemical and structural methods.

PHYLOGENY OF ABC TRANSPORTERS INVOLVED IN GLYCAN EXPORT

The similarities in NBD sequences mean that ABC transporters are readily identified from genome databases, and this has been exploited to examine the distribution of glyco-ABC transporters. Homologs were identified by key word and BLAST searches of the NCBI database, using known NBD proteins involved in oligo- and polysaccharide export. These included Wzt (O-PS), KpsT (group 2 CPS), TagH (teichoic acid), PglK (N-linked glycoprotein), and MsbA (LPS lipid A flippase) homologs. The genomic context of each of the NBD “hits” was examined to ensure that they were affiliated with authentic glycan biosynthesis loci. To do this, the open reading frames surrounding each of the identified NBD homologs were examined for the presence of diagnostic genes encoding glycosyltransferases or enzymes involved in sugar-nucleotide precursor biosynthesis. The NBDs of glyco-ABC transporters analyzed here are listed in Table 1. The conserved ATP-binding portion of each NBD was used to establish the phylogenetic relationships between them. MalK (the well-characterized NBD protein from the maltose transporter [37]) was included as the outgroup. NBD homologs were identified in both Gram-positive and Gram-negative bacteria, as well as in archaea, indicating the widespread involvement of ABC transporters in glycan export in prokaryotes. The phylogenetic analysis revealed seven distinct clusters of NBDs (Fig. 3), and similar phylogenetic tree topologies were observed using both distance and parsimony algorithms, providing support for the analysis. With the exception of NBD proteins found in group G, the glyco-ABC transporters consist of independent TMD and NBD polypeptides. An interesting, and potentially diagnostic, feature of the corresponding NBDs is that most contain a variant signature motif with the sequence YSSGM. In contrast, group G proteins are classical “half-transporters,” functioning as homodimers with each monomer containing one TMD and one NBD, and these contain the conventional LSGGQ signature motif.

TABLE 1.

NBD homologs from oligo- and polysaccharide biosynthesis systems

Species and NBD group	Strain or serotype	NBD	GenBank accession no.	Locus tag	Length (amino acids)	Organism classification
Group A
Aggregatibacter actinomycetemcomitans	Serotype e	ORF11	BAA82537		398	Gammaproteobacteria
Aeromonas salmonicida subsp. salmonicida	A449	AbcA	YP_001141286		438	Gammaproteobacteria
Aneurinibacillus thermoaerophilus	DSM10155/G+	Wzt	AAS49125		435	Firmicutes
	L420-91T	Wzt	AAS55714		408	Firmicutes
Aromatoleum aromaticum	EbN1		YP_157893	ebA1593	483	Betaproteobacteria
Arthrospira maxima	CS-328		ZP_03274399	AmaxDRAFT_3223	472	Cyanobacteria
Aurantimonas manganoxydans	SI85-9A1		ZP_01226056	SI859A1_02283	424	Alphaproteobacteria
Burkholderia ambifaria	MC40-6		YP_001807479	BamMC406_0767	406	Betaproteobacteria
Burkholderia cenocepacia	J2315	Wzt	YP_002232239	BCAL3130	437	Betaproteobacteria
Burkholderia phymatum	STM815		YP_001858534	Bphy_2313	450	Betaproteobacteria
Burkholderia pseudomallei	668		YP_001060111	BURPS668_3099	465	Betaproteobacteria
Clostridium botulinum	B strain Eklund 17B		YP_001887433	CLL_A3247	406	Firmicutes
Enterococcus faecalis	V583		AAO81914	EF_2182	405	Firmicutes
Escherichia coli	O8	Wzt	BAA28325		404	Firmicutes
	O9a	Wzt	BAA28332		431	Gammaproteobacteria
	O99	Wzt	ACV53836		433	Gammaproteobacteria
Geobacillus stearothermophilus	NRS2004/3a	Wzt	AAR99607		409	Firmicutes
Geobacillus tepidamans	GS5-97T	Wzt	ABM68319		395	Firmicutes
Geobacter uraniireducens	Rf4		YP_001232537	Gura_3813	711	Deltaproteobacteria
Hyphomonas neptunium	ATCC 15444		YP_759524	HNE_0796	417	Alphaproteobacteria
Klebsiella pneumoniae	O12	Wzt	AAN06493		440	Gammaproteobacteria
Lyngbya sp.	PCC 8106		ZP_01619961	L8106_24925	468	Cyanobacteria
Mesorhizobium sp.	BNC1		YP_676227	Meso_3694	429	Alphaproteobacteria
Methanosarcina barkeri	Fusaro		YP_303935	Mbar_A0371	410	Euryarchaeota
Nitrosomonas europaea	ATCC 19718	Wzt	NP_840568	NE0483	451	Betaproteobacteria
Nitrosomonas eutropha	C71		YP_746405	Neut_0152	437	Betaproteobacteria
Nitrosospira multiformis	ATCC 25196		YP_413079	Nmul_A2398	488	Betaproteobacteria
Pseudomonas aeruginosa	PAO1	Wzt	AAC16668	PA5450	421	Gammaproteobacteria
Pseudomonas putida	KT2440		NP_743935	PP_1779	405	Gammaproteobacteria
	W619		YP_001748261	PputW619_1387	402	Gammaproteobacteria
Pseudomonas syringae pv. Tomato	DC3000		NP_790909	PSPTO_1075	454	Gammaproteobacteria
Roseiflexus sp.	RS-1		YP_001278383	RoseRS_4089	870	Chloroflexi
Serratia marcescens	O4	Wzt	AAC00182		441	Gammaproteobacteria
Sodalis glossinidius	Morsitans		YP_454797	SG1117	434	Gammaproteobacteria
Stenotrophomonas sp.	SKA14		ZP_05135042	SSKA14_2119	475	Gammaproteobacteria
Sulfurimonas denitrificans	DSM 1251		YP_394240	Suden_1731	410	Epsilonproteobacteria
Vibrio cholerae	AM-19226		ZP_04961260	A33_0258	392	Gammaproteobacteria
	O37		AAM22592		429	Gammaproteobacteria
Xanthomonas axonopodis pv. Dieffenbachiae	LMG695		AAZ08621		329	Gammaproteobacteria
Xanthomonas campestris pv. Campestris	B100	Wzt	AAK53481		428	Gammaproteobacteria
Xanthomonas oryzae pv. Oryzae	BXO8	Wzt	ABI93186		410	Gammaproteobacteria
	MAFF 311018		YP_449745	XOO_0716	437	Gammaproteobacteria
Yersinia frederiksenii	ATCC 33641		ZP_04633619	yfred0001_9870	408	Gammaproteobacteria
Group B
Actinomyces odontolyticus	ATCC 17982		ZP_02045164	ACTODO_02054	430	Actinobacteria
Archaeoglobus fulgidus	DSM 4304		NP_068882	AF0041	237	Euryarchaeota
Cellulomonas flavigena	DSM 20109		ZP_04366067	CflaDRAFT_13740	411	Actinobacteria
Dokdonia donghaensis	MED134		ZP_01051491	MED134_13566	415	Flavobacteria
Flavobacterium johnsoniae	UW101		YP_001192680	Fjoh_0325	422	Flavobacteria
Mariprofundus ferrooxydans	PV-1		ZP_01452528	SPV1_07566	421	Zetaproteobacteria
Methanosaeta thermophila	PT		YP_843490	Mthe_1066	399	Euryarchaeota
Methanosarcina acetivorans	C2A		NP_616118	MA1177	409	Euryarchaeota
Methanosarcina mazei	Gol		NP_632680	MM_0656	421	Euryarchaeota
	Gol		NP_634238	MM_2214	504	Euryarchaeota
Microcoleus chthonoplastes	PCC 7420		ZP_05026777	MC7420_2165	433	Cyanobacteria
Rhizobium etli	CE3	Wzt	AAK51165		443	Alphaproteobacteria
	CIAT 652		YP_001976997	RHECIAT_CH0000832	439	Alphaproteobacteria
Spirosoma linguale	DSM 74		ZP_04492830	SlinDRAFT_59260	419	Bacteroidetes
Streptococcus mutans	UA159	RgpD	NP_721238	SMU.828	405	Firmicutes
Synechocystis sp.	PCC 6803	RfbB	NP_440220	slr0982	430	Cyanobacteria
Group C
Erwinia tasmaniensis	Et1/99	RfbB	YP_001907278	ETA_13390	246	Gammaproteobacteria
Escherichia coil	SMS-3-5	RfbB	YP_001744315	EcSMS35_2267	246	Gammaproteobacteria
Klebsiella pneumoniae	O1	RfbB/Wzt	Q48476		246	Gammaproteobacteria
	O8	RfbB/Wzt	Q48479		246	Gammaproteobacteria
Rhodopseudomonas palustris	BisB18		YP_534033	RPC_4190	246	Alphaproteobacteria
Serratia marcescens	O16	Wzt	AAC98415		246	Gammaproteobacteria
Group D
Aggregatibacter	Serotype a	Orf4/Wzt	BAB03203		250	Gammaproteobacteria
actinomycetemcomitans	Serotype b	Orf11/Wzt	BAA19638		245	Gammaproteobacteria
	Serotype c	Wzt	BAA28136		247	Gammaproteobacteria
	Serotype f	Wzt	AAG49408		248	Gammaproteobacteria
Actinobacillus pleuropneumoniae	O1	Wzt	AAR01223		244	Gammaproteobacteria
Aeromonas hydrophila subsp. hydrophila	ATCC 7966		YP_857403	AHA_2899	252	Gammaproteobacteria
Agrobacterium radiobacter	K84	Wzt	YP_002546388	Arad_4840	248	Alphaproteobacteria
Anaerocellum thermophilum	DSM 6725		YP_002571988	Athe_0053	244	Firmicutes
Anaerotruncus colihominis	DSM 17241		ZP_02442862	ANACOL_02160	239	Firmicutes
Azospirillum brasilense		Wzt	AAS83022	pRhico005	260	Alphaproteobacteria
Bradyrhizobium sp.	BTAi1		YP_001241520	BBta_5660	249	Alphaproteobacteria
Brucella suis	1330	RfbE	NP_697541	BR0519	252	Alphaproteobacteria
Burkholderia cepacia	AMMD		YP_772638	Bamb_0745	249	Betaproteobacteria
Caulobacter crescentus	CB15		NP_419451	CC_0634	250	Alphaproteobacteria
Clostridium methylpentosum	DSM 5476		ZP_03705427	CLOSTMETH_00138	244	Firmicutes
Coprococcus comes	ATCC 27758		ZP_03799480	COPCOM_01739	245	Firmicutes
Corynebacterium glutamicum	ATCC 13032		YP_224499	cg0248	263	Actinobacteria
Coxiella burnetii	RSA 493	RfbI	NP_819734	CBU_0704	258	Gammaproteobacteria
Dichelobacter nodosus	VCS1703A		YP_001209256	DNO_0336	233	Gammaproteobacteria
Dorea formicigenerans	ATCC 27755		ZP_02234094	DORFOR_00952	244	Firmicutes
Dorea longicatena	DSM 13814		ZP_01995719	DORLON_01714	252	Firmicutes
Erwinia chrysanthemi		Wzt	AAM33313		243	Gammaproteobacteria
Escherichia coli	O52	Wzt	AAS99165		324	Gammaproteobacteria
Escherichia hermannii	YS-11	Wzt	BAG06240		248	Gammaproteobacteria
Eubacterium ventriosum	ATCC 27560		ZP_02026283	EUBVEN_01539	249	Firmicutes
Mycobacterium tuberculosis	H37Rv	RfbE	NP_218298	Rv3781	273	Actinobacteria
Pelotamaculum thermopropionicum	SI	TagH	YP_001213125	PTH_2575	266	Firmicutes
Pseudomonas fluorescens	PfO-1		YP_349793	Pfl01_4065	252	Gammaproteobacteria
Ruminococcus gnavus	ATCC 29149		ZP_02042722	RUMGNA_03526	244	Firmicutes
Ruminococcus torques	ATCC 27756		ZP_01967310	RUMTOR_00857	251	Firmicutes
Saccharopolyspora erythraea	NRRL 2338		YP_001102478	SACE_0200	272	Actinobacteria
Salmonella enterica serovar Typhimurium	LT2		NP_459708	STM0723	236	Gammaproteobacteria
Stenotrophomonas maltophilia	R551-3		YP_002026879	Smal_0491	241	Gammaproteobacteria
Vibrio cholerae	O1	RfbI	CAA42140		250	Gammaproteobacteria
Xylella fastidiosa	Dixon		ZP_00651487	XfasaDRAFT_1586	246	Gammaproteobacteria
Yersinia enterocolitica	O:3	RfbE	Q56903		239	Gammaproteobacteria
	O:5	Wzt	AAT91802		241	Gammaproteobacteria
	O:9	Wzt	CAE53859		251	Gammaproteobacteria
Group E
Bacillus anthracis	Ames	TagH	NP_847669	BA5510	549	Firmicutes
Bacillus subtilis	168	TagH	P42954	BSU35700	527	Firmicutes
	W23	TarH	CAJ97406		527	Firmicutes
Enterococcus faecalis	V583	TagH	Q831L8	EF_2486	447	Firmicutes
Lactococcus lactis	Il1403	TagH	Q9CH26	LL0915	466	Firmicutes
Leuconostoc citreum	KM20	TagH	YP_001728599	LCK_01331	365	Firmicutes
Listeria monocytogenes	EGD-e		NP_464600	lmo1075	333	Firmicutes
Staphylococcus aureus subsp. aureus	COL		YP_185576	SACOL0694	264	Firmicutes
Group F
Actinobacillus pleuropneumoniae	5a	CpxA	AAB64445		216	Gammaproteobacteria
Actinobacillus suis	Strain SO4, serotype K1	Wzt	AAO65487		216	Gammaproteobacteria
Aeromonas hydrophila	PPD134/91	KpsT	AAM22566		163a	Gammaproteobacteria
Bordetella bronchiseptica	RB50	KpsT	NP_889466	BB2930	245	Betaproteobacteria
Burkholderia pseudomallei	K96243	Wzt2	YP_109399	BPSL2804	218	Betaproteobacteria
Campylobacter jejuni	NCTC 11168	KpsT	YP_002344829	Cj1447c	220	Epsilonproteobacteria
Chlorobium phaeobacteroides	DSM 266		YP_912756	Cpha266_2344	217	Chlorobi
Citrobacter freundii		VexC	AAK14187		213	Gammaproteobacteria
Escherichia coli	K1	KpsT	B42469		218	Gammaproteobacteria
	K5	KpsT	P24586		224	Gammaproteobacteria
	K10	KpsT	AAD31429		216	Gammaproteobacteria
	K54	KpsT	AAC38079		216	Gammaproteobacteria
Haemophilus influenzae	Serotype b	BexA	CAA38734		217	Gammaproteobacteria
Mannheimia haemolytica	A1	CpxA	AAF08240		215	Gammaproteobacteria
Neisseria meningitidis	Serogroup A strain Z2491	CtrD	YP_002341738	NMA0195	216	Betaproteobacteria
	Serogroup B	CtrD	P32016	NMB0074	216	Betaproteobacteria
Nitrococcus mobilis	Nb-231		ZP_01127549	NB231_02823	221	Gammaproteobacteria
Novosphingobium aromaticivorans	DSM 12444		YP_496029	Saro_0748	232	Alphaproteobacteria
Pasteurella multocida		CexA	AAF67272		218	Gammaproteobacteria
Ralstonia metallidurans	CH34		YP_587859	Rmet_5731	221	Betaproteobacteria
Salmonella enterica subsp. enterica serovar Typhi		VexC/Wzt	BAA03198		246	Gammaproteobacteria
Sinorhizobium meliloti	1021	RkpS	CAC48975	SM_b20832	219	Alphaproteobacteria
Yersinia intermedia	ATCC 29909		ZP_04635682	yinte0001_9460	221	Gammaproteobacteria
Group G
Arcobacter butzleri	RM4018		YP_001489609		566	Epsilonproteobacteria
Bordetella petrii	DSM 12804	MsbA2	YP_001633454		558	Betaproteobacteria
	DSM 12804	MsbA1	YP_001631720		615	Betaproteobacteria
Caulobacter crescentus	CB15		NP_419124	CC_0305	607	Alphaproteobacteria
Campylobacter coli	RM2228	WlaB	ZP_00367153	CCO1205	566	Epsilonproteobacteria
Campylobacter jejuni	NCTC 11168	WlaB/PglK	YP_002344523	Cj1130c	564	Epsilonproteobacteria
	NCTC 11168	MsbA	YP_002344210	Cj0803	580	Epsilonproteobacteria
Cyanothece sp.	PCC 7425		YP_002482894	Cyan7425_2172	610	Cyanobacteria
Escherichia coli		MsbA	CAA77839		582	Gammaproteobacteria
Francisella novicida	U112	MsbA	YP_899220	FTN_1606	609	Gammaproteobacteria
Helicobacter pylori	J99	Wzk	NP_223847		578	Epsilonproteobacteria
Microcoleus chthonoplastes	PCC 7420		ZP_05030053	MC7420_5135	601	Cyanobacteria
Myxococcus xanthus	DK 1622		YP_632878	MXAN_4716	594	Deltaproteobacteria
Nostoc sp.	PCC 7120	HepA	AAC32400		532	Cyanobacteria
Pseudomonas aeruginosa	PAO1	MsbA	AAG08382	PA4997	603	Gammaproteobacteria
Wolinella succinogenes	DSMZ 1740	Wlab	CAE09222	WS0052	563	Epsilonproteobacteria
Others
Bordetella parapertussis	Bpp5	WbmM	ABF72476		393	Betaproteobacteria
	Bpp5	WbmN	ABF72475		388	Betaproteobacteria
Myxococcus xanthus		RfbB	Q50863		437	Deltaproteobacteria

^aThe start site was reannotated.

FIG. 3.

Phylogenetic relationships among glyco-ABC transporters. Neighbor-joining analysis was carried out on 1,000 bootstrapped data sets, using the PHYLIP package (42). Trees were viewed using SplitsTree (59). Branch lengths represent relative distances, and bootstrap values are indicated. The tree is based on data available in the NCBI databases in January 2010.

Groups A and B contain NBD homologs from both O-PS and glycoprotein biosynthesis systems. They include representatives from Gram-positive and Gram-negative bacteria, as well as some archaea, and are unified by the presence of an extended C-terminal domain on the NBD. The sequences of the C-terminal domains suggest that they play diverse roles in cellular physiology. Here we use the term NBD to include ABC proteins that contain only the NBD domain as well as those that include additional functional (or putative functional) domains. Group A is larger than group B and contains mostly NBD homologs from low-G+C Gram-positive bacteria and Gram-negative proteobacteria. Group B is more diverse, containing examples from both high- and low-G+C Gram-positive bacteria as well as from a range of Gram-negative bacteria and archaea. Groups C, D, and F contain known and putative NBD homologs lacking an extended C terminus. Most examples in groups C and D are from O-PS assembly systems, but a few representatives from Gram-positive bacteria are also present. As with groups A and B, the significance of these NBDs clustering into two groups is unclear. Group E represents a relatively tight cluster of NBDs involved in the export of the polyol phosphate teichoic acids in Gram-positive bacteria. While these molecules are not oligo- or polysaccharides per se, their mode of synthesis is quite similar to that of many other bacterial glycoconjugates, and this justifies their inclusion here. Group F includes known and putative NBD homologs involved in the export of the “group 2” CPSs in Gram-negative bacteria. Finally, group G contains homologs of PglK and MsbA. This is the only group of half-transporters identified by our survey. MsbA proteins predominate due to the large number of known and highly conserved homologs in those bacteria that require LPS for viability. PglK is involved in the export of the general N-linked protein glycan of C. jejuni, and recent data have identified homologs from O-PS assembly systems in Helicobacter pylori (see below).

In most cases, information on NBD structure was available for representatives of each group to provide insight into the biosynthesis-export mechanism, as evaluated in the discussion below. Taking into consideration the available information, we have assigned prototype systems to each of the seven NBD groups (Table 2).

TABLE 2.

Representative ABC transporters from each of the seven phylogenetic groups

^aWhere the structure of the substrate includes a repeat unit domain, the relevant region is shown in bold. All sugars are pyranoses in the d configuration, unless otherwise indicated. Ac, acetyl; Ala, alanyl; Alt, 6-deoxy-l-altrose; Bac2,4-diNAc, 2,4-diacetamido-2,4,6-trideoxyglucose (bacillosamine); Fuc, fucose; Gal, galactose; GalNAc, N-acetylgalactosamine; Glc, glucose; GlcA, glucuronic acid; GlcNAc, N-acetylglucosamine; Gro, glycerol; Man, mannose; ManNAc, N-acetylmannosamine; Neu5Ac, N-acetylneuraminic acid (sialic acid); Me, methyl; P, phosphate; QuinNAc, 2-N-acetamido-2,6-dideoxyglucose (N-acetylquinovosamine); Tal, 6-deoxytalose. *, variable (nonstoichiometric) substituents.

ABC TRANSPORTERS WITH NBDs CONTAINING A CARBOHYDRATE-BINDING MODULE (CBM)—THE E. COLI POLYMANNOSE O-PS PARADIGM

ABC transporters involved in O-PS biosynthesis are comprised of separate TMD and NBD polypeptides, namely, Wzm and Wzt, respectively. Recent research with the E. coli O8 and O9a O-PSs has illustrated a pivotal role played by the ABC transporter (specifically the C-terminal extension on the NBD) in coordinating the biosynthesis and export phases of the assembly process (26, 31, 32). These NBDs are found in group A (Fig. 3).

The O-PSs of E. coli serotypes O8, O9, and O9a represent well-established examples of ABC transporter-dependent O-PS assembly systems. Identical O-PS structures are found in Klebsiella pneumoniae O serotypes, reflecting horizontal transfer of the corresponding loci (153). These O-PSs are homopolymers of mannose and differ in the linkage sequence that defines the polysaccharide repeat unit (Fig. 4). The gene clusters responsible for O-PS production in each serotype have a common organization and share many components, reflecting highly conserved (sometimes identical) steps in the biosynthesis of the polymannose O-PSs (70, 152, 154). Biosynthesis of the polymannose O-PS requires the activated precursor GDP-mannose and the activities of three dedicated mannosyltransferase enzymes, WbdA, WbdB, and WbdC (70). WbdB and WbdC are identical in each serotype, while WbdA varies significantly. The acceptor for mannosyltransferase activity is Und-PP-GlcNAc produced by the WecA enzyme, a GlcNAc:Und-P GlcNAc-1-phosphate transferase (61, 70, 126). WecA is involved in the biosynthesis of various O-PSs and enterobacterial common antigen, which, as the name suggests, is present in virtually all Enterobacteriaceae. There remains some debate about the precise order of action of the three mannosyltransferases and which enzymes contribute to the serospecific repeat unit domain (70, 163), but it is clear that WbdA, WbdB, and WbdC are collectively essential and sufficient for biosynthesis of the polymannose structures.

FIG. 4.

Schematics of genetic loci and gene functions for well-characterized examples of glycan assembly systems that feature an ABC transporter. For each ABC transporter, the gene encoding the NBD protein is shown in red and the corresponding gene for the TMD is shown in green. Genes encoding glycan biosynthesis proteins are highlighted in blue, and the products of their activities are shaded in the structures provided in each panel. (A) Group A NBDs. (B) Group F NBDs. (C) Group C NBDs.

The polymannose O-PSs terminate with unique residues (i.e., ones not found within the repeat units) on the nonreducing termini of the polysaccharide chains, and these are added by chain-terminating WbdD enzymes (26). A combination of structural and biochemical data identified these terminating residues as a methyl group (E. coli O8) or a phosphate plus a methyl group in an unknown organization (serotypes O9 and O9a) (26, 86, 163). In serotype O8, WbdD is a methyltransferase, whereas the O9/O9a WbdD homolog is a bifunctional kinase-methyltransferase, with the kinase activity being a prerequisite for the addition of the methyl residues (26). Most O-PSs exhibit a range of chain lengths (in what is often called a “modal” distribution), which can vary quite considerably and give rise to the characteristic ladder patterns seen when LPS extracts are separated by SDS-PAGE. The chain-terminating activity of WbdD is central to the establishment of the relatively limited modal chain-length range seen in the polymannose O-PSs. Overexpression of WbdD enhances chain termination frequency and reduces O-PS chain length in serotypes O8 and O9a, suggesting that the stoichiometry of the various biosynthesis components and their potential interactions may be important in determining O-PS chain length (26). Critical to this process is a physical interaction between WbdA and the C-terminal (noncatalytic) region of WbdD, which targets an otherwise soluble WbdA mannosyltransferase to the membrane, where it must reside for O-PS biosynthesis to occur (27). WbdD-mediated chain termination is also essential for coupling biosynthesis to export. E. coli O9a wbdD mutants defective in kinase-methyltransferase activities can synthesize the O9a polysaccharide but are unable to export it (26). This phenotype is essentially identical to that of an ABC transporter (Δwzm Δwzt) mutant. In both cases, the accumulation of polymer in the absence of export is toxic to the cell, and conditional (GDP-mannose synthesis) mutations are required to avoid second-site mutations that relieve the stress (26, 33).

A key element in understanding the function of the polymannose O-PS ABC transporters came from the observation that the O-PS O8 and O9a transporters are serotype specific. Although the Wzm proteins can be exchanged, albeit with reduced export efficiencies, the Wzt components cannot be exchanged between serotypes (33). The specificity observed in substrate recognition and transport lies within the Wzt components. Comparison of the serotype O8 and O9/O9a Wzt protein sequences and functional data from various chimeric proteins have established that the Wzt proteins contain two functional domains (33). The conserved (and exchangeable) N-terminal domain encompasses the characteristic motifs that define NBDs, whereas the variable C-terminal region confers serospecificity (Fig. 5). Both domains are required for export, but surprisingly, a functional transporter exists when the two domains are expressed separately in chromosomal Δwzt mutants (31). The C-terminal domain provides the essential connection between polymer chain termination and export. Biochemical experiments established that the C-terminal domain of Wzt binds O-PS and, furthermore, that glycan binding requires the chain-terminating residues added by WbdD. The structure of the C-terminal domain of Wzt from serotype O9a was determined by X-ray crystallography (Fig. 5B), revealing a dimer in which the interface between monomers is stabilized by the exchange of one β strand (B9) and the continuation of a single hydrophobic core through both monomers (31). The structure is a β sandwich with an immunoglobulin (Ig)-like topology, a motif seen in a number of carbohydrate-binding modules (CBMs), which serve to bring catalytic domains of carbohydrate-active proteins into close proximity with their cognate substrates. Based on sequence similarity, CBMs are currently divided into more than 50 families, and the number is growing (www.cazy.org/). The β-sandwich fold is the most common fold seen in CBMs, and at least 10 of these families (at current count) have an Ig-like topology (54). The C-terminal domain of Wzt provides a broad substrate-binding face where aromatic residues play a critical role (Fig. 5C), potentially by creating opportunities for ring-stacking interactions (31). Despite a common function shared by the CBMs of E. coli O8 and O9a, their sequences are quite different (21% identity and 43% similarity).

FIG. 5.

Structure and organization of NBDs from glyco-ABC transporters. (A) Three representative NBDs all contain the highly conserved protein sequence motifs that are essential for ABC transporter function. In E. coli O8 and O9a (group A), the NBD contains an additional C-terminal extension (highlighted in red) that contains a CBM. Two characterized representative NBDs that lack extended C termini are shown for comparison. (B) The crystal structure of the C-terminal domain of the O9a NBD reveals a dimer in which each monomer is a modified Ig fold typical of certain CBM families. (C) The CBM contains a pocket essential for O-PS binding that is conserved in closely related homologs. The transport phenotypes of site-directed mutants identify critical roles for aromatic residues, suggesting that the CBM binds its substrate through a ring-stacking mechanism. (Panels B and C were reprinted from reference 31.)

It is important that there is currently no direct evidence that the O-PS actually passes through the TMD of the ABC transporter. Although this is the most plausible explanation for the process and is consistent with relevant mutant phenotypes, it is still possible that the ABC transporter participates indirectly in exporting the polymer. Regardless, two critical unresolved questions exist. The first is how the Und-PP-linked intermediate traverses the inner membrane. The lipid-linked O-PS intermediate is a complex substrate with both hydrophobic and polar regions. It is conceivable that the polar and hydrophobic domains are handled differently. For example, the polar chain could pass through the hydrophilic channel, while the lipid acceptor may remain confined to the adjacent membrane core. A similar solution was proposed to address the same issue in the export of lipid A via MsbA (125). These are difficult questions to test experimentally without a structural snapshot that captures export in mid-progress. The second unresolved question is precisely how the C-terminal CBM of Wzt functions overall in the export process for the polymannose O-PSs. We currently envisage two possibilities. In one model, the binding domain introduces the polymer into the transport channel. However, this scenario is complicated by comparison to known ABC transporters, which predict that the C-terminal binding domain of the NBD is separated from the TMD by a considerable distance (∼25 Å in the case of MalFGK₂ [35]). Establishing the position of the CBM in the context of the whole ABC transporter would provide important insight, but attempts to crystallize the full-length Wzt protein have proved unsuccessful. However, it is also possible that the CBM does not lead the substrate to the export channel. In an alternative model, the CBM may be required to disengage the nascent O-PS from the assembly (glycosyltransferase) enzymes to allow it to enter the export pathway. The recognition of nonreducing terminal residues would ensure that only terminated chains that are competent for export are released. Either scenario is consistent with the phenotypes of cells lacking the chain-terminating methyltransferase-kinase. It is unknown whether the modified nonreducing terminus is the first part of the polymer to enter the export channel or the last. In the latter scenario, export could begin before synthesis is complete. Synthesis and export can be uncoupled temporally in a conditional experimental situation, i.e., polymer completed in the absence of the ABC transporter can be exported once expression of the transporter is activated (77). This would be the equivalent of posttranslational export of secreted proteins. However, the coordination suggests that tight coupling (akin to that of cotranslational export) is more likely to be the true physiological situation.

OTHER SYSTEMS WITH NBDs CONTAINING PUTATIVE CBMs

Several other group A NBDs come from assembly systems for which there is biochemical evidence supporting an assembly process that follows the E. coli polymannose O-PS model. The E. coli O99 and Pseudomonas aeruginosa A-band O-PSs can be considered variants of E. coli O8/O9a O-PSs in terms of their biosynthesis. Both antigens have a backbone of poly-d-rhamnose (not the more usual l-rhamnose), and in O99, this is modified with glucose residues (9, 113). The activated precursor for the polyrhamnose backbone is GDP-d-rhamnose, which is synthesized from GDP-d-mannose in a two-step process involving a dehydratase and a reductase (71). The serotype O99 rhamnosyltransferases share significant levels of similarity with the corresponding mannosyltransferases from E. coli O8/O9a (113). In the structure of the O99 antigen, novel nonreducing terminating residues were not reported. However, the gene cluster encodes a predicted protein (WejH) sharing high levels of similarity (44% identity and 62% similarity) with the WbdD protein from E. coli O9a. The critical methyltransferase and kinase motifs are conserved, suggesting that chain termination follows a model very similar to the one reported for E. coli O9a. The A-band glycan contains methyl substitutions, and available data suggest that the methyl groups may be present at the nonreducing terminus (9). Genes encoding two putative methyltransferases were identified downstream and are transcribed in the opposite orientation from that of Wzt in P. aeruginosa (71). The A band is a conserved LPS-linked antigen and is typically coexpressed with a serospecific O-PS that is formed by a Wzx/Wzy-dependent pathway (71). Rhamnose O-PSs are common in plant-pathogenic species of Pseudomonas, and in at least some cases, nonreducing terminal methyl groups have been identified (181). The NBDs from two strains of Pseudomonas putida were included in the analysis, and they grouped with the Wzt homologs from E. coli O8, O9a, and O99 (group A).

It is not clear whether the O99 genetic locus also encodes the glucosyltransferases that modify the main chain. It would be particularly interesting to know if these residues are added before or after export across the cytoplasmic membrane. For the Wzx/Wzy-dependent O-PSs of Shigella and Salmonella, modifications, including glucosylation and O-acetylation, are encoded by serotype-converting lysogenic phages (3). The glucosylation occurs at the periplasmic face of the membrane while the nascent O-PS remains attached to its Und-PP carrier. The donor for glucosylation is Und-P-glucose, which is exported by a dedicated “flippase” protein via an unknown mechanism (51, 76). We are not aware of any confirmed examples of phage seroconversion involving an O-PS assembled via an ABC transporter-dependent mechanism, but there are no features of the pathway that would necessarily preclude such a process.

An obvious issue arising from the examples above is whether all O-PSs assembled by this type of system (and having a group A NBD) terminate with nonreducing terminal methyl groups. The answer is no. The K. pneumoniae O12 O-PS has a repeat unit of rhamnose and GlcNAc, with a 3-deoxy-d-manno-octulosonic acid (Kdo) residue at its nonreducing terminus (163). It is likely that many more examples exist, but polysaccharide structural data tend to focus on the repeat unit, and the residues found only at the terminus of each chain can easily go undetected.

Another fundamentally important question is whether this type of system extends beyond O-PS biosynthesis. The answer is yes. Nonreducing terminal modifications analogous to those terminating the E. coli O8 and O9a O-PSs have also been identified on the S-layer glycans of a number of species of Bacillaceae (131). Others have already noted obvious similarities in the biosynthesis of some S-layer glycans and ABC transporter-dependent O-PSs (102, 134, 149). The ABC transporters from these S-layer glycan systems have separate TMD and NBD polypeptides resembling Wzm and Wzt, respectively (102). Furthermore, NBD proteins from well-investigated S-layer glycan assembly systems from Geobacillus stearothermophilus NRS2004/3a, Geobacillus tepidamans GS5-97^T (180), and Aneurinibacillus thermoaerophilus strains L420-91^T and DSM10155/G⁺ are found in group A. The S-layer glycan from G. stearothermophilus strain NRS2004/3a contains an l-rhamnose trisaccharide repeat unit, and the chain of 15 repeat units is terminated with a 2-O-methyl group (133). The glycan chain is elongated on a Und-PP-galactose acceptor (149) and is linked to the protein via an O-glycosidic linkage (133). Interestingly, WsaE, the enzyme responsible for the S-adenosylmethionine-dependent methylation reaction, is a multifunctional protein that also has rhamnosyltransferase activity (150). The physical coupling of elongation and termination activities may be advantageous in regulating the glycan chain length. While biosynthesis information is limited for the other representatives, the structural theme is conserved: they possess a glycan capped with a nonreducing terminating residue. In G. tepidamans GS5-97^T, a chain of 20 repeat units containing d-fucose and l-rhamnose is O linked to two serine and threonine residues in the S-layer protein (63). The glycan is terminated by two substitutions, i.e., N-acetylglucosamine and N-acetylmuramic acid. The latter sugar is normally associated only with peptidoglycan. The A. thermoaerophilus L420-91^T S-layer glycan contains an O-linked (to threonine) glycan with a complicated hexasaccharide repeat unit containing a backbone of d-rhamnose with d-N-acetylfucosamine side chains. It is terminated with a 3-O-methyl group (132). All evidence points to a conserved strategy for coupling biosynthesis and export in these systems and in the related O-PSs, such as those from E. coli O8/O9a. It would be interesting to determine whether the extended C-terminal domains on the corresponding S-layer glycan Wzt proteins contain cognate CBMs and thus warrant further investigation.

VARIATIONS ON THE THEME—DIVERSITY IN FUNCTION OF EXTENDED C-TERMINAL DOMAINS IN GROUP B NBDs?

The O-PS of the plant symbiont Rhizobium etli CE3 contains multiple methyl groups, including three at the nonreducing terminus (45, 84). The extent of methylation varies according to environmental conditions, specifically, the presence of the host plant or seed exudate, and the internal methyl groups facilitate symbiosis. Multiple methyltransferase domains have been identified in gene products from the biosynthesis cluster (85, 104). The Wzt homolog from R. etli CE3 is found in group B and contains an extended C terminus, suggesting, at first glance, that this system follows the model established for O-PS and S-layer glycans described above. However, unlike the situation with E. coli O8/O9a O-PS biosynthesis, chain-length regulation and O-PS export are not dependent on nonreducing terminal methylation in R. etli CE3 (104). The role of the R. etli CE3 NBD C-terminal domain must therefore differ. The primary sequence of the C-terminal domain offers no clues to its function; from sequence data alone, it is not possible to rule in, or out, a novel CBM participating in a different means of assembly. Interestingly, R. etli provides an example of O-PSs possessing a homogenous chain length, with 5 repeat units (45), rather than the more typical modal range of chain lengths. It would be interesting to know whether the extended C-terminal domain of the NBD plays a role in establishing this homogeneity.

In some cases, bioinformatic analyses identified additional interesting putative functions for the C-terminal domains of group B NBDs. In three cases, NBD proteins, from Geobacter uraniireducens, Roseiflexus sp., and Methanosarcina mazei, were fused to C-terminal domains containing catalytic motifs. The additional motif in Geobacter uraniireducens is a putative sulfotransferase sharing some similarity (15% identity and 46% similarity) with NodH from Rhizobium tropici (82). NodH is a sulfotransferase (16) involved in modification of the lipochitooligosaccharide nodulation factors in Rhizobium sp. (40, 137). In Roseiflexus sp., the NBD homolog is fused to a predicted methyltransferase domain sharing similarity with RfbT from Vibrio cholerae O1 (151). RfbT is responsible for addition of the serogroup-specific methyl group to the O-PS of V. cholerae O1 (see below). Both NBD homologs contain an extended C-terminal region between the core NBD domain and the predicted catalytic domain. The NBD homolog from Methanosarcina mazei (MM_2214) is also fused at its C terminus to a predicted catalytic domain, in this case an acetyltransferase. It is tempting to speculate that the fused (putative) enzymatic domains found on the G. uraniireducens, Roseiflexus sp., and M. mazei NBD proteins may be involved in the addition of nonreducing terminal modifications to the polysaccharides synthesized. It would be interesting if, in these examples, different domains of the same polypeptide are responsible for both modification of the polysaccharide and its subsequent recognition of the modification prior to polysaccharide export. The biosynthetic consequences and possible advantages of such an arrangement are unclear.

A number of predicted NBDs found in group B are from the Euryarchaeota, including members of the genera Methanosarcina, Archaeoglobus, and Methanosaeta, and they appear to be involved in protein glycosylation. In all cases, the genetic loci contain open reading frames encoding predicted homologs of an STT3 oligosaccharyltransferase homologous to PglB from C. jejuni (21). No information is available concerning the structure of the resulting glycans, so interpretation of the significance and roles of the extended C-terminal domains on the corresponding NBDs is currently not possible.

In Aeromonas salmonicida, the group B NBD appears to provide a direct connection between O-PS and S-layer glycan biosynthesis and illustrates a different activity for the NBD C-terminal extension. This bacterium produces homogenous (nonmodal) O-PS chain lengths (22), and the NBD protein (AbcA) is required for O-PS export (24). AbcA also contains an extended C-terminal domain that contains a putative leucine zipper, which is not essential for O-PS export but is required for proper transcription of the surface layer protein gene vapA (25, 100). Conversely, the activity of the core NBD portion of AbcA is confined to O-PS export. In this organism, the S layer is coupled to the cell surface via the O-PS chains (101), so the coupling of O-PS and S-layer protein expression may play an important physiological role. These NBDs present fascinating subjects for further analysis.

Three group B NBD proteins are particularly intriguing in terms of the current models for O-PS and CPS biosynthesis and assembly. The NBD homologs from Dokdonia donghaensis, Mariprofundus ferrooxydans, and Synechocystis sp. cluster quite closely. In each case, the NBD-encoding gene is located downstream of predicted open reading frames encoding OPX (outer membrane polysaccharide export) and PCP-2a homologs. PCP-2a and OPX proteins are inner and outer membrane components, respectively, of a transenvelope assembly-export machinery for Wzx/Wzy-dependent E. coli CPS (29, 32, 173). Again, it is difficult to interpret these findings in the absence of information concerning the glycoconjugates from these loci. Perhaps these organisms represent completely new twists on existing glycan assembly pathways.

EXPORT OF GLYCANS VIA AN ABC EXPORTER LACKING A C-TERMINAL EXTENSION ON THE NBD PROTEIN—ASSEMBLY OF GROUP 2 CPSs

Group F contains known and putative NBD homologs involved in “group 2” CPS export in E. coli. They lack an extended C-terminal domain (Fig. 5). The group 2 designation identifies several CPSs of E. coli whose export involves an ABC transporter; “group 1” CPSs use a Wzx/Wzy-dependent process (173). Most work has been done with the E. coli K1 and K5 CPS assembly systems. In fact, E. coli K1 carries the first polysaccharide biosynthesis cluster to be cloned (141). The “group 2” name is often extended to CPS assembly systems in other bacteria which share the same spectrum of functional components. Examples include Neisseria meningitidis, H. influenzae, and C. jejuni.

In the E. coli paradigm, KpsM and KpsT form the TMD and NBD of the ABC transporter, respectively, and are required for the transport of the CPS across the inner membrane (80, 81, 108, 109, 116). Initial evidence concerning the role of KpsMT in CPS export came from the observation that mutants in either kpsM or kpsT accumulated CPS polymers within the cytoplasm (108, 109, 116). All of the CPS NBDs assigned to group F lack C-terminal extensions, and the corresponding ABC transporters show no apparent specificity for the repeat unit structures of their glycan substrates. Examples have been reported where the export apparatus is functionally interchangeable between different CPS serotypes, or even between different bacterial species (10, 87, 112, 128, 129, 169). Without molecular recognition, how can coupled polymerization and export be achieved? One possibility is that a presently unknown conserved part of the substrate might be recognized as an export signal. The diacylglycerol phosphate moieties found at the reducing terminus in this class of CPSs represent strong candidates if a molecular signal is indeed required (15). However, an attractive recent model invokes a system where there is no specific polymer recognition event in the coupling of synthesis to export. Instead, it is proposed that the system is coordinated by direct heterotypic protein-protein interactions, obviating the need for any conserved feature in the CPS export substrate (146, 162). Several studies have established the existence of multiprotein complexes involving proteins required for biosynthesis and export (94, 127, 146). In the case of E. coli K1, CPS biosynthesis occurs within a complex that effectively shields the nascent polymer from the cytoplasmic milieu (146). The multiprotein complex provides the structural context for coupling polymerization and export.

Synthesis of the E. coli K1 CPS occurs on the cytoplasmic face of the inner membrane, and several dedicated proteins participate in the assembly pathway (Fig. 4). The mechanism of initiation of CPS synthesis is also not yet resolved, but four gene products, NeuE, NeuS, KpsC, and KpsS, are required for de novo synthesis of cytoplasmic K1 CPS in E. coli (5). The polysialyltransferase (NeuS) is responsible for polymer extension through sequential glycosyl transfers to the nonreducing terminus of the growing glycan (147, 148, 164). The resulting CPSs contain a reducing terminal diacylglycerol. This could be attached through a Kdo linker residue (43), although a reducing terminal Kdo has not been identified on all of the related CPSs (49). It is not known at which point in CPS synthesis the polymer is transferred to the lipid anchor. For this system, the involvement of Und-linked intermediates is still controversial, and it is possible that diacylglycerol phosphate serves as the acceptor for chain initiation and extension (173). The KpsC protein has been implicated in the proper export of the CPS to the outer membrane. It has been proposed at various times that KpsC may be involved in Kdo addition and/or attachment of the lipid moiety to the reducing terminus. However, an N. meningitidis serotype B (polysialic acid) mutant deficient in its KpsC homolog can polymerize lipidated CPS in the absence of export (161). The recent proposal (162) that KpsC provides a pivotal adaptor between the biosynthetic enzymes and the ABC transporter seems a plausible alternative function. Two-hybrid data established that KpsC oligomerizes and interacts with both NeuS and KpsE (PCP-3 protein) (146). KpsC proteins are well distributed in bacteria with “group 2-like” CPSs (162).

In ABC transporter-dependent CPS biosynthesis, the final steps of translocation of the polymer across the periplasm and outer membrane require two dedicated proteins (8, 17, 94, 111, 140). In E. coli, these proteins include KpsE (a PCP-3 protein) and KpsD (an OPX representative). In the working model, these proteins are proposed to come together to form a structure analogous to a tripartite drug efflux pump (32). Interestingly, the clustering of NBD homologs in group F shows remarkable similarity to the phylogenetic relationships of the corresponding PCP-3 proteins (32), suggesting that important molecular recognition elements may be required for functional interactions between this class of ABC transporters and their cognate PCP and OPX (outer membrane channel) partners (142). We examined KpsC proteins and determined that they do not follow the same discrete phylogenetic clustering observed for NBD, PCP, and OPX proteins from the same organisms. While this certainly does not rule out a role for KpsC proteins as adaptors linking biosynthesis and export, it does not offer direct support for a cognate pairing between a given KpsC protein and the ABC transporter.

O-PS EXPORT USING GROUP C AND D NBDs LACKING AN EXTENDED C TERMINUS—IS IT SIMILAR TO THE CPS SYSTEM?

Groups C and D contain NBDs from a variety of different O-PS biosynthesis systems, and they all lack a C-terminal extension (see the group C representative in Fig. 5A), suggesting that the mode of coupling polymerization and export is fundamentally different from the E. coli polymannose O-PS paradigm. Biochemical data are available for one example, the d-galactan I O-PS found in several O serotypes of K. pneumoniae. d-Galactan I is a heteropolysaccharide containing the pyranose and furanose forms of galactose, arranged in a disaccharide repeat unit (73, 74, 174) (Fig. 4). d-Galactan I defines the O2a antigen of K. pneumoniae, and in some isolates, it is the only O-PS. In other isolates, d-galactan I may also be extended (or capped) by an additional polysaccharide domain with a different structure, which can define a new serotype (e.g., O1 and O2c) (174, 175). Further diversification of serotypes can occur through additional modifications (e.g., serotype O8 is an O-acetylated version of serotype O1 [67]). O-PSs isolated from these strains contain a mixture of molecules carrying only d-galactan I and others comprising short chains of d-galactan I capped by the additional antigen (73, 74). The activated precursors for d-galactan I biosynthesis are UDP-galactopyranose and UDP-galactofuranose (75). Synthesis proceeds on a Und-PP-GlcNAc acceptor formed by WecA (28, 52, 163), and it requires three dedicated galactosyltransferases (WbbM, WbbN, and WbbO), whose precise contributions are only partially resolved. These steps essentially parallel the synthesis of the polymannose O-PSs, with the exception that there is no obvious terminating residue at the nonreducing terminus.

In the dendrogram in Fig. 3, NBDs from the K. pneumoniae d-galactan I systems are represented by the homologs from serotype O1 (whose sequence is identical to that of the NBD from O2a) and serotype O8 (which shows some sequence diversity) (67). Serratia marcescens O16 also contains d-galactan I (107), and the genetic locus is well conserved (156). The exception is the glycan synthesized by Rhodopseudomonas palustris, which likely contains l-rhamnose due to the presence of the rml (dTDP-l-rhamnose biosynthesis) genes immediately upstream of wzm and wzt. The lack of an extended C-terminal domain in the NBD of the d-galactan I O-PS exporter (Fig. 5A), together with the absence of chain-terminating modifications at the nonreducing terminus of the corresponding polymer, dictates a mechanism linking chain extension and export that differs from the E. coli O8/O9a paradigm. The model proposed for group 2 CPSs (see above) may be a better representation. Unlike the polymannose systems, where WbdD activity determines O-PS chain length, the chain length of d-galactan I is determined by the stoichiometry of the ABC transporter and the O-PS biosynthesis machinery (77). As a consequence, polymerization and export are obligatorily coupled in this system, analogous to cotranslational protein export. In the absence of the ABC transporter, aberrantly long, “unregulated” O-PS chains are synthesized, but these are not substrates for export if expression of the transporter is activated. It is unknown whether this is due only to the disconnection of synthesis and export or reflects a limitation in the maximum size of nascent polymers that the ABC transporter is able to export. Regardless, the equivalent of posttranslational protein export is not possible, unlike the case for polymannose O-PS systems. The modal distribution of O-PS chain lengths in LPS molecules containing d-galactan I is considerably more heterogeneous than that seen in O8 and O9a strains, and this may reflect the absence of a discrete chain-terminating enzymatic reaction. Interestingly, the disconnection of group 2 CPS synthesis can also affect polymer chain length in Sinorhizobium meliloti Rm1021 (98), so the phenotypes seen in d-galactan I O-PS mutants may reflect a broader phenomenon.

If the E. coli O8 and O9/O9a NBDs impart serospecificity by recognizing the nonreducing terminal modifications on the substrate, what does the K. pneumoniae O2a transport system recognize? Important insight comes from the observation that the K. pneumoniae O2a ABC transporter can complement a Δwzm Δwzt mutant of E. coli O9a to restore polymannose O-PS export (77). This led to the prediction that the K. pneumoniae O2a ABC transporter actually exports a wide variety of O-PS structures from different species, independent of repeat unit structures, like the case for the CPS ABC transporters containing group F NBDs. How is this achieved, given the established critical requirement for coupling d-galactan I O-PS export with polymerization? One possibility is the recognition of a conserved feature or property of the export substrate, and the only conserved element in the d-galactan I and polymannose O-PS export substrates is Und-PP-GlcNAc. However, the CPS model represents a rational alternative, where the coupling is potentially achieved by molecular recognition of proteins in the biosynthesis and export pathways rather than by any specific recognition of the glycan substrate. Nevertheless, the existence of different processes that couple polymerization and export of the various glycans does not necessarily mean that the actual mechanisms driving transport through the Wzm channel are different. Defining those mechanisms represents the next challenge.

Other O-PS NBDs from diverse pathogens of plants and humans are found in group D, and the evolutionary and physiological distinctions between group C and D NBDs are unclear. Three group D NBD representatives from Y. enterocolitica O-PS systems were included in the analysis. The O-PS structures from these bacteria are quite different in terms of structure. The O:3 antigen and O:9 antigens are homopolymers of 6-deoxy-l-atrose (47) and 4,6-dideoxy-4-formamido-d-mannopyranose (N-formyl peros- amine) (19), respectively. In contrast, the O:5 O-PS contains a trisaccharide l-rhamnose backbone with substitutions of d-threo-pent-2-ulose (xylulose) (48). The O:3 serotype deserves particular recognition as the first O-PS system for which an ABC transporter was identified (182). The Yersinia NBDs are relatively well separated within group D. In contrast, five NBDs from the genetic loci for the O-PSs of Aggregatibacter (Actinobacillus) actinomycetemcomitans and Actinobacillus pleuropneumoniae form a close-linked phylogenetic group, despite the structures of the corresponding O-PSs being remarkably different (4, 114, 115). The gene clusters for A. actinomycetemcomitans serotypes b, c, and f are highly conserved at the proximal and distal ends of the cluster, with a central region of lower G+C content that includes two to four genes which are unique to each serotype (64). This suggests that the serotypes evolved from a common ancestor and acquired the unique open reading frames from another source (99, 177). The gene cluster of A. actinomycetemcomitans serotype a is located in a different region of the chromosome from that for the other serotypes and contains several unique open reading frames, but relationships between the various ABC transporter components (40 to 60% amino acid identity) suggest that all of these systems share a common ancestor (155). It would be interesting to explore the exchangeability of the transporters between these serotypes. One would predict that like the case for Klebsiella, the NBDs would show no specificity for a particular repeat unit structure. In contrast to the NBD homologs in A. actinomycetemcomitans serotypes a, b, c, and f, the NBD homolog from serotype e contains an extended C-terminal domain and clusters with similar proteins in group A. Thus, the extent of phylogenetic linkage of the NBDs is not necessarily determined by structural diversity in the O-PS structures within the genus Aggregatibacter/Actinobacillus.

The identification of the V. cholerae O1 NBD in group D provides critical insight that helps to explain contradictory information concerning the biosynthesis and export of the O1 O-PS. Although only one glycan backbone is found in O1 serotypes, two subgroups (Inaba and Ogawa) are distinguished on the basis of O-PS serology. The glycan is a homopolymer consisting of α1,2-linked 4-amino-4,6-dideoxy-d-mannose (perosamine), and the serological differences result from the presence (Ogawa) or absence (Inaba) of terminal 2-O-methyl residues (55). Ogawa serotypes express both the methylated and nonmethylated versions of the O-PS, indicating that the chain termination residue added by RfbT (WbeT) (151) is not an absolute requirement for export via the ABC transporter. This situation resembles the multiply methylated O-PS from R. etli CE3 (described above) and differs from the E. coli O8/O9a paradigm, where chain termination and export are obligatorily coupled. The involvement of an NBD protein lacking a predicted substrate-binding domain in V. cholerae O1 predicts an export system that does not recognize nonreducing terminal modifications on the glycan, akin to the K. pneumoniae d-galactan I prototype, which has no apparent specificity for either the O-PS repeat unit or any terminating residues.

A COMMON MECHANISM FOR THE EXPORT OF SOME GRAM-POSITIVE LIPOGLYCANS?

Group D also contains NBD homologs from Gram-positive bacteria and resembles group A in the sense that it includes representatives from different glycoconjugate systems that may be linked via a unified glycan assembly process. Unfortunately, the precise glycan substrates for the Gram-positive group D NBDs are not always clear. The NBD homolog (RfbE; also known as Rv3781) from Mycobacterium tuberculosis and its cognate TMD (RfbD; also known as Rv3783) are separated on the chromosome by the gene encoding GlfT1. GlfT1 is a bifunctional galactosyltransferase that adds the first two galactofuranose residues to an acceptor in the formation of the galactan core of arabinogalactan (12). GlfT2 subsequently adds up to 20 galactofuranosyl residues (92) generating the polyisoprenoid lipid-linked intermediate (decaprenol-PP-GlcNAc-Rha-Galf_n). Further elaboration of the arabinogalactan molecule involves the addition of arabinosyl residues by the Emb and Aft proteins (13, 158). An NBD homolog similar to Rv3781 is found in Corynebacterium glutamicum, which also synthesizes an arabinogalactan (34). Based on our current understanding of related systems, the most likely role for this ABC transporter appears to be the export of the lipid intermediate containing the galactan structure prior to its arabinosylation. This proposal invokes a spatial separation of galactan formation (at the cytoplasmic face of the membrane) and arabinosylation (at the periplasmic face), but there are supporting precedents from other glycoconjugate assembly systems. The donor for the arabinosyltransferases is Dec-P-arabinofuranose (reviewed in references 13 and 158). In Gram-negative bacteria, the use of Und-P-linked donors (rather than the prevalent Und-PP) generally signals a periplasmic modification process, with the donors first being exported by a dedicated transporter (not an ABC transporter). Examples include the addition of 4-aminoarabinose to LPS lipid A (120) and the phage-encoded modification of O-PSs by glucosylation in species such as Shigella (51, 76). ArnE and ArnF, two small proteins of 111 and 128 amino acids, respectively, have been implicated in flipping the Und-P-4-aminoarabinose precursor involved in LPS modification (176). ArnE and ArnF each contain four predicted transmembrane helices and share both topological and sequence similarity with members of the drug/metabolite transporter superfamily, which includes EmrE from E. coli (176). GtrA is predicted to function in flipping Und-P-linked precursors required for O-PS modifications in Shigella (51). Like ArnE and ArnF, GtrA is a small protein (120 amino acids) with four transmembrane helices (76). Although their sequence similarity is low, GtrA may share structural similarity with EmrE (76). Interestingly, M. tuberculosis encodes a GtrA homolog (Rv3789), within the arabinogalactan biosynthesis locus, that may serve as the flippase for Dec-P-arabinose precursors. The hypothetical function for the Rv3783/Rv3781 ABC transporter in the export of the M. tuberculosis galactan remains to be tested, but there is significant interest in this area of glycobiology, as arabinogalactan is an essential part of the mycobacterial cell wall and is an excellent target for therapeutics. Indeed, the Emb proteins (arabinofuranosyl transferases) represent the target for the antimycobacterial agent ethambutol (11, 50).

THE TEICHOIC ACID EXPORTERS—GROUP E NBDs

Group E represents a relatively tight cluster of NBDs involved in the export of the polyol phosphate teichoic acids in Gram-positive bacteria. While these molecules are clearly not oligo- or polysaccharides per se, their mode of synthesis is quite similar to those for many other bacterial glycoconjugates. It is also important that components such as ribitol, glycerol, and phosphate are found occasionally in the repeat unit structures of CPSs and O-PSs. For example, all three components are found among the CPS structures of Streptococcus pneumoniae, and the genetic basis for the synthesis and diversity of these polymers has been established (1, 91). The best-characterized teichoic acid system is arguably the polyglycerol phosphate polymer from B. subtilis 168, where the cognate NBD is TagH (83). Notably, the TagH homologs from B. subtilis 168 and W23 are very similar, sharing 92% identity and 96% similarity, although the W23 teichoic acid is a polyribitol phosphate (6, 7, 170). It remains to be determined whether these transporters, like some ABC transporters with NBDs found in groups C, D, and F, do not discriminate between the structures of the teichoic acid backbones. In B. subtilis 168, chain-length regulation of polymers is apparently determined by the polymerase TagF, and the membrane association of TagF is important for this activity (135). The potential involvement of the teichoic acid ABC exporter in chain-length regulation has not been examined.

Group E contains NBD proteins of various lengths. Although some of them contain an extended C-terminal domain, topology predictions indicate that this domain is actually periplasmic and therefore unlikely to be involved in the recognition of teichoic acids for export from the cytoplasm (44). For example, the C-terminal regions of the TagH homologs from Enterococcus faecalis and Lactococcus lactis contain a predicted LysM domain following the transmembrane domain, which could be involved in binding to peptidoglycan, to which wall teichoic acids are covalently linked in the periplasm (117). The TagH homolog of Bacillus anthracis contains a predicted SH3b domain following a predicted transmembrane segment. SH3 domains are found in eukaryotes and viruses and are involved in protein-protein interactions (110, 172). SH3b domains are a subset of SH3 proteins found in prokaryotes and are involved in cell wall binding (89, 117, 172). The SH3b domain within the B. anthracis TagH homolog may also be involved in peptidoglycan binding in the periplasm. How these domains contribute to the overall assembly of teichoic acid is uncertain, but there is some evidence to suggest that the periplasmic C-terminal domain is not essential for viability of B. subtilis in the laboratory (83).

HALF-EXPORTERS—NBDs IN GROUP G

Group G contains homologs of PglK and MsbA. These proteins lack additional C-terminal domains. The MsbA (LPS lipid A export) homologs generally form a relatively tight cluster, as one might expect given the highly conserved nature of the lipid A export substrate (120). The exception is the MsbA2 protein from Bordetella petrii, which clusters with the PglK (N-linked glycan export) homologs. One wonders if this representative is improperly annotated and plays a cellular role that is different from lipid A export.

The lipid A-core oligosaccharide portion of LPS is exported across the inner membrane by a process requiring the ABC transporter MsbA (120). While MsbA is widely assumed to be the lipid A exporter, and this interpretation is probably correct, it is important to remember that biochemical data verifying the ATP-dependent flipping of lipid A by MsbA are still lacking. The requirement for the first two Kdo residues of the inner core for cell viability in E. coli is a result of the inefficient export of lipid A without these residues (72). Addition of the first two residues of the core oligosaccharide is required prior to completion of lipid A acylation in E. coli K-12 and for efficient export of lipid A. Overexpression of MsbA is able to suppress the requirement for completion of the lipid A-Kdo₂ molecule for efficient export (65, 95). The possible involvement of MsbA in phospholipid export is less certain. Initial reports of MsbA depletion in E. coli suggested that MsbA might also be involved in phospholipid export (38, 39). However, other studies of an msbA mutant of N. meningitidis (an organism that does not require LPS for viability [145]) indicated that MsbA is not involved in phospholipid transport (160). It has been suggested that the accumulation of phospholipids observed in MsbA-depleted cells of E. coli may have been a secondary effect due to loss of LPS transport (160). Alternatively, the range of substrates for different homologs of MsbA may vary in different bacteria. It is certainly true that MsbA can export more than one substrate, as efflux of amphipathic drugs is recognized and the binding site(s) for these compounds is distinct from the high-affinity lipid A binding site (139). Furthermore, the multidrug transporter LmrA from Lactobacillus lactis can replace MsbA in mutant complementation (124). The recent discovery of the role of the multicomponent Mla system in retrograde transport of phospholipids to maintain the asymmetry in the outer membrane (90) represents a significant step forward in understanding lipid trafficking in Gram-negative bacteria. However, the picture is still incomplete, as the precise mechanism of phospholipid export remains to be identified.

PglK is encoded by the pgl (protein glycosylation) locus in C. jejuni NCTC 11168, and similar clusters have been found in other mucosal pathogens (157). The glycan product of the Pgl proteins is a branched, seven-residue glycan (178). Expression of the C. jejuni pgl locus in E. coli has facilitated study of the functions of the encoded gene products (166). Synthesis of Und-PP-linked glycan intermediates takes place at the cytoplasmic face of the inner membrane. The glycan is synthesized by sequential transfer of glycosyl residues by the glycosyltransferases (PglAHIJ) prior to its export across the inner membrane by the ABC transporter, PglK (2). Once exported, the glycan is ligated to specific consensus sites in the protein carriers by the oligosaccharyltransferase PglB (78). PglB exhibits a strict requirement for an acetamidosugar at the reducing end of the glycan (165) but has relaxed substrate specificity for the remaining sugars in the glycan (41, 150, 165). PglK is able to export O-PS repeat units from the E. coli O7 and O16 O-PSs in a chimeric experimental system. In the native E. coli system, the O7 and O16 Und-PP-linked intermediates are exported by Wzx, as part of a Wzx/Wzy-dependent O-PS pathway (2).

PglK seems to be an effective oligosaccharide transporter, suggesting that it behaves like Wzx in its transport abilities (2). It is not yet clear whether PglK can also export longer-chain Und-PP-linked polysaccharides, but interesting new insight into a PglK homolog (Wzk [also a group G protein]) suggests that this may be the case (58). Wzk is required for the export of O-PSs bearing the Lewis X and Lewis Y antigens, but it shows relaxed specificity with respect to the structure of the Und-PP-linked export substrates. Unlike C. jejuni, H. pylori does not contain the N-glycosylation machinery. The identification of Wzk as the first half-transporter involved in O-PS export further highlights the close functional and evolutionary connections between the export systems used for different glycoconjugates.

PHYLOGENETIC OUTLIERS

Based on the current analysis, three NBD proteins show no clear phylogenetic linkage to any of the seven groups. These may form part of a branch within the tree once more sequence data for more gene clusters become available. WbmM and WbmN from Bordetella parapertussis are encoded in the same O-PS biosynthesis locus (118). The O-PS of B. parapertussis terminates in a derivative of 2,3,4-triamino-2,3,4-trideoxy-l-galacturonamide (119). WbmM and WbmN both contain an extended C-terminal region, but the sequences differ and the roles of the two NBD homologs are not known. RfbB from the O-PS assembly system in Myxococcus xanthus (53) also contains an extended C-terminal domain. Just downstream of RfbB is a protein showing weak similarity to PMT (dolichyl phosphate-mannose-protein mannosyltransferase) proteins, raising the possibility that cell surface proteins in M. xanthus may be glycosylated with O-PS units. This situation has been well documented for some P. aeruginosa isolates, where repeat units of O-PS are used to O-glycosylate a pilin protein (20, 30). Until recently, this phenomenon was confined to O-PSs synthesized via a Wzx/Wzy-dependent process. However, the ABC transporter-dependent O-PS biosynthesis pathway (group D NBD) from A. actinomycetemcomitans serotype b can apparently be exploited in a similar fashion (159). Further work will be required to determine whether the predicted M. xanthus oligosaccharyltransferase is involved in protein glycosylation, but it could provide another example highlighting the flexibility of the biosynthesis machinery in generating polysaccharide substituents for a variety of ligands (i.e., lipids and proteins).

CONCLUSIONS

The bioinformatic survey presented here provides intriguing insight into the structure-function relationships shared by NBDs from ABC transporters involved in the export of oligo- and polysaccharides. In all, seven phylogenetic groups of NBDs were identified, and to the extent that information on structure and biosynthesis is available, the NBD phylogeny is consistent with the biosynthetic mechanism. The majority of the ABC transporters involved in bacterial glycoconjugate assembly follow a format of independent TMD and NBD polypeptides. The exceptions are the half-transporters, such as MsbA and PglK, and it is likely that more examples will be found as the extent of available sequence data increases. Perhaps not surprisingly, these NBDs are well separated from the other representatives. Group F NBDs are from CPS assembly systems. The CPS repeat unit structures are diverse, but many are linked by the presence of diacylglycerol at the reducing terminus, and this may be the part of the molecule recognized by the exporter. If so, this would explain both their exchangeability and the close phylogenetic connections of ABC transporters within this group. The teichoic acid NBDs in group E also form a discrete group, which is perhaps not too surprising given the relatively limited range of teichoic acid structures. NBDs in groups C and D are implicated in export of a wide range of glycans in an array of bacteria. While the majority are Gram-negative O-PSs, the presence of an M. tuberculosis ABC transporter in this group raises interesting questions about arabinogalactan assembly. From the perspective of fundamental microbial physiology and biochemistry, the NBDs in groups A and B are particularly interesting. These groups include NBDs from diverse eubacteria (Gram-negative and Gram-positive) and archaea, reflecting a rich array of biology. All of the group A and B NBDs have extended C-terminal domains. Some are predicted to contain CBMs capable of binding the export substrate (potentially via nonreducing terminal residues), similar to prototype ABC-dependent O-PS systems. However, the observation that these NBDs have significantly more sequence diversity, with a wide variety of putative enzymatic activities contained within the C-terminal domains, was entirely unanticipated. These provide fascinating candidates for further study and may identify new concepts in glycoconjugate assembly. As new genomic information becomes available, analyses such as these can bring valuable initial insight into assembly processes and provide a template for further (directed) experiments.