Molecular aspects of complement-mediated bacterial killing. Periplasmic conversion of C9 from a protoxin to a toxin.

As part of the membrane attack complex complement protein C9 is responsible for direct killing of bacteria. Here we show that in the periplasmic space of an Escherichia coli cell C9 is converted from a protoxin to a toxin by periplasmic conditions missing in spheroplasts. This conversion is independent of the pathway by which C9 enters the periplasm. Both, C9 shocked into the periplasm and plasmid-expressed C9 targeted to the periplasm via a signal sequence are toxic. Toxicity requires disulfide-linked C9 because export into the periplasm of cells defective in disulfide bond synthesis (dsbA and dsbB mutants) is not toxic unless N-acetylcysteine is added externally to promote cystines. A N-terminal fragment, C9[1-144], is not toxic nor is cytoplasmically expressed C9, even in trxB mutants that are able to form disulfide bonds in the cytoplasm. Importantly, expression of full-length C9 in complement-resistant cells has no effect on their viability. Expression and translocation into the periplasm may provide a novel model to identify molecular mechanisms of other bactericidal disulfide-linked proteins and to investigate the nature of bacterial complement resistance.

As part of the membrane attack complex complement protein C9 is responsible for direct killing of bacteria. Here we show that in the periplasmic space of an Escherichia coli cell C9 is converted from a protoxin to a toxin by periplasmic conditions missing in spheroplasts. This conversion is independent of the pathway by which C9 enters the periplasm. Both, C9 shocked into the periplasm and plasmid-expressed C9 targeted to the periplasm via a signal sequence are toxic. Toxicity requires disulfide-linked C9 because export into the periplasm of cells defective in disulfide bond synthesis (dsbA and dsbB mutants) is not toxic unless N-acetylcysteine is added externally to promote cystines. A N-terminal fragment, C9[1-144], is not toxic nor is cytoplasmically expressed C9, even in trxB mutants that are able to form disulfide bonds in the cytoplasm. Importantly, expression of full-length C9 in complement-resistant cells has no effect on their viability. Expression and translocation into the periplasm may provide a novel model to identify molecular mechanisms of other bactericidal disulfide-linked proteins and to investigate the nature of bacterial complement resistance.
Complement is part of the innate immune system and one of the first lines of defense against pathogenic microorganisms. More than 20 blood complement proteins control microbial invasion of the host by two different mechanisms: (i) C3-mediated opsonization of cells and (ii) direct killing as a result of C5b-9 complex formation on the target (1). Much is known about the strategies used by pathogens to avoid opsonization because the basic molecular details of the process are recognized (1)(2)(3). In contrast, the mechanisms of direct complementmediated killing of Gram-negative bacteria have remained elusive, and the means used by bacteria to become resistant to its effects are largely unknown (4). Without doubt, opsonization is of great importance in many infectious diseases. However, epidemiological studies of complement-resistant bacteria in systemic infections, as well as studies on individuals with genetically determined complement deficiencies, have indicated that (C5b-9)-mediated mechanisms play an important role in the control of Gram-negative infections. Most striking is the observation by Lassiter et al. (5) that Escherichia coli strains that are normally killed by adult serum ("serum-sensitive" strains) pose a severe threat to survival of premature or newborn infants because of diminished serum C9 concentrations in neonates.
Microbes escape complement destruction by inhibiting the three pathways of activation by synthesizing surface proteins that mimic complement control proteins or by capturing such proteins (1,3,4). Resistance to the action of C5b-9, also referred to as the membrane attack complex (MAC), 1 is frequently achieved by changes from a rough to a smooth phenotype. Extension of surface carbohydrate structures causes an increase in the envelope hydrophilicity and weakens the anchoring of the MAC. Some strains escape killing by shedding of the MAC. Nevertheless, it is also known that not all smooth strains are serum-resistant and that some resistant strains do not shed the MAC but carry it in a stably bound form. Thus, strains that are truly resistant to complement, in that they are able to replicate in 50% serum (4), have acquired virulence factors in addition to those that change the chemical nature of the OM.
Some molecular details of the processes by which the MAC, once assembled on a bacterium, elicits death are currently understood. Death of sensitive strains following exposure to serum is extremely rapid, and it is thought to occur by dissipation of cellular energy (4,6,7). Killing can be prevented by incubation with membrane potential uncouplers or with inhibitors of oxidative phosphorylation indicating that ATP generated by the target bacterial cell is required. Efficient killing requires stable deposition of the C5b-9 complex on the OM but no other serum components in addition to the terminal complement proteins. Furthermore, it has been shown that formation of poly(C9), a tubular polymerization product of C9, is only incidental to hemolysis and not required for bacterial killing (7)(8)(9). However, whereas the (C5b-8) 1 C9 1 complex is sufficient to lyse erythrocytes fully, it has no bactericidal activity (10). This observation has been extended to show that a (C5b-8) 1 C9 4 complex is strongly bactericidal but that incorporation of additional C9 molecules to form poly(C9) does not improve killing significantly (11). A further important conceptual advance was provided by Tomlinson et al. (12), who demonstrated that transfer of preassembled C5b-9 complexes into the OM by fusion techniques did not elicit cell killing despite the fact that small molecules now had access to the periplasm. Finally, Dankert and Esser (7) showed that the C-terminal half of C9, the C9b fragment, dissipated the membrane potential across respiring inner membrane (IM) vesicles, although the complete molecule had no effect. In addition, the requirement for a C9 receptor (that is, C5b-8 assembly) on the OM could be bypassed if the C9 molecule was osmotically shocked into the periplasmic space (13). This report firmly established that the bactericidal activity of complement is dependent upon C9 because none of the other terminal proteins when shocked into the periplasm elicited any cytotoxic effects. What is currently not understood is how C9, after its binding to the C5b-8 complex on the OM, translocates across the periplasm and dissipates the potential across the IM, whether bacterial envelope proteins are needed for these processes, and, most importantly, where the active (bactericidal) center is located in the C9 molecule. Here we show that periplasmic conditions in sensitive cells are important for C9-mediated killing and that the C5b-8 receptor complex plays no role in the final stages of this process.

EXPERIMENTAL PROCEDURES
Complement Proteins and Assays-Complement proteins C8 and C9 were purified and assayed for hemolytic activity according to standard procedures (14), and ␣-thrombin-cleaved C9 (C9 n ) was prepared as published (7).
Bacterial Strains and Killing Assays-All strains used here are listed in Table I. Serum or C9-mediated bacterial killing of bacteria was measured as published (13). Viable spheroplasts were prepared by the method of Marvin and Witholt (22) and assayed for killing as described for whole cells. YW10 and YW13 are the dsb strains FD596 and NLM100, respectively, lysogenized with DE3 carrying the inducible gene for T7 RNA polymerase ( lysogenization kit, Novagen, Madison, WI).
Plasmids and Constructs-The previously characterized eukaryotic C9 expression vector pSVLHuC9 (23) and a derivative (pYW29) with two amino acid mutations, L532S and K538R, and a hexahistidine tag at the C-terminal end were used as the starting plasmids for the generation of the prokaryotic expression vectors shown in Table II. The gene for mature C9 was transferred in two steps into pET12b. First, a SalI restriction site was introduced into pSVLHuC9 at the end of the C9 signal peptide using the "altered sites mutagenesis system" from Promega (Madison, WI) and then a SalI(66)-BamHI(495) fragment was excised and inserted behind the OmpT signal sequence in pET12b to generate pYW49. Second, a BamHI(495-1703) fragment excised from pSVLHuC9, containing C9[144 -538] was inserted in frame and in the correct orientation into BamHI-digested pET12b producing pYW50. To introduce the hexahistidine tag, pYW51 was generated by digesting pYW50 and pYW29 with NsiI and SacI and replacing the NsiI-SacI fragment in pYW50 with the NsiI(1307)-SacI(1700) fragment in pYW29. The secretion plasmid for full-length C9, pYW52, was made by cutting pYW49 with BamHI and inserting a BamHI(495-1703) fragment excised from pSVLHuC9 in frame and in the correct orientation. For the expression of cytoplasmic C9 first a NdeI(58)-BamHI(495) polymerase chain reaction fragment starting at the beginning of mature C9 was generated using pSVLHuC9 as a template. This fragment was inserted into pET12b to generate pYW53 and then a BamHI(495-1703) fragment from pSVLHuC9 was inserted in frame and in the correct orientation to generate pYW54. Because these pET12b-derived plasmids could not be used for protein expression (see below) various complete C9 genes together with ribosomal binding sites and signal sequences (if any) were transferred as XbaI-SacI fragments into pET22b and pASK75, respectively, to provide better control of transcription. This resulted in the replacement of the pelB and ompA signal peptides in pET22b and pASK75, respectively, with the ompT signal peptide in the secretion vectors, and neither the intrinsic hexahistidine tag in pET22b nor the Strep TM tag in pASK75 were fused to the C9 constructs. All nucleotide changes were verified by sequencing.
Expression of Recombinant C9 and Toxicity Assays-Transformed cells were grown overnight in LB medium supplemented with the appropriate antibiotic. Fresh cultures (20 ml) in LB or M63 medium (supplemented with amino acids, thiamine at 1 mg/ml, 1 mM MgSO 4 , 1 mM sodium citrate, and 0.4% glucose as carbon source and the appropriate antibiotic) (16) were inoculated at 1:100 dilution and incubated at the desired temperature under constant shaking. After reaching mid-exponential growth, phase cultures were split in two, and one half was induced with either 1 mM isopropyl ␤-D-thiogalactoside or 50 ng/ml anhydrotetracycline depending on the expression vector, and the other half served as uninduced control. After about 1 h of induction, strains lacking functional dsbA or dsbB received 5 mM N-acetylcysteine (NaC) to promote disulfide pairing (24). Viability of all cultures was assayed by removing small aliquots at predetermined time points and counting colonies on agar plates prepared with M63 medium supplemented with casamino acids, glucose, and the appropriate antibiotic and incubation at 37°C. Strains lacking functional thioredoxin reductase (trxB) were cultured and induced as described (19).
Complement-resistant strains (LP1395 and 25922) were transformed with pYW60, and ampicillin-resistant colonies were picked and cultured. After induction with anhydrotetracycline (100 ng/ml), cell viability was assayed by colony counting as described above.
Ligand Blotting-Published procedures were used to prepare total cell lysates and periplasmic shock fluids (25,26) for SDS-PAGE analysis and blotting. A polyclonal anti-C9 serum adsorbed on immobilized cell extracts of E. coli BL21(DE3) or a monoclonal antibody (mAb216) against human C9 was used to detect recombinant C9 by Western blotting as described previously (23). Recombinant proteins carrying the hexahistidine tag were also visualized using the nickel-nitrilotriacetic acid alkaline phosphatase conjugate detection system (Qiagen, Valencia, CA).
Proteolysis of Spheroplasts-NLM100 cells were induced for 3 h to express secretable C9 (pYW60) or cytoplasmic C9 (pYW61), collected by centrifugation and resuspended in Tris-EDTA-sucrose buffer and lysozyme at about 1 ϫ 10 9 cells/ml to produce spheroplasts as described (17). The spheroplasts were then incubated at 37°C with trypsin (1 g/ml) for up to 60 min, aliquots were withdrawn, and proteolysis was stopped by addition of SBTI (5 g/ml). The cells were homogenized by sonication and analyzed by SDS-PAGE and ligand blotting.
Flotation Gradient Centrifugation-These experiments were carried out as described by Thom and Randall (25). In brief, spheroplasts were lysed by three cycles of freeze-thawing and sonication, and total lysates were centrifuged for 16 h in a metrizamide gradient (1.27-1.29 g/ml) at 80,000 rpm in a Beckman TLA120.2 rotor to separate protein aggregates and membranes. Gradient fractions were analyzed by SDS-PAGE and ligand blotting.

RESULTS
Effect of C9 on Viability of E. coli Spheroplasts-We previously discovered that when C9 was shocked into the periplasm it killed sensitive E. coli, whereas it had no effect on respiring IM vesicles in contrast to the C9b fragment (7,13). This prompted us to ask whether periplasmic processing was required for cytotoxicity. As shown in Fig. 1, neither native C9 nor C9 cleaved with thrombin (to produce C9a and C9b) had a significant effect on spheroplast viability. This strongly suggested that periplasmic factors are required for toxicity and indicated that proteolytic cleavage alone is inadequate to convert C9 from a protoxin to a toxin.
Expression of Recombinant C9 and Export into the Periplasm Is Toxic-To gain further information on required periplasmic factors, we introduced C9 into the periplasm from the inside of the cell by secretion of recombinant C9, rather than from the outside by osmotic shock. The gene for mature C9 was cloned into the pET12 vector, which provides an OmpT signal peptide DsbA::kan, dsbB::kan J. Beckwith (16) for export of recombinant proteins into the periplasm to generate pYW52. However, because of promoter leakiness it was impossible to transform cells efficiently. When very large amounts of pYW52 DNA were used for transformation, a few colonies could be recovered. Sequencing of the plasmid isolated from four different colonies indicated that in each case the C9 gene was inactivated by insertion of bacterial IS elements (data not shown). Using vectors derived from pET22b or pASK75 that are more tightly controlled (pYW55-pYW61), transformation was possible, and recombinant cells could be cultured. Expression of C9 caused an immediate loss of viability of the cells harboring these plasmids (Fig. 2, A and B). A secreted, N-terminal fragment, C9[1-144], was not toxic, but the Cterminal fragment C9[145-538] was as toxic as the complete molecule (Fig. 2C). Significantly, when C9 was cloned into either vector without any signal peptide (pYW53, pYW54, pYW58, and pYW61), no toxic effects were observed after induction, and C9 accumulated as soluble and insoluble forms in the cytoplasm that were hemolytically inactive (data not shown). Of significance is the fact that expression of secretable C9 in two complement-resistant strains of E. coli (LP1395 and 29025) from pYW60 had no effect on viability (data not shown). Disulfide Bond Formation Is Required for Toxicity of Exported C9 -Although these results suggested that C9 was exported into the periplasm and exhibited its toxic effects in a manner similar to external C9 shocked into the periplasm, we could not exclude the possibility that loss of viability was caused by merely expressing a foreign gene in E. coli. Many gene products are known to express such general cytopathic effects. To test for this possibility we took advantage of the fact that C9 is a heavily disulfide-linked protein (Fig. 3A), which is inactive when reduced. Disulfide bond formation in the periplasm of E. coli is controlled by the dsbA, dsbB, dsbC, dsbD, dsbE, and dsbG genes (28). The DsbA protein is a protein disulfide isomerase in the periplasm that introduces cystines into secreted proteins and thereby becomes reduced. The DsbB protein is an IM protein that reoxidizes reduced DsbA. In contrast to wild type cells, E. coli mutants with defective dsbA or dsbB genes harboring pASK75-derived plasmids were not  killed by induction of C9 expression (Fig. 3, B and C), nor were DE3 lysogenized dsb mutants (YW10-YW13) transformed with pET22b-derived vectors (Fig. 3D). This lack of cytotoxicity is probably related to absence of disulfide bond formation in secreted C9 and, therefore, proper folding of the molecule because when NaC was added to the culture medium an immediate viability loss was observed. Importantly, single and double dsb mutants are not complement-resistant per se because they were killed in a C9-dependent process by serum (Table I). Cells harboring plasmids (pYW58 and pYW61) that express C9 in the cytoplasm were not affected by NaC addition, suggesting that the effect is specific for periplasmic C9 and was not caused by NaC directly.
Localization of Secreted C9 -To provide direct evidence for the cellular location of the protein expressed from plasmids coding for exported C9, we performed protease sensitivity assays. In this assay, the trypsin sensitivity of C9 is assayed in cells that have been stripped of their outer cell wall (spheroplasts) exposing periplasmic but not cytoplasmic contents to the protease. Immunoblots (Fig. 4A) specific for C9 or ligand blots specific for the His 6 tag (Fig. 4B) show almost complete degradation of periplasmic C9 but complete protection of cytoplasmic C9, indicating that Ͼ90% of the exported C9 reached the periplasmic compartment. Flotation gradient centrifugation of spheroplast lysates can provide further evidence for the localization of expressed proteins and their physical state. As shown in Fig. 4 (C and D), C9 secreted by a dsbB strain (NML100) was mostly located at the bottom of the metrizamide gradient, providing evidence that it is either aggregated or associated with another insoluble protein, and the remainder floated to the top of the gradient, indicating that some of it is membrane-bound. When we tested periplasmic shock fluids produced from cells expressing secreted C9 none could be detected (Fig. 5), nor was C9[145-538] detectable. In contrast, a N-terminal fragment (C9[1-144]) was detected easily as was cytoplasmic C9.

Periplasmic Tol and Ton Proteins Are Not Required for C9
Toxicity-Our observations that isolated C9 has no effect on spheroplast viability but is toxic when shocked into the periplasm of whole cells suggested that periplasmic constituents may be required for toxicity. This situation is reminiscent of the action of membrane-active colicins (9), which require either Tol or Ton proteins for translocation across the cell envelope, insertion into the IM, and cell killing (29). However, E. coli strains K17A1 and KP1032, which are defective in TolA or TonB, respectively, were killed by serum complement (Table  I), indicating that these proteins are not involved in the translocation of C9 across the periplasmic space.
Is Cytoplasmic C9 Toxic?-The fact that accumulation of C9 in the cytoplasm of wild type cell is not toxic to the host is not surprising. First, the protein is mostly in inclusion bodies and second, because it contains 12 disulfide bonds, it is highly unlikely that many native conformers will be formed under such circumstances. Recent advances in understanding the genetics of disulfide bond metabolism in E. coli led to the creation of thioredoxin reductase mutants (trxB) in which disulfide bond formation in proteins localized to the cytoplasm takes place (18). When signal sequenceless C9 was expressed in E. coli AD494, it accumulated in the cytoplasm, but its production had no effect on viability (data not shown). Several additional conditions were examined that are known to promote disulfide bond formation in trxB cells, such as growing cells in minimal medium and at lower temperatures (e.g. 23°C or 30°C) and then holding them on ice for a few hours after they are grown, but none impaired cell viability. Another approach to increase disulfide bond formation in a cytoplasmic protein has been used by Schneider et al. (19). These investigators co-expressed the chaperons DnaK and DnaJ together with SPARC, a protein containing seven disulfide bonds, in trxB cells and demonstrated that 15-20% soluble monomeric protein with an electrophoretic mobility characteristic of correctly disulfide-bonded and biologically active SPARC was formed. When signal sequenceless C9 was expressed in FC1 cells from the pYW58 plasmid together with DnaK/J from the pDnaK/J plasmid, it still had no effect on viability of such cells (data not shown), suggesting that cytoplasmically located C9 is not toxic to E. coli cells. DISCUSSION Native complement protein C9 is a soluble glycoprotein devoid of cellular toxicity; yet when it enters the periplasmic space of an E. coli cell it is converted from a protoxin to a toxin. This conversion is independent of the pathway by which C9 FIG. 4. Trypsinization and fractionation of spheroplasts expressing C9. A, NML100 (dsbB) cells, induced for 3 h at 30°C to express periplasmic or cytoplasmic C9, were converted to spheroplasts. A portion was treated with 1 g/ml trypsin at 37°C for the indicated time at which aliquots were removed and assayed for the accessibility of C9 to the protease by SDS-PAGE and Western blottinge (A) or by using the nickel-nitrilotriacetic acid alkaline phosphatase conjugate detection system (B). The remainder was lysed by freeze thawing and sonication. Crude lysates were equilibrated with metrizamide (final density, 1.27 g/ml), placed at the bottom of a centrifuge tube, and overlaid with a metrizamide solution of 1.29 g/ml density. After centrifugation the density of each fraction (recovered from the top) was determined (C) and analyzed by SDS-PAGE and Western blotting for the presence of C9 (D). enters the periplasm. Both, C9 shocked into the periplasmic space and C9 synthesized by expression from a plasmid and targeted to the periplasm via a signal sequence are toxic to the host cell. The toxic effects of plasmid-derived C9 were not caused simply by overexpressing a foreign gene because expression of C9 localized to the cytoplasm was not toxic. The lethal effects were also not due to jamming of the export pathway because of overexpression because the protein remained nontoxic when it was targeted to the periplasm of cells with impaired disulfide-bond catalysis (dsb mutants). In such cells the protein is translocated across the IM and becomes accessible in spheroplasts to externally added trypsin verifying its periplasmic location. The protein, however, is not present in a soluble form in the periplasm but is mostly aggregated, perhaps in so-called periplasmic inclusion bodies (30). A small percentage of translocated C9 is also still associated with the membrane. Wunderlich and Glockshuber (24) have shown that folding of a disulfide-bonded protein in the oxidative environment of the periplasm in well oxygenated cell cultures is influenced by addition of a thiol reagent such as NaC. When this catalyst was added to dsb mutants harboring C9 in their periplasm, it elicited an immediate loss of cell viability. We interpret this result to indicate that some cystine formation and therefore at least partial folding of the translocated C9 is necessary for toxicity. Because C9-mediated killing of bacteria is thought to occur by a single or dual hit process (31), the number of effective molecules is extremely low, and it is not surprising that NaC is useful in this process because very few C9 molecules need to form disulfide bonds and fold appropriately to evoke killing. The effectiveness of a few correctly folded C9 molecules is also demonstrated by the fact that transformation of wild type cells was not possible with plasmids in which expression is driven by promoters that are not tightly controlled.
Truncation of C9 indicated that the N-terminal 144 amino acids are not required for cytotoxicity and a fragment comprised of these residues accumulated in soluble form in the periplasm. The remainder, C9[145-538], however, is as toxic as the complete molecule when targeted to the periplasm. It contains the C9b fragment and, as shown earlier (7), this fragment by itself is able to de-energize IM vesicles and to kill spheroplasts (32). Within this fragment are four disulfide bonds, three in the epidermal growth factor domain located between Lys 487 and Glu 518 , and one in the so-called membrane attack complex/ perforin region, connecting Cys 359 with Cys 384 . Experiments are planned to test which of these cystines are required to fold C9 into a conformer that will elicit killing of dsb mutants with the help of NaC.
When native C9 enters the periplasm from the outside, either by binding first to its receptor, through a C5b-8 complex on the OM, or by being shocked into the periplasm, it is converted to a toxin. Experimental data are currently missing to distinguish whether this conversion involves only a refolding of the molecule or the interaction with an intrinsic periplasmic constituent or both. From our previous work (33) it is known that membrane insertion of C9 is accompanied by the transient appearance of refolding conformers that can be trapped by C9 sequence-specific anti-peptide antibodies. In the context of complement-mediated hemolysis, it is assumed that the interaction of C9 with C5b-8 on the erythrocyte membrane triggers these conformational changes. However, because C9 can kill in the absence of C5b-8 when shocked into the periplasm such a (C5b-8)-mediated trigger mechanism cannot be involved here. One possibility is that the molecule is proteolytically cleaved in the periplasm and a toxic fragment is liberated that inserts into the IM. The basis for this hypothesis is that the C9b fragment can dissipate the membrane potential across IM vesicles, whereas C9 n , in which the C9a and C9b fragments are still noncovalently associated, cannot and that it is not toxic to spheroplasts (Fig. 1). It is tempting to speculate that a helper factor in the cell envelope may interact with the C9b portion of C9 and convert it to a toxin. This hypothesis is consistent with fact that cytoplasmic C9 is not toxic even when it is produced under conditions most favorable to formation of disulfide bonds. One would expect that under such conditions at least as many properly folded C9 molecules are created in the cytoplasm as are created by NaC in the periplasm of dsb mutants and that they should be toxic if no other helper factors were required.
Such a helper-dependent mechanism is similar to the events that prevail during the killing of E. coli by membrane-inserting colicins. These colicins have a receptor binding domain, a translocation domain, and a membrane insertion domain (29). The requirement for receptor binding can be bypassed by shocking colicins into the periplasm. The translocation domain interacts in an energy-dependent manner with either Tol or Ton proteins. Because of the similar requirement for a membrane potential and the inhibitory effects of uncouplers for C9-mediated killing, we tested whether Tol or Ton proteins play a role in C9 killing. However, two mutants with defective ton or tol genes, which are refractory to killing by their respective colicins, are sensitive to C9-mediated killing, indicating that other periplasmic constituents are necessary. Considering the tendency of native C9 to undergo disulfide shuffling 2 and the effect of dsb mutations on the activity of expressed C9, it was conceivable that the DsbA and DsbB proteins themselves are involved in serum C9 toxicity. However, this is not the case because dsb mutants were killed even without NaC by serum complement, that is, when C9 enters the periplasm from the outside. Of interest in this context is the fact that E. coli strains that are resistant to serum complement killing are also resistant to killing by C9 when it is shocked into the periplasm or when it is overexpressed in the cytoplasm and targeted to the periplasm. Previous attempts to generate serum complementresistant mutants from sensitive strains have led to the generation of smooth phenotypes, that is, capsule producing mutants or mutants with increased LPS densities and/or longer O-saccharide chains (34,35). However, unlike patient-derived resistant strains such mutants can frequently be killed by complement when their OM integrity is impaired. In the future it may be possible to design genetic screens to search for helper factors that are required for C9 toxicity by using the targeting of C9 to the periplasm as described here.