|Home | About | Journals | Submit | Contact Us | Français|
Pathogenic meningococci have acquired a 24kb capsule synthesis island (cps) by horizontal gene transfer which consists of a synthetic locus and associated capsule transport genes flanked by repetitive Regions D and D’. Regions D and D’ contain an intact gene encoding a UDP-galactose epimerase (galE1) and a truncated remnant (galE2), respectively. In this study, GalE protein alleles were shown to be either mono-functional, synthesising UDP-galactose (UDP-Gal), or bi-functional, synthesising UDP-Gal and UDP-galactosamine (UDP-GalNAc). Meningococci possessing a capsule null locus (cnl) typically possessed a single bi-functional galE. Separation of functionality between galE1 and galE2 alleles in meningococcal isolates was retained for all serogroups except serogroup E which has a synthetic requirement for UDP-GalNAc. The truncated galE2 remnant in Region D’ was also phylogenetically related to the bi-functional galE of the cnl locus suggesting common ancestry. A model is proposed in which the illegitimate recombination of the cps island into the galE allele of the cnl locus results in the formation of Region D’ containing the truncated galE2 locus and the capture of the cps island en bloc. The retention of the duplicated Regions D and D’ enables inversion of the synthetic locus within the cps island during bacterial growth.
The Neisseria genus contains eleven species that colonize the oropharyngeal and urogenital mucosa of humans and of these, Neisseria gonorrhoeae and N. meningitidis, are of medical importance. N. gonorrhoeae colonises the mucosal surfaces of the urogenital tract in both males and females1. N. meningitidis is predominantly an opportunistic pathogen, asymptomatically colonising the mucosa of the nasopharynx of approximately 10% of the adult population2,3 but occasionally invades the host, resulting in septicaemia and meningitis4.
In both N. meningitidis and N. gonorrhoeae, interactions with host cells are modulated by the glycome on the bacterial surface which includes the lipooligosaccharide (LOS) and the glycosylation status of the type IV pilin5. Meningococci are distinguished from gonococci by the presence of a capsule polysaccharide synthesis (cps) island which encodes a capsule that facilitates systemic infection5. The cps island has a lower G/C-content than the core meningococcal genome consistent with acquisition via horizontal gene transfer (HGT) by recombination6. The general organisation of the cps island is as follows: Region A is responsible for the synthesis of the capsule polymer; Region B (ctrEF) and Region C (ctrABCD) contain genes responsible for transporting the capsule polymer to the bacterial surface while Region D and Region D’ contain an intact and truncated remnant copy of a gene encoding UDP-galactose epimerase, known as galE1 and galE2, respectively. Region E contains a gene of unknown function termed tex which is a homologue of a transcription factor. GalE is necessary for the synthesis of UDP-galactose (UDP-Gal) which is utilised in for the synthesis of capsule polymers from serogroup W/Y, LOS and protein glycosylation5.
Non-pathogenic meningococci frequently possess a capsule null locus (cnl) with a single copy of Region D with an intact galE and Region E (Fig. 1)7. Evolutionary studies have suggested that meningococci acquired the cps island from potential donors such as Pasteurella multocida and Haemophilus influenzae through mosaic HGT events8,9,10. Although the ancestral events leading to the formation of the cps island are unclear, serogroup switching can result in the replacement of the entire synthetic Region A in instances where serogroups B or C switch to serogroups A, W or Y (Fig. 1) indicating recombination events in this region continue to occur11,12.
Gonococci also possess a cnl locus (Fig. 1), and have the additional ability to synthesise UDP-N-acetyl galactosamine (UDP-GalNAc) which is used to terminate the non-reducing terminus of LOS13. In meningococci, GalNAc is not a component of LOS5,14, but is required for the synthesis of meningococcal serogroup E and Z capsules15,16. Although many of the biosynthetic pathways of Neisseria spp. have been elucidated, the mechanism by which UDP-GalNAc is synthesised has not been proposed. Bacteria synthesize UDP-GalNAc via either a bifunctional UDP-galactose 4-epimerase (GalE) or a UDP-GalNAc 4-epimerase (GNE). Since Neisseria spp. do not harbour a GNE homologue, the neisserial GalE1 could be a bi-functional epimerase that synthesizes both UDP-Gal and UDP-GalNAc from the substrates UDP-glucose (UDP-Glc) and UDP- N-acetylglucosamine (UDP-GlcNAc), respectively.
This study examined the functional diversity of the galE locus and found that Neisseria spp. possess both mono-functional and bi-functional UDP-galactose epimerases. While N. gonorrhoeae and meningococci with a cnl locus possess bi-functional galE alleles which are phylogenetically related, meningococci carrying the cps locus possessed both bi- and mono-functional alleles for galE1 and galE2. The phylogenetic relations of the galE alleles and their associative properties with serogroup and clonal complex has provided further evidence for the hypothetical model of the recombination events during HGT of the cps island.
The distribution of galE1 alleles (NEIS0048) across the genus Neisseria was assessed using a defined set of 194 isolates for which genomic data was available in the PubMLST database (www.pubmlst.org/neisseria, Supplementary Table S1). A total of 107 unique galE alleles, comprising 65 galE1 alleles from meningococci and 42 galE alleles from the other species, represented the top 24 most commonly occurring NEIS0048 alleles within the PubMLST database (>0.38% occurrences). An un-rooted neighbour-net tree (Fig. 2) revealed that the alleles formed two main clusters. Cluster A contained encapsulated meningococci (p-distance=0.092), while cluster B contained N. gonorrhoeae and the non-pathogenic Neisseria species, N. lactamica and N. polysaccharea as well as unencapsulated meningococci possessing the cnl locus (overall mean p-distance=0.063) (Fig. 2). Several other less distinct clusters with deeper roots were observed which corresponded with N. subflava, N. oralis, N. elongata, and N. cinerea. Two meningococcal galE1 variants belonging to serogroup E and Z meningococci were found on the same branch as those from N. animalis and N. weaveri which were more distantly related to Cluster A and B (overall p-distance=0.266). Overall, the distribution of the galE1 alleles into at least two distinct clusters suggest that these alleles are under diversifying selection which may be related to function between species.
To investigate whether NEIS0048 alleles from each phylogenetic cluster had different functions, the galE1 of N. gonorrhoeae strain FA1090 (GalE1 allele 17=GalE_17) from cluster B and N. meningitidis strain MC58 (GalE1 allele 2=GalE_2) from cluster A were cloned into pET15b to create GalE_17::Hisx6 and GalE_2::Hisx6. GalE_17::Hisx6 could epimerise both UDP-Glc and UDP-GlcNAc indicating that this variant is bi-functional, with the equilibrium favouring the production of UDP-Gal and UDP-GalNAc (Table 1). In comparison, GalE_2::Hisx6 would only accept UDP-Glc as a substrate and could not epimerise UDP-GlcNAc indicating that this enzyme is mono-functional. Unlike the gonococcal GalE_17 allele, the meningococcal GalE_2 had an equilibrium favouring the formation of UDP-Glc at 63% of all products.
Previous investigations of known mono-functional and bi-functional GalE proteins have shown that a single amino acid residue is the main factor in the ability of the enzyme to possess singular or dual modes of activity. Thoden et al.17 compared the GalE protein from H. sapiens (Group 2) to the E. coli (Group 1) homologue and demonstrated that four conserved residues hold the sugar in a productive conformation (Supplementary Fig. 1) while a fifth residue affected whether UDP-Glc or UDP-GlcNAc could enter the active site cleft. In E. coli GalE and the human GalE, this position is occupied by a tyrosine (Y299) residue or a cysteine (C307), respectively. When the tyrosine residue of E. coli GalE was mutated to cysteine (Y299C) it switched the native enzyme from a mono-functional to a bi-functional mode while the converse occurred in Yersinia enterocolitica GalE18.
To understand the contribution of amino acid residues to the active site of GalE, 107 galE alleles from ten Neisseria spp. (Supplementary Table S1) were translated and aligned using CLUSTALW219. As expected, residues K85, S125, Y150 and N180 which correspond to the four residues co-ordinating the pyranose ring in the active site were absolutely conserved (Supplementary Fig. 1). The fifth position was occupied by a serine (S299) residue in gonococcal GalE amino acid sequences and phenylalanine (F300) in most meningococcal alleles. Since the serine and phenylalanine residues are similar in size to cysteine in bi-functional enzymes and tyrosine in mono-functional enzymes, respectively, we proposed that these residues govern the functionality of GalE. Using PHYRE220, homology models using GalE amino acid sequences of N. gonorrhoeae strain FA1090 and N. meningitidis strain MC58 were built. These models and the crystal structures of the E. coli and human enzymes in the presence of NADH and UDP-GlcNAc are shown in Fig. 3. To address whether variability within the amino acid sequences of neisseria GalE proteins contributed to the binding site pocket, 38 non-conserved residues derived from the alignment of 103 GalE1 alleles (excluding the four most distantly related alleles from N. weaveri GalE_115, N. animalis GalE_110, and N. meningitidis GalE_14 and GalE_15) were mapped to the predicted structure of GalE_2 from N. meningitidis MC58 (Fig. 3A). The non-conserved residues were localized predominantly to the external surface of the protein and were distant from the active site. Conservation, both in structure and in ligand binding residues, was evident for all four enzymes with the exception of positions F300 in meningococci and S299 in gonococci. Comparison of the structures suggests that when a small residue is present (C307 in humans or S299 in gonococci), both the UDP-GlcNAc and UDP-Glc are able to bind productively and hence the enzyme is bi-functional (Fig. 3C,E). In contrast, in the presence of an aromatic side chain (Y299 in E. coli or F300 in meningococci), the UDP-GlcNAc does not bind correctly in the active site to allow efficient hydride transfer from the sugar moiety to the nicotinaimide and, thus the enzyme is mono-functional for only UDP-Glc (Fig. 3B,D).
To test the role of S299 and F300 in determining substrate specificity of neisserial GalE enzymes, GalE_17(S299F)::Hisx6 and GalE_2(F300S)::Hisx6 mutants were constructed using site-directed mutagenesis and these recombinant enzymes were purified. The GalE_17(S299F)::Hisx6 lost the ability to epimerise UDP-GlcNAc while retaining the ability to convert UDP-Glc to UDP-Gal, although the equilibrium was altered to favour the formation of UDP-Glc compared to the wild-type enzyme (from 39% to 59%). Meningococcal GalE_2(F300S)::Hisx6 gained the ability to epimerise both UDP-Glc and UDP-GlcNAc although the observed equilibrium resulted in less UDP-GalNAc for GalE_2(F300S)::Hisx6 than gonococcal GalE_17::Hisx6 (Table 1).
A Neighbor-Joining phylogenetic tree of 89 GalE1 amino acid sequences (Supplementary Table S1, Supplementary Fig. 2) revealed at least three major lineages Clades 1, 2, and 3, in addition to two sets of deeply branched outliers. One group of outliers contained alleles from N. weaveri, N. animalis with serogroup E and serogroup Z expressing meningococci. The GalE1 alleles from the serogroup E and serogroup Z expressing meningococci both have a cysteine in the active site cleft consistent with the proposed bi-functionality of GalE1 and the requirement for UDP-GalNAc for capsule synthesis. The GalE from N. weaveri also contained a cysteine in the active site cleft and was predicted to be bi-functional. The GalE proteins from N. animalis and N. subflava were both predicted to be mono-functional due to the presence of a tyrosine or phenylalanine residue, respectively, in the active site cleft.
Clade 1 containing GalE amino acid alleles from N. elongata and N. mucosa, was predicted to be bi-functional due to the presence of either a valine or a cysteine residue in the active site cleft. Clade 2 contained alleles from N. cinerea, N. polysaccharea, N. lactamica, N. gonorrhoeae, N. bergeri, and five meningococci. All of these were assigned bi-functionality due to an active site serine residue characteristic of the bi-functional alleles of N. gonorrhoeae. GalE alleles 16, 24, 37, 129 and 166 were present in meningococcal cnl isolates. Clade 3 contained only GalE1 amino acid sequences from N. meningitidis which were either mono- (contained a phenylalanine in the active site) or bi-functional (contained a serine or cysteine in the active site). Two distantly related protein alleles from N. meningitidis, GalE_41 and GalE_34, from a serogroup A and a serogroup C isolate, respectively, were most similar to protein alleles from N. subflava. The bi-functional alleles in clade 3 were associated with cc174 isolates expressing serogroup Y capsules (allele 131), a variety of serogroup B isolates from various genetic lineages (alleles 46, 49 and 236), a cc1 isolate expressing serogroup A capsule (allele 50), and serogroup I and K expressing isolates (allele 4 and 5 respectively). GalE_236 was selected as a representative of the bi-functional alleles in clade 3 for assessment of epimerase functionality. It was cloned, expressed as a His-tagged protein, assessed by HPLC and was shown to be bi-functional (Table 1).
In addition to Region D containing the galE1 allele, encapsulated meningococci also possess the inverted repeat Region D’ which contains a truncated remnant galE2 in an operon with rfbBAC’ (Fig. 1). The galE2 pseudogene has lost the first 402 nucleotides but retains the last 615 nucleotides, containing the active site motif which can be used to assign ancestral functionality. To unambiguously assign a function to galE2 alleles, galE1 and galE2 genes were manually curated in 1196 meningococci representing 10 common clonal complexes. This yielded 65 galE1 alleles and 24 galE2 alleles (data not shown). Of the 65 galE1 alleles in 1196 isolates, 75% were associated with mono-functionality (F300) with the remainder being bi-functional (S300). The GalE1 alleles from serogroups B, W and Y were predominantly mono-functional (Fisher’s exact test, two-tailed, p<0.05). However, no significant association of GalE1 functionality was noted for serogroups A and C although the trend in this small dataset was towards mono-functionality (Supplementary Table S2). Bi-functional GalE1 alleles were associated with all serogroup E expressing isolates (Fisher’s exact test, two-tailed, p<0.0001) due to the requirement for UDP-GalNAc for the synthesis of this capsule polymer. Of the 24 galE2 alleles, 71% were associated with bi-functionality with the remainder being mono-functional. Overall, 92.4% of meningococcal strains carried a mono-functional galE1 and a bi-functional remnant galE2 allele. There is a strong retention (Fisher’s exact test, two-tailed, p<0.05 for each serogroup) of separate ancestral functionality between galE1 and galE2 alleles in meningococcal isolates that expressed capsules other than serogroup E.
Since meningococcal galE1 and galE2 are highly conserved at the nucleotide level, homologous recombination between the two genes may randomly switch functionality between galE1 and galE2 loci. To examine this further, recombination events per mutation (r/m) was used to calculate the number of recombination events relative to mutations (ρ/θ) and the recombination events relative to the number of mutations within galE1 and galE2 with respect to other cps island regions (Supplementary Table S3, Fig. 4). Recombination events occurred twice as often as point mutations in galE1, 1.5 times as often in ctrA-D and tex while recombination and mutation equally contributed to diversity within ctrG, galE2 and ctrEF. Diversity within NEIS0044, located on the boundary of Region D containing galE1, was equally driven by recombination and mutation, whereas diversity of NEIS0069 on the boundary of Region B was mostly caused by mutation (Fig. 4A). When recombination rates were corrected for local block boundaries, galE1, tex and galE2 had the highest recombination rates relative to the other genes in the cps locus (Fig. 4B). Therefore, although recombination does contribute to the diversity of galE1 and galE2 alleles, the exchange of functionality between the two loci is infrequent since the functional separation of the two loci is maintained in the clonal complexes that do not possess serogroup E.
The origin of Region D’ containing galE2 is currently unknown. While the majority of encapsulated meningococci contained galE2 with a motif consistent with an ancestral bi-functional activity, most cnl meningococci from various genetic lineages including cc53 and cc198 possessed bi-functional alleles (GalE_37, GalE_16 and GalE_116, GalE_161, GalE_281, GalE_362 alleles, GalE_378, and GalE_397). An alignment of the conserved portions of the galE1 and galE2 alleles revealed a bi-furcated phylogeny with the bi-functional galE1/galE2 alleles and mono-functional alleles forming separate clusters (Fig. 5). Interestingly, the galE alleles from cnl meningococci clustered with the bi-functional galE2 alleles. This observation generated the hypothesis that the galE2 of Region D’ in pathogenic meningococci is a remnant of galE from the original cnl locus of the ancestral recipient genome during the capture of the cps island. The cnl site is always located between two boundaries: NEIS0044 upstream of the galE1-rfbBAC cluster and NEIS0069 downstream of tex (Fig. 1). The capture of the cps island was postulated to lead to the deletion of the 5′ end of galE-tex- NEIS1357 in the cnl locus to create NEIS0044-rfbCAB’-galE2 linked to the cps island ending at the downstream boundary of NEIS0069. In this model, the cps island HGT event would require the introduction of Region A-C-E-D-B by recombination into the galE of the cnl locus.
One outcome of this model would be the expectation of strong linkage disequilibrium between the two galE loci with a strong association with serogroup as these would be linked en bloc during the HGT event. A plot of the association of galE1 and galE2 allelic pairs with serogroup in 1196 isolates representing each of the 10 clonal complexes showed that each clonal complex was characterized by a predominant galE1/serogroup/galE2 combination (Supplementary Fig. S3). In summary, 84.2% of cc1 is represented by GalE139/GalE230, 74.2% of cc11 is represented by GalE110/GalE28, 46.6% of cc22 is represented by GalE18/GalE213, 88.8% of cc23 is represented by GalE127/GalE223, 73.2% of cc41/44 is represented by GalE120/GalE213, 65.3% of cc60 is represented by GalE110/GalE213, and 78.4% of cc269 is represented by GalE1120/GalE27. A strong associative property with clonal complex was observed for galE1/galE2 allelic pairs (Cramer’s V coefficient, V=0.985), tex (V=0.963), ctrE (V=0.948) and ctrF (V=0.770) suggesting that the correlation of alleles in the cps island with a specific clonal complex was formed independently during the acquisition of the cps island into the founder of each clonal complex.
This hypothetical model for the acquisition of the cps island in modern clonal complexes also predicted that the original organization of the cps island upon acquisition could have been NEIS0044-Region D’ (containing galE2 bi-functional)-E-C-A-D (galE1 mono-functional)-B. However, the most common arrangement found in closed genomes of strains MC58 and FAM18 has NEIS0044 followed by Region D containing galE- A- C- E- D’- B- NEIS0069 (Fig. 6) which suggests that the synthesis cassette, A-C-E, may undergo inversion in this site. Colony PCR confirmed that both cps island arrangements were detectable in plate grown cultures of strains MC58, FAM18 and NMB (Supplementary Fig. S4). Complete sequencing of these PCR products did not detect recombination events within the galE1 and galE2 genes which is consistent with potential breakpoints in the flanking rfbBAC gene clusters of Region D and Region D’. However, since the paired rfbBAC and rfbBAC’ regions in each strain was almost identical, the recombination site was not detected.
Mustapha et al.12 have recently described the recombination events detected upon the integration of the serogroup W cassette into the cc11 lineage which originally expressed serogroup C (Fig. 1). They showed that the left hand boundary for the integration of Region A for serogroup W synthesis occurred in galE1 while galE2 remained unchanged. To examine whether this is a common occurrence in serogroup switching, the association of the galE1/galE2 allelic pairs with serogroup was plotted (Supplementary Fig. S3). The dominating serogroup in each lineage is strongly associated with only one galE1/galE2 allelic pair. For example, 93.5% of cc11:serogroup C isolates possessed GalE1_27/GalE2_8 allelic pair whereas 100% of cc11:serogroup W isolates possessed GalE1_10/GalE2_8. Capsule switching events which involved the acquisition of serogroup B into cc11 displayed a similar pattern with 75% retaining GalE2_8 while the allele of GalE1 with Region A had been changed. Similarly, 16 events in cc60:serogroup E strains involving the acquisition of serogroup B synthetic regions showed a similar pattern of variation in GalE1 with conservation of the GalE2 alleles (Supplementary Table S4). Although the library of 1196 isolates contained 65 unique GalE1/GalE2 allelic pairs, only four GalE1/GalE2 pairs were found in multiple clonal complexes (Supplementary Fig. 3) suggesting that these exchanges had recombination boundaries outside of the galE1/galE2 pairs.
The acquisition of the cps island by HGT into an ancestral non-pathogenic meningococcal lineage carrying a cnl locus similar to cc53 and cc198 isolates of N. meningitidis has been proposed by Schoen et al.6 and others21,22,23,24,25. This hypothetical model proposes that the modern arrangement of the cps island is the outcome of at least two recombination events that resulted in the capture of Region A-C and Region B, located upstream and downstream of Region E (tex), respectively, in the capsule null locus of non-pathogenic meningococci8,9,10,26. In this study, we provide evidence for a second hypothetical model which suggests, that once the ancestral cps island was formed, this structure became the donor sequence for en bloc transfer into the modern clonal complexes of N. meningitidis. Evidence for this hypothetical model is based upon the observation that the galE1 and galE2 alleles are phylogenetically distinct due to different enzymatic functions, and that these phylogenetic relationships can be traced through capsule null loci and in cps islands in the various clonal complexes of N. meningitidis.
Gonococci have a requirement for UDP-GalNAc which is a component of the gonococcal LOS27. The gonococcal GalE epimerase was shown to be bi-functional with the capacity to synthesise both UDP-Gal and UDP-GalNAc whereas the meningococcal GalE could only produce UDP-Gal. Exchanging a pivotal amino acid in the active site cleft (F300 in meningococci and S299 in gonococci) enabled the exchange of mono- and bi-functionality of the enzymes suggesting that the identity of the amino acid positioned at the mouth of the active site cleft determined the functional phenotype of the epimerase. Ishiyama et al.28 proposed that changing the amino acid at this position of the active site cleft of GalE epimerase determined the ability of the UDP-linked N-acetyl-sugar to rotate and undergo epimerisation. Although there are other non-conserved amino acid positions within neisserial GalE variants, none of these were shown to influence the cleft leading to the active site (Fig. 3). Therefore, the use of the six amino acid residue motif surrounding the active site cleft residue of F300 (meningococci)/S299 (gonococci) or equivalent in commensal Neisseria spp. was considered sufficient to assign functionality to the galE alleles.
Functional and phylogenetic analysis of galE1 revealed that all Neisseria spp. possessed the gene, but that the functionality of the epimerase was predicted to vary across species (Fig. 2). Although the role of UDP-Gal and UDP-GalNAc is well described for the pathogenic Neisseria spp., very little is known regarding other species. Immunotyping studies have proposed that N. cinerea, N. polysaccharea and N. lactamica, which possess putative bi-functional galE1 alleles, synthesise a LOS containing Gal and GalNAc residues but no structural studies have been done to confirm this29,30,31. The bi-functional meningococcal galE1 alleles (Supplementary Fig. S2), were found in the presence of serogroup A, B, E, Y, I and K synthetic cassettes. A representative of this group, GalE_236 allele, had an equilibrium similar to that of the bi-functional allele of N. gonorrhoeae, thereby suggesting that the synthesis of UDP-GalNAc is the favoured outcome for this reaction (Table 1). This allele was found in N. meningitidis strain NMB which was shown to synthesise UDP-GalNAc by Lee et al.32. Although UDP-GalNAc is not present in LOS of this isolate33, it may be involved in another synthetic pathway such as protein glycosylation which has been recently detected in commensal Neisseria spp.34,35. However, since the GalNAc lipopolysaccharide transferase, lgtD, has been found in some isolates of meningococci, the capacity to add GalNAc to LOS could occur if these strains also possessed a bi-functional galE136. The phylogeny of the meningococcal galE alleles (Fig. 5) suggests that bi-functional galE1 alleles found in serogroups A/B/C/Y/W may have arisen sporadically through recombination with the galE2 remanent or through point mutation of the active site of the galE1 allele (Fig. 4). In contrast, the bi-functional galE1_14 allele found in serogroup E- expressing isolates is phylogenetically distant from these bi-functional galE1 alleles (Fig. 5) suggesting a separate heritage potentially with the closest relative, N. weaveri (Fig. 1).
The phylogenetic relationship of the bi-functional galE2 alleles from pathogenic meningococci and the galE alleles in the cnl of non-pathogenic isolates led to the hypothesis that Region D’ was the original locus into which the cps island recombined (Fig. 6A). Since statistical association analysis demonstrated linkage disequilibrium between galE1 and galE2 as well as galE1/galE2 pairs within each clonal complex, this suggested that the pairing of galE1 with galE2 occurred upon the acquisition of the cps island into the cnl locus of the ancestral strain of each clonal complex. One event that could result in this arrangement is the illegitimate recombination of an ancestral cps island consisting of Region A-C-E-D-B into galE on the left boundary and NEIS0069 on the right boundary of the recipient cnl locus (Fig. 6A). The strong associative properties of the galE1/galE2 (Region A), tex (Region E) and ctrEF (Region B) alleles with clonal complex supports the concept of en bloc transfer and provides an explanation for the creation of Region D’ due to an illegitimate recombination event that truncates the galE allele of the cnl locus to become the remnant galE2 locus (Fig. 6A).
The initial organisation of the cps locus predicted by this model is NEIS0044-region D’(galE2)- E- C- A- (galE1)D- B- NEIS0069 which is the same as that of the closed genome of strain B194023. However, the closed genomes of strains MC58 and FAM18 are reported as NEIS0044-region D (galE1)- A- C- E- (galE2) D’- B- NEIS0069, suggesting that an inversion of the A-C-E block between region D and region D’ could occur. This was confirmed by directional PCR of the cps locus using plate grown strains MC58 and FAM18 (Supplementary Fig. S4). Since single colony PCR revealed that each colony contained a mixture of cps orientations (data not shown), it appears that this process is dynamic, occurring during chromosomal replication when single stranded DNA is present for RecA mediated homologous recombination (Fig. 6B). Unless the duplicated Region D’ is deleted, the locus is not able to be fixed in any particular orientation. Colony immunoblots of strain NMB which is not phase variable for capsule expression did not detect any mixed phenotypes consistent with previous studies on the expression of capsule by this strain (data not shown37,38,39). It is interesting to note that the cps island is inserted 54kb downstream of the origin of replication which is surrounded by a number of genomic inversions40. In E. coli, Ivanova et al.41 have recently shown that inversions surrounding the origin of replication occur to resolve collisions between highly transcribed genes and the direction of chromosomal replication, thus improving fitness. It is currently unclear whether this mechanism applies to this situation in N. meningitidis but is worth considering in future investigations of this region.
The proposed en bloc model for the transfer of the cps island into meningococcal clonal complexes relies upon the concept previously advanced by Schoen et al.6,7,42 that the mosaic structure is created in an ancestral meningococcal isolate. Recently, Harrison et al (personal communication) have detected similar cps islands consisting of regions A-C-E-D-B, and lacking Region D’, in commensal Neisseria spp. which could also be a source of the donor element for the en bloc transfer hypothesis. Currently the PubMLST database does not contain meningococcal isolates lacking Region D’ as the majority of isolates represented in this collection have been isolated from cases of disease. It will be interesting to examine a larger collection of non-disease causing meningococcal isolates for the presence of strains bearing incomplete cps islands as these may provide further evidence for the model. This model was also used to examine serogroup switching, which from the observations of Mustapha et al.12 disrupt the linkage disequilibrium of galE1 and galE2. An examination of other serogroup switching events in the database indicated that this was a general phenomenon associated with this HGT event.
In conclusion, we propose evidence for a hypothetical model in which the acquisition of the cps island via HGT results in the loss of UDP-GalNAc synthesis from UDP-GlcNAc in most pathogenic meningococcal isolates, except for those expressing serogroup E capsules. UDP-GlcNAc is an entry metabolite for the synthesis of cell wall components, peptidoglycan and lipid A, in addition to sialic acid found in serogroup B, C, W and Y capsular types. Presumptively, the inactivation of the bi-functional galE allele of the cnl locus and replacement with a mono-functional galE1 allele releases a pool of UDP-GlcNAc that can be redirected into the sialic acid capsule biosynthesis pathway while minimising the metabolic fitness cost associated with the acquisition of the cps island in pathogenic meningococci. However, other accessory changes to central metabolism are likely to be required. Mustapha et al.12 noted that the capsule switching event in cc11 from serogroup C to W included central metabolic genes such as the pyruvate kinase. In addition, Schoen et al.43 found that the central metabolism of pathogenic meningococci is diverse and that multiple adaptive changes to metabolism have occurred thus providing a framework in which to test our theory regarding the fitness costs associated with the acquisition of the cps synthesis pathway.
Meningococcal strains were cultured under aerobic conditions with 5% CO2 at 37°C on GC agar (GCA) or GC broth (GCB) (Oxoid) supplemented with 0.4% glucose, 0.01% glutamine, 0.2mg of cocarboxylase per litre, and 5mg of Fe(NO3)3 per litre. The wild-type strains and constructed mutants used in this study are shown in Supplementary Table S5. Antibiotic selection for meningococcal mutants was performed on GCA containing 100μg/ml of kanamycin (sulfate salt), 60μg/ml of spectinomycin, 5μg/ml of tetracycline or 2μg/ml of erythromycin (Sigma). Escherichia coli DH5α was used as a host for all DNA manipulations and Rosetta™ (DE3) as a host for all protein expression. E. coli strains were routinely grown on Luria-Bertani broth (LBB) and agar (LBA, Oxoid) which, where appropriate, was supplemented with antibiotics at the following concentrations: ampicillin at 100μg/ml, spectinomycin at 50μg/ml, kanamycin at 50μg/ml, erythromycin at 300μg/ml, tetracycline at 12.5μg/ml and chloramphenicol at 30μg/ml (Sigma).
The plasmids constructed and used in this study are listed in Supplementary Table S5. Genomic DNA was isolated from N. gonorrhoeae strain FA1090 and N. meningitidis strains NMB and MC58 using the PurelinkTM genomic DNA purification kit (Invitrogen) according to manufacturer’s instructions. The gene galE was amplified with the primers KAP580 (5′-GGAATTCCATATGAAAAAAATTCTCGTTACCG-3′) and KAP581 (5′-CGCGGATCCTTAATCGTCGTAGCCATTCGG-3′) for MC58, KAP582 (5′-GGAATTCCATATGACCGTCCTGATTACCG-3′) and KAP583 (5′-CGCGGATCCTTAATCCCCATATCTGCCG-3′) for FA1090 and KAP563 (5′GGAATTCCATATGCCCTATACGGAAGATATG-3′) and KAP564 (5′-CGCGGATCCTTAATCCCCATATCCGTTGGG-3′) for NMB, such that there was a 5′ NdeI site overlapping the start codon and a 3′ BamHI site downstream of the stop codon incorporated into the PCR fragment. Each PCR product was directionally cloned into the NdeI and BamHI sites of pET15b, resulting in the 5′ fusion of the open reading frame with a hexahistidine motif, resulting in plasmids pCMK730, pCMK729 and pCMK771. The galE of N. gonorrhoeae and N. meningitidis that have been cloned into pET15b were sequenced and their sequence found to be identical to the native gene.
The cloned galE of N. gonorrhoeae strain FA1090 and N. meningitidis strain MC58 were mutagenised using site directed mutagenesis to change a single nucleotide. The primers KAP427 (5′-CGACTTGGCGTGTTTCTATGCCGACCC-3′) and KAP428 (5′-GGGTCGGCATAGAAACACGCCAAGTCG-3′) to mutate the gonococcal galE, and KAP429 (5′-GGTGATTTGGCGTGCTCCTATGCCGACCC-3′) and KAP430 (5′-GGGTCGGCATAGGAGCACGCCAAATCACC-3′) to mutate the meningococcal galE were used to amplify the plasmids pCMK729 and pCMK730 respectively using the non-strand displacing polymerase Phusion (New England Biolabs). The template DNA was digested with DpnI, prior to transformation into DH5α. The resulting plasmids were sequenced and confirmed to contain only the introduced mutation. These plasmids were called pCMK733 and pCMK734 respectively.
Expression strains were constructed by transforming the plasmids pCMK729, pCMK730, pCMK771, pCMK733 and pCMK734 into BL21-DE3 Rosetta. Inducible protein expression from these strains were confirmed by growing the strains to mid-log phase in 10ml LBB, splitting the culture in two, and inducing one culture with 0.3mM IPTG for 1hr. The cells from 1ml of each culture was collected by centrifugation and resuspended in 100 μl of distilled water to which 25 μl of 5x Laemelli loading buffer was added. The samples were boiled for 10mins prior to separation on 12% SDS-PAGE at 150V, then stained with Coomassie Blue R250. Expression was determined by the presence of a band of increased intensity present in the induced sample compared to the un-induced sample at the correct molecular weight.
Expression of His-tagged GalE was confirmed by Western immunoblot. Whole cell lysates (750ng) were separated by 15% sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) by standard methods and transferred to nitrocellulose membranes. The membranes were blocked overnight with 2% BSA in TBS, then incubated with the monoclonal mouse anti-His IgG (Sigma) primary antibody at 1:1,000 dilution. Horse radish peroxidase-conjugated anti-Rabbit IgG and anti-Mouse IgG secondary antibodies (Santa Cruz Biotechnology) was used for detection and the membrane was developed with an ECL kit (GE Healthcare).
E. coli strains expressing meningococcal GalE proteins (both native and mutagenised) were grown overnight with shaking in 10ml LBB containing 100μg/ml ampicillin. An aliquot of 100μl was used to inoculate 10ml LBB containing 100 μg/ml ampicillin the following day and the strains grown to OD600 0.6–0.8 at which time protein expression was induced by the addition of IPTG to a final concentration of 0.3mM. Induction occurred for 3hr at 37°C prior to the purification of protein using the Wizard HisLinkTM Spin Protein Purification System (Promega) as per manufacturer’s instructions. The gonococcal GalE could not be purified in this way due to the formation of inclusion bodies. Strains expressing the gonococcal GalE proteins were grown overnight with shaking in 10ml LBB containing 100μg/ml Ampicillin. An aliquot of 1ml was used to inoculate 1L LBB containing 100μg/ml Ampicillin the following day and the strains grown to OD600 0.6–0.8 at which time protein expression was induced by the addition of IPTG to a final concentration of 0.3mM. Induction occurred for 3hr at 37°C prior to the collection of all bacterial cells by centrifugation at 3000 rcf for 15mins at 4°C. The pellet was resuspended in 50ml binding buffer (20mM sodium phosphate, 0.5M NaCl and 20mM imidazole, pH 7.4) and sonicated 1min on, 1min off for 40mins. Cell debris was removed by centrifugation and the supernatant applied to a HisTrapTM FF column. The column was washed with 10 volumes of binding buffer and the protein eluted in 10 volumes of elution buffer (20mM sodium phosphate, 0.5M NaCl and 500mM imidazole, pH 7.4).
The purified proteins were dialysed against 100 volumes of dialysis buffer (10% glycerol, 20mM Tris/HCl, pH 7.9), changed three times. The gonococcal GalE proteins were concentrated to 100 μl using a centricon-10 (Millipore, Germany). Protein concentrations were determined by Bradford assay and glycerol was added to a final concentration of 50% (v/v) to stabilize the enzyme.
Activity of the gonococcal and meningococcal GalE proteins, and the site directed mutagenised proteins was done using methods described in Dong et al.44. Briefly, the activated sugar substrates (UDP-Glc, UDP-Gal, UDP-GlcNAc and UDP-GalNAc, 1mM) were incubated with 200ng of purified protein in 20mM Tris-HCl, 4mM Mg2+, and 1mM NAD+ (pH 8.0) in a final volume of 50μl. The reactions were performed at 37°C for 2h and were terminated by heating at 100°C for 5min. The products from the reaction were run on HPLC using a UV detector and compared to UDP-Glc, UDP-Gal, UDP-GlcNAc and UDP-GalNAc standards.
A dataset of 194 isolates of Neisseria spp. was selected from the Bacterial Isolate Genome Sequence Database (BIGSDB) located at PubMLST (http:pubmlst.org/neisseria)45 and described in Supplementary Table S1. This dataset was selected to include the most common alleles of and provide the maximal diversity for NEIS0048 and NEISS0062 across 133N. meningitidis isolates, 16N. gonorrhoeae isolates, 17N. lactamica isolates, seven N. cinerea isolates, six N. polysaccharea isolates, four N. oralis isolates42,46, four N. elongata isolates, four N. subflava isolates and one isolate each from the Neisseria species N. bergeri, N. animalis and N. weaveri. All of these genomes were sequenced on an Illumina Genome Analyser II platform and assembled according to Jolley and Maiden45. However, because Region D and D’ are duplicated regions, each contig was manually checked and exported as XMFA files containing aligned sequence blocks and then converted to a fastA format for import into MEGA version 5.047 and Splitstree48.
The galE1 and galE2 alleles were downloaded from the Neisseria PubMLST database (http://pubmlst.org/neisseria/)45 as NEIS0048 and NEIS0062 respectively. The phylogeny of the GalE1 peptide sequences was reconstructed using the neighbor-joining method using the MEGA software package47 with the following parameters: Bootstrap (1000 replicates), using a Poisson model and gamma distribution (parameter set to 1). For the phylogeny of the conserved regions of the galE1 and galE2 alleles, the sequences were trimmed using UGENE and the final 615 nucleotides were aligned using MUSCLE. Phylogenetic analysis was conducted by the neighbor-joining method using 1000 bootstrap tests. Evolutionary distances were calculated by the maximum composite likelihood method. The analysis was conducted with MEGA6.06.
The distribution of NEIS0048 and NEIS0062 was assessed in 1196 pathogenic meningococcal isolates summarised in Supplementary Table S2. Aligned sequences were analysed using the default parameters of ClonalFrame49 and the output parsed using the ParseCF script written by Barry Hall. Values calculated by ClonalFrame and ParseCF include the mutation rate (θ), the mutation rate per site (θ/number of sites), the corrected recombination rate for local block boundaries (ρ/number of sites), the estimate of recombination relative to that of mutation (r/m) and the estimate of the number of recombination events relative to mutations (ρ/θ).
KAP727 (5′-GAAGAATACCGCGAACTGACGC-3′) is a forward primer specific to the 5′ end of NEIS0044; KAP730 (5′-AGTTGGAAGAGCGCAAAGCCG-3′) is a forward primer specific to the 5′ end of tex (NEIS0059); KAP735 (5′-TTCAGACGGCAAGAGGGTACG-3′) is a reverse primer specific to the 3′ end of ctrE (NEIS0066). The PCR was performed using Phusion® High-Fidelity DNA Polymerase (New England Biolabs). PCR conditions for the assay were 2min at 98°C followed by 35 cycles of 98°C for 10seconds, 65°C for 30seconds and 72°C for 10minutes with a final extension time of 10minutes at 72°C.
How to cite this article: Bartley, S. N. et al. Acquisition of the capsule locus by horizontal gene transfer in Neisseria meningitidis is often accompanied by the loss of UDP-GalNAc synthesis. Sci. Rep. 7, 44442; doi: 10.1038/srep44442 (2017).
Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
This publication made use of the Neisseria Multi Locus Sequence Typing website (http://pubmlst.org/ neisseria/) which is funded by the Wellcome Trust Grant No. 104992 to MCJM and Keith Jolley. SNB and CAM were supported by PhD scholarships from the Amanda Young Foundation (http://www.amandayoungfoundation.org.au/). SM is supported by an International Postgraduate Research Student award from the University of Western Australia. KAS is funded by the Australian Research Council (FT100100291). CMK and SNB were supported by the National Health and Medical Research Council (APP546003). CMK and AV are supported by the National Health and Medical Research Council (APP1003697 and APP1078642). MCJM and OBH were supported by the Wellcome Trust (Grant 087622). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
The authors declare no competing financial interests.
Author Contributions S.N.B. completed the first draft of the manuscript. S.N.B. and K.A.S. performed the purification and enzymatic analyses contributing to Table 1. AV completed the structural biology and modelling of the enzyme contributing to Fig. 3. O.B.H. and M.J.C.M. completed the analysis of the phylogeny of the galE allele in all neisserial species see in Fig. 1. T.T.P., S.M. and C.A.M. completed the phylogeny, calculated the associative properties of the alleles with clonal complex and developed the final model for horizontal gene transfer as described in Figs 5 and 6. C.M.K. maintained oversight of the entire project, resources and final edit of the manuscript.