Introduction

Bacteria belonging to the genus Salmonella are ubiquitous. They cause a range of human infections including gastroenteritis and bacteremia (usually by foodborne, non-typhoidal strains), and enteric fever (by typhoidal strains)1. The Salmonella lineage contains two species: enterica and bongori. The latter almost exclusively infect cold-blooded animals, although occasional human outbreaks are reported2, while S. enterica contains the predominant human pathogens and is made up of six subspecies (enterica (I), salamae (II), arizonae (IIIa), diarizonae (IIIb), houtenae (IV), and indica (VI). The subspecies are further categorized into serotypes via the Kauffmann–White scheme, based on different combinations of lipopolysaccharide (LPS) O-antigens and flagellar filament H-antigens. These define more that 2600 serovars often named after the geographic location where they were isolated3.

LPS O-antigens (O-polysaccharides; OPSs) are the major surface polysaccharide structures present in Salmonella (and in many other Gram-negative genera). They are composed of repeat units whose variable structures provide distinct antigenic epitopes defining the serotype specific portion of LPS4. Currently, 46 structurally characterized OPS structures have been reported for S. enterica serogroups and the genetic regions driving their biosynthesis have been identified5. However, a recent pangenome study suggests further diversity may exist5,6. The majority of Salmonella O-antigens are synthesized by the Wzx/Wzy-dependent assembly strategy, where OPS repeat units are built on an undecaprenyl phosphate (und-P) acceptor at the interface of the cytoplasm and the cell membrane (reviewed in4). These are subsequently exported by a Wzx flippase and polymerized by the Wzy polymerase at the periplasmic face of the membrane, before being ligated to the conserved portion of the LPS glycolipid (consisting of lipid A and a core oligosaccharide containing the OPS attachment site). Serotype O67 is the only example that uses a different assembly pathway, where the und-linked glycan is polymerized in the cytoplasm and exported to the periplasm via an ATP-binding cassette (ABC) transporter for ligation5. The completed LPS is then translocated to the cell surface.

While OPSs are distinguishing polysaccharide antigens in Salmonella serotypes, they can be co-expressed with other glycans. All isolates produce Enterobacterial Common Antigen, which they share with other members of the Enterobacteriaceae; the precise cellular role(s) for this polysaccharide is uncertain7. Salmonella also make a colanic acid exopolysaccharide, regulated by the Rcs regulon8. This is absent in human-adapted Salmonella Typhi and Paratyphi C, which produce the host immune response-shielding Vi antigen capsular polysaccharide9. Finally, some isolates also produce T antigens (where T denotes their apparent transient nature) but the structure(s) and function(s) of these antigens is enigmatic. The T1 antigen was the first representative identified, found in a strain of Paratyphi B that was not typable by the available O-antisera10. A subsequent study employed methylation analysis to describe an incomplete T1 structure containing a complex mix of galactofuranose (Galf) and ribofuranose (Ribf) residues and it was hypothesized that the T1 glycan might be linked to the LPS lipid A–core11. Further studies described preliminary but inconclusive attempts to identify the precursors for T1-antigen biosynthesis, as well as the consequences of T1 expression on virulence and phage-sensitivity12,13,14,15.

Surprisingly, investigation of the T1 antigen ended in the early 1970s without resolving its complete structure or understanding its relationship to serotype-specific OPS. Furthermore, the genetic locus responsible for its production and the molecular basis for its transient nature was never identified. The goal of this study was to address these open questions to provide a platform for investigating the biological functions of this unusual glycan.

Results

Identification of the Salmonella T1 biosynthetic cluster encoding an ABC transporter-dependent system

We recently reported the first structure and function of a family of ribofuranosyltransferase (Ribf-GT) enzymes involved in polysaccharide bacterial biosynthesis16. The prototypical dual catalytic domain Ribf-GTs contain a C-terminal glycan phosphoribosyltransferase (gPRT), which transfers ribose-5-phosphate from the activated donor, phosphoribosyl-5-phospho-D-ribosyl-α−1-diphosphate (PRPP), to a sugar acceptor. In a second reaction step, the 5′ phosphate is then removed by an N-terminal phosphoribose phosphatase (PRP) catalytic module. Because the T1 antigen was previously reported to contain Ribf11, the biochemically defined Ribf-transferase prototype from Klebsiella pneumoniae16 was exploited in a BLAST search to identify a homolog in Salmonella. A candidate Ribf-transferase gene was found in serovar Paratyphi B, the original T1 source10 but, surprisingly, additional candidates were identified in genomes from a wide range of Salmonella serovars. The candidate Ribf-transferases share 97–100% identity to one another (with 100% coverage) and are encoded by a gene in a well conserved locus located between nei (encoding endonuclease VIII) and pxpA (5-oxoprolinase subunit) (Fig. 1a). The locus also contained genes associated with production of Galf-containing polysaccharides in other bacteria (see below), lending confidence that it was responsible for T1-antigen production. Unfortunately, we were unable to obtain the original T1-producing Paratyphi B isolate, so the 11-orf locus from the well-studied S. enterica serovar Typhimurium LT2 (strain SL3770) was used as a prototype for subsequent investigation. At the start of this study the molecular basis for the transient T1 expression in Salmonella was unknown, thus precluding the use of the natural host for the primary investigations. We therefore chose to perform the analysis in a recombinant E. coli K-12 background with the 11-orf locus introduced on a plasmid. The host glycosylation machineries in E. coli K-12 are clearly understood, allowing confident data interpretation.

Fig. 1: Genetic region and biosynthesis of the Salmonella T1 antigen.
figure 1

a The T1 antigen biosynthesis cluster in Salmonella enterica serovar Typhimurium SL3770 (LT2) (NC_003197.2). The cluster is a highly conserved ~12,500 base pair region between pxpA and nei in Salmonella. STM0726 (in bold) encodes the Ribf-transferase used to identify the cluster. Genes responsible for galactofuranose addition are colored yellow, and genes responsible for ribofuranose addition are shown in pink. The ABC transporter components are shown in green and the tyrosine recombinase (characterized in Fig. 5) is shown in red. b The biosynthetic pathway for T1 antigen biosynthesis established in this study. c Cartoon representation of the biosynthesis and export steps of the T1 antigen. Note that biosynthesis and export are separated for simplicity, but in the cell these processes are expected to be obligatorily coupled.

Full size image

To establish the function of the Salmonella locus, the 11-orf region was cloned into the pACYC184 vector, with the genes downstream of the tetR promoter, permitting expression from the tetR promoter, or from any native promoters within the cloned region. The recombinant plasmid pWQ1128 was transformed into E. coli DH5α, which (like other K-12 strains) lacks a native OPS17. Examination of LPS profiles of whole cell lysates in silver-stained SDS-PAGE (Fig. 2a) revealed that the expression of the 11-orf gene region diminished the amount of lipid A-core species and led to the production of high-molecular-weight molecules with a heterogenous size distribution, typical for the addition of glycan chains (and seen typically with OPS). As described below, eight of the eleven orfs in this region (beginning with STM0719) encode putative proteins sharing similarity to characterized participants in glycan biosynthesis and export in other bacteria (Fig. 1b, c). A DNA fragment containing STM0719–STM0726 (cloned in pWQ1129) produced a similar profile to pWQ1128 (Fig. 2a), so pWQ1129 was used as the basis for other constructs. Addition of OPS to lipid A-core depends on WaaL, the O-antigen ligase enzyme, which attaches OPS to the core oligosaccharide in the periplasm (i.e., post transport)4. The high-molecular-weight product was lacking when the locus was expressed in a waaL mutant (Supplementary Fig. 1a), confirming the T1 antigen is presented like conventional OPS and does not possess a different lipid terminus like that used in the Salmonella Vi antigen glycolipid18.

Fig. 2: Production of the T1 antigen in E. coli K-12.
figure 2

a Silver stained SDS-PAGE gel and western immunoblots demonstrating recombinant expression of the T1 antigen in E. coli DH5α, which lack their native OPS due to mutations in the ancestral O16 gene cluster17. Samples were prepared as whole-cell lysates and similar results were acquired in triplicate. Rabbit T1 antiserum was raised against E. coli DH5α transformed with pWQ1128 (STM0719-STM0726). Ribf-PS antiserum was produced by adsorbing the T1 antiserum against E. coli DH5α expressing pWQ1130 (STM0719-STM0724) to remove antibodies recognizing Galf-PS and therefore also shows no reactivity to the core-oligosaccharide. Source data are provided as a Source Data file. b NMR spectroscopic analysis of the T1 antigen. The spectra reveal a mixture of two Galf-PS species and one Ribf-PS species, with the relevant structures shown above the spectra. Repeat unit assignments are shown on the HSQC spectrum and sugar linkage points are indicated on the HMBC and NOESY spectra. Chemical shifts are reported in Supplementary Table 1.

Full size image

STM0722 and STM0723 are homologs of wzm and wzt (Fig. 1a), encoding the transmembrane and nucleotide-binding polypeptides for an ABC transporter, defining one of the major strategies for surface polysaccharide assembly in bacteria4 (Fig. 1c). These ABC transporters export undecaprenyl diphosphate-linked glycans (including some OPSs) to the periplasm. In most examples, the synthesis of the lipid intermediate is initiated by a homolog of WecA, a UDP-N-acetylglucosamine (GlcNAc)-dependent phosphoglycosyltransferase that transfers GlcNAc-1-P to undecaprenyl-phosphate4 (Fig. 1c). WecA is encoded by a gene in the enterobacterial common antigen cluster located elsewhere on the Salmonella genome7. Consistent with this, no high-molecular-weight products are made in an E. coli wecA mutant transformed with pWQ1129 (Supplementary Fig. 1a). Six genes in the Salmonella locus (STM0719–STM0721 and STM0724–STM0726) encode enzymes required for the biosynthesis of the T1 antigen (Fig. 1b) and their functions are described below.

The T1 antigen contains galactofuranose and ribofuranose homopolymeric domains

To determine the structure of the T1 antigen, LPS containing T1 was produced in E. coli K-12 harboring plasmid pWQ1129 and purified. The T1 antigen was released from the LPS by mild acid hydrolysis and analyzed by a combination of 1- and 2- dimensional 1H and 13C NMR spectroscopic experiments (Fig. 2b, Supplementary Table 1). These data showed that T1 was composed of two homopolymeric components; one containing Galf and the other Ribf, with integration of the anomeric peaks indicating Ribf is present in ~2:1 abundance of Galf. The Galf-PS was a heterogenous mixture of two species. The first contained a disaccharide repeat unit [→3)-β-d-Galf-(1 → 6)-β-d-Galf-(1 → ] (residues A and B), while the second had an additional (1 → 3)-linked Galf residue, resulting in a trisaccharide repeat unit [→3)-β-d-Galf-(1 → 3)-β-d-Galf-(1 → 6)-β-d-Galf-(1 → ] (residues A’ and B’ and C). The Ribf-PS polymer contained a single anomeric signal (residue D) and was determined to be a β-(1 → 2)-linked homopolymer. Briefly, data from 1H–1H-COSY and 1H–1H–TOCSY experiments were used to assign the individual residues. To assign the linkage between the residues, 1H–13C HMBC (Fig. 2 and 1H–1H NOESY data were employed (Fig. 2). The H1 of residue A showed a correlation at δH 5.08/δC 70.2 in the HMBC and δH 5.08/δH 4.21 in the NOESY, corresponding to a linkage at C6 of residue B and B’. In turn, H1 of residues B and B’ showed correlations of δH 5.17/δC 83.6 and δH 5.18/δC 83.6 in the HMBC and δH 5.17/δH 4.17 in the NOESY, demonstrating the linkages to C3 of residue A and C3 of residue C, respectively. Finally, residue C showed a correlation at δH 5.21/δC 83.3 in the HMBC and δH 5.21/δH 4.17 in the NOESY, corresponding to a linkage at C3 of residue A’. Residue D showed no major correlations to any of the Galf-PS residues but instead had a strong correlation between H1/C2 at δH 5.40/δC 82.1 in HMBC and H1/H2 δH 5.40/4.21 in NOESY. Furthermore, the downfield shift of H2/C2 compared to literature data for unsubstituted Ribf glycosides19 confirmed that the Ribf-PS was a (1 → 2)-linked homopolymer. Comparison of all anomeric signals with literature data were consistent with 1,2-trans (β) configurations for all glycosidic linkages (i.e., all appeared as singlets)20. However, the NMR data could not distinguish between two organizational scenarios. In one, the two structural domains are arranged in tandem. Alternatively, the T1 antigen may contain two populations, one with a linked Ribf-PS and the other with Galf-PS. This was resolved by a combination of genetic and biochemical analyses.

The Galf-PS is linked to lipid A-core via an adapter trisaccharide

The STM0719 gene is predicted to be encode a Glf (UDP-galactopyranose mutase) enzyme. The protein shares 75% identity, 85% similarity (96% coverage) with the well characterized K. pneumoniae Glf that produces UDP-Galf from UDP-Galp21. STM0720 (WbbN; 59% identity, 75% similarity, 100% coverage) and STM0721 (WbbO; 70% identity, 85% similarity, 99% coverage) are predicted UDP-Galf- and UDP-Galp-dependent GTs involved in biosynthesis of the K. pneumoniae O2a antigen. In O2a biosynthesis, these three proteins cooperate with WecA to generate an undecaprenyl diphosphate-linked trisaccharide (Galf-Galp-GlcNAc-PP-und)22. The Salmonella proteins possess the same biochemical activities (shown in the pathway scheme in Fig. 1b,c), based on the ability of the Salmonella homologs to restore O2a-antigen biosynthesis in mutants with corresponding deletions in the Klebsiella operon (Supplementary Fig. 1b). Furthermore, transformation of E. coli DH5α with a plasmid containing the STM0719–STM0723 fragment (pWQ1130) results in a new product whose migration in SDS-PAGE is consistent with the addition of the predicted trisaccharide to lipid A-core (Fig. 2a).

In Klebsiella O2a biosynthesis, a dual catalytic module polymerase (WbbM) assembles a glycan with a disaccharide repeat unit [→3)-β-d-Galf-(1 → 3)-α-d-Galp-(1 → ] using Galf-Galp-GlcNAc-PP-und produced by WecA-WbbO-WbbN as an acceptor. The T1-antigen pathway apparently uses the same acceptor for a different structure, and STM0724 was predicted to provide the putative polymerase for the repeat-unit domain based on its similarity to the GlfT2 GT from Mycobacterium tuberculosis. GlfT2 polymerizes Galf residues with alternating β-(1 → 5) and β-(1 → 6) linkages, to build the backbone of the cell wall arabinogalactan, which is essential for viability of mycobacteria23. The predicted STM0724 protein and GlfT2 share 27% sequence identity and 42% similarity (Supplementary Fig. 2). Furthermore, an AlphaFold model of STM0724 shares structural similarity with the experimentally derived structure of GlfT2 (Fig. 3a)23. Transformation of E. coli DH5α with a plasmid containing STM0719–STM0724 resulted in an SDS-PAGE LPS profile that retained a cluster of intermediate molecular weight species (larger than the presumed trisaccharide formed in the absence of STM0724). Compared to transformants containing the full locus, this material showed reduced reactivity with antiserum raised against cells expressing the T1 antigen (Fig. 2a). This would be anticipated if the LPS contained only a partial T1 structure, lacking some antigenic epitopes. These observations led to the hypothesis that STM0724 is the polymerase for the biosynthesis of the Galf-PS component of T1 antigen.

Fig. 3: STM0724 is the T1 Galf-PS polymerase and an ortholog of mycobacterial GlfT2.
figure 3

a Crystal structure of Mycobacterium tuberculosis GlfT2 (pink)(PDB ID: 4fix) and AlphaFold model of STM0724 (blue)(Uniprot Q8ZQV0). The proteins share 21% sequence identity and have a Z-score of 32.5, with an rmsd of 3.3 Å. b HPLC traces of in vitro GT reactions containing STM0724, UDP-Galf (synthesized in situ by Glf from UDP-Galp) and acceptor compound 1. Reactions were performed for 1 h at 30 °C, stopped by the addition of an equal volume of acetonitrile, and the mixture analyzed by HPLC, using detection of the methoxybenzamide tag in the acceptor. The results from triplicate reactions were identical. Also shown is the product purified from large-scale reactions by gel filtration and SepPak elution. c NMR analysis of purified STM0724 product. The identity of the polymer is assigned on the HSQC spectrum and the linkage points of the sugars are shown on the HMBC spectrum. Sugars labeled X,Y and Z represent the reducing end and most of the chemical shifts converged onto the repeat unit. MB stands for methoxybenzamide. Chemical shifts are reported in Supplementary Table 2.

Full size image

To provide conclusive evidence for this activity, STM0724 was purified, and its activity was investigated in vitro using synthetic acceptor 1, which replicates part of the repeat unit structure of the T1 Galf-PS, linked to a methoxybenzamide (MB) group for detection. The donor substrate, UDP-Galf, was synthesized in situ by Glf. The reaction generated a range of polymeric products with differing HPLC elution times, accompanied by consumption of the input acceptor (Fig. 3b). The reaction mix was scaled up and the product was purified by size exclusion chromatography and analyzed by NMR spectroscopy. The resulting spectra revealed that the in vitro product matched the Galf-PS portion of the polysaccharide produced in vivo (Fig. 3c, Supplementary Table 2).

Linkage of Ribf-PS to LPS is dependent on production of Galf-PS

The collective evidence indicates that the genes in the STM0719–STM0724 fragment in the cluster are sufficient to produce Galf-PS linked to the non-reducing terminus of the trisaccharide generated by WecA–WbbN–WbbO (Fig. 1b,c). E. coli DH5α cells producing LPS containing this part of the T1 antigen were used to adsorb the Galf-specific antibodies from T1 antiserum, generating antiserum specific for the Ribf-PS component. The Ribf-specific antiserum recognized bacteria transformed with the STM0719–STM0726 construct and reacted with the highest molecular weight LPS species (Fig. 2a). Deletion of either STM0725 or STM0726 from the T1 locus eliminated production of the putative Ribf-PS, as might be expected because STM0726 is the only identifiable Ribf-transferase (i.e., possessing sequence similarity to the known Ribf-transferase prototype) encoded by the locus. The wildtype profile and immunoreactivity were restored by co-transformation with a plasmid carrying STM0725–STM0726. However, the profile was not restored by these genes when the Galf-PS polymerase (STM0724) was absent, confirming that STM0724-directed synthesis of the Galf-PS was necessary for further extension of the T1 antigen with the immunodominant non-reducing terminal Ribf-PS portion (completing the pathway scheme in Fig. 1b,c). These results eliminated the possibility of a separate population of LPS molecules substituted only with Ribf-PS. Further biochemical experiments described below identified the position within the Galf-PS that provided the linkage site for Ribf-PS.

STM0725 and STM0726 are both required for Ribf-PS polymerization

Mutant complementation experiments with the construct lacking STM0725–STM0726 showed that STM0725 and STM0726 are both required for restoring production of the Ribf-PS (Fig. 2a). A multiple sequence alignment revealed strong sequence conservation (including catalytic residues) between STM0726 and the characterized Ribf-transferase prototype (Supplementary Fig. 3). Consistent with the sequence similarity, an AlphaFold model of STM0726 contains the requisite N-terminal PRP module and a C-terminal gPRT module (Fig. 4a). Sequence data predicted that STM0725 is a GT and the corresponding AlphaFold model and multiple sequence alignment shows a single Rossmann-like fold and a DXD motif involved in the binding of a metal co-factor for coordination of nucleotide-diphosphate linked sugars, characteristic of the widespread GT-A fold (Supplementary Figs. 4, 5)24.

Fig. 4: STM0725 and STM0726 are required for polymerization of T1 Ribf-PS.
figure 4

a AlphaFold models of STM0725 (Uniprot: Q8ZQU9) and STM0726 (Uniprot: Q8ZQU8) resemble the crystal structures of a prototypical GT-A-fold GT enzyme, Bacillus subtilis SpsA (PDB ID: 1h7q), and the Thermobacillus compostii Ribf-transferase (PDB ID: 7shg), respectively. STM0725 shares 15% identity with SpsA, and the structural comparison shows a Z-score of 17.9 and an rmsd of 3.2 Å. STM0726 shares 16% identity with the prototypical Ribf-transferase structure, and the structures have a Z-score of 21.4 with an rmsd of 6.5 Å. The lower level of three-dimensional similarity in this STM0726 comparison arises primarily from differences in the relative orientation of the PRP and gPRT domains in the two proteins. b HPLC traces of in vitro reactions containing STM0725 and STM0726 PRPP donor substrate and acceptor compound 2. Reactions were incubated for 18 h at 4 °C to prevent protein precipitation, which occurred rapidly at warmer temperatures. Reactions were stopped by the addition of an equal volume of acetonitrile and the mixture analyzed by HPLC, which detected the methoxybenzamide tag in the acceptor. Products indicated on the chromatograms were confirmed by MS (Supplementary Fig. 8) and the results from triplicate reactions were identical. c NMR analysis of the purified STM0725/STM0726 product from a scaled-up reaction. The determined structure, including the reducing terminal synthetic acceptor, is shown above the spectra and assignments are shown on the HSQC spectrum. Linkage points of the monosaccharides were determined by an HMBC experiment and are shown. Chemical shifts are reported in Supplementary Table 3.

Full size image

To assign the precise activities of STM0725 and STM0726, six synthetic acceptors (1–6) (Supplementary Fig. 6), representing fragments of Galf-PS with different lengths and terminal linkages, were employed. The acceptors were incubated with purified enzymes and different donors. Because Galf and Ribf were the only sugars found in the T1 antigen carbohydrate structure (as described above)11, UDP-Galf and PRPP were reasoned to be the only donor candidates. The initial working hypothesis was that STM0725 would add a Galf residue (as it lacks the characteristic architecture of Ribf-transferase) thus serving as a transition GT bridging the Galf-PS and the Ribf-PS. However, this proved to be incorrect; in reactions containing STM0725 + STM0726 and UDP-Galf only, none of the 6 acceptors were modified (Supplementary Fig. 7b). In contrast, in the presence of PRPP as the sole donor, compounds with a non-reducing terminal (1 → 3)-linked Galf (2, 4 and 5) all served as acceptors for robust Ribf-transferase activity in reactions containing STM0725 and STM0726, giving a range of products with increasing retention times consistent with the transfer of multiple Ribf residues (Fig. 4b, Supplementary Fig. 7a,8). To validate the structure of the in vitro reaction products, a large-scale reaction was performed with 2 and the product was purified. NMR analysis (Fig. 4b) confirmed the presence of a β-(1 → 2) linked ribofuranose homopolymer linked to 2, with no additional modifications, identical to the Ribf-PS domain in the authentic T1 antigen (Fig. 4c, Supplementary Table 3). Furthermore, an HMBC experiment identified the linkage between the reducing-end Ribf (C) and the 5-position of the non-reducing Galf (B), by a correlation at δH 5.44/δC 77.8, and a characteristic downfield shift of C5 of B consistent with its substitution. Notably, re-analysis of the NMR spectra of T1 PS produced in recombinant E. coli showed the same linkage between the PS domains (Supplementary Fig. 9), but this assignment was only possible with the guidance provided by the data for the in vitro product due to the complexity of the spectra from the natural product. This reinforces the presence of a covalent linkage between the two structural domains and identifies the C5 hydroxyl group of (1 → 3)-linked Galf residues as the attachment site for the Ribf-PS.

The data above also indicate that the combination of STM0725 and STM0726 transfer only Ribf, a result that is hard to reconcile with the GT-A-fold of STM0725. Therefore, in vitro reactions were performed in an attempt to dissect the individual roles of STM0725 and STM0726. STM0725 was unable to catalyze any Ribf additions to 2, while STM0726 was capable of adding a small amount of a single Ribf residue (Fig. 4b, Supplementary Figs. 7, 8). One possible interpretation of this data is that STM0726 is sufficient for both the initiating transfer of Ribf and further extension of the Ribf-PS. In this scenario, STM0725 performs an ancillary non-catalytic role in polymerization. To pursue this, we needed to determine whether STM0726 was capable of transferring Ribf to a terminal Ribf residue, in addition to the terminal Galf shown above. An enzymatic strategy was established to trap the product of a single Ribf addition to 2. PRP modules of Ribf-transferases possess an invariable Asp required for dephosphorylation16. The catalytic Asp14 in the PRP domain of STM0726 was predicted by multiple sequence alignment (Supplementary Fig. 3) and AlphaFold modeling and replaced with alanine. Reactions containing STM0725 and STM0726D14A with 2 resulted in a new product in the chromatogram (Fig. 4b). This reaction product (7) was scaled up and purified for NMR and confirmed Ribf-5-P linked to the 5-position of the non-reducing Galf in 2 (Supplementary Fig. 10, Supplementary Table 4). Incubation of 7 with wildtype STM0726 resulted in complete dephosphorylation and, in the presence of PRPP, products were observed with one or two additional Ribf residues (Supplementary Fig. 11). In contrast, multiple Ribf residues were added when 7 was incubated with both STM0726 and STM0725. To determine the linkage of the second Ribf residue, 7 was dephosphorylated by incubation with STM0726 in the absence of PRPP to generate the trisaccharide 8 (Supplementary Fig. 12, Supplementary Table 5), followed by installation of a single Ribf-5-P residue using STM0726D14A. The NMR data of the resulting phosphorylated tetrasaccharide (9) confirmed the terminal Ribf-5-P was β-(1 → 2)-linked (Supplementary Fig. 13, Supplementary Table 6). Collectively, the results assign no required GT catalytic activity to STM0725 in T1 antigen biosynthesis, despite its apparent GT-A-fold, and implicates STM0726 as the sole (bifunctional) Ribf-transferase, capable of making both β-Ribf-(1 → 2)-β-Ribf and β-Ribf-(1 → 5)-β-Galf linkages.

STM0725 forms a heterocomplex with STM0726

The data above is consistent with STM0725 performing a structural role as a stabilizing chaperone or stimulatory partner for STM0726, rather than being a bone fide GT. Fulfilling this role would require STM0725–STM0726 interaction, which was examined by co-expressing epitope-tagged versions of the proteins. Purification of His6-STM0725 on Ni-NTA resin resulted in the co-elution of STM0725 and STM0726-FLAG, as evident from the corresponding western immunoblots (Supplementary Fig. 14a). The reciprocal purification, applying the Ni-NTA elution fraction to anti-FLAG beads also demonstrated co-elution of STM0725 and STM0726. Taken together, these data show that the two proteins form a heterocomplex. AlphaFold modeling of the complex indicates a high-confidence model in which STM0725 interacts with STM0726 at an interface between the PRP and gPRT domains (Supplementary Fig. 14b), consistent with the interaction of the two proteins. Unfortunately, STM0725 alone, and copurified with STM0726, was consistently unstable in vitro. This instability led to precipitation after purification that was only partially prevented with glycerol. Thus, this characteristic precluded more extensive investigations of isolated STM0725 and its complex with STM0726.

Distribution of the T1 antigen biosynthesis locus in Salmonella

Examination of several Salmonella genomes revealed that some isolates possessed an intact T1 antigen locus, while others apparently did not. To provide more context, we examined the broader distribution of the locus by probing the Salmonella Foodborne Syst-OMICS database (SalFoS)(https://salfos.ibis.ulaval.ca) as a dataset. SalFoS contains genomic sequences from 2850 diverse Salmonella isolates from the environment, plant, and animal food products, as well as from human infections25.

Isolates distributed across the entire Salmonella enterica species were found with intact versions of each of the genes spanning STM0719-STM026 (the raw data is given in Supplementary Data 1 and a summary of the distribution among species and serovars is in Supplementary Table 7). Salmonella enterica serovars Enteritidis and Typhimurium are the dominant causes of non-typhoidal gastrointestinal infections by this species and are closely related. Almost all S. enterica Typhimurium isolates (299/318) had complete T1 clusters, and the loci were also highly prevalent in other infection relevant serovars26,27,28 including Newport (94/107), Javiana (31/38), Derby (12/17) and the monophasic clade (33/34). In contrast, isolates of serovars Enteritidis and Infantis possessed none of the T1 locus genes and no sequence remnants were found between nei and pxpA. The locus in the prevalent Heidelberg serovar (22/75) consistently contained premature stop codons in either glf, wzt or wbbO. Among the human adapted serovars, all 17 Typhi isolates contain a premature stop codon in STM0724 that is expected to abolish T1 production. T1 clusters were well represented and expected to be functional in Paratyphi B (48/67), and Paratyphi A (12/12), but never in Paratyphi C (0/3), which all showed no remnants of T1. Outside of subspecies enterica, T1 was highly present in the IV subspecies (20/28). Parts of the T1 cluster were also found in subspecies IIIB, but none contained complete clusters due to premature stop codons present in one or more genes, depending on the isolate (0/39). Finally, subspecies II (0/3), IIIA (0/14), VI (0/2) and bongori (0/2) showed no hits for any of the T1 genes. To validate the lack of T1 hits (or remnants) in subspecies II, IIIA, VI and species bongori, GenBank was also searched. This also revealed no full T1 clusters in subspecies VI, however unlike the SalFos isolates, remnants of the T1 genes could be detected in ancestral bongori, IIIA and II. The bongori isolates (CP035676.1, CP053417.1) contained premature stop codons in STM0725 and STM0726, or STM0719 respectively, and the IIIA isolates (CP082954.1, LR134156.1) contained premature stop codons in STM0725 and STM0726 or STM0724 respectively. Some II isolates were found with intact T1 clusters (i.e., CP034717.1 and CP034697.1), but other isolates contained premature stop codons in STM0724, STM0725, or STM0726. Notably, the isolates with mutations in STM0725 and/or STM0726 may still have the capacity to produce Galf-PS, as the rest of the T1 cluster is intact.

T1 antigen expression is regulated by a recombinational phase switch

Although stable expression of the T1 antigen was observed with plasmid constructs in an E. coli host, the original transient description of the antigen led to the hypothesis that the cluster must be regulated in some manner in Salmonella. This was supported by varying degrees of anti-T1 immunoreactivity in western immunoblots of whole-cell lysates prepared from several Salmonella isolates with or without the T1 biosynthesis locus (Supplementary Fig. 15). Immunoreactivity was enhanced in a mutant lacking its conventional O antigen, but the lack of silver-stained higher molecular weight material in the PAGE profile indicated that T1 expression was still relatively low (Supplementary Fig. 15).

Upstream of the now established T1 biosynthesis genes are three additional genes: STM0716, STM0717 and STM0718 (Fig. 5a). As shown above, STM0716–STM0718 are not required for T1 antigen biosynthesis in an E. coli background. STM0716 encodes a predicted tyrosine recombinase sharing 60% identity/73% similarity (92% coverage) with FimB from E. coli MG1655. FimB directs the phase-variable transcriptional regulation of type 1 fimbrial adhesins by mediating the inversion of a DNA segment encompassing the fim promoter, thereby flipping the fim operon between OFF and ON positions (reviewed in ref. 29). The switch is composed of inverted nine base-pair repeats, flanking a 314-nucleotide segment that contains the fim promoter. Given the similarity shared by STM0716 and FimB, the region following STM0716 was searched for inverted repeats. A 269 base-pair region flanked by nine base-pair inverted repeats was identified and hypothesized to be a potential phase switch (Fig. 5a).

Fig. 5: The T1 cluster is regulated by a recombination-mediated phase variation mechanism.
figure 5

a Genetic organization of the genomic regions encompassing genetic switches for T1 antigen biosynthesis, and the expression of E. coli type I fimbriae (shown from MG1655; U00096.3), and a putative autotransporter (HyxB) encoded on E. coli pathogenicity island PAI-X (from E. coli O157:H7; EDL933). STM0716 is an ortholog of FimB and FimX, which are tyrosine recombinases responsible for flipping a promoter region preceding the fim and hyxRAB operons. In each case, the invertible region is flanked by nine base-pair inverted repeats (IR1 and IR2). b Schematic showing the experimental strategy to identify ON and OFF states of the invertible switch using qPCR. Primer P1 (in the middle of the switch) serves as the forward primer or reverse primer, depending on the switch orientation and acts with one of the two primers (P2 or P3) located outside the switch to generate a diagnostic amplification product. Standard curves were constructed using synthesized (fixed) templates representing the OFF or ON states, and used to calculate the relative amounts of each state from genomic DNA templates. Bar graphs represent the average amounts detected, with individual data points plotted on top of the bar graph. Samples and standards were analyzed in triplicate and error bars indicating standard deviation is shown using the average DNA quantity detected in each sample as the center point. qPCR was performed on two separate genome preparations with comparable results. c T1 antigen phenotypes resulting from engineered ON and OFF states of the T1 phase switch. Whole-cell lysates were examined by SDS-PAGE and western immunoblotting. pWQ1128 (carrying STM0716-STM0716) was cloned to contain the switch region locked in either the ON or OFF state and STM0716 was deleted to create pWQ1133 (ON) and pWQ1134 (OFF). Each plasmid was also co-transformed with a plasmid carrying STM0716 in trans. The presence of the T1 antigen was then probed by silver staining and western blotting with the specific antisera. These analyses were performed in triplicate. Source data are provided as a Source Data file. d Sanger sequencing results for the switch regions from plasmids recovered from the cultures examined in c. No phase variation was detected in the absence of STM0716, evident in the uniform chromatographic species. Expression of STM0716 resulted in a mixed population of species reflecting both orientations of the switch. Most importantly, expression of STM0716 in the ON state led to a predominant OFF state, whereas expression of STM0716 in the OFF state resulted in a minor amount of ON state.

Full size image

To investigate the physiological consequences of the phase switch candidate for T1 antigen expression, we first sought to determine whether this region is invertible in its native genomic context. qPCR was performed on S. enterica serovar Typhimurium SL3770 using primers designed to probe the two states (Fig. 5b), subsequently determined to be the OFF and ON states (see below). qPCR detected both states, suggesting a dynamic process. Furthermore, quantitation was performed by comparison to a standard curve using synthesized ON or OFF state DNA templates, revealing that the OFF state was approximately 2–3 times more abundant than the ON state (Fig. 5b, Supplementary Fig. 16, Supplementary Table 8). This bias is supported by examination of sequence data for this region in Salmonella isolates deposited in GenBank; all but one isolate (CP037891.1) reflect the predominant OFF state.

Because the switch is located well upstream of the first T1 biosynthesis gene (STM0719) and the intervening region contains STM0717 and STM0718 genes that play no apparent role in T1 production, it was essential to confirm that the switch region directly regulated downstream genes responsible for T1 antigen production. To facilitate this, plasmid pWQ1128 containing STM0716–STM0726 was modified by deleting the vector’s tetR promoter and inserting a transcriptional terminator upstream of STM0716 to prevent any transcriptional read through from the vector. STM0716 was then deleted to prevent DNA inversion, and the invertible promoter region was cloned in both orientations. In the OFF orientation, no T1 antigen could be detected in whole-cell lysates by either silver staining or western immunoblotting. However, when the switch region was in the ON state, robust T1 antigen production was observed (Fig. 5c). Co-transformation of these plasmids with a compatible plasmid carrying STM0716 (encoding the putative recombinase) restored the phase-variation process in E. coli. T1 antigen production was now detected in bacteria containing the OFF-state construct and Sanger sequencing of the switch region showed a mixed population of DNA, consistent with part of the population changing to an ON-state (Fig. 5d). Conversely, when STM0716 was expressed in the presence of the ON-state plasmid, a reduction of the T1 antigen was detected in the silver stain and western immunoblot and Sanger sequencing again showed a mixed population of DNA (Fig. 5d). Together, these results show that the region proceeding STM0716 is indeed a phase-variable promoter region that can be flipped in a bidirectional manner by STM0716 to regulate expression of downstream genes directing biosynthesis of the T1 antigen.

Discussion

In this study, the long-overlooked T1 antigen of Salmonella is reinvestigated and its distribution, structure and relationship to OPS was resolved. We show that the T1 antigen is a widely distributed lipid A–core-linked polysaccharide in Salmonella but its abundance is typically low (and frequently undetectable) compared to the serotype-specific O antigen when wildtype cells are grown under routine laboratory conditions. This is evident by the increased amount of T1 antigen in whole cell lysates of a S. enterica serovar Typhimurium mutant that lacks the serogroup B O antigen due to mutation of the gene for the phosphoglycosyltransferase (wbaP) (Supplementary Fig. 1a). The precise linkage of T1 to lipid A-core oligosaccharide was not pursued here, but the requirement of WaaL for T1 ligation is consistent with the attachment site being the same as native O-antigen to give the structure shown in Supplementary Fig. 18. The organization of a lipid A–core-linked glycan into tandem structural domains is a rare occurrence; the large majority of OPS are composed of a single repeat unit structure. Prototypes for this unusual tandem format are found in K. pneumoniae serotype O1, where the O1 antigen with a disaccharide repeat unit ( → 3)-α-d-Galp-(1 → 3)-β-d-Galp-(1 → ) is added to the non-reducing terminus of the O2 antigen ( → 3)-α-d-Galp-(1 → 3)-β-d-Galf-(1 → )30,31. A similar two domain structure occurs in Klebsiella serotype O2ac31,32. In an intriguing coincidence, biosynthesis of these Klebsiella O antigens and the Salmonella T1 antigen are all assembled on the same undecaprenyl diphosphate linked trisaccharide generated by WecA–WbbN–WbbO (Fig. 1b). This structural motif apparently provides a versatile foundation for glycan diversification.

Biosynthesis of the Galf-PS component of T1 antigen is performed by the STM0724 polymerase, an ortholog of GlfT2. GlfT2 forms the backbone of the mycobacterial cell wall arabinogalactan and possesses a single catalytic-site GT that builds an alternating β-(1 → 5) and β-(1 → 6) Galf galactan23. A query of other Galf-homopolymers was conducted to investigate the distribution of GlfT2-like enzymes. Examples were identified in a diverse set of organisms, and included one, two and three linkage Galf-PSs (Supplementary Fig. 17). In all cases, genes were found for GlfT2 orthologs sharing identities ranging from 23–38% (38–53% similarity). AlphaFold modeling showed that each of the orthologs have the distinct GlfT2 fold indicating all operate with a single catalytic site, regardless of the number of linkage types in the products33. GlfT2 is a processive enzyme, where the glycan is retained by the enzyme during elongation until 30–40 residues have been added. Recent data has provided insight into the basis for dual-linkage (β-(1 → 5) and β-(1 → 5)) specificity34,35 as well as a potential mechanism for translocating the chain through the active site during processive polymerization36. The mixed products in Salmonella T1 antigen suggests STM0724 exhibits less intrinsic fidelity than GlfT2. Establishing the extent to which the mechanistic details apply to other family members that generate products with 1–3 different linkage configurations will require detailed biochemistry and structures of enzyme–substrate complexes but the present study identifies important representatives to pursue.

STM0726 is the polymerase for the biosynthesis of the Ribf-PS. This enzyme possesses the dual gPRT–PRP activity and the prototypical fold found in canonical glycan Ribf-transferases16, but is the first representative known to transfer more than a single Ribf sugar, or possess the ability to make more than one type of glycosyl linkage. In this sense, it is analogous to the STM0724 Galf-PS polymerase. However, STM0726 is able to transfer Ribf to acceptors with either terminal Galf or Ribf residues and, in that regard, is also similar to GlfT1, another bifunctional enzyme involved in mycobacterial arabinogalactan biosynthesis. GlfT1 synthesizes Galf–Galf and Galf–Rhap (rhamnopyranose) linkages, although GlfT1 does not continue polymerization23. How STM0726 synthesizes two distinct glycosidic linkages via a single catalytic site is a topic for further investigation, currently limited by the instability of the purified enzyme. An equally fascinating question is why optimal activity of the STM0726 polymerase depends on another GT-like protein that lacks GT catalytic activity. Further insight into the relationship between STM0725 and STM0726 was obtained by a tBLASTn search using STM0725 as a query. This returned 121 hits (excluding Salmonella) with identities ranging from 54% to 22%. Surprisingly, in 91 of these hits, the genes for the STM0725 orthologs occurred adjacent to a gene encoding an STM0726 ortholog. These were found in genomes of diverse species and genera, as well as in a bacteriophage belonging to the Caudoviricetes. Orthologous pairs were present in Serratia marcescens O1637 and Liberibacter crescens BT-138, which are known to produce Ribf-homopolymers. Interestingly, the STM0725 orthologs were separated into two phylogenetically distinct clades of proteins (Supplementary Fig. 5b). STM0725 itself was grouped with a clade of proteins that all lacked a candidate catalytic base in the position expected in GT-A enzymes. In contrast, the second clade all possessed a conserved Asp typical of functional GT-A enzymes, suggesting these proteins may retain catalytic activity. L. crescens BT-1 produces a complex OPS with two structural components; one involves short spans of poly-Ribf, while the other is a disaccharide repeat composed of Galp and Ribf38. The STM0726 ortholog (B488_06780) is the only identifiable Ribf-transferase encoded the BT-1 genome. Nevertheless, an STM0725 ortholog does not seem to be an essential requirement in all poly-Ribf biosynthesis systems. The Helicobacter pylori SS1 LPS was found to contain short poly-Ribf39, and a single STM0726 ortholog (but no STM0725 ortholog) is encoded by the genome.

Salmonella isolates with intact versions of all the required T1 antigen biosynthetic genes were found distributed across the Salmonella enterica species. The distribution (Supplementary Table 7) implies that the T1 antigen is not required for typhoidal infections and, while widespread in isolates causing non-typhoidal gastrointestinal infections and in environmental isolates, possession of the genetic locus is not universal. The appearance of parts of the locus in S. bongori and S. enterica subspecies IIIA suggests it is not a recent acquisition during the evolution of the genus. A closer look at the S. bongori isolates revealed that both isolates from SalFoS contained a different polysaccharide cluster located between nei and pxpA downstream of copies of STM0716–0718 and an identifiable phase switch (Supplementary Fig. 19). The genes in the cluster closely resemble the well characterized E. coli O9/O9a locus40, which directs production of a polymannose OPS (Supplementary Table 9). The potential production of this glycan in Salmonella has not been investigated. Notably, this O9/O9a like cluster was also identified in 10 other isolates in the SalFoS database including representatives from subspecies salamae, and diarizonae. The number of hits for the O9/O9a-like cluster were substantially less than for the T1 antigen, but this could reflect the bias towards human-related Salmonella isolates (where the T1 antigen is predominant) in the collection.

One of the key questions we had when this study was initiated is why the T1 antigen was originally described as transient and why it has not been detected during the many investigations of O antigen structures in Salmonella. Part of this may be answered by the expected competition between the T1 antigen and the serotype-specific OPS for lipid A–core. However, the other factor is the recombination-mediated phase variation of T1 antigen expression, which is unprecedented for a lipid A-core-linked glycan. Conventional Salmonella O antigens are thought to result from (mostly) constitutive expression of biosynthesis genes. There is some modest fine-tuning of the products through OxyR regulator-mediated on/off transcriptional control of the opvAB locus, which alters the distribution of chain lengths of OPS chains in some way and affects sensitivity to bacteriophage and serum41. In addition, OxyR and the DNA methylase (Dam) are implicated in epigenetic regulation of an OPS-modification process in some Salmonella OPS42. This modification involves adding side chain glucose residues in a periplasmic post-polymerization reaction, directed by enzymes encoded by a lysogenic bacteriophage. A comparable recombinational promoter-inversion mechanism has been described in regulating the multiple cell surface exopolysaccharides produced by Bacteroides fragilis43. The process is directed by a serine site-specific recombinase44 unrelated to STM0716. While the origin of the regulatory system for the T1 antigen is unknown, a similar locus organization is found within E. coli Pathogenicity Island X (PAI-X) (Fig. 5d). The locus in PAI-X possesses a gene encoding an STM0716 ortholog (FimX; 72% identity, 84% similarity, 93% coverage) upstream of a candidate phase switch possessing inverted repeats whose sequence is identical to those in the T1 antigen locus. The switch precedes the downstream genes: hyxA, hyxR, encoding putative transcriptional regulators and hyxB encoding a putative autotransporter adhesin. STM0718 shares 39% identity/60% similarity (81% coverage) with HyxA but the physiological significance (if any) is currently unknown.

The presence of a highly conserved T1-antigen locus with complex regulation in many Salmonella genomes, suggests a potentially important role in fitness. Transposon directed insertion sequencing (TraDIS) in Salmonella Typhimurium revealed preservation of an unknown polysaccharide locus (identified here as a T1 antigen biosynthesis)45,46,47. While this supports a role in fitness, some caution is needed in interpretation because the T1-biosynthesis gene cluster has higher AT content than the rest of the genome. This is a known region of H-NS binding48, which can compromise TraDIS data49. Several transcriptomic and proteomic studies in Salmonella Typhimurium identify no major changes in the abundance of T1 biosynthesis transcripts or proteins under the conditions tested. One study indicates a modest elevation in transcription of the T1 - antigen biosynthesis genes in S. Typhimurium SL1344, following opsonization (12h) by macrophages. This is accompanied by a reduction in transcription of  O-antigen biosynthesis genes, but the biological implications are uncertain because SL1344 is expected to be unable to produce T1 antigen due to a mutation in glf. Further investigation is therefore needed to provide insight50 into possible conditions where a particular ON- or OFF-state would be beneficial51,52,53. Furthermore, it has yet to be established whether the conservation of T1-antigen production reflects importance in the pathogen lifestyle, or if it is required for fitness and survival in an environment outside the host. Nevertheless, the findings of the current study provide the essential foundation to investigate these broader questions about the physiological role of T1 antigen in Salmonella.

Methods

Bacterial strains and DNA methods

Bacterial strains and plasmids used in this study are listed in Supplementary Table 10. All bacteria were cultured in LB supplemented with 50 μg/ml kanamycin, 100 μg/ml ampicillin or 34 μg/ml chloramphenicol, where applicable. Genomic DNA was purified using the PureLink Genomic DNA mini kits from ThermoFisher (Invitrogen). DNA fragments were generated using KOD Hot Start DNA polymerase (Novagen) for cloning, site-directed mutagenesis and gene deletions/manipulations. The sequences of the oligonucleotide primers are provided in Supplementary Table 11. PCR products and plasmids were purified using GeneJET PCR or plasmid purification kit (ThermoScientific), respectively. Plasmid constructs were confirmed using Sanger sequencing in the Advanced Analysis Center at the University of Guelph, or by whole plasmid sequencing from Plasmidsaurus (https://www.plasmidsaurus.com).

Bioinformatics analyses

Initial protein domain predictions were made by BLAST using the Conserved Domain Database. AlphaFold models were obtained from Uniprot or created manually in ColabFold54 when not available in Uniprot. Manually created AlphaFold models are provided in figshare (see data availability). Multiple sequence alignments were performed using ClustalW v.2.055, and alignments were visualized using ESPript v.3.056. 3D structure alignments were performed in Dali57 and protein structures were visualized in Pymol v.2.1. Phylogenetic studies were conducted using Phylogeny.fr58 using a bootstrap value of 100. The Salmonella Foodborne Syst-OMICS (SalFoS – https://salfos.ibis.ulaval.ca) database was used as a data source to investigate T1 antigen distribution across genus Salmonella. Sequencing files were downloaded and queried locally using tBLASTn with an e-value cut-off of x10–20. BLAST searches were performed for each protein in the T1 cluster. Only those isolates that contained full-length homologs for all proteins for T1 production were considered T1 positive, while any isolate with mutation(s) resulting in truncations in one or more of the T1 proteins were considered T1 negative. Bit scores were used as an indicator for intact protein sequence. Bit scores lower than a determined threshold (cutoffs indicated for each protein in Supplementary Data 1) corresponded to a truncated protein product that was considered T1 negative. In a small number of cases premature stop codons resulted in truncated proteins that still had high bit scores (indicated in Supplementary Data 1), and these were manually assessed and designated as T1 negative.

LPS and T1 antigen purification

10 L cultures of E. coli DH5α harboring pWQ1129 were grown overnight at 37 °C. Cells were collected and LPS was purified using the hot-phenol method59. After phase separation, the aqueous layer containing LPS was dialyzed against water to remove residual phenol. The pH was then reduced to 4 to precipitate proteins and nucleic acids, which were removed by centrifugation at 12,000 × g for 20 mins. The supernatant was dialyzed against deionized water until neutral pH was achieved, and then concentrated by a rotovap and lyophilized. OPS was isolated by hydrolysis of purified LPS at 100 °C in 2 % (v/v) acetic acid for two hours, followed by centrifugation to remove the lipid precipitate. The carbohydrate-containing supernatant was then separated on a Sephadex G-50 superfine column (2.5 × 75 cm) in 50 mM pyridinium acetate buffer (pH 4.5) at a flow rate of 0.6 mL/min. Elution was monitored with a Smartline 2300 refractive index detector (Knauer) and fractions were collected at 10 min intervals. Fractions corresponding to peaks were pooled, frozen and lyophilized. The OPS fraction was determined to contain a substantial amount of contaminating enterobacterial common antigen (ECA). Due to the acidic nature of ECA, it was separated from the T1 OPS using anion exchange using a DEAE Fast Flow (GE Healthcare) column eluted with a 5–500 mm sodium phosphate gradient at pH 6.3. The flowthrough containing neutral T1 OPS was collected, concentrated and desalted with a G-50 PD-10 column eluting in water. The final yield was ~16 mg.

Antiserum production

Animal handling and immunization was performed at the University of Guelph Central Animal Facility. Rabbit anti-T1 was produced using formalin killed E. coli DH5α expressing T1 (pWQ1128). Killed cells were resuspended in 0.85% (w/v) NaCl at ~108 cfu/ml and mixed 1:1 with Freund’s incomplete adjuvant (Sigma). A New Zealand white rabbit received an intramuscular injection every two weeks for six weeks total. Blood was collected after six weeks, and the serum was separated and stored at –80 °C. To adsorb antisera and remove antibodies against the T1 Galf-PS, E. coli DH5α cells expressing pWQ1130 (ΔSTM0725–STM0726) were grown overnight and 200 mL of cells were resuspended in 16 mL of sterile PBS. The cell pellet from 4 mL of culture was collected by centrifugation, resuspended in 5.2 mL of antiserum and incubated at 37 °C for 1 h. Cells were then removed, and the process was repeated three more times. All T1 antisera generated here are available upon request.

Expression and purification of proteins

Proteins expressed from pET-vector plasmids were expressed by induction with 0.5 mM 1-thio-β-d-galactopyranoside (IPTG) and those from pBAD-vector plasmids were induced with 0.2% ʟ-arabinose. The induced cultures were grown overnight at 18 °C. The cells were harvested and resuspended in buffer A (50 mM Tris pH 7.5, 500 mM NaCl, 20% glycerol). Cell disruption was performed with an Avestin C3 Emulsiflex. Lysates were cleared by sequential centrifugation at 12,000 × g for 20 min, followed by 100,000 × g for 1 h to remove membranes. The supernatant was applied to Ni-NTA agarose with a 2 mL bed volume. The beads were washed with 10 bed volumes of buffer A containing 10 mM imidazole, followed by 10 bed volumes of buffer A containing 30 mM imidazole. Bound proteins were eluted with buffer A containing 250 mM imidazole. The eluates were concentrated with a 10,000 MWCO concentrator (Sartorius) and buffer exchanged into buffer A using PD-10 Sephadex G-25 columns (Cytiva). Purified protein concentrations were estimated using absorbance at 280 nm, with molecular weights and extinction coefficient estimations from ProtParam.

Protein interaction studies

Protein co-purification was performed by exploiting the epitope tags on His6-STM0725 and STM0726-FLAG. Cultures expressing pWQ1142 were prepared and lysed as described above. The cell-free lysate was passed through a Ni-NTA agarose column and the column was washed as described above. The eluate was concentrated, buffer exchanged to buffer A, and then incubated with 50 µL of anti-FLAG M2 magnetic beads (Sigma) beads for 1 h. The beads were washed three times with 1 mL of buffer A and the protein was eluted by resuspending the beads in 40 µL 100 mM glycine buffer pH 2.5.

Quantitative PCR

Quantitative PCR was performed using the PowerTrack SYBR Green Master Mix (ThermoFisher). Genomic DNA was purified using the PureLink genomic DNA kit (Invitrogen). Amplification reactions were performed using the QuantStudio 3 standard program (Applied Biosystems) and analyzed using the Design and Analysis Software (ThermoFisher Connect). DNA fragments with locked ON or OFF switch orientations were synthesized by IDT. Three technical replicates were performed with each sample and standard, and values reported as the average with standard deviation. Results are representative of two biological replicates from separate genomic preparations.

SDS-PAGE and immunoblotting

LPS samples were analyzed using proteinase-digested whole-cell lysates as described in ref. 60. Cells were collected from a culture volume equivalent to 1 OD600nm unit and resuspended in loading buffer, boiled for 10 min, and subsequently treated with proteinase K at 55 °C for 1 h. Samples were separated on 12% Tris-glycine gels and visualized using silver staining or western blotting after transfer to nitrocellulose membranes (Protran, GE Healthcare). SDS-PAGE of purified proteins was conducted by mixing an equal volume of protein with loading buffer and separating in the way described above and the SeeBlue Plus2 protein ladder provided molecular weight markers (Invitrogen). Coomassie-blue staining was performed using the Pierce Power Stainer. Western immunoblot transfer was conducted with a constant current of 200 mA for 45 min (LPS samples) or 350 mA for 60 min (protein) in buffer containing 25 mM Tris, 150 mM glycine, and 20% (v/v) methanol. The membranes were blocked in 5% skim milk for LPS analysis (BD Difco), or 5% bovine serum albumin for protein analysis (Cytiva); the blocking agent was prepared in TBST (10 mM Tris-Cl, pH 7.5, 150 mM NaCl, 0.005% (v/v) Tween 20). Rabbit T1 antiserum was used as a 1:500 dilution in skim milk-TBST, and mouse anti-His5 (Qiagen - Catalog number: 34660 Lot: 172019951) antibody or mouse anti-FLAG (Sigma - Catalog number: F3165 Lot: SLCM4081) were each used at a 1:1000 dilution in BSA-TBST. The secondary antibodies were goat anti-rabbit-conjugated alkaline phosphatase (1:3000) (Cedarlane - Catalog Number: 111-055-003 Lot: 86-172-030422.) or goat anti-mouse-conjugated alkaline phosphatase (Jackson laboratories - Catalog Number: 115-055-003 Lot: 138204.), and detection was achieved using nitroblue tetrazolium and 5-bromo-4-chloro-3-indolyl phosphate (Roche Applied Science).

In vitro GT reactions

Standard reactions were performed in 20 μL volumes containing 50 mM HEPES pH 7.5, 20 mM MgCl2, 10% glycerol, 1 mM acceptor, 5 mM PRPP, 5 mM UDP-Galp and 50 μm Glf, and 5 mM NADH. Reaction mixtures contained 100 μM enzyme. STM0724 reactions were incubated at 30 °C for 1 h while STM0725 and STM0726 reactions were incubated at 4 °C for 18 h. Reactions were stopped by adding an equal volume of acetonitrile and the resulting protein precipitate was removed by centrifugation at 12,000 × g for 5 min. Large-scale reactions were performed using the same components and concentrations in 1.6 mL volumes, but glycerol was excluded. In the absence of glycerol, STM0724 and STM0725 precipitated quickly, so reactions were performed on freshly purified protein maintained at 4 °C for the entirety of the purification and reaction preparation. For NMR analysis, products with single-sugar additions were purified from the reaction mixture using a SepPak C8 column eluted in 50% methanol. In vitro polysaccharide products made by STM0724 and STM0725 + STM0726 were fractionated by size exclusion chromatography using a Superdex 200 Increase 10/300 column eluted with 100 mM ammonium acetate. The high molecular weight fractions were retained and purified using a SepPak C8 as above. Reactions analyzing STM0725 and STM0726 containing UDP-Galp and NADH were also first applied to a SepPak C8 (as above) to eliminate overlapping peaks from NADH.

The reaction components were separated by HPLC using an Agilent 1260 Infinity II LC system. A 10 μL aliquot of the reaction mixture was injected and separated with a GLYCOSEP N column (4.6 × 250 mm, Prozyme). Buffer A contained 10 mM ammonium formate pH 4.4 in 80% acetonitrile, buffer B contained 30 mM ammonium formate pH 4.4 in 40% acetonitrile and solvent C contained 0.5% formic acid. Separation was accomplished with a linear gradient of 100% A to 100% B over 160 min (0.4 mL/min), followed by a 2 min gradient of 100% B to 100% C.; returning to 100% A over 2 min and holding for 15 min (1 mL/min); followed by 0.4 mL/min for 5 min in A. The column temperature was 30 °C and elution was monitored at 260 nm. All HPLC was analyzed using Agilent OpenLAB revision A.02.16.

Mass spectrometry

LC-MS was performed in the University of Guelph Advanced Analysis Center with an Agilent 1260 HPLC interfaced with an Agilent UHD 6530 Q-TOF mass spectrometer. LC separation was performed on a C18 column (Agilent Poroshell 120, EC-C18 50 × 3.0 mm, 2.7 μm), using water with 0.1% formic acid (A) and acetonitrile with 0.1% formic acid (B). A 10 μL aliquot of reaction mixture was injected and a gradient elution was performed with a flow rate of 0.4 mL/min starting with 10% B for 1 min; increasing to 100% B in 29 min; a column wash at 100% B for 5 min; followed by a 20 min re-equilibration. The mass spectrometer electrospray capillary voltage was maintained at 4.0 kV, and the drying gas temperature at 250 °C, with a flow rate of 8 L/min. Nebulizer pressure was 30 psi and the fragmentor was set to 160. Nozzle, skimmer, and octapole RF voltages were set at 1000 V, 65 V and 750 V, respectively. Nitrogen (purity >98%) was used as nebulizing, drying, and collision gas. Ionization of compounds was performed in negative mode. The mass axis was calibrated using the Agilent tuning mix HP0321 (Agilent technologies) prepared in acetonitrile. All mass spectrometry analysis was performed with Agilent MassHunter Qualitative Analysis version 10.0.

Nuclear magnetic resonance spectroscopy

NMR analyses of OPS and in vitro reaction products were conducted in the University of Guelph Advanced Analysis Center. Analysis was performed using a Bruker Avance III 600 MHz spectrometer, equipped with a cryoprobe. Samples were deuterium exchanged twice by resuspending in 99.0% D2O and lyophilizing. Samples were prepared in 99.96% D2O with chemical shifts referenced to a 3-trimethylsilylpropanoate-2,2,3,3-d4 (δH 0 ppm, δC –1.6 ppm) internal standard. The mixing times for TOCSY and NOESY experiments were 80 ms and 200 ms, respectively. All spectra were recorded at 25 °C. NMR analysis was performed using Bruker TopSpin version 4.2.0.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.