Protein glycosylation has been described in all domains of life. For example, studies have demonstrated that N-linked glycosyltransferases (GTs) are present in prokaryotes and have high similarity to the eukaryotic STT3 family [23, 24]. Furthermore, the discovery of N-linked OSTs in archaebacteria and bacteria increases the understanding of the mechanism of action of OSTs and their potential application [25]. Well-characterized prokaryotic OSTs, oligosaccharyltransferase (AglB) in archaea and PglB in eubacteria, catalyze the transfer of oligosaccharides to the amide moiety of asparagine (N).

The mechanism of N-linked glycosylation is conserved in bacterial species and has been well-characterized in C. jejuni, which possesses a 17-kb pgl gene cluster that encodes proteins for synthesizing and transferring oligosaccharides (Fig. 1A). The process of N-linked glycosylation in C. jejuni could be divided into two steps, namely oligosaccharide synthesis, and transfer (Fig. 1B). Oligosaccharide synthesis begins with modification of UDP-GlcNAc to UDP-2,4-diacetamidoglucosamine (UDP-diNAcBac) by glycosyltransferases: PglF, PglE, and PglD, followed by partial transfer of UDP-diNAcBac by PglC onto undecenol diphosphate. The oligosaccharyltransferase (PglA, PglJ, and PglH enzymes) enzymes subsequently catalyze the formation of a hexasaccharide by the fusion of five units of GalNAc. LLO synthesis is completed by PglI, which adds a branched glucose residue to the hexasaccharide [21, 26, 27, 28]. The LLO is then flipped across the inner membrane into the periplasm by PglK [29, 30], and transferred to the substrate asparagine of the - D/E${}_{\text{X}}$N${}_{\text{X}}$S/T -consensus by PglB, which is the final key step in N-linked glycosylation [31].

Fig. 1.

Mechanism of N-linked glycosylation mediated by the PglB Oligosaccharyltransferase (OST) of C. jejuni. (A) The 17 kb protein glycosylation locus pgl of C. jejuni. N-linked glycosylation in C. jejuni is cooperatively accomplished by the pgl gene cluster. PglB is the core factor for the glycosylation process. (B) Schematic diagram depicting N-linked glycosylation in C. jejuni. The first step of glycosylation involves oligosaccharide synthesis in the cytoplasm, which is mediated by glycosyltransferases (GTs), including PglF/E/D, PglC, PglA/J/H, PglI, and PglK. The transfer of oligosaccharides to the periplasm is carried out by PglK. Subsequently, PglB transfers the hexasaccharide to the asparagine residue of the substrate with the sequence D/ExNxS/T.

The mechanism of bacterial N-linked glycosylation by OSTs has been investigated in detail by determining the structures of the AglB enzyme of Archaeoglobus fulgidus and the PglB enzymes of N. gonorrhoeae, C. jejuni, and C. lari. The full-length structures of AglB and PglB comprise an N-terminal transmembrane domain and a C-terminal periplasmic domain (Fig. 2A). The C-terminal domain of the AglB protein of Ar. fulgidus and the PglB enzyme of C. jejuni share a high structural similarity, with both proteins containing a central core (CC) domain and an inserted $\beta$-structure. In particular, the catalytic sites of the WWDYG motif are conserved in the AglB and PglB enzymes of Ar. fulgidus and C. jejuni, respectively [32, 33]. However, the isolated C-terminal domains of AglB and PglB lack catalytic activity and are unable to bind the acceptor peptide, suggesting that the N-terminal transmembrane domain is essential for catalytic activity. The full-length structure of the PglB protein of C. lari complexed with a hexapeptide has been solved at 3.4 Å resolution. The N-terminal transmembrane domain of the PglB protein of C. lari comprises 13 transmembrane helices, an external EL1 helix, and a long external EL5 loop (Fig. 2B). The structure of the complex also highlighted the specificity of the DQNATF substrate motif and demonstrated that the WWD motif of the periplasmic domain of PglB forms three hydrogen bonds with the $\beta$-hydroxyl group of the +2 Thr residue of the hexapeptide (Fig. 2C) [34].

Fig. 2.

The structure of the PglB enzyme of C. jejuni and the ternary complex of PglB bound to a substrate DQNATF peptide and an lipid-linked oligosaccharide (LLO) analog. (A) Schematic diagram of the PglB enzymes of C. lari and C. jejuni and the AglB protein of Ar. fulgidus. All the enzymes possess a transmembrane domain and a periplasmic domain, while the AglB protein of Ar. fulgidus comprises two more peripheral domains, P1 and P2. (B) Crystal structure of the PglB protein of C. jejuni (PDB ID: 3RCE). The PglB enzyme of C. jejuni comprises 13 transmembrane helices, EL1, and EL5. EL1 is a periplasmic helix and EL5 is a periplasmic loop located between the transmembrane helices. The periplasmic EL1 and EL5 domains are considered to be important for peptide recognition and LLO binding. (C) Crystal structure of the ternary complex of C. jejuni PglB with a substrate DQNATF peptide and an LLO analog (PDB ID: 5OGL). The substrate peptide and the LLO analog in the ternary complex are situated opposite to each other in the platform formed by the transmembrane and periplasmic domains, and the distance between the asparagine residue and LLO analog is 6 Å. The WWD motif in the periplasmic domain is important for peptide binding, and forms three hydrogen bonds with the $\beta$-hydroxyl group of the +2 Thr residue of the bound hexapeptide. The structure of the EL5 monomer comprises N-EL5 and C-EL5, which are essential for peptide and LLO binding, of which the E319 residue of C-EL5 and Y293 of N-EL5 are particularly important. Color scheme: transmembrane domain (green), periplasmic domain (gray), substrate peptide (cyan), LLO analog (blue), N-EL5 (purple), and C-EL5 (red).

Interestingly, the long external EL5 loop is partially ordered, and 25 residues of the EL5 loop are disordered in the electron density map [34]. However, a study demonstrated that EL5 contains a conserved tyrosine residue (Tyr293), and the Y293A mutation leads to an almost complete loss of catalytic activity, confirming that the aromatic side chain of Y293 is essential for catalysis [35]. Moreover, LLO binding induces dynamic changes in the EL5 loop, resulting in movements in the transmembrane domain, which progressively induces the formation of an open conformation of EL5 [36]. In order to obtain further insights into the mechanism of LLO transfer and the function of the EL5 loop, Napiórkowska et al. [37] solved the ternary complex of PglB bound to an acceptor DQNATF peptide and a synthetic nonhydrolyzable ($\omega$ZZZ)-PPC-GlcNAc LLO analog. The ternary structure revealed that the acceptor peptide and the LLO analog were bound to two cavities at opposite sides of the PglB enzyme, and the shortest distance between the peptide and the LLO was approximately 6 Å. The structure revealed that the acceptor peptide interacted with the WWD motif of the periplasmic domain. In addition, E319 of C-EL5 and D56 are located close to the carboxamide group of the acceptor asparagine, and mutations in D56 and E319 are reported to impair the catalytical activity of PglB [34, 37]. Analyses of the interactions of the LLO analog revealed that Y293 of N-EL5 formed a hydrogen bond with the oxygen atom of the second pyrophosphate, while R375 formed a hydrogen bond with the oxygen atom of the first pyrophosphate, which explained the complete abolishment of catalytic activity of PglB in the Y293A mutant [37]. Napiórkowska et al. [38] solved a new ternary complex of PglB in which the distance between the substrates and LLO analog is significantly closer (approximately 3.4 Å). The ternary complex comprises PglB, Dab-containing peptides, and a synthetic $\omega$ZZZ-PP-GlcNAc LLO analog. Structural analysis demonstrated that the reduction in the distance between the substrate and LLO analog was attributed to the fact that the pyrophosphate and GlcNAc moieties shifted closer to the catalytic center [38].

Metal ions are essential for enzymatic activity, and it has been reported that Zn${}^{2+}$ and Mg${}^{2+}$ metal ions are bound to the catalytic center of AglB and PglB, respectively. DXD (Asp-X-Asp) motifs such as Gly-X-Asp${}^{47}$ and Asp-His-His${}^{163}$ in AglB, and the Thr-Asn-Asp${}^{56}$ and Asp-Thr-Asp${}^{156}$ motifs in PglB, are considered to be pivotal for coordinating the metal ions based on the proximity of their side chains [34, 39]. The PglB protein of C. lari possesses a conserved short loop with a DGGK motif in the periplasmic domain of all C. sp., which mediates the binding of the substrate peptide and LLO analog [37, 40].

The PilO enzyme of P. aeruginosa comprises a transmembrane domain that anchors to the periplasmic membrane, and a periplasmic core domain that catalyzes the transfer of glycans to pilin via its Wzy_C domain (Fig. 3A) [47]. Analysis of the transmembrane helix of PilO revealed that the transmembrane region includes residues 1-266, and the residues include nine transmembrane helices. Residues 267-461 of PilO have some amino acids that are less hydrophilic and form four transmembrane domains, in which there are two periplasmic loops and two cytoplasmic loops. A subsequent BLAST search of the Wzy_C structural domain of PilO (residues 275-325) revealed that residues 281-301 are conserved and have no homologous sequences. In order to investigate whether this region affects glycosylation, mutations were introduced at different positions. Residues 281-287 (${}^{281}$RDVLWRD${}^{287}$) were deleted in the first mutant, while residue R281 was mutated to alanine in the second mutant (pR281A). The results demonstrated that mutations in these regions prevented substrate glycosylation, and that these residues are necessary for glycosyltranserases (GT) activity. In addition, other catalytic residues that play a role in the glycosylation of pilin were deleted, and 1-322, 1-430, and 1-453 truncates were constructed. The results demonstrated that none of the truncates exhibited GT activity. Altogether, these findings suggested that residues 267-461 are important for the GT activity of the enzyme [47]. PilO exhibits non-specificity to substrates, only recognizing the C-terminal serine residue of the target peptides, which makes it a potential candidate for glycoprotein vaccine development [48]. Interestingly, the PilO enzyme transfers short glycan chains to an acceptor serine residue, while the PglL enzyme transfers polysaccharides [49].

N. meningitidis is a gram-negative bacterium that causes meningitis and is a common commensal in the human respiratory tract. The genes and phenotype of N. meningitidis have been reported to undergo in vivo alteration within a very short time to evade the host immune system and enhance virulence [50]. The O-linked protein glycosylation system affects polysaccharide diversity owing to selective gene expression. It has been reported that ST-11 strains of N. meningitidis express 7 to 21 different glycans, resulting from the diversity and variation of pgl genes [51]. The PglL enzyme of N. meningitidis contains multiple consecutive transmembrane domains and two periplasmic domains (Fig. 3A). The PglL enzyme has been demonstrated to be responsible for attaching UndPP-glycan to pilin (PilE), and Ser63 is one of the glycosylation sites on PilE [52, 53] (Fig. 3B). The glycosyltransferase (GT) activity of PglB transfers 2,4-diacetamido-2,4,6-trideoxyhexose (DATDH) to undecaprenol diphosphate [49, 54]. The PglA and PglE GTs subsequently add two monosaccharide units to the side chain of the sugar moiety. The trisaccharide is finally flipped by the PglF GT and transferred to the serine residue of its substrate by PglL (Fig. 3B). The final recognition motif of PglL on the exotoxin A (EPA) protein of Pseudomonas aeruginosa was initially identified as ${}^{57}$WPGNNTSAGV${}^{66}$, which was optimized to an 8-residue-long sequence (WPAAASAP) referred to as the minimum optimal O-linked recognition (MOOR) motif [55]. AniA, an outer membrane nitrite reductase, is glycosylated at serine${}^{373}$ and serine${}^{378}$ residues of the L${}_{358}$SDTAYAGNGAAPAASAPAASAPAASASEK${}_{387}$ peptide by PglL. The sequence contains a repetitive AASAP motif with the glycosylated serine residue at the middle of the residue stretch. However, subsequent studies with the ${}_{381}$AASAS${}_{385}$ motif, which contains two serine residues that were not glycosylated by PglL [56]. These results suggested that the AASAP peptide might be the optimum peptide sequence for PglL.

However, full-length structures of the PglL/PilO enzymes or PglL/PilO enzymes bound to the substrate peptide complex have not been solved to date. These structures are necessary for further studies on elucidating the mechanism of peptide binding and glycan transfer by PglL/PilO.

V. cholerae is a gram-negative pathogen causing cholera, a water-borne diarrheal disease in animals and humans [57]. Bioinformatics analysis identified the VC0393 gene, which encodes O-OTases in the O1 E1 strain of V. cholerae, and is referred to as PglL${}_{\text{Vc}}$. When the foreign substrate protein DsbA1 was introduced, the glycosylation site by PglLvc was mapped as VQTSVPADSAPAASAAAAPAGLVEGQNYTVLANPIPQQQAGK [58]. PglL${}_{\text{Vc}}$ glycosylates multiple proteins of V. cholerae, including the outer membrane proteins OmpA and OmpU, the periplasmic chaperones SurA-like and FkpA, and the DegP chaperone/protease. Using DegP as a representative candidate for identifying glycosylation sites, a study demonstrated that the peptides ${}^{47}$KVTPAVVSIAVE${}^{58}$ and ${}^{371}$SLHQGLSGAE${}^{380}$ are modified by glycosylation with a diNAcBac moiety at a single site [59]. However, the specific structural model is unknown thus far.

O-linked glycosylation systems have also been detected in gram-positive bacteria, including L. monocytogenes, Clostridium difficile, Staphylococcus aureus, and Streptococcus agalactiae [13, 60]. The flagellin protein of L. monocytogenes is modified by the addition of O-linked N-acetylglucosamine (GlcNAc) moieties at serine and threonine residues, which is regulated by the GmaR GT [61, 62]. GmaR is a bifunctional protein in which one hand is a temperature sensor that controls the temperature-dependent transcription of flagellar motility genes by interacting with MogR via conformational changes [63, 64]; the other hand is a GT-2 family GT that glycosylates the flagellin protein FlaA. Unlike the PilO or PglL enzymes that transfer oligosaccharides to the substrate peptides, GmaR only transfers a mono-GlcNAc to serine and threonine residues of FlgA. The secondary structure of GmaR comprises an N-terminal domain for GT activity, a C-terminal domain that binds MogR for temperature sensing, and a linker region. Small-angle X-ray scattering (SAXS) analysis revealed that the apo GmaR protein adopts a multidomain shape, and the binding of Mg${}^{2+}$ ions and donor substrate induces conformational changes in GmaR so that the protein adopts an elongated and relatively open conformation. The linker region is essential for binding Mg${}^{2+}$ and the donor substrate, and the N-terminal ${}^{83}$DXD${}^{85}$ motif is critical for metal-dependent enzymatic activity [60].

PglB${}_{Cj}$ can transfer the natural C. jejuni polysaccharide, the antigenic capsular polysaccharide of Gram-positive bacteria, and the antigenic lipopolysaccharide of Gram-negative bacteria to the specific recognition sequence of the substrate [71]. There are currently many recombinant vaccines produced by the C. jejuni N-glycosylation system; for example, Williams AJ et al. [72] used PglB${}_{Cj}$ to produce a sugar-conjugated vaccine against Enterotoxigenic E. coli in non-pathogenic E. coli. These successfully induced an O-antigen polysaccharide-specific response in mice, producing bactericidally active IgG antibodies [72]. In addition, Huttner et al. [73] describe the early development of a conjugate vaccine against extraenteric pathogenic E. coli (ExPEC), from creation in the laboratory to testing in a large first-in-human phase Ib trial (Table 1). Recombinant vaccine candidates against B. pseudomallei were developed by utilizing the O-polysaccharide (OPS II) of B. pseudomallei, along with the PglB enzyme of C. jejuni and the appropriate protein carrier (AcrA) of the C. jejuni. This recombinant vaccine produced an IgG immune response during intranasal challenge and protected against B. pseudomallei [74]. A glycoconjugate vaccine candidate was developed against Actinobacillus pleuropnuemoniae by combining a fragment of the Apx toxin with a conserved glycan, which can potentially confer enhanced protection against all serotypes of Act. pleuropneumoniae [75]. Shigella sp are gram-negative pathogen that causes severe diseases in humans. A new recombinant glycoconjugate vaccine, (EPA-2a), produced by combining S. flexneri 2a with the EPA carrier protein of P. aeruginosa using the PglB OST of C. jejuni enhanced specific protection against S. flexneri 2a [76]. Subsequent trials showed that the vaccine elicited antibodies in a ratmodel and were safe and immunogenic. [77]. Moreover, Riddle et al. [78] verified the safety and bactericidal activity of the vaccine through phase I clinical trials, and induced functional antibodies, which provided a reference for further production of this technology (Table 1).

Although PglB${}_{\text{Cj}}$ has been successfully used in some vaccine production, there are limitations in the production reaction. The sugar chain that can be recognized by C. jejuni PglB has an acetyl group at the C${}_2$-N position of the reducing end and forms a stable hydrogen bond with the carrier protein. If the C${}_2$-N acetyl group at the reducing end is missing or the $\beta$-(1$\to$4) glycan is missing in the two monosaccharides before the reducing end, the glycosylation efficiency will be reduced [79, 80]. For example, Salmonella enterica sv. Typhimurium LT2 and E. coli capsular polysaccharide K30 lack the C${}_2$-N acetyl group at the reducing end, so the signal of glycosylation cannot be detected on the substrate. This limits the selection of antigenic polysaccharides and hinders the large-scale production of recombinant vaccines. While Ihssen J et al. [81] discovered that mutations in the key amino acids (N311V, Y77H, and S80R) in wild-type C. jejuni PglB enhanced glycosylation efficiency, the underlying mechanism remains unclear. Furthermore, these mutations may also result in non-specific glycosylation of PglB [81].

Compared to N-linked glycosylation, there is a greater flexibility of glycan structures and protein acceptors in O-linked glycosylation. Faridmoayer et al. [49] demonstrated that in engineered E. coli and Salmonella cells, PglL can transfer any sugar carried by undecaprenyl pyrophosphate (UndPP) to the pilin substrate. Considering whether the lipid carrier was a limiting factor for PglL, the authors established the authors established a glycosylation system in vitro using purified PglL, pilin, and chemically synthesized lipid farnesyl pyrophosphate (FarPP) with pentasaccharides. The results indicated that synthetic lipid carriers could also transfer polysaccharides to the periplasm and localize polysaccharides around PglL. PglL is also broadly selective for vectors. The flexibility of PglL is likely to play a role in the production of future glycoconjugate vaccines [44]. Therefore, it has a higher application potential in the development of new conjugate vaccines. Salmonella enteritidis (SE) is a zoonotic pathogen. Li M et al. [82] constructed the PglL plasmid, recombinant Pseudomonas aeruginosa exotoxin A (rEPA), and cholera toxin B subunit (CTB) plasmids into S. enteritidis strains. It was detected that when PglL was expressed, rEPA and CTB could undergo glycosylation. The results provide the basis for the SE glycoprotein biosynthesis [82]. A conjugate vaccine against S. enterica serovar Paratyphi A produced by the co-expression of PglL and the CTB4573 carrier protein in vivo evokes a protective and specific immune response against the bacterium [83]. A polysaccharide conjugate vaccine against Klebsiella pneumoniae serotype O2 strain was developed by transfecting a K. pneumoniae waaL mutant with a vector encoding PglL and a recombinant cholera toxin B subunit, and the vaccine conferred effective protection in a murine model (Table 1) [84, 85]. Brucellosis is a common human and animal infectious disease worldwide, endangering human health. There is currently no effective vaccine against this disease. Huang et al. [86] designed a vaccine by replacing the Brucella O-antigen polysaccharide (OPS) with the low-pathogenic Yersinia enterocolitica serotype O:9 (YeO9) OPS via an O-linked glycosylation system. As a result, specific immunity was successfully induced, effectively defending against brucellosis and avoiding large-scale cultivation of pathogenic strains [86]. In order to further mass-produce biological conjugate vaccines against B. abortus, Li S et al. [87] established a new vaccine engineering production process in 2023. The OPS gene cluster was divided into five separate fragments and then reassembled, and subsequently expressed in E. coli. The final product is glycosylated by PglL. A series of experiments proved that the vaccine can induce a strong immune response, which induces the production of specific antibodies [87]. Other glycosyltransferases in the O-linked glycosylation system have also been used in vaccine development in recent years. For example TfpM from Moraxella osloensis. Knoot CJ et al. [88] found that TfpM-derived bioconjugates can trigger specific antibody IgG production in mice and are immunogenic. Because a variety of long-chain polysaccharides can be transferred to the substrate protein carrier, the shortest recognition glycosylation sequence is identified. This provides new options for the production of new sugar-conjugated vaccines in the future [88].

Another glycosyltransferase, PglS, can transfer many different types of bacterial glycans, including those with glucose at the reducing end. For example, Feldman et al. [89] successfully developed a bivalent K1/K2 Klebsiella pneumoniae biological conjugate vaccine with immunogenicity and effectiveness through PglS. Vaccines protect mice against infection by hypervirulent K. pneumoniae (hvKp). Although in vivo work suggests vaccine efficacy against hvKp in the lung, further studies are needed to assess efficacy against hvKp in other sites including the liver, blood, and meninges (Table 1) [89]. Harding CM et al. [90] utilized PglS to successfully produce sugar-conjugated vaccines with glucose polysaccharides at the reducing end, thereby overcoming the limitations imposed by the structural requirements of polysaccharides. Furthermore, PglS demonstrated the ability to glycosylate various vaccine carrier proteins. Inoculating mice with the polyvalent pneumococcal bioconjugate vaccine resulted in the production of antisera that effectively killed bacteria in vitro [90]. Group B Streptococcus (GBS) is a leading cause of neonatal infections and invasive diseases in nonpregnant adults worldwide. Duke et al. [91] expressed GBS types Ia, Ib, and III in recombinant E. coli and studied the generation of bioconjugate vaccines using PglS conjugation with the carrier protein Pseudomonas aeruginosa exotoxin A (EPA). This technology provides a basis for the subsequent development of multivalent GBS bioconjugate vaccines [91]. These results support our expansion to future sugar-conjugate vaccines.