The por gene of Neisseria gonorrhoeae encodes the Protein I porin responsible for serovar specificity. In this study the por genes have been sequenced from clinical isolates which exhibited anomalous serovar reactivity. One group of ‘intermediate’ strains differed significantly from both Protein IA and IB strains, were more closely related to IA but appeared to represent a distinct class of Protein I. Another strain was closely related to Protein IB of serovar IB-6 but contained a deletion of six amino acids in surface exposed loop 6 which removed epitopes recognised by IB specific monoclonal antibodies. The third group of strains, which reacted with both IA and IB specific monoclonal antibodies, expressed hybrid Protein I molecules containing both IA and IB epitopes. These strains appeared to originate from a double crossover between Proteins IA and IB with the amino and carboxy terminal residues homologous to IB while the surface exposed loop 6 demonstrated close homology to IA. This is the first demonstration of naturally occurring gonococci expressing a hybrid Protein IA/IB.
Outer membrane protein
Hybrid porin protein
Protein I (PI) is the major protein present in the outer membrane of Neisseria gonorrhoeae and functions as an anion selective porin . In addition to its essential physiological role, PI is an important surface antigen, having been identified as a potential vaccine candidate [2, 3] and forming the basis of the serological differentiation between strains. The structural gene, which encodes PI (por), exists as one of two alleles, so that any individual strain expresses only one of two different subclasses designated PIA and PIB . These subclasses have been further divided into serovars by their reactivity with two different panels of monoclonal antibodies (mAbs) [4, 5]. This system provides the basis of the differentiation of gonococci into serovars and has been widely used in epidemiological studies .
Sequencing of the por genes from PIA and PIB expressing strains has revealed the structural basis of PI antigenic heterogeneity and led to a model of the organisation of the proteins on the surface of the gonococcus, in which a series of conserved β-sheets traverse the outer membrane generating eight surface exposed loops . Despite the biochemical and structural differences between the two subclasses of PI, considerable similarity exists, with about 70% of the amino acid residues conserved and differences located in the exposed loops [8, 9]. Even greater homology is seen within each PI subclass. Different serovars may have up to 99% of sequence in common and studies with synthetic peptides have mapped the epitopes recognised by serovar specific monoclonal antibodies to localised regions of variation largely located at the apex of the surface exposed loops [8–10].
Although hybrids between PIA and PIB have been constructed in vitro [11, 12], the vast majority of clinical isolates react with either PIA or PIB antibodies, confirming that only one allele of the por gene is being expressed . However a recent study described two clusters of gonococcal isolates which reacted with antibodies to both PIA and PIB and hence appeared to express a hybrid porin . In this study we have sequenced the por gene from representatives of these clusters and from other isolates that show similar behaviour in order to determine the structural basis of the apparent hybrid PI molecules.
2 Materials and methods
2.1 Bacterial strains and growth conditions
N. gonorrhoeae isolates (a) were from male patients attending the genito-urinary clinic St. Mary's Hospital London between 1986 and 1988 and represent less than 0.1% of the total isolates from those submitted for serotyping to the laboratories at St. Mary's Hospital Medical School, London (strains M194, RT117, RT380, M68, M89 and 756) and (b) referred to the Scottish Gonococcal Reference Laboratory, University of Edinburgh Medical School between 1987 and 1992 (strains 1703–1707, 31493 and 97330). For antigen detection and sequencing isolates were grown on proteose peptone agar at 37°C in 5% (v/v) CO2.
2.2 Monoclonal antibodies and serotyping
Serotyping was performed by co-agglutination reaction using the panel of twelve mAbs (Genetic Systems) directed against epitopes on PI as previously described  and also with an alternative panel of mAbs supplied by Pharmacia . Isolates were also reacted in dot blots with the SM series of mAbs, which react with PI from either PIA or PIB expressing strains, and have been described in detail previously [8–10].
2.3 Sequencing of por genes
The sequences of the por genes encoding PI were determined following selective amplification of the gene by polymerase chain reaction (PCR) as previously described . Briefly, gonococcal lysates were subjected to PCR using oligonucleotides corresponding to the previously determined 5′ and 3′ termini of the por gene of gonococcal strain P9. Excess PCR primers were removed and the amplified DNA concentrated 2-fold using Magic™ PCR preps DNA purification system (Promega). The purified DNA was used in sequencing reactions with a Taq DyeDeoxy™ Terminator cycle sequencing kit (ABI) and the resulting products were analysed using a model 373 automated DNA sequenator (ABI). Sequencing of both strands of the por gene was accomplished with a set of custom synthesised oligonucleotide primers.
2.4 Sequence analysis
Sequence comparisons were carried out using DNAstar (Lasergene) and with the GCG sequence analysis programs using the BBSRC Seqnet facility.
3 Results and discussion
3.1 Immunological analysis of isolates
A total of 14 isolates were studied, which could not be serotyped by the normal co-agglutination procedures using the standard Genetic Systems mAbs. The isolates were also subjected to dot blot analysis with the SM series mAbs, which have previously been used for immunochemical studies of PIA and PIB [8–10]. Two patterns of reactivity were observed (Table 1), two isolates reacted with both PIA and PIB specific mAbs from both mAb panels while 12 isolates failed to react with any mAbs. However nine of the latter strains reacted with both IA and IB antibodies from the Pharmacia panel, showing the Av and Bx pattern which has previously been reported .
Serological properties of gonococcal isolates. Isolates were reacted in co-agglutination with the panel of serotyping mAbs (Genetic Systems)  directed against epitopes on PI and also with the SM series of mAbs, which react with PI from either PIA or PIB expressing strains [9, 10].
3.2 Sequence analysis of por genes
In order to investigate the structural basis for the anomalous serological reactivities, the por genes were sequenced from each of the isolates described above. The corresponding translated protein sequences were compared to the previously determined sequences from strains with defined serovars, both PIA and PIB (Fig. 1). A dendrogram of the relationships between the sequences (Fig. 2) showed that they fell into three distinct classes ‘intermediate’, ‘deletion’ and ‘hybrid’.
Comparison of predicted amino acid sequences of PI from gonococcal isolates and previously determined sequences of standard serovar strains. Boxes indicate the regions predicted to form surface exposed loops according to the model of Van der Lay et al. , residue numbers correspond to typical PIB (top) and PIA (bottom) sequences. Numbers in brackets indicate the number of additional strains with an identical sequence (strains 1703–1707 were identical to strain 1702; strains 9733 and 756 were identical to strain 31493). The sequences of the encoding por genes have been deposited in the GenBank database under the accession numbers AF015117–015122.
Dendrogram showing the relationship between gonococcal PI sequences. The sequences shown in Fig. 1 were compared with the UWGCG Pileup program using the BBSRC Seqnet facility. The dendrogram shows a representation of the sequence similarity in which the vertical branch lengths are proportional to the distance between the sequences.
Eleven isolates including those which failed to react with the Genetic Systems or SM mAbs, but which reacted with Pharmacia mAbs as AvBx, formed a distinct group. The protein sequences differed significantly from both PIA and PIB strains but were more closely related to PIA (Fig. 2), with loops 3 and 5 having the four and 17 fewer residues respectively which are characteristic of PIA. These ‘intermediate’ strains could however be readily distinguished from PIA strains by a single amino acid deletion in loop 1 and by unique sequences at the predicted apices of loops 6 and 7. This group could be divided by further minor sequence variations. Six epidemiologically linked isolates, which were referred to SNGRL in 1992 were identical, three unrelated isolates differed by a single amino acid and the two further strains by four additional positions.
The non-typeable strain, M194, was clearly more closely related to PIB than PIA (Fig. 2), showing most similarity with PIB from serovar IB-6. However it showed one major difference, a deletion of nine amino acids at the apex of predicted loop 5.
The two strains, RT117 and RT380, which showed cross-reaction with PIA and PIB mAbs from both the Genetic Systems and SM panels, were identical to each other and closely related to PIB from position 1 to 234. However at that point the region corresponding to loop 6 resembled PIA, possessing one less residue than the typical PIB and containing the DAKLTW sequence found in most PIA but absent from PIB.
3.3 Epitope analysis
The epitopes recognised by a number of the mAbs from the Genetic Systems and SM panels have been previously mapped by the use of synthetic peptides [8–10]. All of the mapped epitopes lie in the regions of the PI molecule predicted to form surface exposed loops. The epitopes recognised by most of the PIB mAbs lie at the apex of loop 5, SM20, SM21, SM22 and SM203 all recognise variable sequences based on a minimum of 192YEXQXY197, the widely cross-reacting SM24 and 3C8 recognise the adjacent conserved sequence 197YSIPS201 while 2D4 recognises the sequence 233KLYQNQLVRD243 present in loop 6 of a minority of strains. Several PIA epitopes also map to the apex of loop 6; SM100, SM102 and SM103 require the presence of the sequence 216WRND219 while 6D9 recognises 211DAKLTWRND219. Additional PIA conformational epitopes have been predicted by comparison of the sequences of reactive and non-reactive strains . Thus 4G5 requires the sequence 118IAQPEE123 from loop 3, 5G9 and SM100 require the loop 8 sequences 291GTEKF295 and 289GKGTEK294 respectively, and 5D1 appears to recognise a discontinuous epitope comprising the sequences 18VAYHG22 and 291GTEKF295 from loops 1 and 8.
The unusual antibody reactivities of the isolates used in the current study may be explained on these epitope specificities. The lack of reactivity of the ‘intermediate strains’ despite some similarity to PIA results from the sequence differences which are concentrated in loops 1, 3, 5 and 8. Thus the sequence ALID in loop 5 which is not found in any other strains, replaces the TWRN required for reaction of 6D9, SM100, SM102 and SM103; the single amino acid deletion in loop 1 and the substitution of GTEK by GTAK in loop 8 remove both the sequences associated with the discontinuous epitope for 5D1, while the GTAK also removes the sequence required for reactivity of both 5G9 and SM101. The identity of the epitopes recognised by the Pharmacia mAbs have not been reported, so the reason for the AvBx serovar of the intermediate strains using this system is not clear.
In contrast to the widely distributed differences between the ‘intermediate’ strains and PIA the lack of reactivity of strain M194 is due entirely to a single deletion from an otherwise typical PIB sequence. The nine amino acid deletion in loop 5 removes the epitopes for 3C8, SM20, SM21, SM22, SM24 and SM203. Loop 6 is retained but like many other PIB serovars does not contain the KYYQNQLVRD epitope required for 2D4 reactivity.
The apparent hybrid behaviour of strains RT117 and RT380 shown by the reactivity with a range of PIA and PIB mAbs can also be explained by sequence variations confined to loop 6. The typical PIB sequence before position 254 contains the conserved sequence 197YSIPS201 in loop 5 which is recognised by SM24 and 3C8 while the adjacent variable region contains 191YDNQSY196 which would not be recognised by SM20, SM21, SM22, SM24 or SM203. In comparison with a typical PIB, loop 6 contains the DAKLTWRN sequence required for reaction of the PIA specific mAb 6D9. This region also contains WRNN similar to the WRND recognised by SM100, SM102 and SM103. The observed reaction with SM100 and SM102 but not with SM103 appears to indicate a difference in the ability of these antibodies to accommodate the D to N change. This difference in sensitivity to a D to N change exactly parallels the effect of a mutation observed in a meningococcal class 1 porin which destroys the epitope seen by one mAb but not by another .
3.4 Origins of sequence diversity
Unlike several other gonococcal surface antigens PI does not undergo antigenic variation during the course of an infection. It therefore represents a major potential target for immune attack against the gonococcus. Evidence suggests that anti-PI antibodies provide serovar specific protection against gonococcal salpingitis  and also partial immunity to uncomplicated infection . Previous studies have demonstrated the occurrence of serovar specific variations in the predicted loop regions of both PIA and PIB, suggesting that these are the regions of the protein most accessible to the immune system.
In the present study, the unusual sequence variations found in each of three categories of strains are also all located in predicted surface exposed loops suggesting that immune attack provides a selection pressure for such variations. However, comparison of both amino acid and corresponding DNA sequences suggests that the three categories of isolates have arisen by distinct genetic mechanisms. Despite their reactivity with both IA and IB antibodies from the Pharmacia panel , the intermediate strains are not true hybrids but represent a distinct group which may have evolved, probably from PIA expressing strains (Fig. 2), over a period of time. In contrast, strain M194 appears to have arisen directly from a PIB expressing strain by a deletion arising from strand mispairing during replication. Comparison of the region of por genes encoding loop 6 of PI from M194 and the PIB-6 strain, shows that they are identical, with the exceptions of an out-of-frame deletion of 27 bases between positions 635 and 661 of strain M194, which generates a deletion of seven amino acids corresponding to positions 193–201 of the PIB-6 strain, and substitution of L for V at the position corresponding to 202 (Fig. 3A).
Comparison of por gene sequences showing the origin of atypical PI sequences. A: Comparison of regions corresponding to the PI loop 6 from the serovar IB-6 strain and ‘deletion’ strain M194. Numbers correspond to positions in por gene from IB-6. B: Comparison of por gene from ‘hybrid’ strain RT117 with corresponding regions of serovars IB-2 and IA-10. Asterisks show positions of identity between RT117 and the other two sequences, arrows indicate the putative origin of the hybrid sequence, numbers correspond to positions in por gene from IB-2.
The amino acid sequences of PI from the two strains which react with both PIA and PIB mAbs appear to be PIA/PIB hybrids although an alternative explanation could be that a two amino acid deletion from loop 6 of PIB would generate the DAKLTWRN sequence recognised by these antibodies (Fig. 1). However, examination of the corresponding por gene sequences reveals that these are true hybrid molecules which appear to have arisen from a double crossover between PIA and PIB (Fig. 3B). These strains show 99% and 98% homology respectively, with PIB-2, between residues 1–681 and 808–1044 at the 3′ end of the gene. In contrast, the intervening sequence 709–777 shows only 80% homology with IB-2 but 99% homology with PI from serovar IA-10. This variable region which encodes the loop 6 sequence containing PIA epitopes is flanked by two regions corresponding to codons 682–708 and 777–807 which show complete identity between por genes of PIA and PIB and hence represent likely sites for recombination.
Given the known ability of gonococci to be transformed by DNA from other strains and the likely occurrence of double infection by PIA and PIB expressing strains, it is surprising that occurrence of genuine PIA/PIB hybrids has not been previously documented. Such strains have also been constructed in vitro by transformation with chromosomal DNA  and by recombinant DNA methods with cloned por genes . The rarity of naturally occurring hybrids suggests that such strains must suffer a selective disadvantage in vivo. This would be in accord with the observation that strains expressing the recombinant por genes are particularly susceptible to the bactericidal effect of human serum .
Despite their relative rarity, the occurrence of the three classes of strains which exhibit anomalous serovar specificity demonstrates the potential of DNA based methods to provide greater epidemiological information than is available from conventional serovar analysis. The location of the sequence variations also reinforces the likely importance of the surface exposed loops of PI to immunity to gonococcal infection.
This work was supported by the Medical Research Council (J.E.H.) and by The Welcome Trust (C.A.I.).
(1990) Vaccination against gonorrhoea: the potential protective effect of immunization with a synthetic peptide containing a conserved epitope of gonococcal outer membrane protein IB. Vaccine 8, 225–230.
(1985) Comparison of two sets of monoclonal antibodies for the serological classification of Neisseria gonorrhoeae. In: The Pathogenic Neisseriae (SchoolnikG.K., BrooksG.F., FalkowS., FraschC.E., KnappJ.S., McCutchanJ.A., MorseS.A., Eds.), pp. 31–36. American Society for Microbiology, Washington, DC.
(1990) Gonococcal outer membrane protein PIB comparative sequence analysis and localization of epitopes which are recognized by type-specific and cross-reacting monoclonal antibodies. J. Gen. Microbiol. 136, 2165–2172.