Summary
Glycosylation of surface structures diversifies cells chemically and physically. Sialic acids commonly serve as glycosyl donors, particularly pseudaminic (Pse) or legionaminic acid (Leg) that prominently decorate eubacterial and archaeal surface layers or appendages. We investigated a new class of FlmG protein glycosyltransferases that modify flagellin, the structural subunit of the flagellar filament. Functional insulation of orthologous Pse and Leg biosynthesis pathways accounted for the flagellin glycosylation specificity and motility conferred by the cognate FlmG in the α-proteobacteria Caulobacter crescentus and Brevundimonas subvibrioides, respectively. Exploiting these functions, we conducted genetic glyco-profiling to classify Pse or Leg biosynthesis pathways and we used heterologous reconstitution experiments to unearth a signature determinant of Leg biosynthesis in eubacteria and archaea. These findings and our chimeric FlmG analyses reveal two modular determinants that govern flagellin glycosyltransferase specificity: a glycosyltransferase domain that accepts either Leg or Pse and that uses specialized flagellin-binding domain to identify the substrate.
INTRODUCTION
Sialic acids, also known are nonulosonic acids (NulO), are nine-carbon (α-keto) acidic sugars featuring acetamido linkages that are found in all domains of life [1]. The most prevalent vertebrate sialic acid, (5-)N-acetylneuraminic acid (Neu), occurs on surface glyco-conjugates like glycolipids or glycoproteins [2, 3]. While meningitis-causing eubacteria also camouflage their surface with Neu, most eubacteria and the archaea typically decorate their cell surface structures with (5-, 7-)di-acetamido derivatives, either pseudaminic acid (Pse) and/or its stereoisomer legionaminic acid (Leg, Figure 1). Pse or Leg are constituents of capsular polysaccharides (CPS or K-antigen)[4] or the O-antigen of lipopolysaccharide (LPS)[5], but they often also occur conjugated to proteinaceous surface appendages, for example on the subunits of S-layer arrays [6], pilus adhesins [7] or flagellar filaments (the H-antigen)[8, 9]. Pse and Leg derivatives synthesized in vitro can be added exogenously in metabolic labeling experiments to be incorporated into bacterial surface structures [10, 11]. Moreover, Pse and Leg are attractive vaccine targets as shown by the recent report that mice immunized with Pse chemically conjugated to a carrier protein were protected against the Pse-containing pathogenic Acinetobacter baumannii strain Ab-00.191 [12].
Pse- or Leg-decorated flagella may also be immunogenic. The flagellum consists of three major parts: an envelope-embedded basal body that houses the rotary engine and the secretion apparatus, a universal joint known as the hook that transmits torque from the motor and that protrudes to the cell surface, and finally a tubular flagellar propeller composed of flagellins that is mounted on the hook (Figure 1)[13, 14]. Once flagellin subunits are translated, they are exported through the flagellar protein secretory apparatus along the hollow flagellar filament for polymerization at its growing tip. Glycosylation typically occurs post-translationally on serine or threonine residues of flagellin by highly specific and flagellin glycosyltransferases (fGTs)[15, 16]. Unlike the pilus-specific glycosyltransferases that execute the glycosylation only after the acceptor protein has been translocated across the membrane [17, 18], the fGTs are soluble enzymes that act on flagellin in the cytoplasm before their secretion through the flagellar apparatus. Two types of fGTs have been described to date, the Maf- and FlmG-type [15]. It is thought that these fGTs accept CMP-activated forms of Pse or Leg as glycosyl donors and then join the Pse or Leg moiety to the flagellin acceptor molecule that they bind directly. Inactivation of the fGT or the corresponding Pse-/Leg-biosynthesis pathway results in failure to modify flagellin and (often) a motility defect [19]. It remains mysterious why these flagellin glycosylation mutants are non-motile and flagellin is often poorly secreted. Such mutants harbor a hook-basal-body (HBB), yet they lack a flagellar filament [19].
The synthesis of CMP-Pse or CMP-Leg proceeds enzymatically by series of steps [20-22], ultimately ending with the condensation of an activated 6-carbon monosaccharide (typically N-acetyl-glucosamine, GlcNAc) with 3-carbon pyruvate (such as phosphoenolpyruvate, PEP) by Pse or Leg synthase paralogs, PseI or LegI, respectively (Figure 1)[23, 24], whereas the sialic acid Neu is synthesized by the NeuB paralog [20-22, 25]. A major difference between the Pse and Leg pathways is that the former uses UDP-GlcNAc as starting material whereas the latter usually builds on GDP-GlcNAc [21, 22]. Pse or Leg must first be activated with CMP by the PseF or LegF enzyme, respectively (Figure 1) for used as glycosyl donors by terminal glycosyltransferases, including fGTs.
The structure-function relationship and specificities of Maf and FlmG fGTs is poorly understood. Some Maf enzymes have been linked to flagella glycosylated with Pse, while other Maf affect modification of flagella with Leg [10, 26-28]. The determinants conferring donor or acceptor specificities in these enzymes have not been elucidated. Recently, FlmG and Pse biosynthesis enzymes from the Gram-negative α-proteobacterium Caulobacter crescentus were shown to be necessary and sufficient for modification of flagellin [19]. C. crescentus encodes six flagellin paralogs [29-31] that are no longer modified in the absence of Pse or FlmG [19]. Conversely, expression of FlmG and the FljK flagellin in heterologous hosts producing Pse resulted in FljK modification [19]. Sequence analysis predicts a simple 2-domain organization for FlmG: an N-terminal tetratrico-peptide-repeat (TPR) domain and a C-terminal GT-B type glycosyltransferase domain [15]. Bacterial-two-hybrid (BACTH) assays revealed that the TPR domain can directly bind flagellin, whereas the GT-B domain cannot [19]. The donor specificity of the GT-B domain remains unexplored in the absence of a FlmG system that links Leg to flagellin.
Here we establish a glyco-profiling platform for functional analysis of Pse and Leg biosynthesis pathways using motility as a proxy and we exploit this set-up to uncover a novel FlmG glycosylation system in Brevundimonas subvibroides that modifies flagellin with Leg. Using the B. subvibroides and C. crescentus Leg and Pse biosynthesis mutants, we show that the two pathways are genetically insulated, defining a first level of specificity. We then reconstitute flagellin glycosylation using the B. subvibroides components in C. crescentus and we reprogram a Pse-dependent FlmG into a Leg-dependent enzyme through domain substitutions in chimeras. Thus, two modular determinants govern specificity in fGTs, with the GT selecting either Leg or Pse as donor and linking it to the correct acceptor identified through a flagellin-binding domain.
RESULTS
Genetic glyco-profiling in C. crescentus ΔpseI cells using motility as proxy
Phylogenomic and functional analyses show that the genes encoding PEP-dependent synthases of sialic acids are wide-spread, present in all domains of life. The PseI and LegI synthases predominate in the eubacterial and archaeal lineages, sometimes co-encoded in the same genomes. As a rare example, Campylobacter jejuni 11168 has three (NeuB, PseI and LegI) synthases [22], while the Pseudomonas sp. Irchel 3E13 genome (NZ_FYDX01000009.1)[32] encodes two predicted synthases, a PseI and LegI homolog, and C. crescentus only encodes only PseI (previously called NeuB)[19, 33]. Our previous heterologous complementation experiments of the motility defect associated with C. crescentus ΔpseI cells showed that of the three C. jejuni 11168 synthases, only PseI could support motility in C. crescentus [19]. These experiments provided strong support for the notion that the Pse synthesis pathway can only function properly with PseI, but not when it is substituted with LegI or NeuB. However, it is known that Pse and Leg often occur in derivatized (modified) forms [1, 3]. Such modifications could occur before the PseI synthase acts or afterwards. In the latter case, most (if not all) synthases would be predicted to produce the same Pse molecule, which is then derivatized once it has been synthesized. If so, then the protein executing a particular enzymatic reaction should be replaceable by an orthologous enzyme executing the same reaction.
To investigate this idea on a comprehensive scale, we individually cloned 21 synthetic (codon-optimized) PseI or LegI coding sequences (CDSs) onto an expression plasmid for genetic glyco-profiling experiments using motility as proxy to report the ability of the candidates to substitute for the endogenous PseI of C. crescentus (Figure 2A and S1). In support of the notion that derivatization occurs after the PEP-dependent condensation reaction to form Pse or Leg, our glyco-profiling analysis revealed that putative PseI proteins (identified by sequence comparisons to C. jejuni 11168, Table S1) conferred motility to C. crescentus ΔpseI cells, whereas putative LegI synthases did not. This stringency for PseI synthase function using the C. crescentus motility readout was not only observed across species (e.g. Shewanella oneidensis vs. Shewanella japonica) or class (Shewanella japonica vs. Magnetospirillum magneticum), but also across the Gram-negative / Gram-positive divide (e.g. Pseudomonas sp. Irchel 3A5 vs. Kurthia sibirica) and, remarkably, across kingdoms (e.g. Leptospira interrogans vs Methanobrevibacter smithii, see Figure 2 and S1). Strikingly, in the case of certain A. baumannii strains, only one synthase least was able to confer motility to C. crescentus ΔpseI cells, suggesting that it is a PseI ortholog, while other two genomes might encode LegI-type synthases (see below).
Immunoblotting with antibodies to C. crescentus FljK [19] (FljKCc, Figure 2B) revealed that all the PseI-type synthases that restored motility, also restored FljK modification. By contrast, the non-orthologous synthases neither supported motility, nor flagellin glycosylation. We conclude from our survey that (heterologous) PseI synthase activity generally confers motility to C. crescentus ΔpseI cells, whereas LegI-type (or NeuB-type) synthases are unable to do so.
Flagellin glycosylation in Brevundimonas subvibrioides is FlmG- and LegI-dependent
An unexpected glyco-profiling result was that the synthase orthologs encoded in the genomes of different Brevundimonas species, that are members of the same family (Caulobacteraceae) as C. crescentus, were unable to replace PseI (Figure 2). A closer look by sequence comparisons revealed that three Brevundimonas orthologs tested are in fact more similar to LegI from C. jejuni 11168 than to PseI. For example, the B. subvibrioides ortholog is 42% identical and 64% similar to C. jejuni 11168 LegI and only 32% identical and 52% similar to PseI (Table S1). On this basis, we speculated that these Brevundimonas species likely synthesize Leg rather than Pse. In support of this idea, our bioinformatic searches using C. jejuni 11168 as reference genome identified all six putative enzymes in the B. subvibrioides ATCC15264 genome (CP002102.1) predicted to execute the synthesis of Leg from GDP-GlcNAc. Importantly, B. subvibrioides also encodes a FlmG ortholog (43 % identity and 59% similarity to C. crescentus FlmG), raising the possibility that it uses FlmG to glycosylate its flagellins as C. crescentus. Yet, no obvious sequence homologs of the six Pse biosynthesis enzymes were found by BlastP searches, whereas orthologs of Leg biosynthesis enzymes are readily discernible. Thus, we reasoned that B. subvibrioides FlmG could be a Leg-specific flagellin glycosyltransferase, rather than a Pse-dependent enzyme as for C. crescentus [19].
To test this idea, we first confirmed that sugar modifications are indeed present on B. subvibrioides and C. crescentus flagella. For C. crescentus, flagellin glycosylation by Pse was inferred, but not yet chemically proven. We purified flagella from supernatants of B. subvibrioides and C. crescentus cultures by ultracentrifugation, dissociated covalently linked sugars by acid-hydrolysis, derivatized them and then analyzed the liberated material by HPLC (Figure S2A). A Pse-like molecule was extracted from C. crescentus flagella, having a retention time (9.8 minutes) that is nearly identical to that (9. 7 minutes) of a Pse standard (harboring a triple acetamido modification, Pse4Ac5Ac7Ac) isolated from an A. baumannii capsule [34]. Co-injection of this Pse-standard along with the material extracted from C. crescentus flagella, revealed a co-eluting peak at 9. 7 minutes of double intensity compared to that of the standard (Figure S2A). When the same procedure was used to liberate a derivatized nonulosonic acid from B. subvibrioides flagella, a major peak was detected by HPLC analysis having a retention time of 9.8 minutes, along with a minor one eluting at 15.3 minutes (Figure S2B). A known Leg standard with a double acetamide modification (Leg5Ac7Ac) isolated from a different A. baumannii capsule [35] eluted at 12.3 minutes, suggesting that B. subvibrioides flagella are modified with a Leg-derivative that is distinct from Leg5Ac7Ac. Indeed, Leg derivatives of different mass or just simply epimers are known with substitutions of the N-acetyl/acetamido groups at the C-5 and C-7 positions, such as N-acetimidoyl or acetamidino, N-formyl and N-hydroxybutyryl groups [1, 3], that are synthesized from a Leg-type biosynthesis pathway requiring LegI.
To determine if the gene predicted to encode the LegI-like synthase of B. subvibrioides (Bresu_0507, henceforth LegIBs) or the FlmG ortholog (Bresu_2406, FlmGBs) are also required for motility in B. subvibrioides ATCC15264, we engineered in-frame deletions in each gene. We then probed the resulting ΔlegIBs and ΔflmGBs single mutants for motility defects in soft agar and analyzed flagellin glycosylation by immunoblotting using antibodies to FljKCc (Figure 3A-3D). Both mutants showed strongly reduced motility on soft agar and increased migration of flagellin through SDS-PAGE compared to WT. While no difference in the abundance of flagellin was observed in extracts from mutant versus WT cells, flagellin was barely detectable in the supernatants of mutant cultures, suggesting flagellar filament formation is defective in these mutants. Moreover, transmission electron microscopy (TEM, Figure 3E) revealed substantially shorter flagella on both mutants (average length 1 or 1.2 μm, Figure 3F) compared to those on WT cells (4 μm), suggesting that LegIBs and FlmGBs govern flagellin glycosylation and export (or stability after export). However, we cannot rule out that LegIBs and FlmGBs also promote filament assembly in addition to flagellin secretion. Flagellins are exported before their assembly into the filament [13, 14, 36], but when the assembly step is blocked they typically accumulate in the supernatant. In this situation, the resulting cells feature only a hook on the surface lacking the filament or possibly a very short stubby filament, similar to the ones revealed in our TEM images.
The impaired flagellar filament assembly observed in our mutants are clearly due the absence of LegIBs or FlmGBs as shown by the fact that introduction of a plasmid harboring either legIBs or flmGBs under control of the IPTG-inducible Plac promoter into the corresponding mutants, restored motility as well as flagellin modification and export, whereas the empty vector (pSRK-Gm [37]) was unable to do so (Figure 3A-3D). Having confirmed the importance of LegIBs and FlmGBs in flagellin modification/secretion and motility in complementation experiments, we asked if C. crescentus FlmG (FlmGCc) can substitute for FlmGBs and vice versa. These heterologous complementation experiments revealed that the FlmG variants are not interchangeable between C. crescentus and B. subvibrioides (Figure S3A), whereas the PseICc substitution experiments described above showed that PseI orthologs are functionally interchangeable (Figure 2). To test if such heterologous complementation is also possible for Leg synthases using the motility defect of B. subvibrioides ΔlegIBs cells as proxy, we conducted the orthologous glyco-profiling for LegI orthologs expressed from pSRK-Gm plasmids as described above (Figure 4). Strikingly, we obtained a near mirror-image of the complementation results from the C. crescentus ΔpseICc glyco-profiling: the orthologs that were unable to restore motility and flagellin glycosylation to C. crescentus ΔpseICc cells, predominantly restored motility (Figure 4A) and flagellin glycosylation (Figure 4B) to B. subvibrioides ΔlegIBs cells. Since these complementing synthases exhibit greater overall sequence similarity to LegI than Pse of C. jejuni 11168 (Table S1), we concluded that B. subvibrioides indeed encodes a Leg-dependent flagellin glycosylation pathway. Thus, while the C. crescentus and B. subvibrioides flagellin glycosylation systems are clearly evolutionarily related, they diverged to exhibit dissimilar donor and acceptor specificities.
LegX, a new molecular marker for Leg biosynthesis pathways
Irrefutable molecular evidence for the complete dissection of glycosylation pathway typically requires the demonstration of sufficiency by reconstitution of glycosylation in a heterologous host expressing a minimal set of the required constituents. Since FlmGCc and FlmGBs are not interchangeable and the glycosyl donor and acceptor specificities must have diverged, we tried to reconstitute the B. subvibrioides flagellin glycosylation system in C. crescentus using heterologously expressed determinants. To this end, we expressed a synthetic operon of the six B. subvibrioides Leg biosynthesis enzymes (i.e. those predicted to be responsible for the production of CMP-Leg from GDP-GlcNAc, Figure 5A) from the C. crescentus xylX locus in cells lacking flagellins, PseICc and FlmGCc. This synthetic Leg biosynthesis operon included the following CDSs of the predicted B. subvibrioides orthologs of the Leg pathway from C. jejuni 11168[22]: Bresu_3266 (LegB), Bresu_0765 (LegC), Bresu_0506 (LegH), Bresu_3264 (LegG), Bresu_0507 (LegI) and Bresu_3265 (LegF)(Figure 1 and 5A). Next, we introduced a plasmid co-expressing FljKBs and FlmGBs and then probed for modification of FljKBs by immunoblotting, asking whether a change in migration of FljKBs was discernible. As shown in Figure 5B, under these conditions the migration of FljKBs was not altered, indicating that i) additional determinants are likely required to execute the glycosylation or that ii) the synthetic CDSs do not express well enough from our plasmids.
Previously we showed that an equivalent synthetic enzyme operon comprising six Pse biosynthesis enzymes was able to direct the synthesis of CMP-Pse from the UDP-GlcNAc precursor in a heterologous host [19]. On the basis of our failure with our corresponding synthetic Leg construct, we considered the possibility that Leg biosynthesis pathway might be incomplete in our heterologous host because the putative precursor, GDP-GlcNAc, is not naturally available in C. crescentus (and other bacteria that do not normally synthesize Leg). If true, then this essential biosynthetic activity might also be encoded in Leg biosynthesis gene clusters of B. subvibrioides or other Leg-producing bacteria. Upon inspection of the predicted Leg biosynthesis clusters in the genomes of the Gram-negative bacteria A. baumannii LAC-4 (GCA_000786735.1)[38] and P. sp. Irchel 3E13 (GCA_900187455.1), as well as that of the Gram-positive bacteria Geobacillus kaustophilus HTA426 [26, 39] and Moorella humiferrea DSM 23265 [40], we noted the presence of one gene encoding an ortholog of PtmE, an enzyme that was used in the enzymatic reconstitution of Leg biosynthesis in vitro using enzymes encoded in C. jejuni 11168[22]. In these experiments PtmE, a putative guanylyltransferase of GlcN-1-P (α-D-glucosamine-1-phosphate), promoted the production of GDP-GlcNAc in vitro (Figure 5A). An ortholog (Bresu_3267, henceforth LegXBs) is also encoded adjacent to the genes encoding LegB, LegG and LegF (Bresu_3266, Bresu_3264 and Bresu_3265) orthologs in the B. subvibrioides genome (see Figure 1 and 5A).
If LegXBs is indeed required for Leg biosynthesis in B. subvibrioides, ΔlegX cells should recapitulate the motility and flagellin glycosylation defect reported above for ΔlegI and ΔflmG cells. We engineered an in-frame deletion mutation in legX and found that the resulting ΔlegX cells suffer from impaired motility (Figure 6A). Moreover, they neither glycosylate, nor export flagellin (Figure 6B) and TEM revealed only short flagellar filaments on the pole (Figure 6C), as for ΔlegI and ΔflmG cells. If LegX indeed acts in Leg biosynthesis, then it might be possible to restore motility to ΔlegX cells by expression of a LegX/PtmE ortholog (Figure 6A, 6B), similarly to the heterologous complementation of ΔlegI cells. This was indeed the case: expression of the M. humiferrea LegX ortholog (MOHU_20790) from pSRK-Gm not only restored motility to ΔlegX cells, but also flagellin glycosylation and export in a manner indistinguishable from the complementation with LegXBs (expressed from pSRK-Gm). As MOHU_20790 exhibits 54% similarity (36% identity) to LegXBs (Table S2), and the predicted fold of LegX (Figure 6D, right) closely resembles that of the nucleotidyltransferase PtmE (Figure 6D, left), we conclude that LegX enzymatic activity is required for motility and Leg biosynthesis in B. subvibrioides and that its function in motility can be conferred by LegX orthologs from phylogenetically distant bacteria, such as the Gram-positive bacterium M. humiferrea.
As the legX gene lies downstream of the predicted legB (Bresu_3266) gene, we also inactivated legB and observed that the motility of the corresponding mutant (ΔlegB) is curbed (Figure 7A) and that flagellin glycosylation and export is defective (Figure 7B). However, complementation analyses with plasmids harboring either legB or legB-legX revealed that the ΔlegB mutation is polar on legX expression, indicating that these two genes indeed form an operon (Figure 7A, 7B). We also inactivated the predicted legH gene (Bresu_0506) that lies upstream of legI (Bresu_0507), but in this case there was no evidence of polarity (Figure 7D-7F), despite a similar apparent translational as inferred from the genome sequence. In summary, our analyses show that LegX, LegB and LegH are necessary for Leg- and FlmG-dependent flagellin glycosylation in B. subvibrioides. Importantly, LegX is an ideal marker to distinguish Leg from Pse biosynthesis pathways, often embedded in flagellar clusters [26] and, owing to its functional conservation, suitable for the establishment of glyco-profiling set-ups that rely on the motility defect of ΔlegX cells as proxy.
Reconstitution and rewiring of FlmG-dependent flagellin glycosylation
Having unveiled LegX as a critical component of the B. subvibrioides Leg-based motility system, we asked whether addition of the LegX enzyme would permit reconstitution of the Leg-dependent flagellin glycosylation by FlmGBs in our recombinant C. crescentus cells expressing the other six Leg biosynthesis enzymes. To this end, we transformed a compatible plasmid harboring B. subvibrioides LegX CDS (pMT375-legX) into the expression C. crescentus strains already described above that lack flagellins, PseI and FlmG and performed immunoblot to determine if FlmGBs can support the modification of FljKBs in Leg- and LegX-dependent manner. As shown in Figure 5B, FljKBs was converted to a substantially slower migrating species, a modification that was dependent on the presence of FlmGBs and all seven Leg biosynthesis enzymes (including LegX). Additionally, we observed a barely detectable change in mobility that is FlmGBs-dependent, but requires neither Pse, nor Leg (see asterisk, Figure 5B). This change in FljKBs mobility may reflect a certain degree of promiscuity of FlmG towards other donor molecules that are transferred to FljKBs.
We hypothesized that the specificity of FlmG enzymes towards Leg versus Pse likely resides in the C-terminal glycosyltransferase (GT-B domain [41]). This hypothesis is based on our previous finding that the N-terminal TPR domain of FlmGCc can bind FljKCc, whereas the GT-B alone cannot [19]. Since FlmGBs shares this modular architecture based on sequence analysis, we wondered if a chimeric version of FlmGCc-Bs in which we substituted the GT-B domain from C. crescentus with that of B. subvibrioides would thus glycosylate C. crescentus flagellins with Leg. To this end, we used C. crescentus ΔflmG mutant cells harboring the synthetic six-gene Leg operon at the xylX locus. We first transformed these cells with pMT375-legXBs and then finally with pSRK-Gm variants expressing either FlmGCc, FlmGBs, or the chimeric FlmGCc-Bs version. As shown in Figure 5C, the chimeric FlmGCc-Bs was able to modify the C. crescentus flagellins in a manner that depended on the presence of LegXBs, but it did not modify flagellin in cells producing only Pse (also observed in Figure S3A). Moreover, WT FlmGBs version did not support efficient flagellin modification in the Leg-producing C. crescentus cells regardless of whether LegXBs was present or not (Figure 5C). As control, FlmGCc also supported flagellin modification in this system (likely with Pse), because these ΔflmG cells produce both Pse and Leg, but only in the presence of pMT375-legXBs.
In summary, exchanging the C-terminal GT-B domain enabled rewiring the glycosyltransferase specificity from Pse-accepting enzyme to a Leg-accepting enzyme, resulting in the modification of C. crescentus flagellins with Leg in cells recombinantly expressing at least seven Leg biosynthesis genes. The fact that such cells are non-motile (Figure S3B) indicates that additional factors exist in the flagellation pathway that exhibit specificity towards the glycosyl group that is joined to flagellins.
DISCUSSION
Insulated Leg- or Pse-dependent glycosylation pathways
The exquisite specificity in cellular glycosylation reactions are predetermined to ensure that the desired structures are decorated with the correct sugars. In as much as the underlying glycosyl donor and acceptor selectivity underlie biological function, biotechnological processes often necessitate relaxing these specificities, for example in engineering promiscuous glycosyltransferase (GT) enzymes that can be used to modify a desired target protein with a sugar of choice [42]. Such long-term goals are achievable, but ideally facilitated by the discovery and dissection of the determinants underpinning the GT specificities, including acceptor and donor. In addition to illuminating the molecular mechanism of FlmG fGTs, our work also opens the door towards biotechnological engineering of flagellin-based bio-glycoconjugates using Pse or Leg for example as simple vaccine [43-45] that could serve to combat Pse/Leg in infections by prior immunization not only for A. baumannii strains or other pathogens that decorate surfaces with Pse, but, importantly, also for those that contain Leg, including most clinical A. baumannii isolates [12, 34, 46].
Our genetic dissection of orthologous FlmG fGTs provided unprecedented insight into the donor sugar and acceptor protein specificities underlying protein glycosylation mechanisms. At the level of the donor, featuring a remarkable stereoisomer selectivity, we showed that the FlmGs from C. crescentus and B. subvibrioides evolved a strong preference for either Leg or Pse (Figures 2 and 4). Additionally, the stereoisomer specificity of the donor is already reflected in the biosynthesis pathway. Inactivation of the defining synthase enzymes for Leg or Pse, LegI and PseI, yields the same motility defects as the inactivation of the corresponding FlmG enzyme. LegI and PseI cannot substitute for one another in the two flagellation systems that we studied, indicating that the corresponding biosynthesis pathways are genetically (and therefore biochemically) insulated. However, the fact that different PseI orthologs can substitute for the endogenous C. crescentus enzyme and, in turn, LegI orthologs can substitute for the endogenous enzyme from B. subvibrioides when probing motility, underscores the specificity of the biosynthesis pathways for the two stereoisomers. This stringency lends itself for in vivo glyco-profiling using ΔpseI and ΔlegI mutant strains of C. crescentus and B. subvibrioides, respectively, to functionally probe for Pse or Leg biosynthesis pathways identified in genome searches. Remarkably, such profiling assays not only permit distinction among strains and species, but are also discriminatory across larger phylogenetic distances, including the Gram-positive to Gram-negative divide and even the boundaries between eubacterial and archaeal kingdoms. By extension, having recognized the LegX/PtmE enzyme as a critical element in the Leg-specific enzymatic biosynthesis step (Figure 6) likewise offers another functional, but also a novel bioinformatic, criterion for the correct assignment and discrimination of predicted stereoisomer biosynthesis routes residing in ever-expanding genome databases. The current era of synthetic biology offers unlimited depth to which such synthetic genetic glyco-profiling approaches can be applied.
Specificity determinants in flagellin glycosyltransferases
Our reconstituted Leg-dependent glycosylation of FljKBs by FlmGBs in C. crescentus ΔpseI cells using a synthetically assembled Leg-biosynthesis operon, complemented with LegX (Figure 5), allowed us to unambiguously establish the minimal set of components that are required to achieve protein glycosylation using a Leg-based system. The FlmG class of fGTs are suitable subjects for molecular dissection of the underlying specificity determinants because of their conspicuous two-domain architecture that is recognizable by simple primary structure (sequence) comparisons, even without tertiary structural analysis. In fact, the (predicted) bilobed FlmG structure [15] had previously prompted us to determine that the N-terminal TPR domain of FlmGCc confers flagellin (acceptor) recognition, whereas the GT-B domain cannot bind flagellin [19]. Hypothesizing that the GT-B domain could act as determinant for the donor, we considered a simple division of labor model between the two parts of FlmG accounting for the bipartite specificity. Proof for this notion came from the analysis of a chimeric form, FlmGCc-Bs, in which the flagellin binding domain from FlmGCc was joined to the GT-B domain of FlmGBs. Expression of this chimeric FlmGCc-Bs variant in C. crescentus ΔflmG cells that had been engineered to synthesize Leg resulted in the modification of the C. crescentus flagellin, whereas the WT version of FlmGBs had poor activity (Figure 5C). Conversely, FlmGCc-Bs was unable to support glycosylation of C. crescentus flagellins with Pse, likely because it no longer possesses the Pse-specific GT-B domain of FlmGCc.
Similar dissection experiments should be conducted with the other class of fGTs that are wide-spread in bacteria, the Mafs [10, 26, 28, 47, 48], to reveal if analogous mechanisms and determinants underpin flagellin glycosylation in these systems. It stands to reason that donor and acceptor specificities exist in Mafs as well, however, the flagellin recognition determinants remain unknown. An X-ray structure determined for the Maf from M. magneticum [28] revealed a tripartite domain architecture with central GT-A domain bearing clear resemblance to the GT29 and GT42 family of sialyltransferases. The GT-A domain is a characteristic of the Mafs (also known as the signature MAF_flag10 domain) and is likely to confer Pse donor specificity. In fact, our glyco-profiling revealed the corresponding synthase of M. magneticum to have PseI activity in our motility assay (Figure S1) and our sequence analysis by BlastP easily discerned a complete (predicted) Pse-biosynthesis pathway encoded in its genome. While the flagellin binding determinant was not evident in the M. magneticum Maf structure, a weak structural similarity with flagellin and flagellin secretion chaperones may point to a C-terminal flagellin recognition determinant. However, it remains to be determined whether this region is necessary and sufficient for flagellin binding.
Khairnar et al. [26] provided evidence of some donor promiscuity in the Maf from G. kaustophilus that is encoded in flagellar locus. When this Maf was expressed in recombinant Escherichia coli cells that synthesize sialic acid, modification of the co-expressed G. kaustophilus flagellin was seen with sialic acid. However, it would be interesting to test if the efficiency of flagellin glycosylation by G. kaustophilus Maf is increased in a heterologous host producing Leg as maf gene is adjacent to Leg biosynthesis genes, including a LegX ortholog (Table S2) and our glyco-profiling in Figure 4 revealed that G. kaustophilus indeed encodes a LegI ortholog. Overall, it remains to be determined whether the Mafs are inherently more promiscuous than the FlmG enzymes.
Leg- and Pse-based glycosylation in the (same) prokaryotic cell
The donor specificity observed with the two FlmG enzymes studies in our work and the possible specificity inferred for Maf-encoding gene clusters may ensure that the correct cellular structure is modified with the right donor. In our experiments when C. crescentus cell synthesizing Leg were used, FlmGCc had a clear preference to modify flagellin with Pse, rather than Leg. Thus, the terminal determinant of a given glycosylation pathway governs selectivity of CMP-Pse over CMP-Leg or vice versa. Our work also indicates that the biosynthesis pathways themselves are kept insulated by dedicated enzymes, perhaps to prevent the formation of Leg/Pse hybrid intermediates that would otherwise create too much chemical variability for the systems to function properly in their biological roles. Pse and Leg glycosylation systems are used for other cell surface structures, not only flagellins [1, 16, 49]. While Leg or Pse biosynthesis enzymes are often encoded in flagellar gene clusters, they can also occur within O-antigen or capsular gene clusters, sometimes even in the same genome. In this situation, a possible enzymatic interference of the Leg and Pse biosynthesis pathways must be avoided. In bacterial cells, it might be possible to restrict Leg or Pse synthesis to specific (mutually exclusive) growth conditions, but a perhaps more common solution is to design each biosynthetic pathway with specific chemical marks. An appealing hypothesis is that this is achieved with the synthesis of Leg from GDP-activated precursors, whereas Pse synthesis occurs from UDP-activated molecules [21, 22]. This dependency on GDP for Leg also comes at a price, since at least one specific conversion enzyme, such as LegX, is required to launch the biosynthesis pathway with the production of the activated precursor bearing the GDP mark (Figure 6).
For such chemical complexity in the biosynthesis of stereoisomers to evolve, these glycosylation systems must be of considerable value to cells, begging the question of their function. Do surface modifications with Pse or Leg just serve to generate epitopes or envelopes with different modifications or are there special physical or chemical properties associated with Pse or Leg. As members of the sialic family of molecules they certainly have the potential to function as innate immune modulators, but Leg or Pse also found on environmental bacteria that are not known to associate with eukaryotic cells. While Pse and Leg play an important role in C. crescentus and B. subvibrioides flagellation, respectively, many other flagellation systems exist that do not require Pse or Leg. On the basis of this fact, it is conceivable that glycosylation does not fulfill a conserved role in flagellar assembly in general, but we cannot exclude that it has been appropriated for regulatory purposes in some bacterial flagellation systems. We observed that modifying FljKCc with Leg in our recombinant C. crescentus system still did not restore motility, suggesting that the type of sugar modification does matter, possibly because other flagellin interacting proteins such as the FlaF secretion chaperone, capping proteins or unknown factors no longer interact or function properly with FljKCc that does not harbor the Pse modification. Alternatively, or additionally, it is conceivable that modification of the flagellar filament with Pse or Leg simply protects against infection by certain flagellotropic phages, in a manner analogous to that reported recently for pilus glycosylation in Pseudomonas aeruginosa [50].
EXPERIMENTAL PROCEDURES
Strains and growth conditions
Bacterial strains used in this study are listed in Table S3. C. crescentus and B. subvibrioides strains were grown at 30°C in peptone-yeast extract (PYE) (2g/L bacto-peptone, 1g/L yeast extract, 1 mM MgSO4 and 0.5 mM CaCl2)[51]. E. coli S17-1 λpir and EC100D were grown at 37°C in LB. Antibiotics were added to the medium at the following concentration (µg/mL in liquid/solid medium for C. crescentus and B. subvibrioides; µg/mL in liquid/solid medium for E. coli): nalidixic acid (20 only in solid medium for C. crescentus and B. subvibrioides), tetracycline (1/1; 10/10), kanamycin (5/20; 20/20), gentamycin (1/1; 25/25). Gene expression was induced when required with 50 µM vanillate, 0.3% D-xylose or 0.5 mM isopropyl-beta-D-thiogalactoside (IPTG) for C. crescentus and B. subvibrioides cultures. Electroporation, bi-parental mating and motility assays were performed as previously described in C. crescentus and B. subvibrioides [51, 52].
For motility assays, 1 µL of overnight cultures were spotted on soft (0.3%) agar plates with the corresponding antibiotics and inducers (IPTG or vanillate) and incubated for 3 days and 7 days for C. crescentus and B. subvibrioides, respectively.
Immunoblots
For immunoblots, protein samples were prepared from cells harvested in the middle of the exponential growth phase (1 mL at OD600nm≈0.4). Proteins samples were separated on SDS polyacrylamide gel, transferred to polyvinylidene difluoride (PVDF) Immobilon-P membranes (Merck Millipore) and blocked in Tris-buffered saline (TBS) 0.1% Tween-20% and 5% dry milk [19]. The anti-FljKCc anti-serum (raised against His6-FljK expressed in E. coli [19]) was used at 1:10,000 dilution. Protein-primary antibody complexes were revealed with horseradish peroxidase-labeled donkey anti-rabbit antibodies (Jackson ImmunoResearch, West Grove, PA) and ECL detection reagents (Amersham, GE Healthcare, Glattbrugg, Switzerland).
Negative stain transmission electron microscopy
Samples for negative stain TEM were prepared by first glow discharging 200-mesh copper, carbon-coated, formvar grids (EM Science, Hatfield, PA) for 1 min. 20 µL of exponential cultures of B. subvibrioides were applied to the grids and allowed to adsorb for 1 min before being washed three times in water, stained with 1% uranyl acetate for 1 min and washed with water for 30 sec. Negatively stained B. subvibrioides were imaged on a Tecnai 20 (FEI Company, Eindhoven, Netherland). Flagellum length measurement was performed using the ImageJ software.
Derivation of nonulosonic acids (NulOs) with DMB and analysis on HPLC
We extracted NulOs from lyophilized purified flagella from culture supernatants. Briefly, 250 mL of an overnight culture (24 h for B. subvibrioides) was spun for 15 min at 8,000 r.p.m. at 4°C to remove cells. Shed flagella were then pelleted from the culture supernatant by ultracentrifugation at (27,000 r.p.m. 30 min, 15°C), washed with 50 mL water and pelleted again by ultracentrifugation. Purified flagella were resuspended in water and frozen at -80°C prior to lyophilization. 1,2-diamino-4,5-methylene dioxybenzene (DMB) was used to derivatize NulOs as previously described [28]. Briefly, dried glycoconjugates were hydrolyzed in 0.1 M trifluoroacetic acid for 2 h at 80°C to release NulOs. NulOs were coupled to DMB for 2 h at 50°C in the dark in a derivation solution (7 mM DMB; 1 M β-mercaptoethanol; 18 mM sodium hydrosulfite; 0.02 mM trifluoroacetic acid). NulO derivatives were separated isocratically on a C18 reverse-phase HPLC column (Thermo Scientific, Hypersil ODS, 4.6 mm by 250 mm, 5 µm) using acetonitrile/methanol/water (7:9:84 vol/vol/vol) mixture solvent and detected by a fluorimeter (Waters 2475, excitation wavelength λexc=373 nm, emission wavelength λem=448 nm.
Strain and plasmid constructions
For in-frame deletions, bi-parental mating was used to deliver the corresponding pNPTS138 derivatives (listed in Table S3) into B. subvibrioides strains. Double recombination was selected by plating bacteria onto PYE plates supplemented with 3% sucrose. Putative mutants were confirmed by PCR using primers (listed in Table S4) external to the DNA fragments used for the pNPTS138 constructs.
pNK562: PCR was used to amplify two fragments flanking the flmG (Bresu_2406) ORF with Bs_flmG_del_1/Bs_flmG_del_2 and Bs_flmG_del_3/Bs_flmG_del_4. The PCR fragments were digested with MfeI/BamHI and BamHI/HindIII, respectively and triple ligated into pNPTS138 restricted with EcoRI/HindIII.
pNK580: PCR was used to amplify two fragments flanking the neuB (Bresu_0507) ORF with Bs_neuB_del_1/Bs_neuB_del_2 and Bs_neuB_del_3/Bs_neuB_del_4. The PCR fragments were digested with EcoRI/BamHI and BamHI/HindIII, respectively and triple ligated into pNPTS138 restricted with EcoRI/HindIII.
pNK926: PCR was used to amplify two fragments flanking the Bresu_3266 ORF with NK339/NK340 and NK341/342. The PCR fragments were digested with HindIII/BamHI and BamHI/EcoRI, respectively and triple ligated into pNPTS138 restricted with EcoRI/HindIII.
pNK1000: PCR was used to amplify two fragments flanking the Bresu_3267 ORF with NK345/NK346 and NK347/348. The PCR fragments were digested with HindIII/KpnI and KpnI/EcoRI, respectively and ligated into pNPTS138 restricted with EcoRI/HindIII.
pNK1002: PCR was used to amplify two fragments flanking the Bresu_0506 ORF with NK366/NK367 and NK368/369. The PCR fragments were digested with HindIII/BamHI and BamHI/EcoRI, respectively and ligated into pNPTS138 restricted with EcoRI/HindIII.
Inducible plasmids were constructed with a NdeI site overlapping the start codon and an XbaI site (or EcoRI site when mentioned) flanking the stop codon were constructed as follows:
pNK660: the flmG ORF was amplified by PCR with Bs-flmG-NdeI/ Bs-flmG-XbaI. The PCR fragment was digested by NdeI/XbaI and ligated into pSRK-Gm [37] restricted with NdeI/XbaI.
pNK631: the synthetic fragment encoding the legI (Bresu_0507) CDS, codon optimized for E. coli (see Table X), was subcloned into pSRK-Gm from pUCIDT plasmid using NdeI/XbaI.
pNK948: the Bresu_3266 CDS was amplified by PCR with NK357/358. The PCR fragment was digested by NdeI/XbaI and ligated into pSRK-Gm restricted with NdeI/XbaI.
pNK950: the Bresu_3267 CDS was amplified by PCR with NK359/360. The PCR fragment was digested by NdeI/XbaI and ligated into pSRK-Gm restricted with NdeI/XbaI.
pNK988: the Bresu_3266-67 CDSs were amplified by PCR with NK357/360. The PCR fragment was digested by NdeI/XbaI and ligated into pSRK-Gm restricted with NdeI/XbaI.
pNK974: the Bresu_0506 CDS was amplified by PCR with NK374/375. The PCR fragment was digested by NdeI/XbaI and ligated into pSRK-Gm restricted with NdeI/XbaI.
pNK957: the Bresu_3267 CDS was amplified by PCR with NK359/360, digested by NdeI/XbaI and cloned into pMT375.
pSA228: the synthetic fragment encoding fljKBs (Bresu_2638, codon optimised for C. crescentus) CDS was digested by NdeI/EcoRI and ligated into pMT463 [53] restricted by NdeI/EcoRI.
pLT2043: the neuB CDS of Pseudomonas irchel 3A5 was amplified with 3A5_PseI_NdeI/3A5_PseI_mfeI. The PCR fragment was digested by NdeI/MfeI and ligated into pMT335 restricted with NdeI/EcoRI.
pLT2036: the legI CDS of B. subvibrioides was amplified with Bs_neuB_nde/Bs_neuB_eco. The PCR fragment was digested by NdeI/EcoRI and ligated into pMT335 [53] restricted with NdeI/EcoRI.
pLT2237: the pseI CDS of Kurthia was amplified with Ku_neuB_nde/Ku_neuB_eco. The PCR fragment was digested by NdeI/EcoRI and ligated into pMT335 restricted with NdeI/EcoRI.
pLT2262: the pseI CDS of M. magneticum was amplified with Mm_neuB_nde/Mm_neuB_eco. The PCR fragment was digested by NdeI/EcoRI and ligated into pMT335 restricted with NdeI/EcoRI.
pLT2263: the pseI CDS of S. oneidensis was amplified with So_neuB_nde/So_neuB_eco. The PCR fragment was digested by NdeI/EcoRI and ligated into pMT335 restricted with NdeI/EcoRI.
To create the vector co-expressing fljKBs and flmGBs, the flmG ORF was amplified by PCR with primers Bs_flmG_rbs_Eco/Bs_flmG_Xba (with ribosome binding site and EcoRI site flanking the flmG start codon and XbaI site flanking the flmG stop codon) and digested by EcoRI/XbaI. The digested fragment was subcloned into pSA228[19] restricted by EcoRI/XbaI.
To express the legX ortholog from Moorella humiferrea, the MOHU_20790 ORF (codon optimized for E. coli) was amplified by PCR using NK361 and M13(−48) primers from pUC-GW-MOHU_20790syn plasmid. After digestion by NdeI/XbaI, the fragment was ligated into pSRK-Gm restricted by NdeI/XbaI to generate pNK955.
pLT2295: to express the six enzyme B. subvibrioides legionaminic acid biosynthesis pathway in C. crescentus, a synthetic operon (codon-optimized for E. coli, see Table S4) encoding Bresu_3266, Bresu_0765, Bresu-0506, Bresu-3264, Bresu-0507 and Bresu-3265 was subcloned from pUC-GW plasmid to pXGFP4 using NdeI/XbaI.
To express pseICj, legICj and neuBCj, the corresponding CDSs were individually subcloned from pSA126, pSA47 and pSA48 [19] to pSRK-Gm using NdeI/XbaI to generate pNK991, pNK992 and pNK994, respectively.
To express heterologous PseI or LegI orthologs, the synthetic fragments harboring the CDSs (codon optimized for E. coli, see Table S4) were subcloned from pUC-GW or pUCIDT plasmids to pSRK-Gm using NdeI/XbaI (Table S3).
FIGURE LEGENDS
ACKNOWLEDGEMENTS
We thank Laurence Degeorges for excellent technical assistance and Silvia Ardissone for the plasmids expressing B. subvibrioides flagellins and FlmG. We also thank Bohumil Maco for the help with TEM experiments. Funding support was from the Swiss National Science Foundation (31003A_182576), the University of Geneva (DIP) and UNITEC (InnogapR23-24) to P.H.V, and a Swisslife Foundation (Jubiläumsstiftung) grant for medical research to N.K..