N. gonorrhoeae MS11 variants were propagated on gonococcal clear typing agar 20. Wild-type variants MS11mk expressing chromosomally encoded Opa proteins were provided by J. Swanson and are designated according to Swanson et al. 21 by capitals, e.g., OpaA–OpaK. Recombinant MS11 Opa variants 5 were a gift of T.F. Meyer (Max-Planck Institut für Biologie, Tübingen, Germany). Opa protein expression was verified by SDS-PAGE and immunoblotting of bacterial lysates followed by detection with anti-Opa antibody 4B12 21. Only nonpiliated bacteria were used. For experiments, bacteria were grown for 3 h in 10 ml Hepes medium (10 mM Hepes, 145 mM NaCl, 5 mM KCl, 0.5 mM MgCl2, 1 mM CaCl2, 5 mM glucose, and 1.5% proteose peptone no. 3 [Difco]), pH 7.4, in a gyratory shaker at 37°C. Bacterial suspensions were pelleted and resuspended in 1 ml Hepes buffer (Hepes medium without proteose peptone).
Construction of Mutant CD66 N-domains.
The construction of 6xHis-tagged N-domains of CD66e and CD66b in the pRSET-A vector was described previously 18. Mutations were introduced in CD66 N-domains by a modification of the procedure of Picard et al. 22. In brief, a mutagenic primer was designed containing the desired base changes flanked by at least 12 perfectly matched bases both upstream and downstream of the mutation (a list of primers is available on request). A megaprimer was generated by PCR using the mutagenic primer and a common vector-based 3′ primer (pRSET-rev) with the CD66 N-domain construct as template. The pRSET-rev primer was removed by passing the reaction mixture through a 100-kD Centricon device (5 min at 3,000 g; Amicon). A second PCR was performed with the same template plus 17 μl of the 50 μl of 100-kD Centricon retentate as 3′ primer and a common vector-based 5′ primer (pRSET-for). The resultant PCR product was cut with EcoRI and HindIII and ligated into pRSET-A. Constructs were electroporated into E. coli strain BL21 (DE3; Novagen). Mutations were verified by DNA sequencing through the entire N-domain insert. Primers were purchased from Genosys and restriction enzymes from New England Biolabs.
Binding of CD66 N-domains by Gonococci.
Cleared lysates of E. coli cells expressing the appropriate CD66 N-domain were prepared as described 18. Gonococci (108) in 200 μl Hepes buffer were incubated with 5–10 μl of cleared lysate for 20 min at 37°C. Bacteria were collected by centrifugation (5 min at 2,000 g), washed twice with 1 ml Hepes buffer, then solubilized in 30 μl SDS-PAGE sample buffer of which 2.5 μl was electrophoresed in 13.5% SDS-PAGE and transferred onto nitrocellulose. Bound CD66 N-domains were detected by anti-His antibody (1:15,000; Amersham Pharmacia Biotech) followed by peroxidase-conjugated protein A (1:20,000; Sigma Chemical Co.). Blots were developed using the enhanced chemiluminescence (ECL) protocol (Amersham Pharmacia Biotech). Documentation and quantification of bands were performed with an AlphaImager® 2000 Imaging system (Alpha Innotech).
Construction of CD66b cDNA Mutants.
CD66b cDNA in pUC118 was a gift from Motomu Kuroki (Fukuoka University, Fukuoka, Japan). The CD66b insert was subcloned into the eukaryotic expression vector pTracer-CMV2 (Invitrogen). All mutations were made first in the CD66b N-domain construct in pRSET-A as outlined above. To introduce mutated N-domains into the full-length CD66b cDNA, the BlpI site present at the start of the N-domain in CD66b cDNA was introduced in the CD66b N-domain construct in the pRSET-A vector by PCR. For introduction of the chimeric CD66e/b N-domain into CD66b, the BlpI site present in CD66b (GCTCAGC) was mutated to the BlpI site found in CD66e (GCTAAGC) by megaprimer PCR, in order to have the N-domain start with the CD66e-derived lysine. The resultant PCR product was cleaved with BlpI and NsiI (NsiI cuts at residue 71 in CD66 N-domains) and substituted with the fragment present in the full-length CD66b cDNA in pTracer-CMV2. To substitute the entire CD66b N-domain with the CD66e N-domain, a silent mutation containing a ClaI site was introduced at residue 113 just downstream of the N-domain in CD66b. The CD66e N-domain in pRSET-A was amplified with primers containing a BlpI and a ClaI site, respectively, cut and ligated into BlpI/ClaI-cut CD66b. The constructs were electroporated into E. coli DH5α. All mutations in the pTracer-CMV2-CD66b constructs were verified by DNA sequencing.
Chinese hamster ovary (CHO) cells (Pro5) were obtained from the American Type Culture Collection and were grown in RPMI 1640/5% FCS in 25-cm2 flasks to 50% confluency. Plasmid preparations of pTracer-CMV2-CD66b mutants were made by the Wizard miniprep procedure (Promega Corp.). Pro5 cells were incubated in 2 ml DMEM/10% Nu-serum (Collaborative Biomedical Products) containing 4 μg plasmid DNA and 0.2 mg/ml DEAE-dextran (M
r 5 × 105; Amersham Pharmacia Biotech) for 4 h at 37°C. Cells were then shocked with 2.5 ml 10% DMSO in PBS for 1 min at room temperature, washed once with HBSS, and subsequently cultured overnight in RPMI 1640/5% FCS 23. The next day, the transfectants were trypsinized and seeded onto 12-mm-diameter circular glass coverslips in 24-well plates (105 cells per coverslip). Cells were cultured for 2–3 d before infection assays were carried out.
Gonococci (1.5 × 107) were added to 24-well plates containing the transfected cell cultures on coverslips, in 1 ml DMEM (without serum) for 45 min at 37°C and 5% CO2. Nonadherent bacteria were removed by three washes with HBSS. Infected cultures were fixed with 2% formaldehyde in PBS for at least 30 min.
Fixed infected cells were incubated with 0.5% Triton X-100 in PBS for 20 min and then blocked with 5% FCS in PBS for 1 h. Antibodies were diluted in PBS/0.05% Tween/0.5% FCS. To detect receptor expression, cells were stained with rabbit anti-CD66 antiserum (1:200; Dako) followed by Alexa 594–conjugated goat anti–rabbit (GAR) IgG (1:400; Molecular Probes, Inc.). To subsequently stain gonococci, coverslips were incubated with a mouse mAb against gonococcal LPS, generated in our laboratory by J. Swanson, followed by FITC-conjugated goat anti–mouse (GAM) IgG (1:400; Sigma Chemical Co.). When only receptor expression was evaluated, the permeabilization step with Triton X-100 was omitted. mAb Kat4C was provided by H. Turley (John Radcliffe Hospital, Oxford, UK). To distinguish extra- and intracellular bacteria, infected cell cultures on coverslips were subjected to a differential staining procedure as described previously 10
24. In brief, fixed cells were incubated successively with an anti-LPS mAb and a protein A–gold conjugate. The gold was enhanced by silver-staining to visualize extracellular bacteria, after which cells and intracellular bacteria were stained with 0.005% crystalviolet in H2O for 10 min.
Identification of CD66e N-domain Residues Critical for Binding of Opa-expressing Gonococci.
The differential recognition of CD66 receptor N-domains by gonococcal Opa variants is shown in Table . CD66e is recognized by the majority of Opa variants, whereas CD66b does not bind any Opa variant. OpaB, OpaC, OpaG, and OpaI variants demonstrate the broadest recognition of CD66 receptors. Assays with previously constructed chimeric receptor N-domains consisting of the NH2-terminal half (residues 1–59) of CD66e fused to the COOH-terminal half of the CD66b N-domain and vice versa, located the critical domain for binding of OpaB, OpaC, OpaG, and OpaI variants to the first 59 residues of CD66e (18; compare binding of the chimeric N-domains CD66e/b and CD66b/e in Table ). Using homologue scanning mutagenesis, we exchanged regions and single residues between the NH2-terminal 59 residues of CD66e and CD66b (for sequence differences, see Fig. 1 A) and measured the ability of the mutant proteins to bind to Opa−, OpaB-, OpaC-, and OpaI-expressing gonococci in order to identify CD66 residues required for Opa protein binding. Since OpaG is nearly identical to OpaB 25, we did not include this Opa variant in our assays.
As is shown in Table and Fig. 2 A, mutations in the first 11 residues of the CD66e N-domain did not influence binding of any tested Opa variant. Mutation of the region comprising residues 27–29 (mut3) resulted in a complete loss of recognition by all variants. Evaluation of the individual residues in this region revealed that the loss of binding was caused by the single F29R mutation (mut6). The single mutation S32N (mut7) also abrogated all Opa binding. The double mutation G41A+Q44R (mut9) clearly diminished Opa binding, but when these mutations were tested individually, a difference between Opa variants was noted. For OpaI interaction, both the G41A and the Q44R mutations were deleterious. This was also the case for interaction with OpaC, although to a lesser extent. The Q44R mutation had a moderate effect on OpaB binding, whereas the G41A change had no effect at all. Mutations in the region between residues 51 and 55 caused only a moderate decrease in Opa binding. Thus, the two residues F29 and S32 in CD66e are critical for binding of all Opa variants, whereas residues G41 and Q44 are important to different extents for the various Opa proteins.
Gain-of-function by the non-Opa Binding CD66 Family Member CD66b.
If the four residues indicated above were indeed responsible for the receptor function of CD66e, it should be possible to impart Opa binding properties to CD66b by introducing those four residues into CD66b. To test this concept, we changed the residues in question in the CD66b N-domain to the corresponding ones of CD66e in different combinations. As can be seen in Table , the presence of F29+S32 (mut13) did not result in detectable binding of any Opa variant. Residue F29 in mut13 is preceded by two CD66b-specific residues, D27 and P28, which may influence the correct conformation of F29. To address whether the F29 residue in a more “CD66e-like” environment would mediate Opa binding, we added residues H27 and L28 to mut13. This molecule (mut14) also failed to bind significantly to any Opa variant. The A41G+ R44Q mutant (mut15) did not bind any Opa variant either. We then added the G41 and Q44 residues individually to mut14, resulting in mut16 and mut17, and found that each addition slightly enhanced binding of all Opa variants compared with mut14. Only when both G41 and Q44 were present (mut18) was full binding activity by all Opa variants evident (Fig. 2 B). We then readdressed the question whether the origin of residues 27 and 28 played any role in Opa binding by construction of mut19–21. The presence of F29+S32+G41+Q44 was sufficient for binding of OpaB and OpaI variants; however, for OpaC binding an additional CD66e-derived residue (L28) was necessary, indicating a difference in binding characteristics among Opa variants (Fig. 2 B, and Table ). These data confirm the key role of four residues in the CD66e N-domain for receptor function of CD66e and again show a difference in binding characteristics within the OpaB, OpaC, and OpaI group of variants.
Expression and Recognition of Full-length CD66b Mutants on the Surface of Epithelial Cells.
To confirm that these results obtained with soluble recombinant His-tagged N-domains reflect interactions taking place at the cell surface, we introduced the identified key residues into full-length CD66b cDNA, transfected CHO cells with these constructs, and performed infection assays to determine Opa receptor activity. All transfectants stained positive with mAb Kat4C, which recognizes an epitope in internal CD66 domains (26; data not shown), as well as with a polyclonal anti-CD66 serum (Fig. 3), indicating that the receptors were expressed at the surface of the cells.
Infection of the various transfectants with MS11 Opa variants was evaluated by double immunofluorescence staining. Bacteria were found only on receptor-positive cells, which comprised 50–75% of the total cell population. The results of these experiments are shown in Table , with Fig. 3 illustrating the designations used in Table . Surprisingly, a single mutation in CD66b (N32S) resulted in significant binding of OpaB (Fig. 3 c) and OpaI variants, but not of OpaC variants. The single mutations A41G and R29F resulted in low level binding of OpaB (Fig. 3 b) and OpaI variants, but not of OpaC variants. The single mutation R44Q conferred no detectable Opa binding properties upon CD66b. In addition, the double mutant R44Q+ N32S was indistinguishable from the N32S mutant, indicating that the R44Q mutation did not contribute to Opa binding. The double mutations R29F+N32S and N32S+ A41G resulted in binding of large numbers of OpaB, OpaC, and OpaI variants (Table , and Fig. 3 d). Addition of mutation P28L to N32S, to R29F+N32S, or to N32S+ A41G did not affect interaction with any Opa variant (data not shown). The triple mutant R29F+N32S+A41G showed strong binding of Opa variants: infections with OpaB and OpaI variants resulted in distinct redistribution of the receptors towards the sites of bacterial adhesion, resulting in footprint-like appearances in the microscope (Fig. 3 e). This phenomenon was not seen with OpaC variants. The triple mutant was indistinguishable from a CD66b receptor containing the chimeric CD66e/b or the native CD66e N-domain, indicating that maximal binding was achieved with the triple mutant (Table ). Opa− variants did not interact with any transfectant (data not shown). To determine whether the transfectants that bound Opa variants were also able to ingest them, we applied a differential staining procedure after infection of the cells. We found that in all transfectants that bound Opa variants, intracellular gonococci could readily be found (Fig. 3 f). This demonstrates that CD66b molecules, containing the appropriate Opa binding domain, can act as complete functional receptors for gonococcal Opa variants. In summary, our findings show that, when introduced in full-length receptors in CHO cells, the mutations R29F, N32S, and G41A each conferred significant receptor activity upon CD66b, with N32S being the most effective single mutation. Replacement of all three residues provided maximal adherence of Opa variants. These gain-of-function experiments largely confirm the results obtained with soluble receptor N-domains, although the requirements for binding appear to be less stringent. Furthermore, again using full-length receptor mutants a difference in binding characteristics was noted for variants expressing OpaC compared with OpaB and OpaI variants.
In this study, we have mapped residues on CD66 receptors that determine Opa protein binding. We have identified three key amino acid residues in CD66e required for maximal binding using homologue scanning mutagenesis and subsequent analysis of both loss-of-function and gain-of-function of CD66 mutants in binding and infection assays. Furthermore, Opa proteins B, C, and I, which bind to an identical spectrum of native CD66 receptors, were found to differ in their recognition of mutant CD66 N-domains and mutant receptors on cells, indicating that their binding characteristics may actually not be identical.
Neisserial Opa proteins are thought to span the bacterial outer membrane in an eight-stranded β-barrel conformation resulting in four extracellular loops. Three of the exposed loops consist of variable sequence domains while a fourth loop near the COOH terminus of the protein is highly conserved. The differential binding of Opa proteins to CD66 receptors is likely a reflection of the ability of these variable domains to interact with the receptors. As we pointed out previously 10, it is remarkable that Opa proteins, such as OpaB, OpaC, and OpaI, that contain heterologous variable domains recognize the same subgroup of CD66 family members. However, our present data showing differences in binding of mutant molecules by these Opa proteins may indicate that they actually bind with slightly different characteristics, which would be expected from proteins with such divergent binding domain sequences. The opa gene family is thought to have arisen from recent gene duplication events and genetic reassortment of variable sequence domains between members of the gene family 25
29. Interestingly, the CD66 family also appears to have arisen recently by duplication of at least one ancestral gene. Sequence comparisons between the rodent and primate CD66 gene families show higher interspecies than intraspecies variation, suggesting that the duplication occurred after mammalian radiation took place 30
31. In addition, the mutation rate in the CD66 N-domain exons is twice as high as that of the adjacent intron, suggesting that the CD66 family is still undergoing rapid evolution 30. It is tempting to speculate that the extensive evolution of the opa gene family in the strictly human pathogen Ngo has taken place in response to the rapidly evolving family of primate-specific CD66 molecules. The observed differential Opa–CD66 binding patterns may be a reflection of this process.
According to the predicted structure of the CD66e N-domain 32, the residues we have identified as important for Opa protein binding, F29, S32, and G41, are located in exposed loops and strands of the GFCC′C′′ face of the CD66e N-domain (Fig. 1 B). This face of the molecule is not covered by carbohydrate, in contrast to the ABED face, as predicted by a low resolution model for CD66e 34, and would therefore be accessible for protein–protein interactions. The key role of S32 and G41 in binding of Opa proteins is supported by the fact that these residues are conserved among Opa-binding CD66 molecules (Fig. 1 A). Residue F29 is conserved in three out of four of the Opa-binding CD66 molecules (CD66a, d, and e), whereas CD66c contains an I at residue 29. Possibly hydrophobic residues such as F or I at that position support Opa protein binding, while a charged residue, such as R present in CD66b, does not. The role of residues L28 and Q44 is probably minor, since they are not required for Opa recognition of the native molecule, although their presence enhances binding of the soluble N-domain. Several ligand and viral binding sites on IgSF members have been found on the GFC face of the ligand-binding domain 35
36, indicating that this domain face is positioned favorably to serve as a ligand-binding platform for IgSF members.
Homologue scanning mutagenesis does not address the role of conserved residues among the two homologous proteins in ligand binding. The finding that single, relatively conservative mutations in CD66b, such as N32S and A41G, were sufficient to induce receptor function in CD66b may suggest that CD66b contains conserved residues participating in Opa protein binding. To test whether the actual binding site is comprised in the first 59 residues of CD66e, we constructed a truncated, His-tagged CD66e molecule (residues 1–59) and tested it for binding to Opa variants. This molecule was well expressed by E. coli but failed to bind significantly to any Opa variant (data not shown), indicating that other sites within the N-domain may be required for binding or that the peptide did not adopt the correct conformation for binding activity. The finding that single mutations in CD66b result in functional receptor activity could suggest that in vivo isoforms of CD66b exist that will be recognized by Opa variants. Evidence to support this concept comes from analysis of CD66b cDNAs cloned from normal white blood cells and leukemic cells. These cDNAs differed in two base pairs in the coding region, resulting in two amino acid differences, one in the N-domain (R80K) and one in the COOH-terminal M-domain (V288L) 37
38. Cloning of another CD66 family member, CD66d, by two different groups resulted in proteins differing in two residues 39
40. These phenomena fit with the notion that CD66 molecules are subject to sequence variation associated with rapid evolution.
CD66 family members are capable of mediating homophilic and heterophilic intercellular adhesion, like many other IgSF proteins 41. Binding between CAMs is usually of very low affinity, but due to the highly multimeric nature of cell–cell adhesion, sufficient avidity can be achieved to allow detection of the interaction between cells. The weakness of CAM interactions is illustrated by the difficulties in detecting binding of purified, monomeric forms of CAMs. This difficulty arises because binding assays require separation and washing steps, during which time weakly interacting molecules dissociate 42. This phenomenon may explain our observation that binding of Opa variants to soluble CD66b N-domain mutants required more CD66e-derived residues than binding to cell surface CD66 receptors. If each mutation introduced into CD66b enhances binding affinity between Opa and CD66b, as suggested by our infection assay data, then the threshold level of affinity necessary for detection will be reached sooner for the infection assay than for binding in solution. Alternatively, the level of multimerization may be important for Opa–CD66 binding, as has been shown for the binding of IgSF member intercellular adhesion molecule 1 (ICAM-1) to its receptor LFA-1. Recombinant ICAM-1 exists as monomers in solution, and direct binding to LFA-1 has been impossible to detect. Only when ICAM-1 was modified to induce dimerization could LFA-1 binding be detected 43
44. CD66 family members can exist as dimers in the plasma membrane of eukaryotic cells 45, and recombinant CD66e N-domains have been shown to form oligomers in solution 46. Receptor dimers will more likely be found on the surface of cells, where the receptor concentration is higher than in solution. Thus, if Opa binding requires a receptor dimer, one would expect Opa variants to bind more readily to receptor-expressing cells than to soluble receptors. Another possible explanation for the discrepancy between the two binding detection methods is that in the solution assay the receptor binding domain exists as an unglycosylated, single domain, while in the infection assay the binding domain is presented in the context of a complete, glycosylated molecule. Although Opa binding does not require carbohydrate 18, the presence of sugar moieties may influence the strength of adhesion, as has been observed for the interaction between CD2 and CD58. Human CD2 has a single carbohydrate addition site in its Ig variable–like N-domain that is absolutely required for binding to its normal ligand, CD58. Evidence from solution structure of this carbohydrate chain in relation to the GFC binding face indicates that the glycan is not itself situated in the binding face but is required to balance an unfavorable negative charge in order to maintain an active binding configuration 47. In fact, CD66 N-domains contain a potential glycosylation site at residue 70 (Fig. 1), which corresponds exactly to the structural position of the glycosylation site affecting ligand binding ability of human CD2 41. Mutation of this site in CD66e influences CD66e homophilic interactions, which are based on protein–protein interactions, indicating that in CD66e also the degree of glycosylation can influence binding events mediated through CD66 protein sequences 41. Our data stress the importance of evaluating binding events in different assays, since differences between assays can reveal further details of the molecular interaction between ligands. Regardless of the molecular basis for the observed discrepancy, both assays show clearly that OpaC binding requires more CD66e-derived residues than OpaB or OpaI, which may indicate that OpaC binding of the receptor is of lower affinity than binding of OpaB or OpaI.
In conclusion, mapping of key residues in CD66 required for recognition by the various gonococcal Opa adhesins indicates that single amino acid residues in CD66 receptors determine Opa protein binding. These results may provide a first step towards resolving the structural requirements for the Opa–CD66 receptor interaction and thereby help the development of infection inhibitory strategies, and may provide insights into the function of CD66 molecules in normal tissue and in carcinogenesis.