Functional Identification of Proteus mirabilis eptC Gene Encoding a Core Lipopolysaccharide Phosphoethanolamine Transferase

By comparison of the Proteus mirabilis HI4320 genome with known lipopolysaccharide (LPS) phosphoethanolamine transferases, three putative candidates (PMI3040, PMI3576, and PMI3104) were identified. One of them, eptC (PMI3104) was able to modify the LPS of two defined non-polar core LPS mutants of Klebsiella pneumoniae that we use as surrogate substrates. Mass spectrometry and nuclear magnetic resonance showed that eptC directs the incorporation of phosphoethanolamine to the O-6 of l-glycero-d-mano-heptose II. The eptC gene is found in all the P. mirabilis strains analyzed in this study. Putative eptC homologues were found for only two additional genera of the Enterobacteriaceae family, Photobacterium and Providencia. The data obtained in this work supports the role of the eptC (PMI3104) product in the transfer of PEtN to the O-6 of l,d-HepII in P. mirabilis strains.


Introduction
Gram-negative motile and frequently swarming bacteria of the genus Proteus from the family Enterobacteriaceae are usually found in soil, water, and the intestines of human and animals. The genus includes five species (P. mirabilis, P. penneri, P. vulgaris, P. myxofaciens, and P. hauseri) and three genomospecies [1]. Among these species, P. mirabilis, P. vulgaris, and P. penneri are the most common opportunistic pathogens [2]. P. mirabilis is frequently associated with urinary tract infections (UTI) in individuals with functional or anatomical abnormalities of the urinary tract or long-term catheterized patients. Complications arising from P. mirabilis infections include bladder and kidney stone formation, catheter obstruction by encrusting biofilms, and bacteremia [3]. Identified virulence factors include swarming, fimbriae, urease, hemolysin, and iron acquisition systems. Signature-tagged mutagenesis [4] has allowed the additional identification of an extracellular metalloprotease, several putative DNA binding regulatory, cell-envelope related, and plasmid encoded proteins.
In the lipopolysaccharide (LPS) of P. mirabilis, as in other members of the family Enterobacteriaceae, three domains can be recognized: lipid A, an endotoxic glycolipid; an O-polysaccharide chain or O-antigen (O-PS); and a core oligosaccharide (OS) domain, linking lipid A and O-PS. The chemical structure of the P. mirabilis lipid A has been determined for one strain showing the presence of a residue of 4-amino-4-deoxy-L-arabinose (L-Ara4N) substituting the phosphate at position 1 of lipid A, this L-AraN modification has been related to the high intrinsic P. mirabilis resistance to polymyxin B and related cationic antimicrobial peptides.
Among P. mirabilis and P. vulgaris, 60 O-serogroups have been recognized and the gene clusters involved in the biosynthesis of P. mirabilis serogroups O3, O10, O23, O27, and O47 have been reported [5]. In most P. mirabilis O-serogroups studied so far, the O-PSs contain acidic or both acidic and basic components, such as uronic acids, and their amides with amino acids, including lysine, and phosphate groups (reviewed in [6]).
The core OS structure of the P. mirabilis genome strain HI4320 has been recently reported [7] (Figure 1). This structure up to the first outer-core galacturonic acid residue (D-GalA I) is shared by 11 P. mirabilis strains and also by several P. vulgaris and P. penneri serogroups. This common fragment is also found in the core LPS of Klebsiella pneumoniae and Serratia marcescens, but without the Ara4N and phosphoethanolamine (PEtN) residues [8,9] (Figure 1). Some P. mirabilis core-OS structures contain unusual residues such as quinovosamine, an open-chain form of N-acetyl-galatosamine, or amide linked amino acids. Recently, most of the genes involved in the biosynthesis of the sugar backbone of the core LPS from several P. mirabilis strains have been identified and characterized [10,11]. Little is known about the role of the non-sugar charged residues or groups in the core OS. In this study we identify a gene, eptC, essential for core LPS modification with PEtN.
PMI3104 showed high levels of identity (75% to 77%) along the entire protein with putative homologues from Photorhabdus asymbiotica, Pho. luminescens, Providencia rettgeri, Prov. alcalifaciens, and Prov. rustigianii (Figures 2 and 3), but not to other members of the Enterobacteriaceae family. The three genera containing PMI3104 or its homologues are closely related at the phylogenetic level.

Identification of a Core LPS PEtN Transferase
Since the core LPS isolated from P. mirabilis strains present a moiety of PEtN substituting the O-6 position of L,D-Hep II (HepII6-PEtN) [6], it was hypothesized that PMI3104 would be the enzyme involved in this PEtN transfer and hence named eptC, for ethanolamine phosphate transferase C. On the basis of carbohydrate backbone composition and structure identity between the core LPS of K. pneumoniae and P. mirabilis up to the first residue of the outer core [9,10] (Figure 1), we decided to use the LPS from non-polar mutants K. pneumoniae 52145∆wabH and K. pneumoniae 52145∆wabG [9,16] as surrogate LPS acceptors to test eptC activity. This new approach has the advantage that we look for a positive trait as the addition to the mutant LPS of a particular residue, instead of the mutagenesis studies to determine the function of core LPS biosynthetic genes usually employed. This is only possible, as in our case, if a proper surrogate LPS acceptor is available.
To guarantee the inducible expression of the eptC gene, plasmid pBAD18-Cm-eptC was constructed and electroporated into the two K. pneumoniae mutants 52145∆wabH [9] and K. pneumoniae 52145∆wabG [16]. LPS from K. pneumoniae 52145∆wabG (+pBAD18-Cm-eptC grown under arabinose promoter inducing conditions) migration was slower than that of LPS from K. pneumoniae 52145∆wabG grown under the same conditions ( Figure 4, lanes 2 and 3). Similar results were obtained when comparing LPS from K. pneumoniae 52145∆wabH (+pBAD18-Cm-eptC) and K. pneumoniae 52145∆wabH always grown under arabinose promoter inducing conditions (  The spectrometry (ESI-MS) spectrum of K. pneumoniae 52145∆wabG core oligosaccharide showed a pseudo-molecular ion (M-H)-at m/z 783.37 which was in agreement with the calculated average molecular weight (783.67) of the expected molecular structure, with one hexose, two heptoses and one Kdo unit as previously reported [16]. The identical result was obtained when the core oligosaccharide of K. pneumoniae 52145∆wabG (+pBAD18-Cm-eptC) was obtained from cells grown under non-induced conditions. To unambiguously test the eptC requirement for PEtN transfer, LPS from cells of 52145∆wabG (+pBAD18-Cm-eptC) grown under arabinose promoter inducing conditions was isolated from dried cells by phenol-water extraction and purified (see Experimental Section). Mild acid degradation of this LPS resulted in a core fraction purified by gel-permeation chromatography on Sephadex G-50. High-resolution negative ion electrospray ionization mass spectrum of this fraction showed a major [M-H] − ion peak at m/z 906.2499 for a compound corresponding to an OS composed of two heptose residues and one residue each of hexose, Kdo and PEtN (Hex 1 Hep 2 Kdo 1 P 1 EtN 1 ) (calculated ion mass 906.2497 amu). In addition, a number of lower-molecular mass ions were observed in the spectrum at m/z 888.2416, 862.2590, and 818.2335 which were assigned to various Kdo artifacts generated by loss of H 2 O, CO 2 , and C 3 H 4 O 3 , respectively ( Figure 5). No non-phosphorylated OS was detected.   The 1 H-NMR spectrum of the OS was assigned using two-dimensional COSY and TOCSY experiments, and then the 13 C-NMR spectrum was assigned using a two-dimensional 1 H-, 13 C-HSQC experiment. The assigned 1 H-and 13 C-NMR chemical shifts (Table 1) were in full agreement with the carbohydrate backbone composition and structure expected for the core OS derived from K. pneumoniae wabG deletion mutant [16]. The signals for H-6 and C-6 of the terminal heptose residue (L,D-Hep II) were shifted downfield to  4.57 and 74.6, as compared with their positions in the non-substituted heptose at  4.04 and 70.4, respectively.   13 C-NMR chemical shifts (, ppm) of the modified core OS from K. pneumoniae 52145∆wabG plus pBAD18-Cm-eptC. Assignment was performed using a two-dimensional 1 H, 13 C-HSQC experiment. n.f., not found. The 31 P NMR signal is at  0.5.

C-8; H-8 (H-8a,8b)
These data defined the site of phosphorylation at position O-6 of L,D-Hep II and, hence, the OS has the following structure:

eptC Gene Distribution
Since the presence of a PEtN moiety linked to the O-6 of the L,D-Hep is a common feature of the core LPS of the studied P. mirabilis strains [17] the eptC gene should be present in these strains. To confirm the above hypothesis, a collection of P. mirabilis strains obtained from Z. Sydorckzyk (National Microbiology Institute, Warsaw, Poland) were used. Genomic DNA from these strains was used as template for amplification with oligonucleotides F1eptC and R1eptC. Fragments of about 2000 bp were amplified for all the strains studied, as exemplified by the results for the four strains shown in Figure 7. Sequencing of the amplified fragments confirmed the presence of the eptC gene. In addition, in all the amplified fragments the presence of a sequence coding for the C-terminal amino acid residues of celY (PMI3103) gene, encoding a putative cellulase, were found adjacent to eptC, suggesting that in the analyzed strains the eptC location in their genome is the same as that found in genome strain HI4320 (Figure 1). In agreement with the EptC function as the transferase responsible for the linkage of a PEtN moiety to the O-6 position of L,D-Hep II, the eptC gene is found in all the P. mirabilis strains analyzed in this study. Within the Enterobacteriaceae, eptC homologues appear to be limited to species of the Proteus phylogenetic related genus Photobacterium and Providencia (Figure 2), and recently the presence of a PEtN moiety linked to the O-6 position of L,D-Hep II was shown in Prov. alcalifaciens O8 and O35, and Prov. stuartii O49 [18]. Further work will be necessary to understand the reasons for the presence of this common feature in these three closely related genera.

General DNA Methods
General DNA manipulations were done essentially as previously described [21]. DNA restriction endonucleases, T4 DNA ligase, E. coli DNA polymerase (Klenow fragment), and alkaline phosphatase were used as recommended by the Sigma-Aldrich (St. Louis, MO, USA).

DNA Sequencing and Computer Analysis of Sequence Data
Double-stranded DNA sequencing was performed by using the dideoxy-chain termination method [22] from PCR amplified DNA fragments with the ABI Prism dye terminator cycle sequencing kit (PerkinElmer, Barcelona, Spain). Oligonucleotides used for genomic DNA amplifications and DNA sequencing were purchased from Sigma-Aldrich (St. Louis, MO, USA). 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 [23] network service at the National Center for Biotechnology Information and the European Biotechnology Information. ClustalW was used for multiple-sequence alignments [24].
Recombinant plasmid pBAD18-Cm-eptC was obtained by pGEMT-eptC double digestion with SalI and PvuII and subcloning into pBAD18-Cm doubly digested with SalI and SmaI. This construct was transformed into K. pneumoniae core LPS mutants and eptC was expressed from the arabinose-inducible and glucose-repressible pBAD18-Cm promoter [20]. Repression from the araC promoter was achieved by growth in medium containing 0.2% (w/v) D-glucose, and induction was obtained by adding L-arabinose to a final concentration of 0.2% (w/v). The cultures were grown for 18 h at 37 °C in LB medium supplemented with chloramphenicol and 0.2% glucose, diluted 1:100 in fresh medium (without glucose), and grown until they reached an A 600 of about 0.2. Then, L-arabinose was added, and the cultures were grown for another 8 h. Repressed controls were maintained in glucose-containing medium.

Large-Scale Isolation and Mild-Acid Degradation of LPS
Dried bacterial cells in 25 mM Tris-HCl buffer containing 2 mM CaCl 2 , pH 7.63 (10 mL·g −1 ), were treated at 37 °C with RNase and DNase (24 h, 1 mg·g −1 each), and then with proteinase K (36 h, 1 mg·g −1 ). The suspension was dialyzed and lyophilized, and the LPS was extracted with aqueous phenol [26]. After dialysis of combined water and phenol layers, contaminants were precipitated by adding 50% aqueous CCl 3 CO 2 H at 4 °C, the supernatant was dialyzed and lyophilized to give the LPS. A portion of the LPS was degraded with 2% acetic acid for 2 h at 100 °C, a precipitate was removed by centrifugation (13,000× g, 20 min), and the supernatant was fractionated on a column (50 × 2.5 cm) of Sephadex G-50 Superfine in 0.05 M pyridinium acetate buffer (4 mL pyridine and 10 mL acetic acid in 1 L of water) with monitoring using a differential refractometer (Knauer, Berlin, Germany).

Mass Spectrometry and NMR Studies
High-resolution electrospray ionization mass spectrometry was performed in the negative ion mode using a microTOF II instrument (Bruker Daltonics, Billerica, MA, USA). A sample of the OS (~50 ng·µL −1 ) was dissolved in a 1:1 (v/v) water-acetonitrile mixture and sprayed at a flow rate of 3 µL·min −1 . Capillary entrance voltage was set to 4.5 kV and exit voltage to −150 V; drying gas temperature was 180 °C.

Conclusions
A phylogenetic tree constructed from an alignment between PMI3040, PMI3576, PMI3104 (eptC), and representative known and putative PEtN transferases allowed us the identification of putative PEtN transferases in a proteomic search of the genome strain. The chemical data obtained in this work ( Figures 5 and 6, and Table 1 The other two putative PEtN transferases PMI3576 and PMI3040 need further investigation. PMI3576 appears phylogenetically related to EptB and to a lesser degree to EptA of several Enterobacteriaceae species (Figure 2). EptA, also known as PmrC, transfers a PEtN moiety to the phosphate at the 1 and/or 4' positions of lipid A [14]. No PEtN substituting lipid A was found in P. mirabilis grown in standard conditions. EptB transfers PEtN to the Kdo II residue of the inner core in E. coli [13], but no such Kdo II modification has been described in P. mirabilis core LPS [6,16]. The presence of a moiety of PEtN linked to Kdo II in minor P. mirabilis core molecules or when these strains are grown in non-standard conditions cannot be ruled out. The above considerations suggest that PMI3576 could be a PEtN transferase more likely involved in Kdo than in lipid A modification. At this time no function can be hypothesized for PMI3040.