INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

The genus Neisseria comprises a number of species associated with mucosal carriage and disease in animals and humans (3, 26); these species include Neisseria meningitidis, a major cause of meningitis and septicemia worldwide (6), and Neisseria gonorrhoeae, the etiological agent of gonorrhoea (31). Comparative studies of these two pathogenic species with others not frequently associated with disease, some of which are closely related (16), are helpful in understanding both the pathogenesis of meningococcal and gonococcal infections and the evolutionary events that led to their emergence. Such studies are particularly valuable when they target cell components potentially involved in immunity or other host-pathogen interactions, such as outer membrane proteins (OMPs) (35-37).

The Neisseria porins, a distinct class within the porin superfamily (20), are major components of the outer membranes of all members of the genus Neisseria (11, 35, 37). Most Neisseria species express only one, referred to as Por, the meningococcus exceptionally expressing two, PorA and PorB (17). The gonococcus is the only other Neisseria species known to have a porA gene, which is not expressed due to frameshift and promoter mutations (11). The meningococcal and gonococcal porins are targets for serological typing schemes (14, 15), candidates for inclusion in vaccines (13), and have been implicated in pathogenesis (29).

Interest in the antigenic variability of the Neisseria porins, from both typing and vaccination perspectives, has resulted in the generation of many sequences of their genes. Detailed interpretation of this database of sequences has been hindered by the lack of a molecular structure for the gene products, although a two-dimensional model has been proposed on the basis of sequence similarity (23, 36). This model comprises a -barrel porin structure, similar to that of the Escherichia coli porins (9), with regions that are relatively conserved among Neisseria sequences forming the  barrel. Other regions, which are less well conserved among and within species, form surface-exposed loops which protrude from the bacterial surface into the surrounding environment (23, 36). While this model was useful, it was not three-dimensional, being proposed before the molecular structures of the E. coli porins were established; the assignment of residues to membrane-embedded or surface-exposed environments was somewhat uncertain.

In this work, we have used molecular modelling techniques to generate a three-dimensional homology model of the Neisseria porins using E. coli porin structures, enabling us to identify more precisely the structural roles of individual residues. Phylogenetic analyses of the porins from different Neisseria species were enhanced when the model was used to assist sequence alignment and to identify structurally conserved parts of the proteins, identifying features likely to be important in the emergence of the pathogenic Neisseria species.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Neisseria strains and sequences from databases. The Neisseria porin sequences were obtained by translation of nucleotide sequences obtained by direct nucleotide sequence determination in this study or from GenBank, as detailed in Table 1. The porin of Neisseria polysaccharea was excluded from the analysis, as it was essentially identical to the Neisseria lactamica porin. Multiple examples of the porins were not included, with the exception of five additional N. gonorrhoeae porins classified elsewhere as "intermediate" and "hybrid "(8) and both Neisseria sicca porins, which were not closely similar.

Nucleotide sequence determination of Neisseria porin genes. Chromosomal DNA was extracted from each of the Neisseria strains by the Isoquick DNA extraction procedure (Orca Research). The porin genes were amplified with PCR primers 27 and 28 (12), purified as described previously (10), and their sequences were determined on both strands with Big Dye terminators (PE Applied Biosystems). The reaction products were separated with an Applied Biosystems Prism 377 automated sequencer, and the sequences were assembled with the Staden sequence analysis package (33).

Generation of PorA and PorB structural models. The initial model of the PorA monomer was generated by MODELLER (release 3) (30), a sophisticated software package that calculates protein three-dimensional structure based on sequence alignments, spatial restraints, and stereochemical considerations. The models for the Neisseria PorA and PorB sequences were constructed with two E. coli structures, OmpF (Protein Database [PDB] accession no. 2omf) and PhoE (PDB accession no. 1pho), as templates. All parameters within MODELLER were set to their default values and, in order to generate the trimer, the loop regions corresponding to Gly 23 to Ile 59, Ala 188 to Ser 221, and His 248 to Asn 281 (numbers as in Fig. 1) were removed from the model. The model was subjected to 50 steps of conjugate gradient energy refinement, as implemented within X-plor version 3.1 (4), to remove poor contacts. This step was performed by inserting the model for the PorA monomer into the OmpF unit cell (P321 a,b = 118.50 Å; c = 52.70Å), effectively generating the full PorA trimer by crystallographic symmetry. After refinement, coordinates for the full trimer were assembled with PDBSET, as implemented with the CCP4 suite of crystallographic programs (7).

View larger version (96K): [in this window] [in a new window]   FIG. 1.   Alignment of Neisseria and E. coli porins included in this study. Eco OmpF, E. coli OmpF protein; Eco PhoE, E. coli PhoE protein; the designations of Neisseria porin sequences are as given in Table 1. A question mark indicates a region of the protein for which the sequence is unknown, and a dash indicates an alignment gap. The regions of the E. coli porins known from structural data to form the  strands and  turns of the  barrel are shown by underlining and bold text, respectively. < and >, putative loop regions (I to VIII) of the Neisseria porins; *, regions corresponding to parts of the  barrel for which data were available for all of the Neisseria porins and were used in phylogenetic analysis of only structural regions.

Alignment of Neisseria porin gene sequences. The PorA and PorB models were used in the alignment of the translated porin genes (Fig. 1) with the program SeqLab, part of the GCG sequence analysis software package (Wisconsin Package version 9.1; Genetics Computer Group, Madison, Wis.). Alignments were performed manually so as to maintain the reading frame and to ensure that the  strands identified in the models of PorA and PorB were minimally disrupted. Many alignment gaps were required to accommodate the differences in the lengths of the porins, the majority of these occurring in regions of the proteins predicted by the model to form surface loops. As a consequence of high sequence diversity, many of the loop region alignments were essentially arbitrary.

Phylogenetic analyses. Nucleotide sequence alignments derived from the peptide sequence alignments shown in Fig. 1 were analyzed with MEGA (22), which was used to generate nonsynonymous distances, estimated by the method of Nei and Gojobori (27), with deletions excluded in pairwise comparisons. In some analyses, only sites within regions predicted to form the  barrel of the protein were included to minimize the effect of the large diversity present in the loop regions. The distance matrices obtained were used to reconstruct phylogenetic trees by the neighbor-joining method; these phylogenies were represented as unrooted radial trees with the program TREEVIEW (data not shown). The distance data were also visualized as split graphs generated by use of the split decomposition analysis technique with the program SPLITSTREE version 2.4 (18).

Nucleotide sequence accession numbers. The novel sequences determined in this study have been deposited in GenBank under accession no. AF121870 to AF121876 (Table 1).

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Novel porin sequences and alignments. The novel porin sequences obtained were consistent with those previously described for various Neisseria porins and are shown in Fig. 1, aligned with the sequences for the two E. coli porins with known three-dimensional structures and one example of each previously sequenced porin gene. The alignment was based on that of Jeanteur et al. (20), modified to take account of the additional structural information available (9). The initial alignment made was that of the E. coli porins and the meningococcal PorA and PorB proteins; this alignment was refined and the other Neisseria porins were added after construction of the structural homology models. Correct alignment of meningococcal PorA and PorB with the E. coli porins was critical for the generation of accurate models, and each putatively conserved residue in the E. coli crystal structures was examined to verify that there was a sound structural basis for its conservation between Neisseria and E. coli porins. The most striking feature of the alignment was that the Neisseria sequences began at the second  strand, within the -barrel framework, effectively truncating the first  strand, unlike the situation for E. coli. This alignment provided more convincing homology models and was also justified from sequence variations within the Neisseria porins that would otherwise have occurred within the -barrel framework.

Homology models. Given the high degree of sequence divergence between the Neisseria and E. coli porins, the -barrel framework of the modelled porins was constructed remarkably well. Both the PorA and the PorB models formed a well-defined main-chain hydrogen-bonding pattern characteristic of the -sheet motif, although no hydrogen-bonding or backbone dihedral-angle restraints were used in the calculation. Larger loop regions were less well modelled by this procedure and showed a variety of conformations, depending on the starting alignment and other restraints. Loop regions within proteins are notoriously difficult to model, and a more reliable calculation of their conformations requires further experimental data. The PorA monomer model is illustrated in Fig. 2.

View larger version (78K): [in this window] [in a new window]   FIG. 2.   Molecular surface of a porin eyelet from the PorA model. The PorA model was calculated as described in Materials and Methods, and the diagram was generated with MOLSCRIPT (21). The positions of the external loop regions are indicated as I, II, and so forth, with the exception of loop III, which lies inside the protein and is shown by the broken line. The positions in space of the beginning and end residues of the longer loops (I, IV, and V), which were too long to be modelled reliably and were excluded from the final model, are marked with a solid sphere at the position of the C atom in the residue.

View larger version (65K): [in this window] [in a new window]   FIG. 3.   Model of a PorA trimer, viewed from above. The positions of the external loop regions are marked as for Fig. 2, with the subunits designated a, b, and c. Other details are also as for Fig. 2.

Phylogenetic relationships of Neisseria porin genes. A number of phylogenetic reconstructions were made with nucleotide sequence alignments based on the peptide sequence alignments shown in Fig. 1. The final phylogeny, shown in Fig. 4, was reconstructed with the following considerations. Due to the strategy used in sequencing, some of the porin genes were truncated; regions corresponding to the first 12 and last 8 amino acids were missing (12, 23). These regions were therefore excluded from all sequences for the phylogenetic analyses. As these regions were relatively small and located in parts of the protein that are relatively conserved, the consequent loss of phylogenetic resolution was minimal. Given the large distances among sequences, nonsynonymous distances, estimated as described previously (27), were used. Further, only those regions of the gene encoding the  barrel were included to remove any distortions of the phylogeny introduced by signals from the rapidly evolving antigenically variable loops (32). Finally, split decomposition, which does not assume a tree-like phylogenetic structure (1), was used to ensure that any signal implying that recombination had occurred in the evolution of the Neisseria porin genes was identified.

View larger version (15K): [in this window] [in a new window]   FIG. 4.   Phylogenetic analyses of the Neisseria porins. A refined split graph was generated with a distance matrix obtained by the method of Nei and Gojobori (27) from nucleotide sequence alignments derived from Fig. 1, including only residues corresponding to the amino acids marked with an asterisk. Branch lengths are drawn to scale, and the fit parameter is shown.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

The antigenic variation of human pathogens remains incompletely understood, impeding vaccine development and complicating molecular typing, which relies on the characterization of such variation to distinguish isolates. An improved understanding of the complex interactions of the host immune system with variable antigens requires a multidisciplinary approach that integrates the inferences available from structural, immunological, and evolutionary data. At present, this integration is most readily achieved for protein antigens, where the relationships among genetic change in the pathogen, antigen structure, and interaction with the host immune system are relatively easily investigated. The porins of the genus Neisseria are an instructive example for a number of reasons, including their intrinsic interest, the large number of sequences available, and the fact that the genus includes both pathogenic and commensal species, whose lifestyle imposes constant selection pressure from the host immune response.

Neisseria porin antigens comprise regions of relatively conserved sequence, which are predicted to form the -barrel structure of the proteins, interspersed with more variable regions, which form the putative surface-exposed loops. With two exceptions, these loops are highly variable within and among the Neisseria species. These exceptions, putative loops II and III, have structural roles in the three-dimensional model. Putative loop II is important in monomer-monomer interactions within the porin trimer, while putative loop III is sequestered in the pore of each monomer, potentially influencing pore function (Fig. 3). The other loops, which exhibit diversity in both length and sequence, are the determinants of antigenic variability in species that have been extensively studied. Loops I, IV, V, and VI (Fig. 1) show the greatest length variation. The hypervariable regions of meningococcal PorA, VR1 and VR2 (24), correspond to loop regions I and IV of the model (36). In the monomer PorA model, these loops appear to be widely spaced (Fig. 2), making it unlikely that the hypervariable regions would come into direct contact across a single monomer. However, the view of the PorA trimer from above (Fig. 3) indicates that the proposed locations of VR1 and VR2 (loops I and IV) are in close proximity between adjacent monomers, raising the possibility that these regions may interact to some extent across the monomer-monomer interface. In principle, therefore, a PorA epitope could extend across more than one polypeptide chain in a PorA trimer.

Comparison of the pairing of  strands and the positioning of the loop regions present in the three-dimensional PorA model with those predicted in the two-dimensional model of van der Ley et al. (36) showed good agreement for the locations of loops I, IV, V, VIII and for the  strands, with the exception of strands 1, 2, 4, 5, and 14. The deviations observed were due at least in part to the fact that the lengths of the  strands vary in the OmpF crystal structure, illustrating the additional information that a three-dimensional structure can add to a model based on sequence alignments.

The variability of the loop sequences, which is likely to be the result of strong positive selection (32), may distort the phylogenetic signal present in the sequences. One result would be to make closely related porins appear more diverse. For example, a recent study proposed hybrid and intermediate porin classes for the gonococcal porins (8). Including porins from other species, limiting the analysis only to those parts of the protein that form the  barrel, and applying more sensitive phylogenetic methods to the alignments generated here show that these proteins cluster closely with other members of their respective porin classes, the apparent differences being introduced mainly by the highly variable loop sequences (Fig. 4).

The phylogenetic relationships of the porin genes from most of the commensal and animal Neisseria species (N. animalis, N. canis, N. cinerea, N. dentirificans, N. flavescens, N. flava, and N. sicca) were reconstructed as a "star" phylogeny (Fig. 4), with many lineages emerging near the root. This configuration may be due to rapid population growth during the divergence of the different Neisseria species, positive selection as they specialized into different niches, high rates of intraspecies recombination, or simply a lack of phylogenetically informative changes for the appropriate branching events. High rates of recombination among species are unlikely, given the sequence diversity of the porin genes. Selection is also an unlikely explanation given that the star phylogeny is still present when only the structural regions of the proteins are included. Furthermore, as the lack of resolution is present when all sites are included (data not shown) or just those from the -barrel-encoding regions (Fig. 4), it is unlikely that the star phylogeny is caused by the lack of a phylogenetic signal. We therefore propose that rapid population growth is the most likely explanation for this observation.

Whatever the cause of the star phylogeny, the porins of N. canis, N. elongata, and N. mucosa appear to share a common ancestor, occupying one branch of the star, as do those of N. denitrificans and N. animalis. The two N. sicca variants are diverse, perhaps indicating difficulties in the definition of this species. The two separate porin genes present in one N. flavescens isolate are sufficiently diverse that they may have been assembled in the same species relatively recently by horizontal genetic exchange.

Meningococcal PorA and gonococcal PorA are the most diverse members of the family of Neisseria porins and might constitute an outgroup relative to the other porins. The PorB porins of N. meningitidis and N. gonorrhoeae and the Por porin of N. lactamica (and the closely related N. polysaccharea), with the exception of meningococcal PorB2, form a well-resolved and supported bifurcating tree. The interrelationships of these porins indicate that interspecies genetic exchange of porin genes has occurred during the emergence of the present-day species. It is possible that the common ancestor of the meningococcus and the gonococcus possessed a gene ancestral to the porB1b and porB3 genes. The porB1b gonococcal porin gene appears to have arisen by the replacement of a porB1a allele with a gene sharing a common ancestor with the N. lactamica porin gene.

The meningococcal porB2 gene is a special case which appears to have arisen as a result of intragenic recombination during the evolution of the pathogenic Neisseria species. The split graph shows that PorB2 occupies an intermediate position between N. gonorrhoeae and N. meningitidis PorB and the N. lactamica porin on the one hand and the commensal-animal Neisseria porins on the other, with a network of relationships suggesting a phylogenetic signal that is inconsistent with a bifurcating tree. The data contributing to these relationships are illustrated in Fig. 5, which shows the amino acid identities within the -barrel-forming regions among PorB2, PorB3, and the N. flavescens Por2 protein (similar results were obtained in comparisons of other representative porins from the commensal-animal and pathogenic groups) (data not shown). PorB2 was most similar to the N. flavescens porin (19 amino acid identities in structural regions) but was also related to PorB3 (17 amino acid identities). However, while the identities between PorB2 and the N. flavescens porin were distributed throughout the sequence, those between PorB2 and PorB3 were mainly localized in  sheets on either side of loop II. This region includes the three lysine residues (positions 66, 82, and 113 in Fig. 1) putatively identified as being involved in the binding of ATP and GTP. This binding provides a gating mechanism that down-regulates pore size and changes voltage dependence and ion selectivity when PorB proteins are inserted into mammalian cells (29). In the PorB model, the side chains of these three residues protrude into the center of the pore, providing a plausible steric explanation of the role of GTP in obstructing the passage of solutes and changing the selectivity of the porin. As this gating mechanism, which is only observed in PorB2, PorB3, PorB1a, PorB1b, and the N. lactamica and N. polysaccharea porins, is conserved, it is tempting to suppose that PorB proteins with this characteristic were important in the evolution of the pathogenic Neisseria species, perhaps by promoting intracellular invasion and thereby improving carriage in humans. The fact that there is a network of interrelationships among PorB2, the porins of the pathogenic Neisseria species, and the porins of the commensal-animal group suggests that PorB2 may have evolved by recombination of two proteins ancestral to the pathogenic and commensal porins, resulting in an antigenically different porin that retained the GTP binding site.

View larger version (22K): [in this window] [in a new window]   FIG. 5.   Comparison of variable amino acids present in the -barrel-forming regions of the N. meningitidis PorB2 (Nme PorB2) and PorB3 (Nme PorB3) and N. flavescens Por2 (Nfl Por2) proteins. The numbers printed vertically above each residue in the alignment correspond to the position in the alignment shown in Fig. 1. The 18 residues which are identical between Nfl Por2 and Nme PorB2 are shown by black text on a grey background, the 16 residues which are identical between the two N. meningitidis porins are shown by white text on a black background, and the 11 residues which are identical between Nfl Por2 and Nme PorB3 are shown by white text on a grey background. The three lysine residues which form the PorB2 and PorB3 GTP binding sites are marked by asterisks.

While advances in automated sequence technology have enabled large numbers of nucleotide sequences of variable antigens to be sequenced rapidly, interpretation of these data continues to be limited by the lack of structural data, which remain comparatively difficult to obtain. The insights available from structural biology can enhance understanding of the evolutionary, immunological, and pathological significance of the gene sequence data. The approach taken in the present work, of combining experimental observations, structural modelling, and phylogenetic analyses, is valid, notwithstanding the wide sequence variation between the Neisseria and E. coli porins, as the -barrel porin structure is well conserved in a wide variety of species. The porin model presented here not only extends the two-dimensional models proposed earlier, enabling the design of further structural and immunological analyses, but also permits more accurate phylogenetic analyses, providing insights into the biochemical mechanisms that may have contributed to the evolution of the pathogenic Neisseria species.