Genetic and Structural Characterization of the Core Region of the Lipopolysaccharide from Serratia marcescens N28b (Serovar O4)

Abstract

The gene cluster (waa) involved in Serratia marcescens N28b core lipopolysaccharide (LPS) biosynthesis was identified, cloned, and sequenced. Complementation analysis of known waa mutants from Escherichia coli K-12, Salmonella enterica, and Klebsiella pneumoniae led to the identification of five genes coding for products involved in the biosynthesis of a shared inner core structure: [l,d-HeppIIIα(1→7)-l,d-HeppIIα(1→3)-l,d-HeppIα(1→5)-KdopI(4←2)αKdopII] (l,d-Hepp, l-glycero-d-manno-heptopyranose; Kdo, 3-deoxy-d-manno-oct-2-ulosonic acid). Complementation and/or chemical analysis of several nonpolar mutants within the S. marcescens waa gene cluster suggested that in addition, three waa genes were shared by S. marcescens and K. pneumoniae, indicating that the core region of the LPS of S. marcescens and K. pneumoniae possesses additional common features. Chemical and structural analysis of the major oligosaccharide from the core region of LPS of an O-antigen-deficient mutant of S. marcescens N28b as well as complementation analysis led to the following proposed structure: β-Glc-(1→6)-α-Glc-(1→4))-α-d-GlcN-(1→4)-α-d-GalA-[(2←1)-α-d,d-Hep-(2←1)-α-Hep]-(1→3)-α-l,d-Hep[(7←1)-α-l,d-Hep]-(1→3)-α-l,d-Hep-[(4←1)-β-d-Glc]-(1→5)-Kdo. The D configuration of the β-Glc, α-GclN, and α-GalA residues was deduced from genetic data and thus is tentative. Furthermore, other oligosaccharides were identified by ion cyclotron resonance-Fourier-transformed electrospray ionization mass spectrometry, which presumably contained in addition one residue of d-glycero-d-talo-oct-2-ulosonic acid (Ko) or of a hexuronic acid. Several ions were identified that differed from others by a mass of +80 Da, suggesting a nonstoichiometric substitution by a monophosphate residue. However, none of these molecular species could be isolated in substantial amounts and structurally analyzed. On the basis of the structure shown above and the analysis of nonpolar mutants, functions are suggested for the genes involved in core biosynthesis.

In gram-negative bacteria, the lipopolysaccharide (LPS) is one of the major structural and immunodominant molecules of the outer membrane. It consists of three domains: lipid A, core oligosaccharide, and O-specific polysaccharide, or O antigen. The genetics of O-antigen biosynthesis has been intensively studied in members of the family Enterobacteriaceae and other gram-negative bacteria. Studies on characterization of the genes involved in LPS core biosynthesis in Escherichia coli, Salmonella enterica, and Klebsiella pneumoniae have shown that these genes are usually found clustered in a region of the chromosome, the waa (rfa) gene cluster (19, 38). This gene arrangement is not always present in other gram-negative bacteria (35): e.g., in Bordetella, it was shown that genes involved in the biosynthesis of the O antigen and core region are present in the same gene cluster (2). (The nomenclature proposed in 1996 by Reeves et al. [37] for proteins and genes involved in core LPS biosynthesis is used in this work, with the names originally reported given in parentheses.)

All core regions identified (23, 24) contain at least one residue of 3-deoxy-d-manno-oct-2-ulosonic acid (Kdo), which links this region to the lipid A moiety (Kdo I). The second characteristic sugar of the core region is l-glycero-d-manno-heptose (l,d-Hep), in addition to which, d-glycero-d-manno-heptopyranose (d,d-Hep) is present in a few LPSs. This sugar was identified as the biosynthetic precursor of l,d-Hep. Also, there are few LPSs that contain only d,d-Hep and others that lack heptose completely. Either Kdo I (in Acinetobacter) or Kdo II (in Burkholderia cepacia and Yersinia pestis) (23, 24) can be replaced by d-glycero-d-talo-oct-2-ulopyranosonic acid (Ko). Its biosynthesis and the regulation of the exchange between Kdo and Ko are still unknown.

In those cases in which l,d-Hep is present, the presence of one Hep-α-(1→5)-Kdo moiety is a characteristic feature. Kdo I may further be substituted at O-4 by a second Kdo residue (Kdo II; e.g., in S. enterica and E. coli).

Serratia marcescens is a recognized nosocomial pathogen that causes pneumonia, septicemia, meningitis, and urinary tract infections (1, 7). S. marcescens N28b (O4) produces a bacteriocin able to kill E. coli K-12 (48); this bacteriocin binds to the core of LPS and to the outer membrane proteins OmpA and OmpF of sensitive E. coli cells (11). It was expected that expression of foreign genes in E. coli K-12 leading to alterations of the relative amounts or composition of the outer membrane molecules that interact with bacteriocin 28b would confer a bacteriocin-resistant phenotype. We have shown that bacteriocin 28b is a useful tool to identify recombinant plasmids or cosmids harboring structural genes for small Ail-like outer membrane proteins that, when expressed in E. coli K-12, lead to a decrease in the outer membrane proteins OmpA and OmpF (16). Similarly, we have shown that expression in E. coli K-12 of genes coding for enzymes involved in S. marcescens O-antigen (40, 41) and core region (17, 38) biosyntheses confer a bacteriocin-resistant phenotype. This approach allowed the identification and characterization of the S. marcescens waaA (kdtA) gene, coding for Kdo transferase, and the adjacent waaE (kdtX) gene (17, 25).

In this work, the characterization of the complete waa gene cluster involved in S. marcescens LPS core biosynthesis and a structural investigation of the core region are presented.

MATERIALS AND METHODS

Bacterial strains, plasmids, and growth conditions.

Bacterial strains (Table 1) were grown in Luria-Bertani (LB) broth and on LB agar (30). LB medium was supplemented with kanamycin (50 μg ml⁻¹), ampicillin (100 μg ml⁻¹), chloramphenicol (30 μg ml⁻¹), and tetracycline (25 μg ml⁻¹) when needed. The physical maps of the plasmids used in this study are shown in Fig. 1.

TABLE 1.

Bacterial strains and plasmids used in this study

Strain, cosmid, or plasmid	Relevant characteristic(s)a	Source or reference
Strains
S. marcescens	Serotype O:4	14
N28b
N28b4	N28b-derived wbbA mutant (lacks O:4 antigen)	41
N28b16	N28b-derived waaE mutant	This work
N28b20	N28b-derived double orf9 orf10 mutant	This work
N28b30	N28b-derived orf7 mutant	This work
K. pneumoniae		32
52145	O1:K2
NC16	52145-derived waaE mutant	38
NC19	52145-derived waaQ mutant	38
E. coli		36
NM554	recA13 araD139 Δ(ara-leu)7696 Δ(lac)X74 galE15 galK16 hsdR2 rpsL mcrA mcrB
DH5α	F−endA hsdR17(rK− mK+) supE44 thi-1 recA1 gyrA96 φ80lacZM15 Δ(argF lacZYA)U169	18
X11 Blue	recA1 endA1 gyrA96 thi-1 hsdR17 supE44 relA1 lac (F′ proAB lacIθZΔM15 Tn10)	Stratagene
CJB26	waaA::kan recA-harboring plasmid pJSC2	5
S. enterica serovar Typhimurium
SA1377	waaC630	8
SL3789	waaF511	39
SL3749	waaL	29
Cosmids and plasmids
SuperCos1	Tetr Kmr cosmid vector	Stratagene
CosFGR2	Recombinant SuperCos1 harboring the S. marcescens orf10, orf11, and waaA, waaE, coaD, and fpg genes	17
CosFGR16	Recombinant SuperCos1 harboring S. marcescens waaC, orf4, waaL, orf6, orf7, waaQ, orf9, orf10, orf11, waaA, waaE, and coaD	This work
pKO3	CmrsacB temperature-sensitive replication	28
pKO3Km	KmrsacB temperature-sensitive replicon	This work
pKO3Δorf7	pKO3 containing the engineered orf7 deletion	This work
pKO3Δorf9-10	pKO3 containing the engineered double orf9 orf10 deletion	This work
pKO3ΔwaaE	pKO3 containing the engineered waaE deletion	This work
pGEMT	Ampr plasmid vector	Promega
pGEMT-WaaCSm	S. marcescens waaC gene in pGEMT	This work
pGEMT-WaaFSm	S. marcescens waaF gene in pGEMT	This work
pGEMT-WaaASm	S. marcescens waaA gene in pGEMT	This work
pGEMT-WaaESm	S. marcescens waaE gene in pGEMT	This work
pGEMT-WaaEKp	K. pneumoniae waaE gene in pGEMT	38
pGEMT-WaaLSM	S. marcescens waaL gene in pGEMT	This work
pGEMT-WaaQSm	S. marcescens waaQ gene in pGEMT	This work
pGEMT-Orf9-10Sm	S. marcescens orf9 and −10 in pGEMT	This work
pGEMT-Orf8-9Kp	K. pneumoniae orf8 and −9 in pGEMT	38
pGEMT-Orf7Sm	S. marcescens orf7 in pGEMT	This work
pGLU	LgtF from H. ducreyi	12
pJSC2	Cmr temperature-sensitive replication plasmid containing E. coli waaA	5

^aAmp^r, ampicillin resistance; Cm^r, chloramphenicol resistance; Km^r, kanamycin resistance; Tet^r, tetracycline resistance.

FIG. 1.

Diagram of the S. marcescens N28b waa region and comparison of this cluster with those of E. coli K-12, S. enterica serovar Typhimurium, and K. pneumoniae C3. The physical maps of plasmids used in this study showing only insert DNA are shown. Small arrows denote primers used to amplify and characterize the gmhD-waaF region. Inner core genes common to all known Enterobacteriaceae (black arrows), inner core genes common to E. coli and S. enterica serovar Typhimurium (arrows with horizontal bars), core genes common to S. marcescens N28b (O4) and K. pneumoniae (striped arrows), hypothetical lipooligosaccharide or capsule-related genes (checkerboard arrows), and the O-antigen ligase gene (gray arrows) are also shown.

Bacteriocin 28b production and sensitivity assay.

Bacteriocin 28b was prepared as previously described (48). The overlay test for qualitative bacteriocin sensitivity assays and quantitative bacteriocin sensitivity assays were performed as previously described (11).

Southern blot hybridization.

The DNA fragment containing the waaA and waaE genes from S. marcescens was labeled with digoxigenin as described by the manufacturer (Boehringer Mannheim). BamHI-digested CosFGR16 DNA was electrophoresed, denatured, and transferred to Hybond B membrane. After baking, the membrane was prehybridized and hybridized in 5× SSC (1× SSC is 0.15 M NaCl plus 0.015 M sodium citrate)-0.5% blocking reagent (Boehringer Mannheim)-0.1% Sarkosyl-0.02% sodium dodecyl sulfate (SDS). Washing, antibody incubation, and signal detection with p-nitroblue tetrazolium chloride and 5-bromo-4-chloro-3-indolylphosphate were done as recommended by the manufacturer (Boehringer Mannheim).

General DNA methods.

General DNA manipulations were done essentially as described previously(42). DNA restriction endonucleases, T4 DNA ligase, E. coli DNA polymerase (Klenow fragment), and alkaline phosphatase were used as recommended by the suppliers. The S. marcescens N28b genomic DNA used was described previously (41). The following plasmids subcloned in vector pGEMT were constructed by ligation to the vector of PCR-amplified products: pGEMT-WaaC_Sm (5′-GTTTAATGCACGTTGCCGCA-3′ and 5′-CCCAGGTTGATAATGTGCAG-3′), pGEMT-WaaF_Sm (5′-ACAAAAAAGGCAGCATCGAGTA-3′ and 5′-TGTCGCTGCCGAACCAGTTT-3′), pGEMT-WaaL_Sm (5′-GCTGTTGTCGCATATCGACT-3′ and 5′-TGCATGCTGCAGGCCGACATT-3′), pGEMT-WaaQ_Sm (5′-GCGAACTCGACGTAAGCC-3′ and 5′-TGCACGCCCATAAAGTGAA-3′), pGEMT-orf9-10_Sm (5′-TCAAATGCTGGAGCGAAGAG-3′ and 5′-TGTTCTTTGGCGATACCGATA-3′), pGEMT-orf11 (5′-AATCCGCCGCAGATAAATCA-3′ and 5′-GATCACCAGCTTGGGATTCA-3′), pGEMT-WaaA_Sm (5′-AGGCGTGGTGCAAACAAGAT-3′ and 5′-AAGACTTTGGCGCCCAGACT-3′), and pGEMT-WaaE_Sm (5′-ACCTTCAACTTTAAAGACA-3′ and 5′-AAAGTCAGACACCGCCCG-3′).

DNA sequencing and computer analysis of sequence data.

Double-stranded DNA sequencing was performed by using the Sanger dideoxy-chain termination method (43) with the Abi Prism dye terminator cycle sequencing kit (Perkin-Elmer). The relevant parts of CosFGR2 and CosFGR16 inserts were sequenced with oligonucleotide T3 (5′-AATTAACCCTCACTAAAGGG-3′), which binds to the T3 promoter region on vector SuperCos1; oligonucleotide F412 (5′-TTTGCACCACGCCTCTGA-3′) to extend the sequence from the previously reported waaA gene (17); and oligonucleotide FB1 (5′-CGGCTTCCTCGACGGTAAA-3′) to obtain sequence data downstream of the known waaE gene sequence (17). Other sequence-derived oligonucleotides were used to extend the nucleotide sequence. Primers used for DNA sequencing were purchased from Amersham Pharmacia Biotech. The DNA sequence was translated in all six frames, and all open reading frames (ORFs) greater than 100 bp were inspected. Deduced amino acid sequences were compared with those of DNA translated in all six frames from nonredundant GenBank and EMBL databases by using the BLAST (3, 4) and FASTA (33) network service at the National Center for Biotechnology Information and the European Biotechnology Information, respectively. The Genetics Computer Group (Madison, Wis.) package Terminator program was used for prediction of possible terminator sequences. Clustal W (46) was used for multiple sequence alignments. Hydropathy profiles were calculated according to Kyte and Doolitle (26). The TopPredII program (9) was used to identify predicted protein transmembrane domains.

LPS isolation and electrophoresis.

Cultures for analysis of LPS were grown in Trypticase soy broth at 37°C. LPSs were extracted with either hot phenol-water (51) or phenol-chloroform-light petroleum (13). For screening purposes, LPS was obtained after proteinase K digestion of whole cells (21). LPS samples were separated by SDS-polyacrylamide gel electrophoresis (PAGE) or SDS-Tricine-PAGE and visualized by silver staining as previously described (34, 47).

Isolation of oligosaccharide 3.

The LPS was hydrolyzed in 1% acetic acid (100°C for 90 min), and the precipitate was removed by centrifugation (2,500 × g for 1 h). The supernatant was evaporated to dryness, dissolved in water, centrifuged (100,000 × g at 4°C for 4 h), and separated by gel permeation chromatography on a column (2.5 by 70 cm) of Sephadex G-50. The core oligosaccharides were then separated by high-performance anion-exchange chromatography (HPAEC) on a column (4 by 250 mm; Dionex Corp.) of CarboPack PA100, which was eluted at 1 ml min⁻¹ with a linear gradient program of 15 to 40% 1 M sodium acetate in 0.1 M NaOH over 70 min, and isolated fractions were desalted by gel-permeation chromatography on a column (1 by 70 cm) of Sephadex G-10 in 10 mM aqueous NH₄HCO₃.

Compositional analyses.

Neutral sugar and uronic acid (as neutral sugars, after reduction of the carboxyl group) analyses, fatty acid analyses, determination of organic bound phosphate, and Kdo and GlcN quantification were performed as described previously (31).

NMR spectroscopy.

For structural assignments, NMR spectra were recorded on a solution (0.5 ml) of oligosaccharide 3 (2 mg) in ²H₂O with a Bruker AMX 600 spectrometer (¹H NMR, 600.13 MHz; ¹³C NMR, 125.77 MHz) and a Bruker Digital Avance 800 instrument at 27 or 47°C. The resonances were measured relative to internal acetone: (CH₃)₂CO δ_H, 2.225; δ_C, 31.07. The correlation (COSY), total correlation spectroscopy (TOCSY), double-quantum-filtered COSY (DQFCOSY), as well as the ¹H-¹³C-heteronuclear multiple-quantum coherence (HMQC) and nuclear Overhauser enhancement spectroscopy (NOESY) experiments were all measured with standard Bruker software.

Mass spectrometry.

Ion cyclotron resonance Fourier-transformed electrospray ionization mass spectrometry (ESI FT-ICR-MS) was performed in the negative-ion mode with an APEX II instrument (Bruker Daltonics) equipped with a 7-T actively shielded magnet and an Apollo ion source. Mass spectra were acquired with standard experimental sequences as provided by the manufacturer. Samples were dissolved at a concentration of ∼10 ng · μl⁻¹ in a 50:50:0.001 (vol/vol/vol) mixture of 2-propanol, water, and triethylamine and sprayed at a flow rate of 2 μl · min⁻¹. Capillary entrance voltage was set to 3.8 kV, and the dry gas temperature was set to 150°C. To facilitate the interpretation, mass spectra, which showed several charge states for each component, were charge deconvoluted, and the mass numbers given refer exclusively to the monoisotopic molecular masses.

S. marcescens orf7, orf9-10, and waaE mutant construction.

To obtain S. marcescens mutant strains N28b30, N28b20, and N28b16, the method of Link et al. (28) was used to create chromosomal in-frame waa deletions. Briefly, CosFGR16 and primer pairs A (5′-CGCGGATCCCCGTTGGGCGTTCAACGAAT-3′), B (5′-CCCATCCACTAAACTTAAACAGAACCAGTCGGCAACCTTAAT-3′), and C (5′-TGTTTAAGTTTAGTGGATGGGATTCAGCCGCAGCGGATTTAT-3′), D (5′-CGCGGATCCGCAGGGGAAACGTTCGAAGA-3′) were used in two sets of asymmetric PCRs to amplify DNA fragments of 602 (AB) and 546 (CD) bp, respectively. DNA fragment AB contains from nucleotide 7302 (corresponding to the third base of the 13th codon of orf7) to nucleotide 7904. DNA fragment CD contains from nucleotide 5756 (inside orf6) to nucleotide 6302 (corresponding to the first base of codon 346 of orf7). DNA fragments AB and CD were annealed at their overlapping region (underlined letters in primers B and C) and amplified by PCR as a single fragment with primers A and D. The fusion product was purified, BamHI digested (BamHI site shown as double-underlined letters in primers A and D), ligated into BamHI-digested and phosphatase-treated pKO3Km vector, electroporated into E. coli DH5α, and plated on kanamycin plates at 30°C to obtain plasmid pKO3KmΔorf7. Plasmid pSKF41 (17) and primer pairs A1 (5′-CGCGGATCCCACCGCAAGCTGCTGGAAAA-3′) and B1 (5′-CCCATCCACTAAACTTAAACAGCTTTTGCGGCTGCTCATTC-3′) and C1 (5′-TGTTTAAGTTTAGTGGATGGGGTGGTCAACGCGCAATATAC-3′) and D1 (5′-CGCGGATCCTCCTTCACCAGTGATGAGGA-3′) were used to obtain plasmid pKO3KmΔwaaE, which contains an internally deleted waaE gene (the first 6 codons, a 7-codon tag, and the last 24 codons) flanked by 541 bp upstream and 409 bp downstream. Plasmid CosFGR16 and primer pairs A2 (5′-CGCGGATCCAAATGCTGGAGCGAAGAGA-3′) and B2 (5′-CCCATCCACTAAACTTAAACACGCCAAAAGAAATGCTTTC-3′) and C2 (5′-TGTTTAAGTTTAGTGGATGGGAATCCGCCGCAGATAAATCA-3′) and D2 (5′-CGCGGATCCTTGGGCACGAAAGATATTCA-3′) (BamHI site shown as double-underlined letters in primers A2 and D2) were used to obtain plasmid pKO3KmΔorf9-10, which contains a double-deleted orf9-orf10 pair (the first 7 codons of orf9, a 7-codon tag, and the last 20 codons of orf10) flanked by 480 bp upstream and 787 bp downstream. Plasmids pKO3KmΔorf7, pKO3KmΔwaaE, and pKO3KmΔorf9-10 were used to construct mutants N28b30, N28b16, and N28b20, respectively.

Nucleotide sequence accession number.

The nucleotide sequence of the S. marcescens N28b waa gene cluster containing the gmhD, waaF, waaC, orf4, waaL, orf6, orf7, waaQ, orf9, orf10, orf11, waaA, waaE, coaD, and fpg genes has been deposited in GenBank under accession no. U 52844 (Fig. 1).

RESULTS

Cloning and sequence determination of the S. marcescens waa gene cluster. We have previously reported the isolation and characterization of S. marcescens N28b waaA, coding for Kdo transferase, and waaE (kdtX) genes from recombinant CosFGR2 (17). DNA fragments containing waaQ (rfaQ) or waaGPSBI (rfaGPSBI) genes from E. coli K-12 were found to hybridize to recombinant CosFGR2, suggesting that other core LPS biosynthesis-encoding genes were located in CosFGR2 besides waaA and waaE. The nucleotide sequence of the whole insert in CosFGR2 revealed only two putative complete open reading frames (ORFs) upstream of the waaA gene (Fig. 1). To find further upstream genes, an S. marcescens N28b genomic library was screened again by a previously described approach (17). A recombinant cosmid was isolated (CosFGR16) that overlapped with CosFGR2 and contained six additional complete ORFs (Fig. 1). In all of the waa gene clusters characterized so far in the Enterobacteriaceae, the cluster begins with a gmhD gene (19, 38). To characterize this region, two oligonucleotides were used. The GMHD1 oligonucleotide (5′-TGAAAGSCGGCACCAAGTTT-3′) was designed from the known K. pneumoniae gmhD gene sequence (38), and the WAAF1 oligonucleotide (5′-GTGTCGCTGCCGAACCAGTTT-3′) was designed from the sequence determined from the insert of CosFGR16. Using genomic DNA from the N28b strain as a template, these oligonucleotides generated a PCR-amplified DNA fragment of about 2 kbp. The nucleotide sequence determined for this fragment overlapped that of the CosFGR16 insert (Fig. 1).

Analysis of the S. marcescens waa gene sequence.

Analysis of the 20,693-bp nucleotide sequence revealed 14 ORFs. Sequences corresponding to putative ribosome binding sites were found upstream of each of the ORF start codons. Data summarizing the location of the ORFs and the characteristics of the putative encoded proteins are shown in Table 2. The analysis of the intergenic regions between the successive ORF pairs revealed distances of 10 (orf1-orf2), 3 (orf2-orf3), and 29 (orf4-orf5) bp and overlapping stop and start codons for the orf3-orf4 and orf5-orf6 pairs. Since no sequences similar to Rho-independent transcription termination sequences were found between orf4 and orf5, this organization suggests that the first six ORFs constitute a transcriptional unit. A similar analysis of the intergenic regions between the other ORFs suggests that orf8, -9, and -10 and orf12, -13, and -14 constitute two additional transcription units, while orf7 and -11 apparently correspond to monocistronic genes transcribed in the opposite direction.

TABLE 2.

S. marcescens N28b waa gene cluster and downstream coaD (kdtB) and fpg genes

Locus	Base positionsd	No. of aminoacids	Molecular mass (kDa)	pIa	GRAVYb
orf1 (gmhD)c	2-775	257	29.3	4.87	−0.444
orf2 (waaF)	785-1831	348	38.7	9.11	−0.123
orf3 (waaC)	1834-2799	321	35.9	8.44	−0.318
orf4	2796-3962	388	43.4	8.39	−0.104
orf5 (waaL)	3991-5232	413	46.0	9.40	0.586
orf6	5229-6209	326	37.4	8.89	−0.277
orf7	6246-7340c	364	41.2	9.02	−0.161
orf8 (waaQ)	8516-9598	360	40.1	7.22	−0.037
orf9	9595-10722	375	42.3	9.47	−0.201
orf10	10725-11825	366	40.3	9.14	−0.114
orf11	11870-12883c	337	38.2	8.97	−0.239
orf12 (waaA)	13009-14286	425	47.6	9.53	−0.055
orf13 (waaE)	14287-15060	257	29.2	8.82	−0.283
orf14 (coaD) (15)	15064-15549	161	17.6	5.55	0.242
orf15 (fpg) (44)	15553-16368c	271	30.5	8.56	−0.339

^aIsoelectric point of the protein calculated using ProtParam at the ExPassy server.

^bGrand average hydropathicity of the protein calculated by the Kyte and Doolitle method (26).

^cTruncated ORF.

^dc indicates that base positions correspond to the complementary DNA strand.

S. marcescens waa genes shared by all known Enterobacteriaceae.

In the LPSs of the Enterobacteriaceae studied so far, the inner core region contains one to three Kdo residues and at least three residues of l-glycero-d-manno-heptopyranose (l,d-HeppI, l,d-HeppII, and l,d-HeppIII). These residues constitute the structure l-α-d-HeppIII-(1→7)-l-α-d-HeppII-(1→3)-l-α-d-HeppI-(1→5)-[α-KdopII-(2→4)-]-α-KdopI (23, 24). The 5′-truncated orf1 and orf2, orf3, orf8, and orf12 had high levels of amino acid identity to the known enterobacterial GmhD (ADP-d-glycero-d-manno-heptose epimerase) (95 to 96%), WaaF (ADP-heptose-LPS heptosyltransferase II) (82 to 88%), WaaC (ADP-heptose-LPS heptosyltransferase I) (82 to 88%), WaaQ (ADP-heptose-LPS heptosyltransferase III) (82 to 88%), and WaaA (bifunctional CMP-Kdo:lipidA Kdo transferase (83 to 88%) proteins, respectively (10, 19, 20, 38) (Fig. 1).

Complementation analyses of known inner core backbone mutants were performed to confirm the functions of these genes. E. coli strain CJB26 harbors a kanamycin resistance determinant inserted in the waaA gene and a wild-type waaA gene in a temperature-sensitive plasmid (pJSC2), leading to a temperature-sensitive phenotype. A plasmid containing the S. marcescens waaA gene (pGEMT-WaaA_Sm) was found to restore growth at 44°C in the E. coli CJB26 mutant.

S. enterica serovar Typhimurium mutant strains SA1377 (waaC630) and SL3789 (waaF511) were complemented by plasmids pGEMT-WaaC_Sm and pGEMT-WaaF_Sm, respectively, as judged by analysis of LPS by SDS-Tricine-PAGE. These results strongly suggested that orf3 and orf2 coded for the ADP-heptose-LPS heptosyltransferase I and ADP-heptose-LPS heptosyltransferase II, respectively.

To identify the gene encoding ADP-heptose-LPS heptosyltransferase III, the K. pneumoniae waaQ mutant strain NC19 was used. LPS obtained from mutant NC19 contained O antigen and migrated slightly faster than that of the parent 52145 strain. Strain NC19 was transformed with a plasmid containing the S. marcescens orf8 (pGEMT-WaaQ_Sm). LPS from the transformed strain showed an electrophoretic banding pattern identical to that of the wild-type strain (data not shown). These results suggested that orf8 codes for ADP-heptose-LPS heptosyltransferase III.

Three other waa genes shared by S. marcescens N28b and K. pneumoniae C3.

We have shown that in both S. marcescens and K. pneumoniae, a similar gene (waaE) is present downstream from the waaA gene (17, 38) (Fig. 1). Recently, we presented evidence suggesting that the K. pneumoniae waaE gene is involved in the addition of a branched d-Glc residue to l,d-HeppI by a β-(1→4) linkage (25). Both WaaE proteins share high levels of similarity and identity (70 and 80%), suggesting that they perform the same function. To test this hypothesis, an S. marcescens waaE nonpolar mutant was constructed essentially as previously described (28, 38). LPSs from strains N28b (wild type) and N28b16 (waaE) were extracted and analyzed by SDS-Tricine-PAGE. The result obtained (Fig. 2, lanes 1 and 2) showed that the core LPS from strain N28b16 migrated faster than that of the wild-type strain, and it appeared that the mutant LPS still contained O antigen, although in smaller amounts than wild-type LPS. Plasmids containing either the S. marcescens waaE gene (pGEMT-WaaE_Sm) or the K. pneumoniae waaE homologue (pGEMT-WaaE_Kp) were transformed into NC16 and N28b16 mutants. Analysis of LPS by SDS-Tricine-PAGE showed that both waaE homologues were able to complement waaE mutations in both strains (Fig. 2, lane 3). In addition, the lgtF gene from Haemophilus ducreyi (Fig. 2, line 4) complemented strain N28b16. These results suggested that, similarly to K. pneumoniae and H. ducreyi, a substitution of position O-4 of l,d-HeppI by a β-d-glucopyranose [β-d-Glcp-(1→4)-l,d-HeppI] should be present in the S. marcescens LPS inner core.

FIG. 2.

SDS-Tricine-PAGE analysis of LPS from S. marcescens. N28b (wild type) (lane 1), N28b16 (waaE) (lane 2), N28b16(pGEMT-WaaE_Kp) (lane 3), and N28b16(pGLU) (lane 4).

The deduced 375-amino-acid protein encoded by orf9 was found to share high levels of similarity (80%) and identity (70%) to the protein encoded by orf8_Kp in the K. pneumoniae waa gene cluster (38) (Fig. 1). Furthermore, both proteins shared the same number of residues and also showed limited similarity to WaaG proteins from Pseudomonas aeruginosa (accession no. O33426) and S. enterica (19) and E. coli core types K-12 (44) and R2, R3, and R4 (19, 20). WaaG protein was reported to be a glucosyltransferase involved in the α1→3 linkage of d-GlcpI to l,d-HeppII in E. coli and S. enterica (19). The deduced 366-amino-acid protein encoded by orf10_Sm showed similarity (68%) and identity (58%) to the protein encoded by orf9_Kp in the K. pneumoniae waa gene cluster (38) (Fig. 1). The proteins encoded by orf8_Kp, orf9_Kp, orf9_Sm, and orf10_Sm belong to the retaining glycosyltransferase family 4 (http://afmb.cnrs-mrs.fr/∼pedro/CAZY/db.html).

The above analyses suggested that S. marcescens orf9 and orf10 were involved in biosynthesis of the core region. To test this hypothesis, a double nonpolar orf9 orf10 mutant, strain N28b20, was constructed (see Materials and Methods). LPS was extracted from the mutant and wild-type strains, and analysis of LPS preparations by SDS-Tricine-PAGE showed a faster-migrating band for the mutant strain N28b20 LPS (Fig. 3, lanes 1 and 2). As expected, the mutant phenotype was complemented by introduction of plasmid pGEMT-orf9_Sm-10_Sm (Fig. 3, lane 3), while neither orf9 nor orf10 alone was able to complement the double mutation. The similarity of S. marcescens orf9 and orf10 to K. pneumoniae orf8 and orf9 indicated that these ORFs could perform similar functions in the inner core biosynthesis in both species. Further support for this suggestion was obtained by complementation of the N28b20 mutant by pGEMT-orf8_Kp-9_Kp, as judged by SDS-Tricine-PAGE analysis (Fig. 3, lane 4).

FIG. 3.

SDS-Tricine-PAGE analysis of LPS from S. marcescens N28b (lane 1), N28b20 (double orf9 orf10 mutant) (lane 2), N28b20(pGEMT-Orf9-10_Sm) (lane 3), and N28b20(pGEMT-Orf8-9_Kp) (lane 4).

Core structure of LPS from S. marcescens strain N28b4.

These genetic data indicated that the core region of the LPS from S. marcescens N28b should share with the LPS from K. pneumoniae three residues in addition to the common inner core features shared by all known cores from LPS of Enterobacteriaceae. To test this hypothesis, the structure of the core region from LPS of the O-antigen-deficient mutant N28b4 was investigated. Sugar analyses of the LPS revealed the presence of Glc, GlcN, GalA, Kdo, and l,d- and d,d-Hep residues (Fig. 4). In addition, small amounts of Ko were detected in Glc-MS analysis.

FIG. 4.

Proposed structure of oligosaccharide 3 isolated from the LPS from S. marcescens N28b4. The D configuration of residues B, F, and I was deduced from genetic data and thus is tentative, as is the partial structure D-(I-)-C. In oligosaccharide 1, residues H and K could not be identified. The proposed functions for the S. marcescens waa genes and the effect of the double orf9 orf10 mutant (N28b20) are also shown.

The core oligosaccharide fraction was isolated from LPS after acetic acid hydrolysis and gel permeation chromatography on Sephadex G-50 (see Materials and Methods) and further separated by preparative HPAEC, yielding a complex mixture of oligosaccharides, from which oligosaccharides 1, 2, and 3 were obtained. Monosaccharide analysis of these three isolates revealed that all contained Glc, GlcN, GalA, l,d-Hep, and d,d-Hep; however, the latter was present in oligosaccharide 2 only in traces. Since ¹H NMR analyses (not shown) of oligosaccharide 2 suggested that this was not a pure compound and since ¹H NMR analyses (not shown) of oligosaccharide 1 suggested that it represented a smaller variant of oligosaccharide 3, only the structure of oligosaccharide 3 was studied in detail by ¹H and ¹³C NMR spectroscopy.

The anomeric region of the ¹H NMR spectrum region of oligosaccharide 3 contained 10 signals (Fig. 5 and Table 3). Additionally, signals of a deoxy compound (residue K) were identified between 2.34 and 3.96 ppm. The 11 residues could be identified as 5 heptose residues, 3 hexose residues, 1 aminohexose residue, 1 hexosuronic acid residue, and, presumably, 1 Kdo derivative. Their identification was possible by the assignment of most chemical shifts, utilizing two-dimensional homonuclear ¹H, ¹H COSY, TOCSY, and NOESY and heteronuclear ¹H, ¹³C HMQC experiments. Two hexose residues (J and I, Fig. 5) possessed the β-gluco-configuration, as identified by their chemical shift data and J_2,3 coupling constants. Another hexose (residue A) and the hexosamine (F) possessed the α-gluco configuration. The hexuronic acid was α-galacto configured, and the heptoses (C, D, E, G, and H) were α-manno configured, which in the latter cases was established by the chemical shifts of the anomeric protons and their small J_2,3 coupling constants of about 1 Hz. According to the anomeric proton shifts and the coupling constants of H-1, H-2, H-3, and H-4, the heptose residues could be classified into three groups. Residues G and H gave very similar chemical shifts and possessed anomeric proton resonances at high field, in agreement with their proposed terminal nature. The anomeric proton of the proposed trisubstituted heptose C was the one possessing a chemical shift at relative low field. In heptoses D and E, the chemical shifts of the anomeric protons were found in intermediate positions. Finally, residue K resembled Kdo: however, with some unidentified modification. Probably, it represented an artifact that was formed from Kdo during alkaline HPAEC. The chemical shifts at 2.34 and 2.81 ppm (probably of protons H-3_ax and H-3_eq) were not consistent with those published for pyranosidic Kdo (6). In the ¹³C NMR spectrum, the chemical shift of a carbonyl group was identified at 180.35 ppm. Chemical shifts at 2.34 and 2.81 ppm possessed intensities similar to those of the anomeric proton signals of the other residues. Thus, it was supposed that residue K was part of oligosaccharide 3. Other proton chemical shifts of K were at 3.64, 3.87, and 3.96 ppm.

FIG. 5.

Anomeric regions of the ¹H NMR spectra of oligosaccharide 3. The letters refer to the carbohydrate residues and denote the anomeric proton of each residue.

TABLE 3.

¹H NMR and ¹³C NMR chemical shift data for oligosaccharide 3 derived from LPS of the rough mutant S. marcescens N28b4^a

Residue	Chemical shift (ppm)
	H-1	H-2	H-3	H-4	H-5	H-6a	H-7a/ H-6b	H-7b
	C-1	C-2	C-3	C-4	C-5	C-6	C-7
B, α-GalA	5.45	4.08	4.25	4.49	4.53
	100.1	74.0	68.8	80.3	71.96	175.7
A, α-Glc	5.49	3.64	3.76	3.57	3.96	3.93	4.23
	100.1	72.5	73.1	69.6	72.6	69.5
C, α-l,d-Hep	5.47	4.08	4.11	4.3	3.82	4.11
	101.6	70.7	73.0	75.2	72.5	69.6
D, α-l,d-Hep	5.39	4.2	4.03	3.81	3.67	4.24	3.85	3.76
	101.8	70.8	80.0	66.2	73.9	68.7	69.9
E, α-d,d-Hep	5.32	4.02	4.01	3.95	3.86	4.05
	96.6	81.2	71.2	70.9	71.5	73.0
F, α-GlcN	5.14	2.90	4.09	3.78	4.34	3.86	3.98
	98.3	52.0	70.6	77.0	71.4	61.2
G, α-Hep	5.08	4.07	3.89	3.88	3.73	4.07	3.78	3.75
	103.0	71.0	72.0	67.0	74.0	73.4	64.0
H, α-l,d-Hep	5.04	4.06	3.91	3.94	3.67	4.1	3.82	3.74
	101.0	70.5	72.2	67.2	72.2	71.0	64.0
I, β-Glc	4.61	3.34	3.56	3.4	3.44	3.77	3.93
	103.9	73.6	76.2	70.9	77.2	62.5
J, β-Glc	4.52	3.36	3.54	3.45	3.5	3.76	3.96
	103.8	74.4	76.2	70.7	76.5	62.0
K			2.34ax, 2.81eq	3.96	3.88	3.65
	180.4		42.0	74.6	75.6	78/76.8

^aMonosaccharide residues are as shown in Fig. 5.

The ¹³C NMR chemical shifts were assigned by heteronuclear ¹H, ¹³C HMQC experiments using the interpreted ¹H NMR spectrum. Nine anomeric signals were identified in the ¹³C NMR spectrum of oligosaccharide 3. There was no anomeric signal, but a chemical shift (probably of C-3 of the modified Kdo residue) could be detected. Low-field shifted signals indicated substitutions at O-2 (residue E), O-4 (residue F), O-6 (residue A), O-2 and O-4 (residue B), O-3 and O-4 (residue C), and O-3 and O-7 (residue D). Residues J, G, I, and H were terminal sugars.

The sequence of the monosaccharide residues was determined by using data obtained from a NOESY experiment (Table 4). Interresidual NOE contacts were identified from H-1 of GalA residue B to protons H-3 (strong) and H-2 of heptose D and to H-1 of heptose E; H-1 of residue A and H-4 (strong) of residue F; H-1 of residue D and H-3 (strong), H-4, and H-2 of residue C; H-1 of residue E, H-1 (strong) and H-2 (strong) of residue B, and H-1 (weak) of residue G; H-1 of residue F and H-4 (weak) and H-5 (strong) of residue B; H-1 of residue G and H-2 and H-3 (both strong) of residue E; H-1 of residue H and H-7 (strong) and H-6 of residue D; H-1 of residue I and H-2 (weak) and H-6 (strong) of residue C; and H-1 of residue J and H-6 (strong) of residue A. Finally, a NOE contact could be determined between H-1 of heptose residue C and the putative deoxy protons of K. With regard to residue I, no NOE connectivity could be identified between its H-1 and H-4 of residue C. However, C was substituted at O-3 and O-4 (¹³C chemical shifts of C-3 and C-4 at 73.0 and 75.2 ppm, respectively). According to genetic data and in analogy to other enterobacterial core structures, it is highly likely that residue C was substituted at O-3 and O-4 by residues D and I, respectively. C itself should be linked to Kdo, probably at O-5.

TABLE 4.

NOE signals of oligosaccharide 3 observed in the NOESY spectrum^a

Residue	Proton	Signal
		Interunit	Intraunit
B, α-GalA	B1	D3(s), D2(w), E1(s)	B5(m)
	B5	D2
A, α-Glc	A1	F4(s)	A3(s), A2(m)
	A6	J2(w)
D, α-l,d-Hep	D1	C3/C6(s), C2(m), C4(m), B5(w)	D3(m), D2(w), D7(w)
E, α-d,d-Hep	E1	B1(s), B2(s), B3(m), G1(s), G5(s)	E1(s), E2(m)
F, α-GlcN	F1	B5(s), B4(w)
G, α-Hep	G1	E2(s), E3(s)	G2(s), G3(w)
H, α-l,d-Hep	H1	D7(s), D6(s)	H3(s), H2(s), H5(w)
I, β-Glc	I1	C6/C3(s), C2(w)	I5(s), I4(m), I3(m)
J, β-Glc	J1	A6, A5	J5(s), J4(m), J3(m)
C, α-l,d-Hep	C1	K 3.65 ppm(s), K3.88 ppm(m)
Kdob	2.811	C2, C4

^aShown are signals that were important for the structural determination. Monosaccharide residues are as depicted in Fig. 5. w, weak; m, medium; s, strong.

^bUnknown Kdo modification.

Although the absolute configuration of the monosaccharide residues was not determined, the successful complementation of the N28b16 and N28b20 mutants by the K. pneumoniae waaE and orf8 and orf9 genes, respectively, and the known K. pneumoniae core structure suggested that residues F, B, and I are D configured. The l,d configuration of the inner core heptose residues C, D, and H was deduced from compositional analysis of this and the LPS of strain N28b20 (see below).

ESI FT-ICR-MS.

In order to check whether the chemical structure determined from the isolated oligosaccharide 3 was representative, ESI FT-ICR-MS was performed from the core oligosaccharide fraction (see above). The charge-deconvoluted mass spectrum (Fig. 6) comprised a complex pattern of mass peaks. The peaks with the highest intensity (2,021.67 and 2,003.66 Da) were in excellent agreement with a molecule consisting of KdoHep₅Hex₃HexAHexN and its anhydro form (calculated masses of 2,021.65 and 2,003.63 Da), respectively. Thus, core oligosaccharide 3 represented the most abundant compound in the mixture of core oligosaccharides. Additional peaks (summarized in Table 5) could be attributed to molecular species comprising (i) oligosaccharide 3 substituted by an additional Ko residue (peak 2 in Fig. 6 [2,257.72 Da]), (ii) oligosaccharide 3 substituted by an additional HexA (peak 5, 2,197.70 Da), and (iii) oligosaccharide 3 lacking two Hep residues but containing an additional HexA (peak 4, 1,813.56 Da). Of peaks 1, 2, and 4, species with an 80-Da-higher mass were identified (peaks 3, 7, and 6, respectively), which could represent monophosphorylated derivatives. None of these molecular species representing variants of oligosaccharide 3 could be isolated.

FIG. 6.

ESI FT-ICR-mass spectrum of the core oligosaccharide fraction isolated from the LPS of strain N28b4 after acetic acid hydrolysis and gel permeation chromatography. The mass numbers refer to the monoisotopic neutral molecular peak.

TABLE 5.

Mass peaks identified in the charge-deconvoluted ESI FT-ICR mass spectrum of the core oligosaccharide fraction isolated from the LPS of strain N28b4 after acetic acid hydrolysis and gel permeation chromatography^a

Peak no.	Mass (Da)		Proposed structure
	Identified	Calculated
1	2,021.67	2,021.65	Oligosaccharide 3
	2,003.66	2,003.63	Anhydro form
2	2,257.72	2,257.71	Oligosaccharide 3 plus one Ko residue
3	2,101.63	2,101.61	Oligosaccharide 3 plus one phosphate residue
	2,083.69	2,083.60	Anhydro form
4	1,813.56	1,813.55	KdoHep3HexA2Hex3HexN
	1,795.55	1,795.54	Anhydro form
5	2,197.70	2,197.68	KdoHep5HexA2Hex3HexN
	2,179.68	2,179.67	Anhydro form
6	1,893.43	1,893.52	KdoHep3HexA2Hex3HexNP
	1,875.51	1,875.51	Anhydro form
7	2,337.69	2,337.68	Oligosaccharide 3 plus 1 Ko residue and P residue

^aListed are monoisotopic masses of the neutral molecule. The peaks are given in descending order of intensity. The peak numbers are as depicted in Fig. 6. Additional sodium and potassium adducts (+22 and +38 Da, respectively) found in the spectrum are not listed.

Charge-deconvoluted ESI FT-ICR-MS analysis of the core fraction of the LPS from the wild-type strain N28b gave a pattern of peaks comprising those of the O-antigen-deficient core, described above, plus one peak of low intensity that represented oligosaccharide 3 lacking two Hep residues (1,637.53 Da). Thus, the LPS from both the wild-type strain and the O-antigen-deficient mutant share the same core structures.

Proposed waa gene functions.

The genetic complementation experiments allowed the identification of the S. marcescens N28b genes coding for the enzymes involved in the transfer of the Kdo (waaA), Hep I (waaC), Hep II (waaF), Hep III (waaQ), and Glc (waaE) inner core residues. The elucidation of the S. marcescens N28b LPS core structure allowed us to hypothesize functions for the remaining genes of the cluster.

The core fraction of the LPS from strain N28b20 (orf9_Sm orf10_Sm double mutant) was obtained by mild acetic acid hydrolysis and recovered by gel permeation chromatography, as described for the fraction of the LPS from strain N28b4. Compositional analysis of this core fraction revealed the absence of GlcN, GalA, and d,d-Hep residues and a strong reduction in the Glc content. Furthermore, charge-deconvoluted ESI FT-ICR-MS analysis gave major peaks at 622.19 and 604.17 Da that corresponded to the oligosaccharide Hep₂Kdo (calculated, 622.20 Da) and its anhydro product (calculated, 604.19 Da). Another peak at 858.25 Da corresponded to the oligosaccharide Hep₂KdoKo (calculated, 858.26 Da), of which the (Na-H) and (K-H) adducts were also identified (peaks at 870.24 and 896.20 Da, respectively). Other peaks with smaller intensities were present but could not be attributed. Since no d,d-Hep was detected in the compositional analysis of this core fraction, it was concluded that, as in other Enterobacteriaceae, the first two inner core heptose residues were in the l,d configuration. According to the proposed structure of the core region from the strain N28b4 O-deficient mutant (Fig. 4), it could be predicted that these two genes (orf9_Sm orf10_Sm) should code for the enzymatic activities involved in the transfer of d-GlcN and d-GalA. However, further work will be necessary to identify the predicted d-GalA- and d-GlcN-transferases between these two genes.

BLAST and position-iterated BLAST searches of the orf7- and orf11-encoded proteins showed that both proteins shared similarities to known ADP-heptose-LPS heptosyltransferases, and both proteins contained a domain characteristic of glycosyltransferase family 4. The orf7- and orf11-encoded proteins were similar to known WaaF (22 to 23% identity and 39 to 40% similarity) and WaaQ (23 to 25% identity and 39 to 40% similarity) proteins of the Enterobacteriaceae, respectively. This analysis was in agreement with the presence of a α-Hep-(1→2)-d,d-Hep disaccharide in the outer core of the LPS from S. marcescens N28b (Fig. 4). An S. marcescens nonpolar orf7 mutant, strain N22b30, was constructed (see Materials and Methods). LPSs from S. marcescens N28b30 and from the mutant complemented by the orf7 gene were isolated, and their chemical compositions were determined. The results showed a drastic reduction in the content of d,d-Hepp in the core region of the mutant LPS, suggesting that the orf7-encoded protein was involved in the transfer of the d,d-Hep residue. It may also be hypothesized that the orf11-encoded protein will be involved in the transfer of the terminal outer core Hep residue.

In the core region of the LPS from S. marcescens N28b, a β-Glc-(1→6)-α-Glc disaccharide attached to O-4 of GlcN was identified (Fig. 4). In addition, the deduced 326- and 388-amino-acid proteins encoded by orf6 and orf4, respectively, were similar to several putative glycosyltransferases of different bacteria involved in the biosynthesis of capsule, O antigen, or core region. Thus, it can be hypothesized that orf6- and orf4-encoded products are involved in the transfer of the two outer core Glc residues.

Finally, the deduced 388-amino-acid protein encoded by orf5 showed significant levels of amino acid similarity (39 to 40%) and identity (22 to 23%) to WaaL proteins from E. coli core types K-12, R1, R2, and R4. In addition, TopPred2 analysis of the orf5-encoded protein predicted 10 membrane-spanning domains, suggesting an integral membrane location. The distribution of these putative transmembrane domains along the protein sequence and the protein hydropathy profile were found to be very similar to those of WaaL proteins, suggesting that orf5 corresponds to the S. marcescens N28b waaL gene.

DISCUSSION

In this work, the waa gene cluster and the structure of the core region of the LPS from S. marcescens strain N28b were investigated. The proposed structure for the major core oligosaccharide (oligosaccharide 3) of S. marcescens N28b differs from that of K. pneumoniae by two main features: the substitution of the α-d-GlcNp at the O-4 position by the β-Glc-(1→6)-α-Glc disaccharide and the substitution of the α-d-GalpA at the O-2 position by the α-Hepp-(1→2)-α-d,d-Hepp disaccharide (Fig. 4). Although no residue of Ko could be identified in the main isolated compound (oligosaccharide 3) obtained from LPS of an O-antigen-deficient mutant, small amounts of this residue were detected in the mixture of core oligosaccharides isolated from the LPS. Furthermore, ESI FT-ICR-MS analysis of similar core oligosaccharide fractions obtained from the N28b wild-type strain, an O-antigen-deficient mutant, and a double orf9 orf10 mutant strongly suggested that Ko is present in some of the strain N28b core oligosaccharides. The presence of Ko residues, linked in nonstoichiometric amounts to O-4 of the Kdo residue, has been described in the core regions of the LPS from S. marcescens strains 111R (serotype O:29) and 3735 (E. V. Vinogradov, B. Lindner, G. Seltmann, J. Radziejewska-Lebrecht, and O. Holst, XXIst Int. Carbohydr. Symp., p. 279, 2002). Ko residues have not been reported in the core region of LPS from K. pneumoniae.

A comparison of the known waa gene clusters from members of the Enterobacteriaceae revealed similarities as well as differences in their organization (Fig. 1). In all known cases, the genes gmhD and coaD are located at the 5′ and 3′ ends, respectively Four genes involved in epimerization (gmhD) and transfer of l,d-Hepp I (waaC), l,d-Hepp II (waaF), and l,d-Hepp III (waaQ) and a fifth gene (waaA) coding for the transfer of the Kdo residue were identified in S. marcescens N28b. The presence of these genes correlated with the structure l-α-d-HeppIII-(1→7)-l-α-d-HeppII-(1→3)-l-α-d-HeppI-(1→5)-Kdop, found in the LPS inner core region of the Enterobacteriaceae studied (23, 24) (Fig. 4). In agreement with the elucidated S. marcescens N28b major core structure, no genes similar to waaP and waaY involved in phosphoryl modification of l,d-HeppI and l,d-HepII, respectively, were found in the S. marcescens N28b waa gene cluster. Instead, genes that could be involved in the transfer of the inner core d-Glc (waaE) and outer core d-GalA and d-GlcN residues (orf9_Sm orf10_Sm and orf8_Kp orf9_Kp) were identified in both S. marcescens and K. pneumoniae. The presence of orf9_Sm orf10_Sm homologues may be expected in Enterobacteriaceae species containing the α-GlcN (1→4)-α-GalA disaccharide. Consistent with this, the core region of the LPS from Proteus mirabilis serotype O3 contains this structural feature (50), and it also contains orf9_Sm orf10_Sm homologues (M. Regué et al., unpublished results). Two additional putative heptosyltransferases were found in the S. marcescens waa gene cluster (i.e., orf7 and orf11). The reduced content of d,d-HepII in the LPS of an S. marcescens orf7 mutant strongly suggested that the orf7-encoded protein was involved in the transfer of one of the branching Hepp residues, probably that possessing the d,d configuration. The remaining two genes (orf4 and orf6) from the S. marcescens waa gene cluster encoded proteins that possessed characteristic features of glycosyltranferases. Since the core region of the LPS of strain N28b contained a β-Glc-(1→6)-α-Glc disaccharide attached to O-4 of the d-GlcN residue, it is tempting to speculate that these two genes are involved in the sequential transfer of the outer core Glc residues.

The JUMPStart sequence (for just upstream of many polysaccharide-associated starts) (22, 27) was found to be located upstream of the waaQ operon in E. coli, containing 9 genes, and that in S. enterica, containing 10 genes (19). No such sequence was found in the 126-bp intergenic region between orf11 and the waaA operon, as expected from the monocistronic nature of orf11. Furthermore, no JUMPStart similar sequences were found upstream of the three operons of the S. marcescens waa gene cluster, similarly to what has been found in the K. pneumoniae waa gene cluster (38). The significance of this feature will require further studies.

Initial analysis using different strains from S. marcescens belonging to several serovars seemed to indicate that the genetic organization of the waa gene cluster reported in this work was shared by all of them. If further studies confirm this preliminary result, this means that a main core type is present in the LPS of S. marcescens, as was described for LPS from K. pneumoniae (38, 45, 49).