Allelic Variation in a Simple Sequence Repeat Element of Neisserial pglB2 and Its Consequences for Protein Expression and Protein Glycosylation

Abstract

Neisseria species express an O-linked glycosylation system in which functionally distinct proteins are elaborated with variable glycans. A major source of glycan diversity in N. meningitidis results from two distinct pglB alleles responsible for the synthesis of either N,N′-diacetylbacillosamine or glyceramido-acetamido trideoxyhexose that occupy the reducing end of the oligosaccharides. Alternative modifications at C-4 of the precursor UDP-4-amino are attributable to distinct C-terminal domains that dictate either acetyltransferase or glyceramidotransferase activity, encoded by pglB and pglB2, respectively. Naturally occurring alleles of pglB2 have homopolymeric tracts of either 7 or 8 adenosines (As) bridging the C-terminal open reading frame (ORF) and the ORF encompassing the conserved N-terminal domain associated with phosphoglycosyltransferase activity. In the work presented here, we explored the consequences of such pglB2 allele variation and found that, although both alleles are functional vis-à-vis glycosylation, the 7A form results in the expression of a single, multidomain protein, while the 8A variant elicits two single-domain proteins. We also found that the glyceramidotransferase activity-encoding domain is essential to protein glycosylation, showing the critical role of the C-4 modification of the precursor UDP-4-amino in the pathway. These findings were further extended and confirmed by examining the phenotypic consequences of extended poly(A) tract length variation. Although ORFs related to those of pglB2 are broadly distributed in eubacteria, they are primarily found as two distinct, juxtaposed ORFs. Thus, the neisserial pglB2 system provides novel insights into the potential influence of hypermutability on modular evolution of proteins by providing a unique snapshot of the progression of ongoing gene fusion.

INTRODUCTION

Diverse arrays of glycoconjugates varying in structure and composition dominate the surfaces of bacteria. Despite their importance and predominance, the origins of polysaccharide diversity at the genetic level and what forces drive glycan diversification remain poorly understood. Protein glycosylation utilizing both N- and O-linked modifications occurs in many bacterial species, and surface-localized proteins are overrepresented among substrates (1). The broad dissemination of these glycosylation systems suggests that they are advantageous and affect overall fitness. Detecting selection at the genetic level in these systems is complicated because complex carbohydrates are secondary gene products and because precise relationships between gene content and glycoform phenotypes have been difficult to define. As such, the evolutionary trends and long-term dynamics of protein glycosylation systems remain poorly understood. Attempts at reconciling the glycosylation gene repertoire with particular bacterial populations and ecotypes are similarly problematic.

Inter- and intrastrain glycan variability has been observed in many species, suggesting that glycoform diversification is under positive selection. For example, protein glycosylation in flagellar systems of Campylobacter jejuni and Clostridium species exhibits plasticity in glycoform expression (2–4). The broad-spectrum O-linked protein glycosylation (pgl) systems found in species within the genus Neisseria present a unique opportunity to study bacterial glycoconjugate diversification and evolution. Three highly related neisserial species of importance to humans are Neisseria gonorrhoeae (the agent of gonorrhea), Neisseria meningitidis (an agent of epidemic meningitis), and Neisseria lactamica (a commensal colonizing the oropharynx of young children). The biochemistry and genetics of glycan biosynthesis, modification, and transfer to protein have been well characterized in these species (5–8). PglB, PglC, and PglD are required in the synthesis of undecaprenyl diphosphate (UndPP)-linked N,N′-diacetylbacillosamine (diNAcBac) that in turn can be further elaborated into disaccharide forms by the action of two distinct glycosyltransferases (Fig. 1A). The pglA gene encodes a galactosyltransferase and is present in all strains, while pglH encodes a glucosyltransferase and is more limited in its distribution (7). PglE is a galactosyltransferase that solely extends the PglA-generated UndPP-linked disaccharide to a trisaccharide glycoform (5, 7). Both the di- and trisaccharide forms can be further modified via O-acetylation mediated by PglI (7). These UndPP-linked glycoforms are subsequently translocated into the periplasm by the action of the PglF flippase, where they serve as donors for protein glycosylation mediated by the PglO oligosaccharyltransferase (7).

Fig 1.

O-linked glycosylation pathway in Neisseria. (A) Current model of the general O-linked glycosylation pathway in Neisseria. (B) Chemical structure of the reducing-end sugar diNAcBac, where the acetyl group is added by the C-terminal part of PglB (PglB-AcT) and the transfer to the undecaprenyl phosphate is mediated by the phosphoglycosyltransferase, the N-terminal part of PglB (PglB-PGT). (C) Chemical structure of GATDH, where the glyceramido group is attached by the action of the C-terminal region of PglB2 that contains the ATP-grasp domain (PglB2-ATP-grasp). The sugar is then transferred to the undecaprenyl phosphate by the phosphoglycosyltransferase, the N-terminal part of PglB2 (PglB2-PGT). OM, outer membrane; IM, inner membrane.

Strains within these species are thought to display glycoform variation as a result of phase-variable, slipped-strand mispairing events within the open reading frames (ORFs) of the pglA, pglE, pglH, and pglI genes (7, 9). The phase-variable character of pglA, pglH, pglE, and pglI results from the presence of simple sequence repeats that undergo expansion and contraction, altering the reading frame at high frequencies. In this way, single strains can express up to seven exclusive protein-associated glycoforms and, because of the overlapping activities of PglA and PglH, two distinct glycoforms simultaneously (6). Such phase-variable alleles have been termed contingency loci, as they are hypermutable, stochastic, and heritable genotypes that maintain a preemptive strategy for adaptation (10). Contingency loci are abundant in the genomes of many human mucosal pathogens, including N. gonorrhoeae, N. meningitidis, Haemophilus influenzae, and others (10). The neisserial pgl contingency loci and their associated biology are most similar in action and character to those associated with the lipooligosaccharide (LOS) biosynthetic pathway in Neisseria, Vibrio, and Haemophilus species (10–12).

PglB is a multidomain, bifunctional enzyme that catalyzes both the amino acetylation of the UDP-4-amino sugar precursor to form UDP-diNAcBac and the transfer of the phospho-diNAcBac to undecaprenyl phosphate (13) (Fig. 1B). Equivalent activities in the C. jejuniN-linked protein glycosylation system, which also utilizes UndPP-diNAcBac, are carried out by the PglD and PglC protein, respectively (14, 15). Based on these commonalities and synteny between the loci, these two systems undoubtedly share ancestry. Moreover, the single pglB ORF of Neisseria is likely the result of a fusion event between the equivalents of Campylobacter pglD and pglC. Nearly one-half of N. meningitidis isolates and some commensal species are reported to carry a variant allele of pglB, designated pglB2 (9, 16). Alleles of pglB2 appear to have been imported into a pglB background from an as-yet-unidentified source outside the genus Neisseria (17). PglB2 possesses an N-terminal domain that is virtually identical to the corresponding segment of PglB (which has a phosphoglycosyltransferase activity) and a novel C-terminal segment with an ATP-grasp domain in place of the amino acetylation domain that is present in PglB (17). The introduction of pglB2 into a defined gonococcal background leads to the expression of an altered sugar at the reducing end of the oligosaccharide termed glyceramido-acetamido trideoxyhexose (GATDH) (6, 17) that bears a glyceroyl moiety rather than the acetyl moiety found in diNAcBac at the amino group at position C-4 (Fig. 1C). ATP-grasp domains are associated with enzymes that carry out the formation of a peptide-like bond between an amino, imino, or thiol group of one substrate and a carboxylate group of another substrate (18). Thus, the altered C-terminal domain of PglB2 and its associated activity are consistent with the chemical features distinguishing GATDH from diNAcBac.

Comparisons of pglB2 alleles within N. meningitidis and commensals reveal an overall high degree of conservation save for polymorphisms within a homopolymeric tract of adenosines occurring within the ORF. Within N. meningitidis strains, tracts of 7 and 8 adenosines predominate, with the former leading to a single polypeptide while the latter predicts two ORFs overlapping at their C and N termini. The potential consequences of alterations within the poly(A) tract are unknown, as only alleles with an in-frame configuration (7 As) have been characterized. According to the paradigm establishing contingency loci as simple on-off translational switches, out-of-frame configurations could lead to a significant reduction in the expression or synthesis of only the N-terminal domain (19). Furthermore, the potential consequences of expressing only the N-terminal domain remain unknown, as the importance of the modification at C-4 to the integrity of protein glycosylation has never been directly tested.

Here, we examined the consequences of altering the length and structure of the poly(A) tract region in pglB2 on protein expression and protein glycosylation phenotype. The results reveal a surprising degree of complexity with regard to the levels of products and glycosylation proficiency, countering a model for simple on-off switching. The findings also highlight the role of nonstandard decoding in controlling gene expression and for contingency loci in driving high-frequency gene fusion and gene fission events.

MATERIALS AND METHODS

Bacterial strains and culture conditions.

The bacterial strains used in this study are described in Table 1 and were grown on conventional GC medium as has been described previously (20). Protein glycosylation mutations were introduced into various backgrounds by using transformation as previously described (5, 6). Antibiotics for selection of transformants were used at the following concentrations: streptomycin, 750 μg/ml; kanamycin, 50 μg/ml; and gradient chloramphenicol, 25 μg/ml.

Table 1.

N. gonorrhoeae N400 mutant strains used in this study

Strain	Parental strain	Relevant genotype	Reference
KS100	N400	recA6	34
KS142	KS100	pglEon	5
KS310	KS142	pglEon pglB28013	6
KS367	KS142	pglEon pglB2FAM18	7
KS686	KS367	pglEon pglB2FAM18::kan	This study
KS704	KS367	pglEon pglB2bFAM18::kan	This study
KS687	KS686	pglEon pglB2FAM18::kan lctP::pglB2FAM18	This study
KS688	KS686	pglEon pglB2FAM18::kan lctP::pglB2bFAM18	This study
KS693	KS686	pglEon pglB2FAM18::kan lctP::pglB28A ATG605GCG	This study
KS700	KS686	pglEon pglB2FAM18::kan lctP::pglB27A ATG605GCG	This study
KS692	KS686	pglEon pglB2FAM18::kan lctP::pglB28A ATG731GCG	This study
KS701	KS686	pglEon pglB2FAM18::kan lctP:: pglB2AAGAAGAA ATG605GCG	This study
KS709	KS686	pglEon pglB2FAM18::kan lctP:: pglB2AAGAAGAA ATG605GCG GTG656GCG	This study
KS750	KS686	pglEon pglB2FAM18::kan lctP::pglB28A GTG656GCG	This study
KS695	KS686	pglEon pglB2FAM18::kan lctP::pglB23A	This study
KS703	KS686	pglEon pglB2FAM18::kan lctP::pglB24A	This study
KS702	KS686	pglEon pglB2FAM18::kan lctP::pglB25A	This study
KS696	KS686	pglEon pglB2FAM18::kan lctP::pglB26A	This study
KS689	KS686	pglEon pglB2FAM18::kan lctP::pglB27A	This study
KS691	KS686	pglEon pglB2FAM18::kan lctP:: pglB2AAGAAGAA	This study
KS694	KS686	pglEon pglB2FAM18::kan lctP::pglB29A	This study
KS697	KS686	pglEon pglB2FAM18::kan lctP::pglB23(AAG)	This study
KS690	KS686	pglEon pglB2FAM18::kan lctP::pglB211A	This study
KS698	KS686	pglEon pglB2FAM18::kan lctP::pglB212A	This study
KS699	KS686	pglEon pglB2FAM18::kan lctP::pglB24(AAG)	This study
KS705	KS704	pglEon pglB2bFAM18::kan lctP::pglB2bFAM18	This study
KS706	KS704	pglEon pglB2bFAM18::kan lctP::pglB2b3(AAG)	This study
KS707	KS704	pglEon pglB2bFAM18::kan lctP::pglB2b9A	This study
KS708	KS704	pglEon pglB2bFAM18::kan lctP::pglB2b3A	This study
KS710	KS367	pglEon pglB2a9A pglB2bFAM18::kan lctP::pglBbFAM18	This study
KS711	KS367	pglEon pglB2a3A pglB2bFAM18::kan lctP::pglBbFAM18	This study
KS712	KS367	pglEon pglB2a3(AAG) pglB2bFAM18::kan lctP::pglBbFAM18	This study

Construction of pglB2 mutants.

The pCRII-pglB2::kan plasmid (KP20) was constructed by amplifying the whole pglB2 gene and surrounding sequences (2,682 bp) from N. meningitidis strain FAM18 (pglB2 of strain FAM18 is denoted hereinafter as pglB2_FAM18) using primers RB68 and RB69 and inserting the PCR product into the pCRII-TOPO vector (Invitrogen). All primer sequences are shown in Table S1 in the supplemental material. Part of the pglB2 gene (1,470 bp) was cut out with ClaI and replaced with the kanamycin cassette from the pKAN vector (on a HincII fragment), resulting in the pCRII-pglB2::kan plasmid. N. gonorrhoeae strain N400 pglE_on pglB2_FAM18 (KS367) was then transformed with pCRII-pglB2::kan, and kanamycin-resistant mutants were selected. The pglB2::kan locus in strain N400 pglE_on pglB2_FAM18::kan (KS686) was verified by sequencing. N400 pglE_on carries the phase-on allele of pglE from strain FA1090 and thereby synthesizes trisaccharide.

To generate a pglB2b null mutant, the pCRII-pglB2a-pglB2b::kan plasmid (KP21) was constructed by splicing by overlap extension (SOEing) PCR. The pglB2a ORF encodes the N-terminal phosphoglycosyltransferase domain, and the pglB2b ORF encodes the C-terminal ATP-grasp domain. We used primers RB74 and RB75 to amplify pglB2a and primers RB72 and RB73 to amplify the downstream region of pglB2b. These primers introduced a PmeI restriction site at the end of pglB2a. A third pair of primers (RB76 and RB77) was used to join and amplify the two PCR fragments together. The final PCR fragment of 1,974 bp was then ligated into the pCRII-TOPO vector. Next, the pCRII-pglB2a-pglB2b plasmid was opened with PmeI, and the kanamycin cassette (HincII fragment of pKAN) was subsequently cloned into the PmeI site, resulting in the pCRII-pglB2a-pglB2b::kan plasmid (KP21). N400 pglE_on pglB2_FAM18 (KS367) was transformed with pCRII-pglB2a-pglB2b::kan, and kanamycin-resistant pglB2b mutants were selected. The pglB2b_FAM18::kan locus of strain N400 pglE_on pglB2b_FAM18::kan (KS704) was verified by sequencing.

Mutations at the endogenous pglB2a_FAM18 gene were introduced into plasmid pCRII-pglB2a-pglB2b::kan by using the commercial QuikChange XL site-directed mutagenesis kit as described in the manual (2009; Agilent Technologies, Inc.). The specific primers used to introduce mutations and the resulting strains are summarized in Table S2 in the supplemental material.

Complementation analysis.

To construct an ectopically expressed pglB2 locus, the pglB2_FAM18 gene was amplified with the specific primers RB70 and RB71, containing PacI and NsiI restriction sites, respectively. The PCR product was then digested and subcloned into the NsiI and PacI fragment of pGCC6, resulting in the pGCC6::pglB2_FAM18 (KP22) plasmid. Likewise, pglB2b_FAM18 was inserted into pGCC6 by using primers RB78 and RB79, resulting in the plasmid pGCC6::pglB2b_FAM18 (KP23).

All mutations in the pGCC6::pglB2_FAM18 and pGCC6::pglB2b_FAM18 plasmids were introduced by using the commercial QuikChange XL site-directed mutagenesis kit as described in the manual. The specific primers used to introduce mutations into the pGCC6::pglB2_FAM18 and pGCC6::pglB2b_FAM18 plasmids are summarized in Table S2 in the supplemental material. The pGCC6-based constructs were then transformed into gonococci, and transformants were selected for on plates containing a chloramphenicol gradient. The ectopically expressed genes were inserted at an intergenic chromosomal site located between the lctP and aspC genes, and all strains were verified by PCR and sequencing.

SDS-PAGE and immunoblotting.

The procedures used for SDS-PAGE and immunoblotting have been described previously (20). Briefly, whole-cell lysates were prepared from equivalent numbers of cells by heating cell suspensions at 65°C for 20 min in SDS-sample loading buffer. The rabbit monoclonal antibody npg3 (1:20,000 dilution) was used to detect the trisaccharide GATDH-galactose-galactose (6). PglB2a and PglB2b proteins were detected by immunoblotting using the pB2a and pB2b polyclonal antibodies described below. Immunoactive proteins were detected by immunoblotting using alkaline phosphatase-coupled goat anti-rabbit secondary antibodies (Sigma).

Development of PglB2-specific rabbit polyclonal antibodies.

The polyclonal antibodies (pB2a and pB2b) were generated by GenScript. Both pglB2a and pglB2b genes were synthesized de novo by GenScript and subcloned into a bacterial expression vector containing a 6His tag added to the N terminus of pglB2a and pglB2b. These recombinant proteins were used to immunize rabbits. After three immunizations, the resulting polyclonal pB2a and pB2b antibodies were used to detect the respective proteins by immunoblotting (1:1,000 dilutions).

Sequence data analysis and manipulation.

Nucleotide sequence data were obtained from NCBI using pglB2_FAM18 (NCBI GI 121634264) as a query sequence. Sequences were then aligned using the Molecular Evolutionary Genetics Analysis MEGA5 software package (21). The resulting alignment was manually curated. Strains with fused ORFs and poly(A) tracts had gaps inserted to ensure correct reading frame maintenance in the pglB2b gene. For the remaining alignment, we manually removed regions containing gaps. Selection of the protein substitution model was done using PROTTEST3 on the translated protein alignment (22). The evolutionary history was then inferred by using the maximum likelihood method based on the JTT matrix-based model (as reported from PROTTEST3) with 500 bootstrap replicates (23). In the original tree, the Flavobacterium psychrophilum JIP02/86 branch collapsed with two other clades, so we chose to use it as an artificial root. The tree branches were modified using proportional transformation. Both tree modifications were performed using FigTree (version 1.3.1) software (http//tree.bio.ed.ac.uk/software/figtree). Bootstrap values lower than 75 are not shown on the tree.

RESULTS

A pglB2 allele with a two-ORF configuration is functional in GATDH glycoform expression.

We first sought to determine if a pglB2 allele with a two-ORF configuration was functional in glycosylation. As detailed in the NCBI database, the pglB2 locus from N. meningitidis strain FAM18 contains an 8A tract (NCBI GI 121634264) that would disrupt a single large ORF (see Fig. S1 and S2 in the supplemental material). Prior to our recognition of this anomaly, we reported that replacement of the pglB allele of N. gonorrhoeae strain N400 with the pglB2_FAM18 allele switched glycan synthesis from diNAcBac-based to GATDH-based glycoforms (as seen by top-down mass spectrometric analyses) (7). Sequencing the pglB2 allele from the FAM18 strain used by us and as expressed in the N400 background confirmed the presence of the 8A tract (data not shown). Using a host background with active alleles of pglA and pglE, we were also able to confirm the pilin glycosylation status by virtue of reactivity with the npg3 monoclonal antibody (Fig. 2A) that recognizes both diNAcBac- and GATDH-based trisaccharide glycoforms (6).

Fig 2.

The pglB2_FAM18 allele encodes two distinct polypeptides. (A) Glycosylation of the major pilin protein PilE is detected in an immunoblot using the trisaccharide-specific monoclonal antibody npg3. Only glycosylated PilE is recognized by npg3. (B) An ∼21-kDa protein is recognized as the PglB2a protein in an immunoblot by using the polyclonal antibody pB2a. Both glycosylated PilE (upper band) and nonglycosylated PilE (lower band) bands are seen due to nonspecific cross-reactivity with the pB2a antibody. (C) An ∼36-kDa protein is recognized as the PglB2b protein in an immunoblot by using the polyclonal antibody pB2b. Neither of the proteins was detected in the null mutant or in the in-frame pglB2₈₀₁₃ construct. +, denotes presence of pglB2 variant in either endogenous or ectopic location; -, denotes absence or pglB2::kan. The designation pglB2 refers to the allele from N. meningitidis strain FAM18 if not otherwise indicated. All strains were made in the N. gonorrhoeae N400 pglE_on background that synthesizes trisaccharide. Genotypes of strains used were pglB2_FAM18 (KS367), pglB2_FAM18::kan (KS686), pglB2_FAM18::kan lctP::pglB2_FAM18 (KS687), and pglB2₈₀₁₃ (KS310).

Two distinct proteins are encoded by the pglB2_FAM18 allele.

To follow up this finding in more detail, we used antibodies generated against recombinant nonoverlapping polypeptides encompassing either the phosphoglycosyltransferase or ATP-grasp domain to examine patterns of protein expression. A product with a relative mobility of 21 kDa, corresponding to the N-terminal polypeptide, was detected (Fig. 2B), and a product corresponding to the C-terminal-related polypeptide was detected at 36 kDa (Fig. 2C) in the strain carrying the pglB2_FAM18 allele at the endogenous locus (within the core pglF, -B, -C, and -D region), as well as when solely expressed from an ectopic site. Moreover, these products were not seen in association with the pglB2 allele from strain 8013 (pglB2₈₀₁₃) that contains the 7A tract predicting a single ORF configuration. We refer to these proteins hereinafter as PglB2a and PglB2b. Neither of the antibodies recognized the PglB2 protein from strain 8013. We were not able to increase immunoblot specificity despite extensive efforts with various antibody dilutions and affinity purification. Fortuitously, antibodies directed at the N-terminal domain also cross-reacted with PilE, the major subunit of neisserial type IV pili (Fig. 2B). This provided a further control of glycosylation proficiency, as nonglycosylated PilE had significantly increased mobility in SDS-PAGE relative to its glycosylated form. Using this phenotype, we were unable to discern any differences in glycosylation proficiency between the pglB2_FAM18 and pglB2₈₀₁₃ alleles.

Characterization of two pglB2_FAM18 ORFs essential for protein glycosylation.

Having demonstrated that the pglB2_FAM18 allele was associated with the expression of two polypeptides, we wished to define the corresponding ORFs. Based on the presence of the 8A tract, the N-terminal polypeptide would terminate with the sequence NKTDEKK (terminated just after the homopolymeric tract), giving rise to a protein predicted to be approximately 23 kDa. This ORF corresponded well with the 21-kDa species detected by immunoblotting. The 36-kDa species could be reconciled best with a large C-terminal ORF overlapping the N-terminal ORF by 4 codons but in a plus 1 reading frame. While the latter ORF had a potential ATG start codon at position 605 (ATG₆₀₅), two other potential start codons were found 3′ to it at GTG₆₅₆ and ATG₇₃₁ (positions are relative to the pglB2_FAM18 allele, detailed in Fig. S2 in the supplemental material). Based on Pfam data, all these potential start sites are located outside the ATP-grasp domain (Fig. 3A). Mutating the first potential start site into that coding for alanine ( ATG₆₀₅GCG ) in an 8A configuration resulted in reduced PilE glycosylation, as seen by reduced reactivity with the npg3 monoclonal antibody (detecting the trisaccharide epitope) and increasing levels of unmodified PilE (Fig. 3B and C). These phenotypes were associated with the inability to detect PglB2b, while PglB2a was detected at reduced levels relative to its level in the wild-type 8A background. The same mutation in the 7A context did not affect glycosylation (Fig. 3B), and both antibodies reacted with a new protein of approximately 60 kDa, likely corresponding to the multidomain PglB2 protein. Analogous mutation of the other two potential start sites ( GTG₆₅₆GCG and ATG₇₃₁GCG ) had no effect on the glycosylation or detection of PglB2a, while a reduced amount of PglB2b was seen in the GTG₆₅₆GCG but not in the ATG₇₃₁GCG background. We hypothesized that the low-level, residual glycosylation seen in the ATG₆₀₅GCG/8A background might be due to nonstandard decoding associated with the homopolymeric tract (24). Therefore, we mutated the tract so as to disrupt its repetitive nature but not perturb the amino acid sequence (the 8A tract was mutated into AAGAAGAA ), but this did not discernibly diminish the relative glycosylation levels (Fig. 3B, lanes 3 and 6). For reasons that are as yet unclear, the PglB2a polypeptide was not detected in this case. We also examined the combined effects of the ATG₆₀₅GCG and GTG₆₅₆GCG changes in the AAGAAGAA background and found that glycosylation was abolished (Fig. 3B, lane 7). The latter observation suggested that the residual activity in the ATG₆₀₅GCG/AAGAAGAA background might be due to low levels of translation due to initiation at the GTG₆₅₆ codon. Finally, we expressed each of the two ORFs independently of one another to confirm their associations with the 21-kDa and 36-kDa proteins and the simultaneous requirement for each in glycosylation (Fig. 3B, lane 11). Taken together, these findings indicate that the 8A configuration leads to the expression of two juxtaposed, overlapping ORFs and that each is required for GATDH-based glycosylation.

Fig 3.

Characterization of the PglB2b ORF. (A) Cartoon of the pglB2 gene with two ORFs, designated pglB2a and pglB2b. The three possible start codons for PglB2b are shown. The pglB2_FAM18 gene sequence and the predicted PglB2a and PglB2b protein sequences in the vicinity of the homopolymeric tract are shown (for more information, see Fig. S1 in the supplemental material). Pfam-predicted domains are indicated; slanted lines show the bacterial sugar transferase domain (glycosyltransferase [GT]), and slanted dotted lines show the ATP-grasp domain. (B) Glycosylation of PilE was detected in an immunoblot using the trisaccharide-specific monoclonal antibody npg3 against strains with the pglB2::kan allele at the endogenous locus and with ectopically expressed pglB2. Only glycosylated PilE is recognized by npg3. (C) PglB2a protein was detected in an immunoblot using the polyclonal antibody pB2a. Both glycosylated PilE (upper band) and nonglycosylated PilE (lower band) are seen due to nonspecific cross-reactivity with the pB2a antibody. (D) PglB2b protein was detected in an immunoblot using the polyclonal antibody pB2b. The designation pglB2 refers to the allele from N. meningitidis strain FAM18. Mutant strains had pglB2 expressed ectopically in lane 1 and lanes 3 to 8, pglB2a expressed endogenously (end.) in lanes 9 and 11, and pglB2b expressed ectopically (ect.) in lanes 10 and 11. -, denotes absence or pglB2::kan as indicated; #nt, indicates number of adenosines in the homopolymeric tract; 8*, denotes AAGAAGAA. The arrow points to a glycoprotein (seen due to nonspecific cross-reactivity with the pB2a antibody) in which migration varies according to whether it is glycosylated (upper band) or not (lower band). All strains were made in the N. gonorrhoeae N400 pglE_on background that synthesizes trisaccharide. Genotypes of the strains used were pglB2_FAM18::kan lctP::pglB2_FAM18 (KS687), pglB2_FAM18::kan (KS686), pglB2_FAM18::kan lctP::pglB2_8A ATG₆₀₅GCG (KS693), pglB2_FAM18::kan lctP::pglB2_7A ATG₆₀₅GCG (KS700), pglB2_FAM18::kan lctP::pglB2_8A ATG₇₃₁GCG (KS692), pglB2_FAM18::kan lctP:: pglB2_AAGAAGAA ATG₆₀₅GCG (KS701), pglB2_FAM18::kan lctP:: pglB2_AAGAAGAA ATG₆₀₅GCG GTG₆₅₆GCG (KS709), pglB2_FAM18::kan lctP::pglB2_8A GTG₆₅₆GCG (KS750), pglB2b_FAM18::kan (KS704), pglB2_FAM18::kan lctP::pglB2b_FAM18 (KS688), and pglB2b_FAM18::kan lctP::pglBb_FAM18 (KS705).

Phenotypic consequences of pglB2 homopolymeric tract variation.

To further probe the relationships between poly(A) tract length, pglB2 gene products, and glycosylation, we generated pglB2 alleles carrying from 3 to 12 adenosines. These mutations thus mimic the potential hypervariability that might occur naturally in vivo by canonical slipped-strand mispairing. To facilitate strain construction, these alterations were made in an ectopically expressed pglB2_FAM18 gene copy in a background that carried a null allele at the endogenous locus. As anticipated from prior findings, variants with either a 4A or 7A tract supported high-level glycosylation and expressed the single, multidomain protein of 60 kDa (Fig. 4). Those with either a 5A or 8A tract also supported high-level glycosylation and expressed both PglB2a and PglB2b, while variants with either a 3A or 6A tract were grossly glycosylation defective and failed to express detectable levels of either of the pglB2-related proteins. Although the glycosylation defect and the inability to express PglB2b in the latter case were consistent with the fact that these configurations place a TGA stop codon in frame with the PglB2b ORF (see Fig. S3 in the supplemental material), it is unclear how these configurations preclude the detection of PglB2a. Interestingly, the 9A variant expressed intermediate levels of glycosylation with nearly equal levels of glycosylated and nonglycosylated PilE (Fig. 4C), although neither PglB2a nor PglB2b was detected. This result was unanticipated as, like the case with the 3A and 6A variants, it places the TGA stop codon in frame with the PglB2b ORF and should preclude its translation. Although the variant with the 9A tract should be more mutable than those with 3 As and 6 As, the level of glycosylation observed was too high to be accounted for by the presence of rare variants arising within the population with frame-correcting alterations. We therefore considered the possibility that the pglB2-associated activity here might be due to recoding (nonstandard decoding) associated with the extended poly(A) tract. Here, the two most likely mechanisms operating would be programmed ribosomal frameshifting and programmed transcriptional realignment (24). As presently modeled, both activities rely on the homopolymeric nature of the tract. To examine this matter in more detail, the 9A tract was replaced with the sequence AAGAAGAAG that would obviate recoding, as well as slipped-strand mispairing, but otherwise maintain the amino acids encoded. In this background, glycosylation was undetectable, as were both PglB2a and PglB2b. In light of this finding, we again tested whether recoding might influence the activity associated with the 8A tract by changing it to AAGAAGAA but found that neither glycosylation nor PglB2a/PglB2b expression were compromised. We then tested the effects of further extending the poly(A) tract and found that an 11A configuration gave a phenotype equivalent to those of the 5A and 8A configurations, while unexpectedly, a fully proficient glycosylation phenotype was seen for the 12A configuration. In contrast to the situation seen for the 3A-, 6A-, and 9A-containing alleles, PglB2a and PglB2b were detected, albeit at very low levels. To again assess whether these phenotypes were related to the susceptibility of the 12A tract to transcriptional/translational slippage, we mutated the sequence to 4× AAG and found that glycosylation was abolished, as was detection of PglB2a and PglB2b. Taken together, these findings support the conclusion that some pglB2 alleles with out-of-frame configurations allow significant protein levels to be achieved through nonstandard decoding and that this activity is dependent on the poly(A) tract length.

Fig 4.

Phenotypic consequences of homopolymeric tract length variation in pglB2_FAM18. (A) Cartoon of the pglB2 gene with two ORFs, designated pglB2a and pglB2b. The poly(A) tract with 8 As is shown. (B) Glycosylation of PilE was detected in an immunoblot using the trisaccharide-specific monoclonal antibody npg3. Only glycosylated PilE is recognized by npg3. (C) PglB2a protein was detected in an immunoblot using the polyclonal antibody pB2a. Both glycosylated PilE (upper band) and nonglycosylated PilE (lower band) are seen due to nonspecific cross-reactivity with the pB2a antibody. (D) PglB2b protein was detected in an immunoblot using the polyclonal antibody pB2b. The designation pglB2 refers to the allele from N. meningitidis strain FAM18. All mutant alleles were expressed ectopically. -, denotes pglB2_FAM18::kan; #nt, indicates number of adenosines in the homopolymeric tract; 8*, 9*, and 12*, indicate homopolymeric tracts composed of AAGAAGAA , 3(AAG), and 4(AAG), respectively. The arrow points at a glycoprotein (seen due to nonspecific cross-reactivity with the pB2a antibody) in which migration varies according to whether it is glycosylated (upper band) or not (lower band). All strains were made in the N. gonorrhoeae N400 pglE_on background that synthesizes trisaccharide. Genotypes of strains used are pglB2_FAM18 (KS367), pglB2_FAM18::kan (KS686), pglB2_FAM18::kan lctP::pglB2_3A (KS695), pglB2_FAM18::kan lctP::pglB2_4A (KS703), pglB2_FAM18::kan lctP::pglB2_5A (KS702), pglB2_FAM18::kan lctP::pglB2_6A (KS696), pglB2_FAM18::kan lctP::pglB2_7A (KS689), pglB2_FAM18::kan lctP::pglB2 (KS687), pglB2_FAM18::kan lctP:: pglB2_AAGAAGAA (KS691), pglB2_FAM18::kan lctP::pglB2_9A (KS694), pglB2_FAM18::kan lctP::pglB2_3(AAG) (KS697), pglB2_FAM18::kan lctP::pglB2_11A (KS690), pglB2_FAM18::kan lctP::pglB2_12A (KS698), and pglB2_FAM18::kan lctP::pglB2_4(AAG) (KS699).

A potentially confounding aspect of this approach relates to the fact that changes in poly(A) tract length would also alter the C-terminal sequence of PglB2a and the N-terminal sequence of PglB2b. Moreover, interactions between the nascent single-domain proteins might be influenced by their cotranslation from a single mRNA. To examine these possibilities in alleles associated with defective glycosylation, we expressed the single-domain ORFs from two distinct loci such that the potential influence of altering the C terminus of PglB2a and N terminus of PglB2b could be directly assessed independently of one another. As shown in Figure 5, strains carrying the 3A, 9A, or 3× AAG configuration in the PglB2a allele in conjunction with an ectopic PglB2b (8A) allele were glycosylation proficient. Curiously, PglB2a was not detectable in these backgrounds. In contrast, strains carrying the same alterations in the PglB2b-expressing allele together with an active PglB2a (8A)-expressing allele were grossly defective in glycosylation due to an inability to produce PglB2b [see Fig. S3 in the supplemental material, showing poly(A) tract configurations and corresponding predicted protein sequences]. Moreover, the low level of residual glycosylation seen for the 9A configuration was not seen with the 3× AAG configuration allele. Thus, the defects seen in the strains carrying these alterations in the endogenous, intact allele can be attributed to the absence of PglB2b and not to altered function of PglB2a.

Fig 5.

Influence of homopolymeric tract length variation on PglB2a and PglB2b function. (A) Cartoon of strains constructed for this experiment to interpret mutations made at the homopolymeric tract on separately expressed proteins. In all strains, pglB2a_FAM18 is expressed endogenously while pglB2b_FAM18 is expressed ectopically. (The endogenous pglB2b allele is replaced by a kanamycin cassette.) (B) In the first lane, both pglB2a and pglB2b have the poly(A) intact with 8 As. Results for mutations introduced into the endogenously expressed pglB2a_FAM18 are shown in lanes 2 to 4, as indicated, while results for poly(A) mutations introduced in the ectopically expressed pglB2b_FAM18 are shown in lanes 5 to 7. Mutations were introduced to only at one locus at a time, either endogenously or ectopically. Glycosylation of PilE was detected in an immunoblot using the trisaccharide-specific monoclonal antibody npg3. Only glycosylated PilE is recognized by npg3. (C) PglB2a protein was detected in an immunoblot using the polyclonal antibody pB2a. (D) PglB2b protein was detected in an immunoblot using the polyclonal antibody pB2b. #nt, indicates number of adenosines at the homopolymeric tract. All strains were made in the N. gonorrhoeae N400 pglE_on background that synthesizes trisaccharide. Genotypes of strains used are pglB2b_FAM18::kan lctP::pglB2b_FAM18 (KS705), pglB2a_3A pglB2b_FAM18::kan lctP::pglB2b_FAM18 (KS711), pglB2a_9A pglB2b_FAM18::kan lctP::pglB2b_FAM18 (KS710), pglB2a_3(AAG) pglB2b_FAM18::kan lctP::pglB2b_FAM18 (KS712), pglB2b_FAM18::kan lctP::pglB2b_3A (KS708), pglB2b_FAM18::kan lctP::pglB2b_9A (KS707), and pglB2b_FAM18::kan lctP::pglB2b_3(AAG) (KS706).

Evolution of pglB2 and its orthologs.

To better understand the forces shaping the variable status of pglB2 alleles, we examined the occurrence and organization of orthologous ORFs in a broader sampling within the eubacteria. Using sequence data obtained from the NCBI database using tBlastN with pglB2_FAM18 as a query, we found that pglB2 orthologs are present in at least seven other species in the genus Neisseria (Fig. 6). Moreover, orthologs are widely but sparsely disseminated within species outside the genus Neisseria. We did sequence alignment of pglB2 gene homologs and then model tested the translated protein alignment to find the appropriate protein substitution model. Finally, the maximum likelihood phylogenetic tree was produced using MEGA5 software (21–23). The resulting phylogenetic tree displays a solid grouping of Neisseria strains indicated by a high bootstrap value. Based on the available DNA sequences, the pglB2 homologs from various bacterial strains, as well as Neisseria pglB2 genes, can be divided into four groups (Fig. 6). In the first group, the pglB2 homologs have two ORFs that are separated by a gap (Fig. 6, green circles). This gap can be as few as four nucleotides in Acholeplasma laidlawii and as many as a whole-gene insertion in Parabacteroides merdae. The second group contains bacteria in which pglB2 is configured as two overlapping ORFs (Fig. 6, green squares). The two functional-domain-encoding regions overlap by a number of nucleotides that ranges from 1 nucleotide in Mannheimia succiniciproducens to 11 nucleotides in Bacteroides plebeius and 14 nucleotides in Neisseria sp. oral taxon 014. Bacteria that possess two overlapping ORFs encompassing a homopolymeric tract constitute the third category (Fig. 6, green diamonds). Within this group, there are strains of Bacillus cereus and several species of the Neisseria genus. Interestingly, the homopolymeric tracts in B. cereus are composed of guanosine nucleotides, poly(G), while in all Neisseria species, we find poly(A) tracts. In all these groups, there was significant concordance in the localization of the C terminus of the phosphoglycosyltransferase domain polypeptide and potential translational start sites of the ATP-grasp domain polypeptide (see Fig. S4 and S5, respectively, in the supplemental material). These findings were also consistent with our mutagenesis data on these segments of the two neisserial ORFs (Fig. 3). The last group (Fig. 6, red circles), which is also strongly supported by a high bootstrap value, comprises 12 N. meningitidis strains that contain pglB2 configured as a single ORF. Moreover, the putative start codon of pglB2b [Met residue just upstream from the poly(A)] is present in all but one strain, N. meningitidis strain 053442. It is also relevant to note that in many of the loci, there is significant conservation in both the organization and predicted functions of genes flanking pglB2 and its orthologs (see Table S3 in the supplemental material). For example, the vast majority of loci in organisms outside the genus Neisseria carry three ORFs immediately 3′ of the ATP-grasp domain ORF that are predicted to encode a haloacid dehalogenase-like hydrolase and orthologs of neisserial PglC and PglD. Likewise, many of these loci carry an ortholog of the neisserial PglH glucosyltransferase just 5′ of the phosphoglycosyltransferase domain polypeptide. This broad conservation in synteny and gene annotation clearly demonstrates that all of these loci descend from a single common ancestor. Moreover, the phylogenetic tree topology showing that the one-ORF situation is only found in N. meningitidis (with one exception) indicates that the fused-ORF (7A) allele has arisen recently and been disseminated in diverse N. meningitidis strains. It follows then that the overlapping ORF (8A) configuration represents the ancestral form of pglB2.

Fig 6.

Phylogenetic analysis of PglB2 and highly related orthologs. PglB2 orthologous sequences were identified through a tBlastN search and aligned. The translated protein alignment was model tested for protein evolution, and the evolutionary history was inferred. The tree with the highest log likelihood is shown. The pglB2 genes can be divided into four major groups. Green dots represent strains that possess pglB2 homologs with two ORFs separated by a gap. Green squares represent strains that contain pglB2 homologs with two ORFs which overlap without a homopolymeric tract. Green diamonds represent strains containing pglB2 homologs that have two overlapping ORFs and homopolymeric tracts; the numbers of mononucleotides (A, adenosines; G, guanosines) are indicated in the phylogenetic tree. Red dots represent a separate lineage that is solely composed of pglB2 homologs with one ORF. N. meningitidis strains ES14902 (shown here), K1207, S0108, and M6140 have identical amino acid sequences. Vibrio parahaemolyticus strains AQ3810 (shown here), AQ4037, K5030, and RIMD2210633 have identical amino acid sequences. Bacillus cereus strains Rock1-3 (shown here), Rock3-28, Rock3-29, and Rock4-18 have identical amino acid sequences.

DISCUSSION

Prior studies of N. meningitidis pglB2 have shown that this gene contributes significantly to the glycoform diversification associated with O-linked protein glycosylation (6, 17). They also raised a number of issues relating to the mechanisms of glycoform evolution, whether pglB2 might behave characteristically as a contingency locus, and how pglB2 polymorphisms might affect the integrity of the glycosylation process. In the work reported here, we addressed these concerns by examining the phenotypes associated with distinct pglB2 alleles in an in vivo-reconstituted glycosylation system. The initial important observation made here was that a natural allele containing an 8A tract was functional and associated with the predominant expression of two polypeptides while one containing a 7A tract was associated with the predominant expression of a single, multidomain protein. Further analyses and mutagenesis of the 8A tract allele identified the source of the ATP-grasp domain protein as an ORF starting at a putative initiating codon 5′ of the stop codon for the phosphoglycosyltransferase ORF. This finding was further supported by the high degree of conservation of the predicted N termini of orthologous ATP-grasp ORFs overlapping or localized just 3′ of phosphoglycosyltransferase ORFs (see Fig. S4 and S5 in the supplemental material). Further manipulation of the poly(A) tract defined four distinct mechanisms by which tract variation could affect the expression of pglB2-encoded proteins and glycosylation proficiency: (i) glycosylation via the expression of a single, multidomain protein (tracts of 4 and 7 As), (ii) glycosylation via the expression of two single-domain proteins (tracts of 5, 8, and 11 As), (iii) no glycosylation due to the failure to express PglB2b (tracts of 3 or 6 As), and (iv) glycosylation resulting from recoding at extended tracts (tracts of 9 or 12 As). These findings thus reveal a level of complexity not seen in classical contingency loci. That said, it remains unclear whether tract configurations other than 7 or 8 As occur naturally in pglB2. Configurations of 6 As are seen in some alleles within commensal Neisseria, but these genes also have compensatory changes in codons flanking the poly(A) tract that maintain an in-frame configuration of the two ORFs (Fig. 6).

A curious observation here relates to the inability to detect PglB2a in backgrounds carrying poly(A) tracts of 3, 6, and 9 As. With regard to the PglB2a polypeptide, these configurations differ from those with 5A, 8A, and 11A tracts only by a 3-amino-acid extension (IIY) due to the use of a stop codon in the altered reading frame (see Fig. S3 in the supplemental material). We suggest that the reduced levels of PglB2a in these instances may be due to intrinsic decreased stability of the extended protein itself or differential targeting for proteolysis mediated by SmpB-tmRNA trans-translation (19, 25). This situation may also be exacerbated by the relatively poor specificity of the antibody used, as well as by the fact that that the pglB/B2 loci are transcribed at moderately low levels, as seen by RNA-Seq (Nadja Heidrich and Jörg Vogel, personal communication).

An important finding stemming from these studies is that the ATP-grasp domain polypeptide (expressed either separately or in the fused state) is essential for protein glycosylation. By inference, this implies that the modification at C-4 of the sugar is essential for glycosylation. The requirement for this modification is unlikely to reflect incompatibility with the PglO oligosaccharyltransferase, as this enzyme exhibits considerable promiscuity with regard to the use of undecaprenyl-phosphate-linked donors (26). Rather, it likely stems from the inability of the phosphoglycosyltransferase to utilize the sugar lacking a C-4 modification, as enzymes of the undecaprenyl phosphate-N-acetylhexosamine-1-phosphate transferase family often display restricted UDP-d-amino sugar specificity (27). It is perhaps relevant to recall that PglB comprises a single polypeptide with an N-terminal phosphoglycosyltransferase domain and a C-terminal N-acetyltransferase domain targeting C-4 of mono-N-acetyl-bacillosamine and that the phosphoglycosyltransferase domain of PglB is highly substrate specific and exclusively transfers phospho-diNAcBac in radioactivity-based in vitro assays (13). These situations contrast with that of the N-linked protein glycosylation system of C. jejuni (that also utilizes diNAcBac), where neither the activity of PglD (the N-acetyltransferase targeting C-4 of mono-N-acetyl-bacillosamine) nor the modification at C-4 is essential (28–30).

Domain fusions involving genes wherein the products act sequentially in a biosynthetic pathway are thought to confer a selective advantage by increasing overall efficiency and tightening coordinated expression (31). As yet, we have not discerned any differences in glycosylation proficiency in the isogenic backgrounds expressing the fused versus split PglB2 forms. However, indirect evidence for such an advantage may be inferred from the apparent dissemination of the fused configuration within diverse strains of N. meningitidis (Fig. 6).

Modular domain reorganization involving gene fusion is generally accepted as playing important evolutionary roles, particularly within eubacteria. Although the scope of gene fusion (and fission), as well as potential evolutionary signals, can be identified through bioinformatic approaches, the extents to which specific molecular genetic mechanisms contribute to gene fusion are largely unknown. This situation stems ostensibly from the fact that gene fusion events can only be modeled retrospectively, based on hypothetical pre- and postfusion configurations. Moreover, these data are often derived from lineages that are broadly divergent from one another, and in the absence of selective pressure, important sequence signatures of the seminal event may rapidly decay. Perhaps the simplest event involves the straightforward fusion of adjacent, tandemly arrayed genes by virtue of rare indels or recombination events. Even in these cases, however, such a fusion requires both the loss of signals for termination of the upstream ORF and the loss of signals for initiation of the downstream ORF (32). As such, it has been proposed that gene fusion might entail an intermediate state in which the nascent fusion components coexist as distinct, overlapping genes (31). One example exists in ORFs of the rpoB and rpoC genes that are found in either overlapping or fused configurations within genera within the epsilonproteobacteria but are separated by a 20- to 100-bp gap in most bacteria (33). It has been suggested that an insertion or deletion event occurring within a shared ancestor could have led to the fusion seen in the Helicobacteraceae. However, sequence comparisons of the fused forms seen in Helicobacteraceae with the overlapping forms in Campylobacteraceae fail to provide evidence for a simple, seminal event. Thus, the data from pglB2 provide unique insights into the incremental nature of real-time domain fusion events. Moreover, the results reveal how the rates of gene fusion might be accelerated due to the hypermutability of simple sequence repeats. In addition, the findings suggest that recoding might facilitate the coordinated expression of overlapping genes prior to their fusion.