The Neisseria meningitidis outer membrane lipoprotein FrpD binds the RTX protein FrpC.

At conditions of low iron availability, Neisseria meningitidis produces a family of FrpC-like, type I-secreted RTX proteins of unknown role in meningococcal lifestyle. It is shown here that iron starvation also induces production of FrpD, the other protein expressed from a gene located immediately upstream of the frpC gene in a predicted iron-regulated frpDC operon. We found that FrpD is highly conserved in a set of meningococcal strains representative of all serogroups and does not exhibit any similarity to known sequences of other organisms. Subcellular localization and [3H]palmitic acid labeling in Escherichia coli revealed that FrpD is synthesized with a type II signal peptide for export across the cytoplasmic membrane and is, upon processing to a lipoprotein, sorted to the outer bacterial membrane. Furthermore, the biological function of FrpD appears to be linked to that of the RTX protein FrpC, because FrpD was found to bind the amino-proximal portion of FrpC (first 300 residues) with very high affinity (apparent Kd approximately 0.2 nM). These results suggest that FrpD represents an rtx loci-encoded accessory lipoprotein that could be involved in anchoring of the secreted RTX protein to the outer bacterial membrane.

Under conditions of limited iron availability, N. meningitidis was shown to produce RTX 1 (repeat in toxin) family proteins secreted through the type I pathway, the so-called FrpC-like proteins (6 -10). These are characterized by the presence of a variable number of carboxyl-proximal glycine and aspartaterich repetitions of a nonapeptide RTX consensus motif (L/I/ F)XGGXG(D/N)DX (11). While the biological activity of meningococcal FrpC-like proteins remains unknown, a number of other RTX proteins was already shown to act as exotoxins and to play an important role in virulence of Gram-negative pathogens of the genera Escherichia (HlyA), Actinobacillus (LtxA, Apx proteins), Bordetella (CyaA), Vibrio (RtxA), or Mannheimia (LktA) (12-17). While HlyA, CyaA, LtxA, and some Apx toxins exhibit pore-forming activity capable of perturbing bactericidal activities of leukocytes by membrane permeablization and colloid-osmotic cell lysis (at high concentrations), the Bordetella CyaA also harbors an adenylate cyclase enzyme activity penetrating myeloid phagocytic cells, paralyzing them by cAMP intoxication, and Vibrio cholerae RtxA causes actin cytoskeleton perturbation and cell morphology changes by catalyzing cross-linking of actin subunits (12-17). Furthermore, RTX proteins were found to exhibit lipase, protease, bacteriocin, or nodulation-inducing activities or form S-layers on bacterial cell surface (18 -22).
Genetic organization of most of the characterized rtx determinants is similar to that of the locus accounting for production of the E. coli ␣-hemolysin (HlyA). This consists of an operon of four structural genes (hlyCABD) encoding the RTX protein itself (HlyA), an acyltransferase activating the toxin by posttranslational fatty acylation (HlyC), and two inner membrane components of the type I secretory apparatus (HlyB and HlyD), required for extracellular secretion of the toxin (23)(24)(25)(26)(27)(28). Some rtx loci also contain a gene for the outer membrane TolC-like channel tunnel component of the type I secretion machinery (29), whereas the loci encoding the RTX proteases contain a gene for an inhibitor of the respective protease, in place of the acyltransferase gene (30 -32). In N. meningitidis, however, neither the gene for the acyltransferase, nor the genes for components of the secretion machinery are present within the frpC locus. Instead, the frpC gene is preceded by an open reading frame, orf1 (GenBank TM acc. no. L06299) encoding a hypothetical protein of unknown function (7) that does not exhibit similarity to any other known protein sequence. Moreover, the 1 The abbreviations used are: RTX, repeat in toxin; BCIP, 5-bromo-4-chloro-3-indolylphosphate; CBD, chitin binding domain; Frp, Fe-regulated protein; IPTG, isopropyl-␤-D-thiogalactopyranoside; OMP, outer membrane protein; DTT, dithiothreitol; CHAPS, 3-[(3-cholamidopropyl)dimethylammonio]-1-propanesulfonic acid; PBS, phosphate-buffered saline; MALDI-TOF, matrix-assisted laser desorption/ionization time-of-flight; MS, mass spectrometry; BSA, bovine serum albumin. genetic organization suggests that the orf1 and frpC genes form an operon transcribed from an iron-regulated promoter located ϳ300-bp upstream of orf1, where a predicted binding site for the ferric uptake regulator protein Fur could be identified (7).
Here we show that Orf1 is an outer membrane lipoprotein that binds the N-terminal portion of FrpC with high affinity and represents, thereby, a new class of accessory proteins of the RTX determinants. We propose to call it FrpD.
pTYB2frpD 271 -frpC and pTYB2frpD 250 -frpC, for production of FrpD 271 or FrpD 250 together with the FrpC-intein-CBD fusion protein, were obtained by inserting the frpD 271 and frpD 250 reading frames as XbaI fragments into pTYB2frpC (8). The pTYB2frpC⌬X-Y plasmids, where X and Y stand for the numbers of the first and of the last amino acid residues of the FrpC segment deleted in the given construct, were prepared by PCR mutagenesis and/or by using the naturally occurring restriction sites in frpC. Construction of all plasmids and the absence of undesired mutations in PCR-amplified DNA were systematically verified by DNA sequencing, and plasmid maps with complete sequences will be provided upon request.
Cloning and Sequencing of frpD Alleles-1186-bp long PCR products comprising the frpD alleles were amplified with the orf1-frpC primer pair (8) from chromosomal DNA of nine representative meningococcal isolates and cloned. Equimolar pools of plasmids from five independent frpD clones for each meningococcal isolate were used in sequencing reactions to avoid potential substitutions introduced during the PCR reaction.
Cultures of Meningococci-The N. meningitidis isolate 10/96 (8) was grown on GCB agar plates (Difco) at 37°C in a humidified 5% CO 2 atmosphere for 10 h. Bacteria were washed and resuspended with RPMI medium to an OD 600 of 0.2, allowed to adapt for 2 h at 37°C, and seeded into a static 60-ml culture in T75 flasks at an OD 600 of 0.02. Upon growth in RPMI (without fetal calf serum) for 10 h until OD 600 reached 0.2-0.3 (maximal viability), 100 M desferrioxamine B was added to chelate free ferric ions where applicable, or 7 M Fe(NO 3 ) 3 was added as iron source for growth under iron-replete conditions, respectively.
Two-dimensional Electrophoresis-Bacterial pellets (2.5 mg of protein) were resuspended in 1 ml of 50 mM Tris-HCl, pH 8.0, 200 mM DTT, 0.3% SDS buffer with Complete TM proteinase inhibitor (Roche Applied Science). Proteins were precipitated overnight with 4 volumes of cold acetone (Ϫ20°C) containing 0.2% DTT and solubilized in 2 ml of 8 M urea, 4% CHAPS, 2.25% v/v Pharmalyte 3-10, and 65 mM DTT. Protein sample (0.5 mg) was soaked into 18-cm immobilized pH 4 -7 gradient (IPG) strips (Amersham Biosciences), and the proteins were separated by isoelectric focusing (IEF) at 3500 V, using a gradual voltage increase from 150 to 3500 V over the first 9 h, followed by focusing at 3500 V for 31 h. IPG strips were equilibrated in SDS-PAGE loading buffer and placed on the top of 22 ϫ 22 ϫ 0.1 cm polyacrylamide (12.5%) slab gels for protein separation by SDS-PAGE for 6 h at 14 watts per gel prior to staining by colloidal Coomassie Blue G-250. Image analysis of sets of stained separation gels (n ϭ 6) was used to identify iron-regulated proteins.
MALDI-TOF MS Analysis-Protein spots were excised from stained SDS-PAGE gels and identified by MALDI-TOF mass spectrometry of tryptic digests as previously described (9).
Purification of FrpD 250 -6xHis-Cultures of E. coli BL21(DE3) carrying pET28frpD 250 were grown to an OD 600 of 1.2 at 37°C in 500 ml of LB medium with 100 g/ml of kanamycin and synthesis of FrpD 250 -6xHis was induced by IPTG (1 mM) for 3 h. Cells were resuspended in 50 mM Tris-HCl, pH 7.4, 150 mM NaCl buffer (TN buffer), and disrupted by sonication. The cell extract was cleared at 15,000 ϫ g for 20 min and loaded on a Ni-nitrilotriacetic acid-agarose column (Qiagen). Upon washing with 10 bed volumes of TN, the FrpD 250 -6xHis protein was eluted with 250 mM imidazole in TN buffer and stored frozen.
Rabbit Sera against FrpD-Slices containing purified FrpD 250 -6xHis protein were excised from 10% SDS-PAGE gels, grinded, mixed with 1 mg/ml of aluminum hydroxide in PBS, and used to immunize a rabbit by three doses of ϳ0.4 mg of FrpD 250 -6xHis.
PhoAЈ Fusions-The phoAЈ open reading frame lacking a signal sequence was excised as an XbaI fragment from the pSWFII plasmid (35) and inserted in-frame with the first 93 codons of frpD 271 , or with the first 72 codons of frpD 250, respectively, using the SmaI site in frpD on the pT7-7frpD 271 and pT7-7frpD 250 plasmids. Sequence-verified frpD-phoAЈ fusions were then placed under the control of the lacZp promoter on pTZ19R (Amersham Biosciences). Expression of alkaline phosphatase activity was assessed on LB agar plates containing 40 g/ml of 5-bromo-4-chloro-3-indolylphosphate as PhoA substrate (36). Production of the FrpD-PhoAЈ fusion proteins was assessed in whole cell extracts by immunoblotting with monoclonal anti-PhoA antibody (Sigma).
Membrane Flotation in Sucrose Density Gradients-E. coli BL21 (Novagen) carrying the frpD 271 -6xHis fusion allele on a reduced copy number vector pSU39 (34) were grown in 100 ml of M9 medium (37) with 100 g/ml of kanamycin and 0.4% maltose to midexponential phase and FrpD protein production was induced by 1 mM IPTG for 3 h. Cells were resuspended in 1 ml of 100 mM Tris-Cl, pH 8.0, 10 mM EDTA, 0.5 M sucrose, incubated for 10 min at 4°C with 10 mg/ml of lysozyme, lysed hypotonically by addition of 10 volumes of ice-cold water, and passed three times through a French pressure cell at 1500 psi. 10 l of Benzonase TM (Merck) was added to digest nucleic acids, and unbroken bacteria were removed by centrifugation at 4000 rpm for 10 min. The lysate was overlaid on a sucrose cushion (0.4 ml of 30% and 0.3 ml of 65% (w/w) sucrose) and centrifuged for 2 h at 40,000 rpm in a Sorvall TH-641 rotor at 4°C. Membrane suspension was recovered at the interphase of 30 and 65% sucrose solutions, supplemented with solid sucrose to 65% (w/w), and placed below a 5-ml sucrose step gradient (62-30% w/w) in 50 mM HEPES-Na, pH 7.5. Membranes were floated to equilibrium at 59,000 rpm for 40 h at 10°C in a Sorvall TST 60.4 rotor, and 200-l gradient fractions were analyzed for the presence of the FrpD-6xHis and LamB proteins by immunoblotting with anti-polyhistidine monoclonal antibody (Sigma) and rabbit polyclonal serum against LamB (gift of A. Charbit), respectively. NADH oxidase activity was determined as previously described (38).
Purification of the FrpC-derived Proteins-Soluble recombinant FrpC and its mutant variants were purified from cytosolic fractions of producing E. coli BL21(DE3) cells by chitin affinity chromatography using a C-terminal intein chitin binding domain (intein-CBD, New England Biolabs), as previously described (8). DTT was removed from the eluted protein samples by gel filtration on Sephadex G-25 in TN buffer. Purity and integrity of the proteins were verified by SDS-PAGE, and the proteins were stored frozen at Ϫ20°C.
FrpD to FrpC Binding Assay-Wells of PolySorp ELISA plates (Nunc, Roskilde, Denmark) were coated at 4°C overnight with 100 l of purified FrpC at 20 g/ml (or its mutant derivatives) in PBS, pH 7.4. Upon washing with PBS-0.05% Tween-20 (PBST), wells were saturated with bovine serum albumin (1%) in PBST (BSA-PBST) for 1 h at 37°C. Plates were washed three times with PBST and serially diluted FrpD 250 , or FrpD 271 protein in BSA-PBST was added at indicated concentrations and allowed to bind to adsorbed FrpC for 1 h at 37°C. Plate wells were washed three times with 200 l of PBST, and the amounts of formed FrpC-FrpD complex were determined by colorimetric immunodetection in an ELISA format using rabbit polyclonal serum against FrpD (1:500 dilution) and anti-rabbit IgG-peroxidase conjugate (1:2000), with o-phenylenediamine as peroxidase substrate. Absorbance at 492 nm was measured, and the curves represent the best fit of obtained data to a sigmoidal dose-response curve following nonlinear regression analysis of mean Ϯ S.D. values from four independent determinations (n ϭ 4). The apparent K d values were determined as FrpD concentrations giving a half-maximal A 492 value for the fitted FrpD to FrpC binding curve. Because of stringent washing and blocking of plate wells, no nonspecific signal from binding of the primary or secondary antibodies was detected when either one of the binding partners was omitted from the assay or when a mock E. coli cell extract was coated on plate wells instead of FrpC.

RESULTS
FrpD Is Conserved among Clinical Isolates of N. meningitidis-We have previously examined the distribution of frpD alleles in a set of 65 N. meningitidis isolates and shown that the frpD gene was present in all 38 examined invasive isolates as well as in 24 of 27 noninvasive isolates, respectively (8). Conservation of the frpD sequence was, therefore, analyzed for nine alleles from clinical isolates representing all meningococcal serogroups, and the deduced amino acid sequences of these alleles exhibited 95.6 -100.0% identity. Six of nine sequences were identical, as shown in Fig. 1, whereas three had a limited set of amino acid substitutions. In one isolate (175/96), an internal deletion of 14 residues was found, whereas in another isolate (263/96) the decapeptide NRINQTEEDS encompassing residues 210 -219 of the predominant FrpD sequence was replaced by an unrelated octapeptide YHRYGEND (Fig. 1). Altogether these results suggest that the sequence of FrpD is quite conserved across serogroups of meningococci.
FrpD Is Produced under Conditions of Low Iron Availability-As illustrated in Fig. 2A, the frpD and frpC genes can be predicted to form an iron-regulated frpDC operon, transcribed from a promoter containing a binding site for the ferric uptake regulator protein Fur at ϳ300-bp upstream the start of frpD gene (7). To examine whether production of FrpD also depends on iron availability, meningococci were grown in medium containing 100 M desferrioxamine B to chelate free ferric ions, or in medium supplemented with 7 M ferric nitrate as iron source, and the specific presence of iron-regulated proteins was assessed by two-dimensional gel electrophoresis (IEF/SDS-PAGE). Among the protein species present only in extracts from iron-limited cultures, 2 the electrophoretic mobility of one protein matched the predicted FrpD molecular mass of ϳ30 kDa and pI of ϳ5,6 ( Fig. 2B). This protein was unambiguously identified as FrpD by peptide mass fingerprint mapping (64% coverage of the FrpD sequence) as documented in Table I.
FrpD Protein May Be Produced in Several Forms-Inspection of the 271-residue long FrpD sequence (FrpD 271 ), indicated that FrpD could be produced with a somewhat atypical Nterminal signal peptide, possibly promoting FrpD export across the cytoplasmic membrane. Moreover, as illustrated in Fig. 3A, the signal peptide of FrpD 271 could potentially be processed at two different sites, with a cleavage site for type I general signal peptidase predicted between residues Thr 21 and Met 22 and an alternative cleavage site predicted between residues Ser 24 and Cys 25 , possibly recognized by the type II signal peptidase processing bacterial lipoproteins. As, however, already noted by Thompson et al. (7), the frpD gene sequence also contains two potential translation initiation sites located 21 codons apart, which could be giving rise to production of two differentially located forms of FrpD (Fig. 3A). Remarkably, initiation of FrpD synthesis at the second site would yield a cytosolic protein of 250 residues, lacking the signal peptide (FrpD 250 ), but having the same N-terminal sequence as the polypeptide resulting from processing of FrpD 271 at the predicted signal peptidase I cleavage site (Fig. 3A). A conclusive characterization of this possibly complex pattern of expressed FrpD forms was, however, not possible in N. meningitidis, because of the very low FrpD amounts produced ( Fig. 2B and data not shown). The functionality of the FrpD signal peptide and subcellular localization of recombinant FrpD 271 were, therefore, examined upon expression in E. coli.
FrpD Is Exported to the Cell Envelope-To allow straightforward detection and purification of FrpD, a C-terminal His 6 affinity tag was fused to the recombinant FrpD 271 and FrpD 250 proteins. Upon expression of the frpD 271 -6xHis and frpD 250 -6xHis constructs in E. coli under the control of strong T7 phage transcription and translation initiation signals, however, only the truncated FrpD 250 -6xHis protein (ϳ30 kDa), lacking the predicted signal peptide, was found to be overproduced in E. coli (Fig. 3B). The FrpD 271 -6xHis protein with the putative signal peptide was produced at much lower levels, as confirmed by immunodetection (Fig. 3B). Such result could be expected if  2. FrpD production by N. meningitidis is induced by ironstarvation. A, schematic representation of the predicted frpD-frpC operon of N. meningitidis (7). B, proteomic detection of iron-regulated expression of FrpD. N. meningitidis 10/96 (serogroup C) was grown under iron-depleted or iron-replete conditions, and total cell proteins (0.5 mg) were separated by two-dimensional electrophoresis (IEF/SDS-PAGE) as described under "Experimental Procedures" (pI 4 -7; 12,5% SDS-PAGE). The gels were fixed and stained by colloidal Coomassie Blue G-250. The FrpD protein spot (identified by MALDI mass spectrometry) is indicated by an arrow. Only the relevant corresponding sections of the paired two-dimensional gels are shown. the predicted signal peptide of FrpD 271 was functional and directed FrpD 271 for export from cytoplasm, since massive overproduction of exported proteins is difficult to achieve due to deleterious jamming of the protein export apparatus (39 -42). FrpD 271 appeared, indeed, to accumulate mainly as an unprocessed precursor, since only traces of a processed FrpD species differing from the major protein product by the size of a signal peptide (2-3 kDa) were detected (Fig. 3B).
The capacity of the putative FrpD 271 signal peptide to direct protein export in E. coli was, therefore, examined by the phoAЈ reporter approach (35). The 93 N-terminal residue portion of FrpD 271 , or the corresponding first 72 residues of FrpD 250 , were fused to the signal-less enzymatic moiety of alkaline phosphatase (PhoAЈ), as depicted in Fig. 3C, and the two FrpD-PhoAЈ chimeras were tested for its capacity to translocate across the bacterial inner membrane and reach the oxidative extracytoplasmic environment where enzymatically active PhoA dimers can form (35). As shown in Fig. 3D, cells expressing the FrpD 1-93 -PhoAЈ construct produced an active PhoA enzyme and yielded dark blue-green colonies (Fig. 3D) on plates with the chromogenic PhoA substrate 5-bromo-4-chloro-3-indolylphosphate (BCIP). In contrast, transformants producing the FrpD 22-93 -PhoAЈ fusion at even higher levels (data not shown), yielded only white colonies on BCIP plates. Hence, the first 21 residues of FrpD 271 appear to be part of a functional N-terminal signal peptide of FrpD 271 promoting PhoAЈ export across the cytoplasmic membrane of E. coli.
FrpD Is a Lipoprotein-We next examined whether FrpD 271 is processed to a lipoprotein. Production of FrpD 271 or the FrpD 250 proteins was induced in exponentially growing E. coli BL21(DE3) cells, and [9,10(n)-3 H]palmitic acid was added to radioactively label the lipid moieties of newly synthesized bacterial lipoproteins. As shown in Fig. 4, three specifically [ 3 H]labeled protein species were detected in cells producing FrpD 271 , and their size matched closely the predicted molecular masses (ϳ30 kDa) of products and intermediates of processing and lipidation of FrpD 271 . No such labeled proteins were detected in samples of mock cells, or in cells overproducing the FrpD 250 protein lacking the signal peptide (Fig. 4, compare with Fig. 3B). This demonstrated that the [ 3 H]palmitic acid was specifically used for labeling of FrpD 271 by covalent lipidation of FrpD, which was processed as a conventional bacterial lipoprotein (43)(44)(45).
FrpD Is Exported to the Outer Bacterial Membrane-To analyze the membrane localization of FrpD in more detail, it was particularly important to avoid overproduction of FrpD precursors that could be overwhelming the export machinery and cause a biased subcellular localization of FrpD. Therefore, the FrpD 271 -6xHis protein was expressed from the moderate copy number pSU39 vector at low levels, and cellular envelopes containing moderate but detectable FrpD amounts were separated into the inner and outer membrane fractions by equilibrium flotation in sucrose density gradients.
As documented in Fig. 5 by a typical result of the obtained membrane separation, the lightest gradient fractions with sucrose concentrations ranging from 30 to 40% (w/w), containing inner membrane vesicles, and exhibiting the highest activity of the cytoplasmic membrane marker enzyme NADH-oxidase, contained two FrpD forms. Edman degradation performed on the larger form of FrpD, affinity-purified from the inner membrane fractions, yielded an N-terminal sequence of (M)RPYAT corresponding to the FrpD 271 precursor with an unprocessed signal peptide. This FrpD 271 precursor accounted for about 25-30% of the total membrane-associated FrpD, judged from the semiquantitative immunoblot analysis. Most of the membrane-associated FrpD protein (ϳ60%) was, however, detected as a processed protein of lower molecular mass. This was partly present in inner membrane fractions and mostly floated at sucrose concentrations ranging from 44 to 50% (w/w), at which typically the outer membrane fraction of E. coli is recovered. These fractions were, indeed, also enriched for the outer membrane marker protein LamB. While identity of the FrpD protein affinity-purified from these gradient fractions could be unambiguously confirmed by MALDI-TOF MS analysis (not shown), Edman degradation failed to yield an N-terminal sequence for this FrpD protein form, suggesting that its N terminus was, indeed, blocked by lipidation (see above). Altogether, these results show that the 271-residue-long FrpD protein is produced with a functional lipoprotein signal peptide, and its mature form is sorted to the outer membrane.
FrpD Binds with High Affinity to the N-terminal Part of FrpC-In search for the unknown biological activity of FrpC, we expressed it together with FrpD in E. coli cells, because other accessory proteins of certain rtx determinants were shown to activate RTX protoxins by post-translational modification. While no activation of FrpC by FrpD could be detected (data not shown), it was observed that both FrpD 271 and FrpD 250 co-purified with FrpC under the rather stringent conditions of chitin affinity chromatography of the FrpC chitin binding domain fusion protein (Fig. 6). This suggested a strong interaction between the two proteins, and the purified FrpD and FrpC proteins could, indeed, be cross-linked with an apparent 2:1 stoichiometry by the short-arm bifunctional crosslinking reagent DSP, which can cross-link only proteins forming stable complexes (data not shown). Therefore, a solid phase binding assay was designed to further characterize the interaction of FrpD with FrpC.
Affinity-purified FrpC (8) or FrpD 250 -6xHis proteins, respectively, were allowed to bind the other Frp ligand coated on microplate wells over a range of concentrations, and the levels of FrpD-FrpC complexes formed were determined by immunodetection. As documented in Fig. 7, FrpD bound to adsorbed FrpC with a high apparent affinity, in a highly concentrationdependent and saturable manner, and the same was also true in a reverse assay setup for binding of FrpC to coated FrpD. The apparent K d dissociation constant of the FrpD-FrpC complex, as derived following non-linear regression fitting of the obtained data, was found to be in the subnanomolar range between 0.16 and 0.24 nM (Fig. 7). Furthermore, no significant a Meningoccocal proteins produced under iron-limited growth conditions were separated by two-dimensional electrophoresis, the stained protein spot matching the predicted pI and M r values of FrpD was excised from SDS-PAGE gels, subjected to in-gel tryptic digestion, and the protonated masses ͓MϩH͔ ϩ of the extracted peptides were determined by MALDI-TOF MS analysis.
b Masses computed for FrpD peptide fragments expected to result from trypsin digestion of N. meningitidis FrpD (Swissprot. acc. no. S35026). effect of the lipidation of FrpD, or of the presence of the unprocessed signal peptide on the affinity of the protein toward FrpC, was observed (data not shown).
To localize the portion of the 1829-residue long multidomain FrpC protein that was involved in the binding interaction with FrpD, a set of intein-CBD-tagged deletion mutants of FrpC (Fig. 8) was constructed, purified by chitin affinity chromatography, and examined for FrpD binding. As summarized in Table II, besides the protein lacking the entire non-repetitive N-terminal portion of FrpC (FrpC⌬1-862), two of the deletion mutants, the FrpC⌬1-199 and FrpC⌬108 -300 constructs, har-boring deletions within the first 300 amino acid residues of FrpC, failed to bind FrpD. The construct with deletion of residues 200 -397 also exhibited an importantly reduced affinity for FrpD (K d ϳ200 nM). In contrast, the proteins with deletions of adjacent portions beyond residue 400 (⌬400 -448 and ⌬451-861) were much less affected in binding to FrpD (K d ϳ 4 nM and K d ϳ 10 nM, respectively), suggesting that the FrpD binding site was localized within the first 300 residues of FrpC.
FrpD Binding Does Not Affect the Calcium-induced Autocatalytic Processing of FrpC-We have recently shown that in the presence of calcium ions the FrpC protein undergoes an auto- FIG. 3. FrpD has a functional N-terminal secretion signal. A, location of the two possible ribosome binding sites (RBS) and translation initiation sites found in frpD that might allow production of two different protein products of FrpD. The 271-amino acid residue long FrpD 271 precursor would be produced with a signal peptide containing predicted cleavage sites for both type I and type II signal peptidases (indicated by arrows). The putative 250-residue long cytosolic protein FrpD 250 would be devoid of a signal peptide. B, expression of FrpD in E. coli. Recombinant FrpD 271 and FrpD 250 proteins were expressed in E. coli BL21(DE3), crude cell extracts were separated by SDS-PAGE (10%) and stained with Coomassie Blue, or transferred to a polyvinylidene difluoride membrane and immunodetected using an antibody recognizing the C-terminal His 6 tag of both the FrpD 271 and FrpD 250 proteins, respectively. An extract of mock E. coli BL21(DE3) was used as a negative control. C, schematic representation of the FrpD-PhoAЈ fusions generated by fusing the first 93-amino acid residues of FrpD 271 , or the first 72 residues of FrpD 250 to a signal-less PhoAЈ moiety , respectively. D, phenotypes of E. coli XL1-Blue cells expressing the FrpD 1-93 -PhoAЈ or FrpD 22-93 -PhoAЈ fusions, the FrpD 271 protein alone, and the PhoAЈ moiety alone, respectively. Bacteria were grown on Luria-Bertani agar plates containing the chromogenic PhoA substrate BCIP. Alkaline phosphatase activity was detected as a blue-green (dark) phenotype of colonies. catalytic processing between residues Asp 414 and Pro 415 , which is followed by formation of high molecular weight (HMW) FrpC forms that are covalently cross-linked by an isopeptide bond forming between the C-terminal carboxyl group of the liberated Asp 414 and an ⑀-amino group of a nearby lysine residue (9). It was, hence, important to examine whether FrpD binding would affect the autoprocessing activity of FrpC. As, however, documented in Fig. 9, no effect of the presence of FrpD on calciuminduced processing of FrpC was observed, either when the processed and lipidated FrpD 271 , or the FrpD 251 proteins were incubated with FrpC at 2 mM concentrations of calcium ions for 1 h (Fig. 9) or when the kinetics of FrpC autoprocessing in the presence of FrpD was analyzed (not shown). Binding of FrpD to FrpC and the calcium-dependent autoprocessing of FrpC appear, hence, to be two independent and unrelated activities of FrpC. Interestingly, however, as verified both by immunoblotting and MALDI-TOF mass spectrometric identification (not shown), the processed N-terminal FrpC fragment formed a very stable and possibly also covalently cross-linked complex with FrpD 250 (Fig. 9), which could not be dissociated under the harsh denaturing conditions of SDS-PAGE sample preparation and separation.

DISCUSSION
Many meningococcal proteins expressed at low iron conditions are exported into or across the cell envelope to mediate iron uptake or host colonization (46 -48). We show here that the iron-regulated protein FrpD, lacking homology to any known sequences in public data bases, is a lipoprotein localizing to bacterial outer membrane. FrpD binds with high affinity to the first 300-residue portion of the RTX protein FrpC, which strongly suggests that FrpD is a new type of accessory protein of rtx genetic determinants. The biological role of FrpD in meningococcal infections is, hence, most likely linked to the thus far unknown biological activity of the RTX protein FrpC. No hints of the biological activity of FrpC could, however, thus far be obtained by examination of phenotypes of meningococcal strains deficient in FrpC production in the in vivo infant rat model of meningococcal sepsis (49) or in various in vitro models involving various cultured cell types (data not shown). Despite intense efforts and unlike that of many other RTX proteins, no activity of purified recombinant FrpC on phagocytic cells could as yet be revealed. At present it can, hence, only be speculated The frpC open reading frame was next fused in-frame to a C-terminal intein (int) chitin binding domain (cbd) self-excisable tag, which allowed affinity purification of FrpC on chitin beads as previously described (8). B, copurification of FrpD 250 with FrpC-intein-CBD on chitin beads. FrpC-intein-CBD was purified on chitin beads columns from the cytosolic fraction of E. coli BL21(DE3) carrying pTYB2-frpD 250 -frpC. Purification steps were analyzed using a 7.5-12% gradient SDS-PAGE gel and in parallel to Coomassie Blue staining (Total protein), the purified FrpC and copurified FrpD proteins were immunodetected in the eluted fractions using specific polyclonal antisera (Immunodetection). The loaded fractions were: CL, clarified E. coli lysate; FT, chitin column flow through; W, column wash; E, eluted FrpC fraction.
that FrpD plays, together with FrpC, some role in meningococcal lifestyle in colonization of humans. This would, be suggested by conservation of the FrpD sequence across meningo-coccal isolates, as reported here, and by the presence of two alleles encoding FrpC-like proteins and of up to four copies of genes for the FrpD genes in the as yet sequenced genomes (50,51) of serogroups A and B meningococci (e.g. a single frpD gene copy in Z2491 and four frpD alleles in MC58 genomes, respectively). In contrast to FrpC, which was found to elicit substantial levels of specific antibodies in colonized and diseased subjects, no significant levels of antibodies against FrpD could, however, be detected in sera of healthy meningococcal carriers and/or in the sera of patients that underwent invasive meningococcal disease (data not shown). This indirectly suggests that besides the low level of FrpD expression in meningococci, the poor in vivo immunogenicity of FrpD might also be caused by occlusion of extracellular portions of FrpD for antibody binding by tightly associated FrpC.
Previously, some open reading frames located immediately upstream of the structural genes for RTX cytolysins and leukotoxins were shown to encode accessory protein acyltrans- FIG. 7. FrpD strongly binds FrpC in a concentration-dependent and saturable manner. A, increasing concentrations of purified FrpD 250 protein were added to ELISA plate wells coated with constant amounts of purified recombinant FrpC (20 g/ml), and complex formation was allowed to proceed for 1 h at 37°C. Upon washing out unbound FrpD 250 , the amounts of formed FrpC-FrpD complex were determined by colorimetric immunodetection using rabbit polyclonal serum against FrpD and anti-rabbit IgG-peroxidase conjugate. B, plate wells were coated with constant amounts of FrpD 250 (20 g/ml), and increasing concentrations of FrpC were added to plate wells. Following incubation at 37°C for 1 h and plate washing, the amounts of formed FrpC-FrpD complex were determined by rabbit polyclonal serum against FrpC and anti-rabbit IgG-peroxidase conjugate. The curves represent the best fit of obtained data to a sigmoidal dose-response curve obtained by nonlinear regression analysis of mean Ϯ S.D. values from four independent determinations (n ϭ 4). The apparent K d values were determined as FrpD or FrpC concentrations giving a half-maximal A 492 value for the fitted binding curve.   9. FrpD does not inhibit the calcium-dependent autoprocessing of FrpC. Affinity-purified FrpC was incubated for 1 h at 37°C and 2 mM free Ca 2ϩ ion concentration in the presence (ϩ) or absence (Ϫ) of purified FrpD 271 or FrpD 250 proteins, respectively, at a final FrpC: FrpD molar ratio of 1:2. The protein samples were separated by SDS-PAGE (7.5%) and stained with Coomassie Blue. HMW forms, high molecular weight forms of FrpC; FrpC, the full-length 198-kDa FrpC; FrpC-C ter , C-terminal fragment of FrpC; FrpC-N ter , N-terminal fragment of FrpC resulting from calcium-induced processing of the peptide bond between residues Asp 414 and Pro 415 of FrpC (9); FrpD, FrpD 250 or FrpD 271 , respectively. ferases catalyzing post-translational activation of RTX protoxins by post-translational fatty acylation at ⑀-amino groups of lysine residues at conserved palmitoylation sites (52-55). As mentioned above, however, FrpD does not exhibit any homology to these enzymes, and no activation of FrpC in the presence of FrpD toward any apparent activity on cells could be detected. Neither does FrpD appear to be similar to the protein ORF1 encoded by a gene preceding the structural gene for the RTX protein ApxIVA of Actinobacillus pleuropneumoniae, which does contain a calcium-dependent autoprocessing module similar to that of FrpC (9).
It is intriguing that the frpD gene has the potential to encode two protein variants with two different subcellular localizations, one being a lipoprotein exported to the outer membrane and one possibly remaining in the cytoplasm. We were, however, unable to detect any accumulation of the cytoplasmic FrpD 250 form in E. coli cells expressing the frpD 271 allele at physiologically reasonable levels that would still allow a meaningful analysis of the subcellular localization of the protein. The cytoplasmic FrpD 250 form synthesized from the second possible translational start of frpD could be detected in producing E. coli cells, and its presence was sequence-confirmed only upon expression of the frpD gene from a high copy number plasmid pT7-7frpD 271 under the control of the very strong phage T7 expression signals (not shown). It could, however, not be excluded that under physiological conditions, the cytoplasmic FrpD 250 might still be produced to some extent in meningococci. As shown here the exported lipidated FrpD, as well as its recombinant non-exported FrpD 250 protein form, both were able to bind the N-terminal portion of FrpC with very high affinity. This raises intriguing questions on what could be the biological role of such a strong binding interaction. The cytosolic and/or exported FrpD forms could, for example, be serving as inhibitors of FrpC activity in bacterial cytoplasm and/or cell envelope, safeguarding the producing cell from FrpC activity in the case of inefficient secretion and/or mislocalization. A precedent for genes encoding accessory inhibitor proteins encoded by rtx loci is found within loci for production of the secreted RTX proteases of Photorhabdus, Erwinia, or Serratia. These contain genes encoding protease inhibitors produced both as cytosolic proteins and also as proteins exported to bacterial periplasm by means of a classical signal peptide (56). The FrpD protein, however, does not appear to serve as a protease inhibitor, because no inhibition of the self-processing activity of FrpC could be observed following binding of FrpD in vitro and no other protease activity of FrpC, beside its capacity of calcium-dependent autoprocessing at the Asp 414 -Pro 415 bond, could as yet be demonstrated (9). FrpD could also be potentially serving in FrpC biogenesis and/or activity by chaperoning or mediating FrpC secretion from meningococci and by anchoring FrpC on bacterial outer surface. Some outer membrane lipoproteins, such as the protein PulS involved in secretion of Klebsiella pullulanase were, indeed, found to exhibit a chaperone activity (57,58). Experiments are underway to explore these possibilities.