Glycoengineering of host mimicking type-2 LacNAc polymers and Lewis X antigens on bacterial cell surfaces


For correspondence. E-mail; Tel. (+41) 44 632 64 13; Fax (+41) 44 632 13 75.


Bacterial carbohydrate structures play a central role in mediating a variety of host–pathogen interactions. Glycans can either elicit protective immune response or lead to escape of immune surveillance by mimicking host structures. Lipopolysaccharide (LPS), a major component on the surface of Gram-negative bacteria, is composed of a lipid A-core and the O-antigen polysaccharide. Pathogens like Neisseria meningitidis expose a lipooligosaccharide (LOS), which outermost glycans mimick mammalian epitopes to avoid immune recognition. Lewis X (Galβ1–4(Fucα1–3)GlcNAc) antigens of Helicobacter pylori or of the helminth Schistosoma mansoni modulate the immune response by interacting with receptors on human dendritic cells. In a glycoengineering approach we generate human carbohydrate structures on the surface of recombinant Gram-negative bacteria, such as Escherichia coli and Salmonella enterica sv. Typhimurium that lack O-antigen. A ubiquitous building block in mammalian N-linked protein glycans is Galβ1-4GlcNAc, referred to as a type-2 N-acetyllactosamine, LacNAc, sequence. Strains displaying polymeric LacNAc were generated by introducing a combination of glycosyltransferases that act on modified lipid A-cores, resulting in efficient expression of the carbohydrate epitope on bacterial cell surfaces. The poly-LacNAc scaffold was used as an acceptor for fucosylation leading to polymers of Lewis X antigens. We analysed the distribution of the carbohydrate epitopes by FACS, microscopy and ELISA and confirmed engineered LOS containing LacNAc and Lewis X repeats by MALDI-TOF and NMR analysis. Glycoengineered LOS induced pro-inflammatory response in murine dendritic cells. These bacterial strains can thus serve as tools to analyse the role of defined carbohydrate structures in different biological processes.


Prokaryotic surfaces contain diverse carbohydrate structures and represent one of the first interfaces between bacteria and the mammalian host. These bacterial surface molecules include capsular polysaccharides, lipopolysaccharides (LPS), lipoteichoic acids, peptidoglycans and glycoproteins of Gram-negative bacteria and/or Gram-positive bacteria. Variations in carbohydrate structures contribute to differences in the immunological epitopes and consequently to the immune response itself (Comstock and Kasper, 2006)

Lipopolysaccharide, a major component on the surface of Gram-negative bacteria, is composed of the O-antigen polysaccharide (O-Ag), covalently linked to the lipid A-core. The lipid A-core region and the O-Ag are synthesized as separate units at the cytoplasmic leaflet of the inner membrane. The lipid A-core moiety is transported to the periplasmic side of the inner membrane by the ABC transporter MsbA. In the polymerase dependent pathway, the O-Ag subunit is assembled on the undecaprenol pyrophosphate (UndPP) lipid carrier in the cytoplasm, which is then flipped to the periplasm by Wzx. WaaL ligates the UndPP linked O-Ag subunit to the preassembled lipid A-core whereas Wzy polymerizes O-Ag subunits, which chain length is controlled by Wzz (Raetz and Whitfield, 2002). The fully assembled LPS is transported through the periplasm and across the outer membrane (OM) by the Lpt pathway (Ruiz et al., 2009). The core oligosaccharide linking the lipid A and the variable O-Ag is formed by non-repetitive hetero-oligosaccharides, with mostly hexoses and N-acetylhexoses (Heinrichs et al., 1998). The inner lipid A-core is typically composed of heptoses and 3-deoxy-d-manno-oct-2-ulopyranosonic acid (Kdo) with variations in phosphorylation and additions of monosaccharides (Raetz and Whitfield, 2002). Integrated in the OM as innermost part of the LPS, the lipid A can vary in substituents and acylation (Mayer et al., 1990). Diverse modifications of the lipid A moiety can occur during its translocation to the OM and provides resistance to cationic antimicrobial peptides (Raetz et al., 2007). Lipid A or ‘endotoxin’ is recognized by pattern recognition receptors (PRRs), namely Toll-like receptor 4 (TLR4) of the innate immune system. LPS-induced signalling of TLR4 triggers the activation of NF-kB dependent genes and the release of pro-inflammatory cytokines as well as nitric oxide (Miller et al., 2005).

Some Gram-negative bacteria associated to mammalian hosts such as Campylobacter jejuni, Haemophilus influenzae or Neisseria meningitidis are devoid of O-Ag repeats but they expose lipooligosaccharide (LOS) (Preston et al., 1996). The core oligosaccharide is directly assembled onto the lipid A in the cytoplasm without requirement of UndPP for its biosynthesis. The outermost glycans of LOS containing mammalian epitopes including LacNAc units are shown to play a role in immune modulation (van Vliet et al., 2009). Sialylation of Neisseria and Haemophilus LOS may further protect the bacterium through masking (Mandrell and Apicella, 1993; van Putten and Robertson, 1995). The glycosyltransferases that modify these surface glycan structures are switched on and off by high frequent phase variation as a result of slipped strand mispairing, presumably allowing immune evasion during chronic or recurrent infection and/or the generation of variants with altered ability to colonize niches in the host (van der Woude and Baumler, 2004). Mucosal pathogens do not only escape innate immune recognition and resistance against complement (Schneider et al., 2007), but it is also conceivable that they modulate surface marker expression and cytokine production of immune cells by interaction with C-type lectins (CLRs) (van Vliet et al., 2009). The innate immune system also recognizes carbohydrate structures by glycan-binding receptors including Sialic-acid binding Ig-like lectins (Siglecs) (Crocker et al., 2007), galectins (Rabinovich and Toscano, 2009) and CLRs (Garcia-Vallejo and van Kooyk, 2009). Some of these receptors are signalling receptors and function as PRRs whereas others mediate cell adhesion and antigen uptake (van Kooyk and Rabinovich, 2008).

Mammalian glycans contain type-2 units, which are composed of the disaccharide Galβ1-4GlcNAc, N-acetyllactosamine (LacNAc), which can be recognized with high affinity by galectins. Poly-LacNAc chains can also serve as acceptors for subsequent glycosylations, including fucosylation and sialylation. The Lewis blood group family includes the Lewis X and Lewis Y determinants (Marionneau et al., 2001), which can be sialylated and/or sulphated, modifications important for human sperm-egg binding and in selectin-dependent leucocyte and tumour cell adhesion processes (Phillips et al., 1990; Varki, 1994; Pang et al., 2011). Lewis X antigen (Galβ1–4(Fucα1–3)GlcNAc) is highly expressed on the embryo cell surface during mammalian development (Muramatsu, 1988). Lewis X is also found in the pathogen H. pylori, which adheres to the gastric mucosa displaying Lewis antigens (Edwards et al., 2000). H. pylori itself displays Lewis antigens within its phase-variable O-Ag as molecular mimicry (Appelmelk et al., 1998; Wang et al., 2000; Moran, 2008). Through its interaction of Lewis X with the CLR DC-SIGN, H. pylori has been shown to modulate the Th1/Th2 balance, leading to enhanced anti-inflammatory IL-10 production and to inhibition of a pro-inflammatory Th1 response (Bergman et al., 2004; 2006). Strikingly, binding of pathogens to DC-SIGN promoted both Th1 as well as Th2-mediated responses as a result of different signalling cascades induced by either mannose or fucose containing glycans (Geijtenbeek and Gringhuis, 2009; Gringhuis et al., 2009). The glycolipids of the parasitic helminth Schistosoma mansoni containing Lewis X induce dendritic cell (DC) activation mediated by TLR4, which requires the binding of DC-SIGN to fucose moieties, hence leading to a Th1 response, a predominant response elicited before the egg laying life stage (van Stijn et al., 2010).

As a first step towards the functional analysis of a specific carbohydrate epitope present in LOS, we used a glycoengineering approach to modify well-defined lipid A-core mutants of Salmonella enterica sv Typhimurium with mammalian glycan epitopes by expression of specific glycosyltransferases. We generated polymeric LacNAc and Lewis X antigens, which were efficiently displayed on the bacterial surface.


Engineering of a LacNAc polymer on lipid A-cores of Escherichia coli

The lipid A-outer core oligosaccharides are truncated by deleting the corresponding glycosyltransferases to provide an assembly platform for glycoengineering (Paton et al., 2005). Derivatives have been generated in E. coli K12 and S. Typhimurium having glucose I as acceptor (Fig. 1) to assemble galactose, which in turn is sialylated resulting in a GM3 epitope (Ilg et al., 2010). We aimed at analysing activities of glycosyltransferases derived from LOS biogenesis pathways in Neisseria sp. for their requirements of heterologous recombinant lipid A-cores (Fig. 1). To generate a Lacto-N-neotetraose (LNnT) motif, we constructed expression vectors encoding lgtA, a β-1,3-N-acetylglucosamine transferase,lgtB, a β-1,4-galactosyltransferase from Neisseria meningitidis (Paton et al., 2005) and lgtE β-1,4-galactosyltransferase from Neisseria gonorrhoeae (Paton et al., 2000). We expressed these genes inserted in compatible low copy vectors (Table 1) alone or in different combinations (not shown). We determined the assembly of LacNAc first by visualization using periodic acid oxidation-silver staining of crude LOS preparations and by a fungal galectin, CGL2 from Coprinopsis cinerea, a beta-galactoside binding lectin (Walser et al., 2004). We investigated expression of lgtA and lgtB in trans of an E. coli K12 derivative without O-Ag repeating units due to lack of the rhamnosyltransferase wbbL catalysing the attachment of rhamnose during assembly of the O-Ag on UndPP. This strain has a complete lipid A-core and due to the activity of UDP-GlcNAc:UndP GlcNAc-1-P transferase WecA, only a single GlcNAc residue is ligated onto the HepIV. It was sufficient to express lgtAB in E. coli wbbL to generate a polymer as determined by size shift on silver stained crude LOS extracts, which corresponded to signals on blots probed with CGL2 (Fig. 2A). Therefore, we conclude that lgtA and lgtB are sufficient to modify nascent lipid A-cores with CGL2 reactive glycan additions. When lgtA and lgtB were expressed in E. coli with a truncated lipid A-core (waaB waaO mutant), the terminating Glc on the lipid A-core was used as acceptor and a polymeric ladder pattern was observed (Fig. 2A). We suggest that lgtA lgtB assembled LacNAc structures on native and truncated lipid A-cores of E. coli as polymers (Fig. 1). We additionally had to repostulate that lgtB had a relaxed specificity for its acceptor sugar, being either Glc or GlcNAc and initiated the assembly, which would then lead to a type II LacNAc polymer in the presence of lgtA. Next, we asked if UndPP-linked polymer was an intermediate in the synthesis of the novel LOS. We isolated crude LOS and performed a mild acid hydrolysis, but we did not observe any difference on crude LOS with regard to polymeric LOS (data not shown). We then used E. coli waaL, lacking O-antigen ligase WaaL and expressed lgtAB. Poly-LacNAc could be observed in the ligase mutant as well as in the isogenic O-Ag negative strain, not mutated in waaL but lacking wecA (Fig. 2A). From these data we concluded that LacNAc units were assembled independently of wecA and the generation of the polymeric LOS did not require WaaL ligase.

Figure 1.

LOS glycoengineering. Schematic representation of E. coli and Salmonella enterica sv Typhimurium native and truncated lipid A-core structures with glycoengineered LacNAc and Lewis X structures. Major forms of truncated core structures were found to be Hep2P3. O-Antigens assembled on the UndPP carrier and transferred by WaaL to the core site are labelled with an arrow with asterisk. Glycosyltransferases involved in LacNAc and Lewis X synthesis are indicated by an arrow.

Figure 2.

Synthesis of polymeric LacNAc containing LOS is independent of O-Ag biosynthesis and motive is surface displayed in E. coli.

A. Silver staining (top) and CGL2 lectin blot (bottom) of crude LOS extracts separated by Tris-Tricine gels containing 17% polyacrylamide of E. coli wbbL and E. coli waaO waaB expressing either lgtA or lgtB as indicated or E. coli wbbL, E. coli waaL or E. coli wecA mutants harbouring lgtAB or vector control.

B. Single-parameter histogram of FACS sorting represents fluorescence intensity plotted against number of bacteria (% of max) in one representative of three independent experiments. Percentages indicated are CGL2 labelled E. coli strains encoding lgtAB (black line) within gate of > 2 × 102 of fluorescence intensity (arrow). Bacteria containing vector control labelled with CGL2 shown as black dashed line, Streptavidin-647 labelled bacteria as grey lines and light grey filled area show unstained bacteria.

Table 1. Strains and plasmids
Strain or plasmidGenotype and/or descriptionReference or source
Bacterial strains
E. coli  

wbbL F− φ80lacZΔM15 Δ(lacZYA-argF)U169 recA1 endA1 hsdR17(rk−, mk+) phoA supE44 thi-1 gyrA96 relA1 λ-

K12 derivative. Native lipid A-core. Deletion in wbbL (O-Ag-)


wcaJ nanA waaO waaB supE D(lac-proAB) hsdR4 F0 traD36 proABþ lacI q lacZDM15

K12 derivative (JM107). Truncated lipid A-core/deletion of waaO waaB (O-Ag-)

Ilg et al. (2010)

lacZ trp (sbcB-rfb) upp rel rpsL wecA-G

K12 derivative. Native lipid A-core

Deletion of complete O-Ag, ECA and colanic acid gene clusters. Deletion in waaL (O-Ag-)

Perez et al. (2008)

lacZ trp (sbcB-rfb) upp rel rpsL wecA-G

K12 derivative. Native lipidA-core

Deletion of complete O-Ag, ECA and colanic acid gene clusters. Deletion of wecA. (O-Ag-)

Alaimo et al. (2006)
Top10wbbL F- mcrA Δ(mrr-hsdRMS-mcrBC) φ80lacZΔM15 ΔlacX74 recA1 araD139 Δ(araleu)7697 galU galK rpsL, endA1 nupG. (O-Ag-)Invitrogen
Salmonella enterica sv Typhimurium
M525Pwild type strain, serogroup B.Mastroeni et al. (2000)
SKI12Derivative of SL1344 wild type strain, serogroup B. Native lipid A-core, deletion in wbaP (O-Ag-)Ilg et al. (2009)

M525P nanA waaI waaB

Truncated lipid A-core. (O-Ag-)

Ilg et al. (2010)
SMM6SKI22 wcaJ (O-Ag-)This study
LxSMM6 harbouring pMMZ10 and pMMZ14This study
LNSMM6 harbouring pMMZ10 and pBSIIKS(+)This study
Core 3SMM6 harbouring pEXT21 and pMMZ14This study
Core 4SMM6 harbouring pEXT21 and pBSIIKS(+)This study
pBSIIKS (+)ColE1 ori, AprStratagene
pBSfutAfutA inserted into the BamHI SacI sites of pBSIIKS.Dumon et al. (2006)
pACT3CamR, p15A ori, tac promoterDykxhoorn et al. (1996)
pEXT21SpecR, IncW ori, tac promoterDykxhoorn et al. (1996)
pKI3*N. meningitidis siaBmyc and lstFLAG in pACT3Ilg et al. (2010)
pLNTlgtA lgtB lgtEPaton et al. (2005)
pMLBADTmpR, pBBR ori, arabinose inducible promoterLefebre and Valvano (2002)
pMMZ 2lgtA lgtB lgtE were amplified from pLNT by PCR from pLNT using primers 551 + 549 containing 5′ SacI extensions. Cut product was inserted into corresponding site of pKI3*. Orientation was determined by Colony PCR.This study
pMMZ 10lgtA lgtB was amplified from pMMZ 2 using 551 + 262 and PCR product was digested with SacI XbaI and inserted into corresponding sites of pEXT21. Reverse primer used for PCR incorporated codons for a C-terminal myc tag in frame with lgtB.This study
pMMZ 11lgtB was amplified from pMMZ 2 using 261 + 262 latter incorporating codons for a myc tag in frame with the coding sequence of lgtB at the C-terminus. 5′ extensions contained restriction sites for SacI and XbaI. PCR product was cut with and inserted into corresponding sites of pEXT21.This study
pMMZ 14

futAC69T,C72T,C75T ColE1, AmpR.

pBSfutA was used as template with primers 703 + 704 with a standard Quikchange protocol for exchanging 3 nucleotides using Pfu Turbo polymerase (C69T, C72T, C75T of futA CDS)

This study
pLLM 1lgtA was amplified using primers 551 + 705, whereas the reverse primer incorporated six histidine codons allowing the insertion of the lgtA coding sequence in frame with a C-terminal His tag. The PCR product was digested with SacI and XbaI and inserted into the same sites of pACT3.This study

To test whether the modified LOS molecules were displayed on the bacterial cell surface, we assayed its surface localization. We used a quantitative analysis of strains expressing lgtAB for displaying LacNAc motives, by using fixed but not permeabilized bacteria stained with biotinylated CGL2 and subsequently probed with Streptavidin coupled to a fluorophore (e.g. Alexa 647). The number of fluorescent bacteria of total 50 000 cells was quantified by FACS with respect to unstained control bacteria (Fig. 2B). We obtained surface labelling for all E. coli strains expressing lgtAB in three independent experiments. A lower percentage of the bacterial population displayed LacNAc motives in E. coli wecA (10–30%), waaL (10–30%) or wbbL (20–60%) in comparison to truncated lipid A-core mutant (waaOB) for which 67–80% of the total population assembled LacNAc epitopes. We concluded that surface-localized LacNAc epitopes were produced by E. coli strains expressing lgtA and lgtB.

Efficient assembly of LacNAc on truncated lipid A-cores of Salmonella enterica sv Typhimurium

We next targeted S. enterica sv Typhimurium to modify its lipid A-core with LacNAc eptitopes, offering the possibility to use these strains in infection models.

When O-Ag negative Salmonella strains with a native or a truncated lipid A-core (wbaP or waaB waaI respectively) were analysed (Fig. 1), CGL2 positive signals were detected in crude LOS extracts depending on lgtAB expression. Consistent with the analyses of E. coli strains, Salmonella harbouring different plasmid combinations indicated that lgtA and lgtB were necessary and sufficient to build LacNAc polymers (Fig. 3A), which were also resistant to mild acid hydrolysis (data not shown). Having only one acceptor site (GlcI) of the truncated lipid A-core available (Fig. 1), a very efficient surface display of LacNAc subunits was observed in S. Typhimurium truncated core mutant (waaBI mutant, 87–96%), whereas in the O-Ag mutant wbaP only 13–22% of the bacterial population displayed LacNAc (Fig. 3B). Taken together, these data showed the efficient surface display of LacNAc epitopes on truncated lipid A-cores of S. Typhimurium.

Figure 3.

Polymeric LacNAc is synthesized in Salmonella waaBI and efficiently displayed on the bacterial surface.

A. Silver staining (left) and CGL2 lectin blot (right) of crude LOS extracts separated by Tris-Tricine gels containing 17% polyacrylamide of S. Typhimurium containing either plasmid encoding lgtAB or vector control.

B. Histogram of FACS sorting showing fluorescence intensity plotted against number of bacteria (% of max) in one representative of three independent experiments. Percentages indicate bacteria in the gate (arrow) of CGL2 labelled Salmonella strains encoding lgtAB (black line). Same bacteria labelled with Streptavidin-647 as control are shown as grey lines. Bacteria containing vector control labelled with CGL2 are shown as black dashed line and Streptavidin control labelling as grey dashed line. Light grey filled area represents unstained bacteria.

Engineered Lewis X antigens on Salmonella Typhimurium

To use polymeric LacNAc as a scaffold for fucosylation in order to obtain blood group determinant Lewis X (Galβ1–4(Fucα1–3)GlcNAc) (Fig. 1), we deleted a glucose 1-phosphate transferase encoded by wcaJ within the colanic acid biosynthesis pathway (Dumon et al., 2001). Colanic acid is an extracellular polysaccharide consisting of a polyanionic heteropolysaccharide repeat that includes l-fucose (Meredith et al., 2007). The wcaJ deletion completely abolished the mucoid colony phenotype of the truncated lipid A-core Salmonella and we hypothesized to have GDP-fucose available as a donor for fucosylation.

We then expressed the FutA fucosyltransferase from H. pylori, mutated within a stretch of 13 cytosines by exchanging three nucleotides C69T, C72T, C75T of futA to stabilize expression and to prevent phase variation. Salmonella waaBI wcaJ was transformed with the plasmids encoding for lgtA, lgtB and futA that allowed for assembly of LacNAc and Lewis X epitopes (‘Lx strain’). The plasmid encoding lgtAB in trans of Salmonella waaBI wcaJ lead to the strain called ‘LN’, while strains harbouring vector controls were named ‘core 3’ and ‘core 4’ (Table 1). We monitored the display efficiency of bacteria induced for 4–6 h by FACS using CGL2 (LacNAc labelling), monoclonal anti-Lewis X antibody or by RSL, a lectin detecting fucose residues to confirm the percentage of the populations displaying LacNAc and LewisX epitopes (Fig. 4A). For Lx and LN strains, CGL2 labelling was 73% (± 20.5%) and 79% (± 16%) respectively (Fig. 4B). The high percentages (0–12%) of core mutants stained with CGL2 could be assigned to background signal, as only the truncated core mutants but not the Lx and LN strains bound Streptavidin-Alexa647, without CGL2 being present (data not shown). For detection of the Lewis X epitope, 41% (± 11%), the Lx strain population was stained using anti-Lewis X monoclonal antibody, whereas LN strain or core mutants had only 1.5–2% signal (background). Fucose specific RSL lectin stained 55% (± 7%) of the Lx population and only 0.4–1.2% of the LN and core strains (Fig. 4B). Bacteria induced for 14–16 h were tested for Lewis X or LacNAc production; 85% (± 3%) of Lx and 85% (± 10%) of LN populations were stained for LacNAc. Lewis X detection increased to 75% (± 0.5%) and RSL labelled 54% (± 17%) in the Lx strain (data not shown). Using an ELISA with live cell labelling, we determined reaction velocities of aforementioned strains and confirmed the binding of CGL2 to Lx and LN strains but not to the core mutants. Anti-Lewis X antibodies only bound to the Lx strain proving the presence of Lewis X epitopes on the surface of live Salmonella (Fig. 4C).

Figure 4.

LacNAc containing LOS can be modified with fucose residues leading to Lewis X antigens on Salmonella surfaces.

A. Histogram of FACS sorting with fluorescence intensity plotted against number of bacteria (% of max) in one representative of three independent experiments. Percentages indicated below are bacteria within gate (arrow) of labelled Salmonella with CGL2 (left panel), anti-Lewis X (middle panel) and RSL (right panel).

B. Percentage of labelled bacteria of three independent experiments are shown as mean with standard deviation (SD).

C. ELISA based detection of surface labelling represented as the highest reaction velocity of ΔA405 nm after substrate addition normalized against total bacteria (OD600). Bars show the mean including SD of triplicate measurement of one representative of three independent experiments.

D. ELISA based binding of Gal1-GST (left) or GST control (right) to bacterial strains indicated, represented as the highest reaction velocity of ΔA405 nm after substrate addition. Bars show the mean ± SD of three and two independent experiments without or with addition of 1 mM LacNAc respectively.

Next, we tested binding to human galectin 1 (Gal-1) in vitro and both, Lx and LN strains, were recognized by Gal-1-GST but not by GST alone in an ELISA assay. This carbohydrate dependent binding could be out-competed by addition of free LacNAc (Fig. 4D) further demonstrating surface-localized LacNAc epitopes on Salmonella. We concluded that Lewis X antigens and LacNAc subunits were displayed at the surface of recombinant Salmonella as chimeric LOS.

Lewis X and LacNac polymers associate to the outer membrane

We then analysed crude LOS extracts separated on Tris-Tricine gels with anti-Lewis X, which confirmed the presence of a Lewis X polymer in the Lx strain. To test whether the polymer was associated to the outer or inner membrane, selective detergent solubilization using sarcosyl and sucrose sedimentation was performed. When LOS extracts or sarcosyl resistant OM preparations were probed with CGL2, a polymer was observed for the LN strain but not for Lx. In Lx strain extracts only bands in the lower MW range slightly stained with CGL2 suggesting that all forms modified with fucose decreased CGL2 recognition. Both, Lewis X and LacNAc polymers were observed in sarcosyl resistant OM preparations (Fig. 5A). Upon cell fractionation using a linear sucrose gradient, we determined the presence of OMs with a polyclonal serum against outer membrane proteins (OMP) and assayed the NADH activity for the fractions containing the inner membrane. We detected the major Lewis X polymers associated with the OM fractions (Fig. 5B). We therefore suggest that the chimeric LOS structures are flipped across the plasma membrane and are translocated to the OM.

Figure 5.

Lewis X polymer is associated to the OM.

A. Silver staining (left), CGL2 lectin blot (middle) and anti-Lewis X immunoblot (right) of crude LOS extracts (‘PK‘) and detergent resistant OM preparations (‘sarcosyl’) separated by Tris-Tricine gels containing 17% polyacrylamide of S. Typhimurium strains encoding indicated genes.

B. Sucrose gradient fractionation of LX material was analysed with anti-OMP for the presence of outer membrane (‘OM‘) and by silver staining of proteinase K digested fractions on SDS gels together with crude LOS extracts (‘PK‘) or whole cell extracts (‘WC‘), cytoplasmic fraction (‘CM‘) or total membrane fraction (‘TM‘). NADH activity relative to protein content was measured as decrease in the absorption 340 nm to indicate the inner membrane (‘IM‘).

Double staining reveals epitope distributions within a population

As we observed surface staining with either CGL2 (LacNAc) or anti-Lewis X of the Lx strain, we wanted to analyse the distribution of the epitopes at single cell level. We performed a double staining with CGL2 and RSL, simultaneously detecting LacNAc and fucose, respectively, and analysed the population by FACS and by confocal microscopy (Fig. 6). In five independent experiments, 29.75% of Lx bacteria had LacNAc and fucose signals, 39.62% were positive only for LacNAc, 13.53% stained only for fucose and 17.10% remained unlabelled. LN bacteria were 73.65% positive for LacNAc but not labelled for LacNAc plus fucose (1.98%), or for fucose (0.3%) while 24.06% remained unstained (Fig. 6B). Strains induced for 14–16 h (n = 2) increased the level of epitopes as seen for Lx strain, which had 72.25% LacNAc plus fucose signals, 20.85% positive only for LacNAc, 4.32% stained for fucose and 2.48% were unlabelled. For LacNAc expressing bacteria, 93.05% were positive for LacNAc using CGL2, while 1.66% were positive for RSL and CGL2. Only 0.01% of LN strain were stained with RSL and 4.78% were unlabelled (data not shown). Taken together, this suggested a variation in epitopes within a population after LgtA and LgtB expression. We therefore performed confocal microscopy with either fixed or live immobilized bacteria, which demonstrated the same observations as in FACS population analyses. Interestingly, the signal for fucose was dispersed around the outer surface whereas LacNAc labelled CGL2 appeared in fluorescent patches (Fig. 6D). Lx strain was distributed as doubly, singly or unlabelled bacteria similarly as quantified by FACS. The monoclonal anti-Lewis X antibody in a single staining clearly demonstrated the presence of surface-localized Lewis X epitope on 1/3 to 1/2 of the bacteria in agreement to FACS counting (Fig. 6E).

Figure 6.

Distribution of LacNAc and Lewis X containing LOS on Salmonella surfaces.

A. Histogram of FACS sorting with fluorescence intensities of FITC (RSL) plotted against Alexa Fluor 647 (CGL2) in one representative of five independent experiments after compensation. Percentages indicated are bacteria from each quadrant of labelled Salmonella.

B. Percentages of labelled bacteria of experiment shown in panel A.

C. Percentages shown as mean from five independent experiments.

D. Confocal microscopy of labelled strains with one representative of three independent experiments shown as overlay of phase contrast, infrared- (CGL2) and FITC- (RSL) channel and (E) overlay of phase contrast and anti-Lewis X signal (infrared channel).

Matrix-assisted laser desorption/ionization–time of flight mass spectrometry analysis of purified and de-O-acylated LOS confirms glycan additions corresponding to poly-LacNAc and Lewis X antigens

To obtain structural information on the engineered LOS, we performed MALDI-TOF measurements of de-O-acylated LOS purified from strains Lx, LN, core 3 and core 4 (Fig. 7). Mass (m/z) values obtained were in agreement of LacNAc units (= 365.13 a.m.u.) added on a lactosyl motif in the LN strain, which contained on its truncated core only two heptoses, the major core structure in waaBI mutants. While the form containing three phosphates had four LacNAc attached, the core with an additional PPEtn had two LacNAc (Fig. 7A). The modification in Lx LOS was consistent with one deoxy hexose corresponding to fucose on the Hep2P3 cores on the first LacNAc (m/z = 2693.78) or on the second LacNAc (m/z = 3058.52) (Fig. 7B). Core 3 and core 4 structures showed variations in additions of P or PPEtn as well as appearance of minor forms containing three heptoses (Fig. 7C and D). In Lx and LN LOS, only Hep2 forms appeared to be modified. Addition of LacNAc disaccharides and fucose residues from MS profile are in agreement with our predicted LOS structures.

Figure 7.

MALDI-TOF analysis of purified LOS reveals additions of polymeric disaccharide units (LN) with additional deoxy hexose modifications (LX). MALDI-TOF analysis of LOS purified from (A) LN, (B) Lx, (C) core 3 and (D) core 4 S. Typhimurium strains. Asterisks indicate sodium adducts. Masses (m/z) are indicated with relevant predicted core structures and masses of corresponding glycan additions are indicated with arrows.

Nuclear magnetic resonance (NMR) spectroscopy reveal LacNAc and Lewis X in chimeric LOS

The genetically modified material was also analysed by NMR spectroscopy to investigate LacNAc and Lewis X epitopes in LOS. LOS of the LN and Lx strains, in which LgtA and LgtB as well as FutA in the latter case had been active on the truncated lipid A-core, were delipidated under acidic conditions thereby leaving a single Kdo residue in the polysaccharide. The polysaccharide preparations were purified by gel permeation chromatography and the two materials are referred to LN (anticipated LacNAc epitope in the oligosaccharide) and LX (anticipated Lewis X epitope in the oligosaccharide). The one-dimensional 1H NMR spectra (Fig. 8A and B) revealed material of high complexity. However, the most conspicuous difference between the two spectra was the presence of resonances at ∼ 1.15 ppm in the LX material whereas these were absent in the spectrum of the LN material. This 1H chemical shift is typical for an α-linked fucosyl residue being part of a Lewis X epitope (Knirel et al., 1999; Moran et al., 2002). This observation indicates the presence of Lewis X epitopes in the oligosacharide as a result of the action of FutA.

Figure 8.

1D NMR spectra reveal fucosylation and 2D NMR spectra confirm Lewis X epitopes. 1H NMR spectra with selected resonances annotated of the purified polysaccharides of S. Typhimurium mutant strain (A) displaying the LacNAc epitope (LN strain) and (B) displaying the Lewis X epitope (Lx strain), in addition to the LacNAc epitope. In the latter spectrum annotations refer to resonances of the Lewis X epitope. Residual peaks from sodium acetate and methanol are present at 1.90 and 3.35 ppm respectively. Overlay of the multiplicity-edited 1H,13C-HSQC spectra of the polysaccharides of Lx strain (red) and the polysaccharide of LN strain (blue): (C) anomeric region and (D) region of H2/C2 resonances from N-acetyl-d-glucosamine residues. Selected spectral regions: (E) the 1H,1H-TOCSY spectrum (mixing time 100 ms) showing correlations from H1 up to H4 in the α-l-Fucp residue and (F) the 1H,1H-NOESY spectrum (mixing time 100 ms) showing intra-residue correlations from H5 to H4 and H3 as well as the inter-residue NOE to H2 in the d-galactosyl residue of the Lewis X epitope in the polysaccharide of Lx strain.

The average molecular masses of the LN and LX materials were estimated using NMR translational diffusion experiments at 25°C in D2O, giving Dt = 1.55 × 10−10 m2 s−1 for LX and Dt = 1.45 × 10−10 m2 s−1 for LN with a standard deviation of 0.01 × 10−10 m2 s−1 in both cases. The molecular masses were calculated using a relationship developed for polysaccharides (Viel et al., 2003) which resulted in a Mw ∼ 2.9 kDa for LX and a Mw ∼ 3.3 kDa for LN. In order to obtain the average chain lengths of the repeats the anticipated molecular mass of the truncated core with an additional β-(1→4)-linked d-galactosyl residue (Fig. 1) was estimated. In the core of the S. Typhimurium mutant strain phosphomonoester substituents and ethanolamine substituents linked via diphosphodiesters are anticipated (Fig. 1). Their presence were supported by signals in the 31P NMR spectrum at 4.5, −10. 8 and −11.2 ppm (Masoud et al., 1991) in the LX material as well as by signals in the 1H spectrum at 3.24 LN (3.25 LX) and 4.18 LN (4.19 LX) ppm and in the 13C spectrum at 41.0 LN (40.9 LX) and 63.8 LN (63.7 LX) ppm for ethanolamine as part of a phosphoester (Masoud et al., 1994; Stewart et al., 1998). In the calculation of the Mw of the modified core region we therefore used as substituents one phosphomonoester and one ethanolamine residue linked via a diphosphodiester and estimated the Mw to be ∼ 1.4 kDa. Based on this result the oligosaccharides were found to be relatively short with only a few (∼ 5 in LN and ∼ 4 in LX) repeating units. These results are in contrast to SDS-PAGE analysis (Fig. 5A), which showed longer polymers. The MALDI-MS data (Fig. 7) showed the presence of a significant amount of truncated core, which suggest that mainly these were detected but may also indicate a substantial amount of non-substituted core in the preparations.

We continued with a detailed analysis of the NMR spectra employing the CASPER program (Jansson et al., 1989; Lundborg and Widmalm, 2011), which is able to predict 1H and 13C NMR chemical shifts of oligo- and polysaccharides. The 1H chemical shifts of the repeating unit →3)-β-d-Galp-(1→4)-β-d-GlcpNAc-(1→ in the polymer were predicted and the anomeric protons were calculated to resonate at 4.48 and 4.77 ppm, respectively, which were consistent with experimental observations (Fig. 8) and the proposed structure. Using this information the presence of polymeric LacNAc structures in the LX material was confirmed.

The α-(1→3)-linked d-glucosyl residue (Fig. 1) substituting the inner core was predicted to have its anomeric proton at 5.24 ppm in the truncated core and at 5.27 ppm when substituted by a β-(1→4)-linked d-galactosyl group. Analysis of the 1H spectrum (Fig. 8A) in the spectral region 5.22–5.30 ppm in conjunction with 1H,1H-TOCSY and 1H,13C-HSQC spectra indicated that the LN material consisted of both truncated cores and substituted glucosyl residues, indicating polymeric material in the latter case.

The LX material was analysed by NMR spectroscopy in a corresponding way to that of the LN material. The fucose H1, H5 and H6 resonances in LX were predicted by CASPER to resonate at 5.11, 4.77 and 1.19 ppm respectively (Fig. 8B) indicating that the Lewis X epitope is indeed present. The LX material was analysed using 1H,1H-TOCSY, 1H,13C-HSQC and 1H,1H-NOESY experiments. From the 1H,13C-HSQC spectrum it was evident that the LX material was heterogeneous having one-bond heteronuclear cross-peaks from the anticipated Lewis X epitope but also from the LacNAc epitopes (Fig. 8C and D). Analysis of 1H,1H-TOCSY (Fig. 8E) and 1H,1H-NOESY (Fig. 8F) spectra together with the 1H,13C-HSQC spectrum facilitated the 1H and 13C chemical shift assignments of the Lewis X trisaccharide (Table 2) as part of the oligosaccharide of the LX material. The Lewis X structure in the oligosaccharide is further supported by the excellent agreement between experimentally determined 1H and 13C chemical shifts and those predicted by CASPER showing average absolute deviations of only 0.02 ppm per signal for 1H resonances and 0.20 ppm per signal for 13C resonances. The trisaccharide structure was confirmed by a key 1H,1H-NOE correlation between H5 in α-l-Fucp and H2 in β-d-Galp (Fig. 8F). The relative NOEs between H5 and H4 in Fuc (1.0), H5 and H3 in Fuc (0.46) and H5 in Fuc and H2 in Gal (0.42) were similar to those observed in a detailed conformational analysis of the Lewis X containing pentasaccharide LNF-3 (1.0, 0.68 and 0.60 respectively) in which the Lewis X determinant is non-substituted and present at the non-reducing end of the oligosaccharide (Miller et al., 1992).

Table 2. 1H and 13C NMR chemical shifts (ppm) of the Lewis X epitope residues of the polysaccharide of Salmonella enterica sv Typhimurium mutant strain displaying Lewis X epitope
Sugar residue1H/13C
123456 NAc
  1. a Tentative assignment.
  2. Chemical shift differences as compared with corresponding monosaccharides are given in parentheses.
(0.01)(0.30)(0.32)(0.48)(0.12)  (−0.04)
103.3256.7875.5973.76a75.88a60.68a 23.03a
(7.47)(−1.08)(0.78)(2.70)(−0.94)(−1.17) (−0.07)

The oligosaccharide chain length analysis on the LX material was then performed in the same way as for LN. The LX material was found to be a mixture between truncated core and polysaccharides with longer repeats substituted on the extended core. By integration of the resonances in the 1H NMR spectrum at ∼ 1.15 ppm (H6 in Fuc) and ∼ 2.0 ppm (methyl group of the N-acetyl group in GlcNAc) it was concluded that Lewis X epitopes were present in about half of the repeating LacNAc units in the LX material. Thus, the NMR results showed the presence of Lewis X epitopes in the prepared OS material, as a consequence of the action of the FutA glycosyl transferase.

Glycoengineered LOS induce pro-inflammatory cytokines in murine dendritic cells

We then tested the effects of the glycoengineered LOS for its properties to stimulate cytokines and nitric oxide in murine dendritic cells. We stimulated murine bone marrow derived cells (BMDCs) with formalin-inactivated bacteria or with isolated LOS for 24 h and found that Lx, LN and core strains induced release of similar amounts of TNF-α and nitric oxide (Fig. 9A). The response was dependent on TLR4 because TLR4-deficient BMDCs produced almost no TNF-α or nitric oxide when stimulated with LOS and only little proinflammatory signal when incubated with formalin-inactivated bacteria (Fig. 9B). Again, no difference between the different strains was observed. We thus concluded that the modified glyco-epitopes had no impact on the lipid recognition via TLR4.

Figure 9.

TNF-α and nitric oxide (NO) release from murine dendritic cells stimulated with glycoengineered S. Typhimurium mutant strains. BMDCs derived from (A) C57BL/6 or (B) tlr4−/− mice were either mock infected or stimulated with formalin-inactivated bacteria at a bacteria to BMDC ratio of 50:1, 20:1 and 10:1 or treated with 10 or 100 ng ml−1 isolated LOS of indicated strains for 24 h. Cell-free culture supernatants were collected. TNF-α and nitric oxide (nitrite) levels were measured in the cell-free supernantant by ELISA and Griess reagent respectively. Shown is the mean from three independent experiments which were performed in duplicates. Error bars represent the SD of the mean.


In this study we characterized the display efficiency of engineered glycan epitopes of E. coli and S. Typhimurium. First, we demonstrated that expressing two glycosyltransferases LgtA and LgtB from Neisseria sp. involved in assembling LNnT structures lead to LacNAc repeats (Figs 2 and 3). LacNAc appeared polymerized in all E. coli and Salmonella strains used in this study. Interestingly, polymers are usually observed for LPS O-Ag in contrast to Neisseria sp. short-chain LOS. It is likely that the glycosyltransferases heterologously expressed act in a processive way due to expression levels or different acceptor structures. With several possible attachment sites in native lipid A-cores, GlcNAc residues may thus be modified with Gal by LgtB leading to a lactosaminosyl motif, which could subsequently be polymerized by LgtA and LgtB generating LacNAc repeats. We observed that galactosyltransferase LgtB, was transferring Gal not only to a GlcNAc as expected but also to a Glc, observed as efficient assembly of LacNAc polymers in truncated lipid A-core strains with a Glc at the terminus (Fig. 2).

If glycan structures mimicking mammalian epitopes can be transferred to proteins using bacterial protein glycosylation systems, new bioactive glyconconjugates could be produced (Wacker et al., 2002; Szymanski and Wren, 2005; Schwarz et al., 2010; Hug et al., 2011; Valderrama-Rincon et al., 2012). Glycans transferred by bacterial oligosaccharyltransferases need to be assembled on the UndPP carrier; therefore, we tested the possibility of LgtAB acting on UndPP-linked GlcNAc initiated by WecA. The LacNAc repeat would then be ligated to the core region of LPS by WaaL. Mild acid hydrolysis treatments and phenotype analysis of waaL and wecA mutants, showed that these enzymes were not essential in synthesis of LacNAc structures. Hence, we suggest that LacNAc units are assembled on the lipid A-core, independent on wecA or waaL. Using surface staining techniques we observed the highest efficiency of surface exposed epitopes in truncated lipid A-core mutants. However, twofold more wbbL bacteria displayed the epitope on the surface as compared with waaL and wecA mutants, hinting at a decrease due to the lack of the UndPP dependent pathway, despite of no visible difference in crude LOS and acid hydrolysed extracts. In a recent study, only a combined approach resulted in a Lewis X glycoconjugate by using in vivo transfer of a LacNAc containing tetrasaccharide on the lipid carrier UndPP to a protein acceptor followed by an in vitro fucosylation step (Hug et al., 2011).

Analyses of LOS extracts by silver staining, lectin blots, MALDI-TOF and NMR provide evidence that poly LacNAc units are synthesized but these data do not prove that this structure is flipped to the cell surface. In Neisseria, LptD has been identified to be responsible for the translocation of LPS through the external membrane (Bos et al., 2004; Ruiz et al., 2009). Our data showed that engineered LacNAc and Lewis X structures were located in the OM and that these epitopes were detectable on the surface of E. coli and Salmonella, suggesting that parental transport systems are sufficient to translocate and surface expose chimeric LOS. However, these surface staining techniques do not allow differentiating between one LacNAc subunit and a LacNAc polymer even though polymeric LacNAc is predicted to be a better ligand of CGL2 as suggested from glycan arrays and other studies (Sauerzapfe et al., 2009). Applying membrane fractionation we observed the Lewis X polymer associated to the OM, which suggests that endogenous MsbA is able to flip longer core structures and that these molecules are transported to the OM.

The length of the host poly-LacNAc chain is an important factor in immune responses, such as inhibition of NK cell-mediated cytotoxicity by its effects on the cell-binding process (Gilbert et al., 1988) and influences basal levels of lymphocyte and macrophage activation (Togayachi et al., 2007). Chemo-enzymatically synthesized poly-LacNAc mixtures covalently bound to surfaces are efficient signals for targeting galectins. These bio-functional materials can be used for galectin-mediated immobilization of ECM glycoproteins and cell adhesion (Sauerzapfe et al., 2009). Poly-LacNAc can be modified by a combination of sialyl and fucosyl residues resulting in Lewis and sialyl-Lewis X antigens. Lewis X trisaccharide can be synthesized in metabolically engineered E. coli to produce soluble oligosaccharides or modified LOS surfaces (Dumon et al., 2004; 2006; Yavuz et al., 2011). One major limitation of this technique is the appearance of unwanted glycoforms like fucose linked to Glc. Taking advantage of our engineered poly-LacNAc LOS, it was successfully used as a scaffold for fucosylation to generate Lewis X antigens. In this study, we use bacteria to expose mammalian polymeric LacNAc and LewisX as chimeric LOS, which provide novel tools to address functional consequences of molecular mimicry in immune response during microbial–host interactions. Stimulation of murine macrophages or dendritic with inactivated bacteria or isolated LOS resulted in pro-inflammatory cytokines release, which was independent of the glycan modification (Fig. 9 and data not shown). The strong TLR4-dependent response may, however, ‘mask’ any TLR4-independent component. Some pathogenic and commensal bacteria have evolved strategies to avoid TLR4-dependent signalling by lipid A modifications (Miller et al., 2005). However, our data suggest that molecular mimicry with host-like glycans on LOS structures is not sufficient to prevent innate immune activation. Strategies to detoxify lipid A (Ingram et al., 2010) could facilitate studies on the effect of engineered bacterial surface glycans on the immune response. Furthermore, cells deficient in CLRs or galectins may allow addressing functional consequences of specific glycan–lectin recognition between host–pathogen interactions.

NMR analyses proved the existence of polymeric Lewis X LOS but also highlighted heterogeneity and we speculate that not every LacNAc unit is fucosylated within a polymer. This is not unprecedented as heptad repeats in FutA are proposed to have a ruler function to determine the position of fucose additions within the poly LacNAc O-Ag in H. pylori, where one heptad repeat in the enzyme corresponds to one LacNAc unit (Nilsson et al., 2006). This model implies that two heptads present in the FutA used in the current study would allow fucosylation of two to four LacNAc units given FutA dimerized in the heptad repeat. This would be in agreement to our finding of three to four Lewis X repeats. The apparently rather short polysaccharide chains in LN and LX materials indicated by the translational diffusion measurements is, however, suggested to be a mixture of low molecular mass core oligosaccharides and higher molecular mass polysaccharides. This interpretation is consistent with the MALDI-MS spectrum of LacNAc epitope-containing LOS (Fig. 7) and with the fact that SDS-PAGE analysis shows the presence of polymeric material as a ladder type pattern (Fig. 5A).

MALDI-TOF analysis proved variations not only in the P or PPEtn but also in heptose substitutions, which is in agreement to previous observations (Yethon et al., 1998; 2000; Ilg et al., 2010). In Lx and LN LOS, only Hep2 forms appeared to be modified with poly-LacNAc. Moreover, the additional PPEtn substitution seems to be unfavourable for fucose modifications suggesting a specific requirement of FutA regarding the core structure. It was, however, shown that periplasmic Kdo hydrolase activity is necessary for subsequent lipid A modifications in H. pylori (Stead et al., 2010). Interestingly, while Lewis Y expression was unaffected in the Kdo hydrolase mutants, Lewis X expression was completely absent, an effect that cannot be simply explained by core recognition of FutA, as it is also involved in Lewis Y modification.

With the investigation on distribution patterns of engineered surface glycans, glycoengineering of bacterial surfaces serves as a tool for functional experiments. In vitro Gal-1 binding gives a first indication that these strains can be used to identify receptor–ligand interactions in cell culture or in vivo models to address consequences on downstream signalling. Moreover, these glycolipids of LOS type can well be used for producing bioactive materials in the future to facilitate cell adhesion thereby imitating natural micro-environments.

Experimental procedures

Bacterial strains, growth conditions, and selective agents

Bacterial strains and plasmids are listed in Table 1. Oligonucleotides are found in Table S1. Bacteria were routinely grown in Luria–Bertani (LB) medium and LB agar plates contained 1.5% (w/v) agar. For selection, antibiotics were used at the following final concentrations: Ampicillin (Amp) 100 μg ml−1, chloramphenicol (Cam) 25 μg ml−1, kanamycin (Kan) 50 μg ml−1, spectinomycin (Spec) 80 μg ml−1, trimethoprim (Tmp) 100 μg ml−1.

Bacteria grown o/n in LB containing appropriate antibiotics at 37°C were diluted in LB to OD600 of 0.1, and IPTG was added at 0.1 mM final concentration after OD600 reached 0.4 to 0.6, to induce for 4–6 h at 37°C. In stated experiments, induction was performed for 14–16 h.

Deletion of wcaJ in Salmonella enterica sv. Typhimurium

Strain SMM6 was generated using Lambda Red recombination (Datsenko and Wanner, 2000). Briefly, a PCR generated cat cassette with 50 bp flanking homology sites to wcaJ was transformed in SKI22 primed for Lambda Red recombination. Integration of the wcaJ deletion cassette was confirmed by primers 664 + 285 and selected clones were transformed with pCP20 to out-recombine the cat cassette. The resulting Cam-sensitive strain was mutated between nt position 2185571 and 2186957 according to GenBank entry AE006468.1 S. Typhimurium strain LT2 with a 84 bp scar site. Deletion was confirmed by PCR using primers 665 + 664 and PCR product was sequenced. Slimy viscous colony morphology of SKI22 was not observed in wcaJ deletion strain, consistent with absence of colanic acid.

Tris-Tricine SDS-PAGE and immunoblot analysis of glycoconjugates

The equivalent of 5 × 108 cfu (colony-forming units) of bacterial cultures was pelleted at 16 000 g for 2 min. Pellets were lysed in 50 μl Lämmli buffer [0.065 M Tris-HCl pH 6.8, 2% SDS (w/v), 5% β-mercaptoethanol (v/v), 10% glycerine (v/v), 0.05% bromophenol blue (w/v)] for 15 min at 99°C. Proteinase K (Roche) was added to a final concentration of 0.4 mg ml−1 and incubated for 1 h at 60°C and equal amounts were separated on 17% Tris-Tricine gels (Schagger, 2006). Glycans were visualized either by silver staining (Tsai and Frasch, 1982) or blotted onto PVDF (polyvinylidene difluoride) membranes. Lectin blots were blocked in 1% BSA PBS-Tween 0.2% o/n at 4°C, probed with biotinylated CGL2 at 1 μg ml−1 final concentration in 1% BSA or biotinylated RSL at 0.5 μg ml−1. Biotin was detected by 0.5 μg ml−1 streptavidin-HRP (Vectorlabs). Lewis X was detected using murine monoclonal antibody (IgM isotype, clone P12, Abcam), at a concentration of 1:500 and IgM was probed with anti-mouse IgM-HRP (Santa Cruz) at 0.2 μg ml−1. ECL reagent (Amersham) for visualization was used as recommended by the manufacturer.

Outer membrane preparation by selective detergent solubilization

Bacteria resuspended in PBS containing DNase and RNase (10 μg ml−1), were sonicated on ice. Unbroken cells were removed at 3000 g for 15 min, and total membranes were collected at 20 000 g for 30 min at 4°C. The membranes were resuspended in PBS and sarcosyl (N-Lauroylsarcosine sodium salt, Sigma) was added to a final concentration of 1% (v/v). After incubation on ice for 1 h, membranes were collected at 20 000 g for 30 min and resuspended in electrophoresis sample buffer and analysed by Tris-Tricine gels.

Membrane fractionation using sucrose gradient sedimentation

Bacteria corresponding to 800 OD were resuspended in PBS containing 10 μg ml−1 DNase and RNAse and were sonicated on ice. Unbroken cells were removed at 5000 g for 10 min and sterile filtered supernatant was spun using a 45Ti rotor at 100 000 g for 1 h at 4°C to collect total membranes. The membrane pellet was carefully resuspended in 25% sucrose, 5 mM EDTA and 30 mM Tris pH 7.5 and loaded on linear sucrose gradients consisting of 1.8 ml each of 55%, 50%, 45%, 40%, 35% and 30% sucrose with 5 mM EDTA, which were then spun at 256 000 g at 4°C in a SW 40 Ti rotor for 19 h. The ultraclear centrifugations tubes were then punctured at the bottom and the gradient was collected by gravity flow in 500 μl steps. Fractions were analysed for NADH activity by monitoring Vmax as the decrease in 340 nm over 5 min in 0.12 mM NADH, 0.2 mM DTT and 40 mM TrisCl pH 7.5 reaction buffer in triplicate measurements. Protein content was measured using A280. Fractions, 10 μl loaded each, were separated by SDS-PAGE and probed with rabbit polyclonal serum raised against the major OMP from E. coli cross-reactive with Salmonella OMP, a kind gift from Jörg Vogel's lab, for the presence of OM. The sucrose gradient fractions were proteinase K digested, equally loaded on 12% SDS-PAGE and stained with silver for LOS detection.

Quantification of surface glycan epitopes by FACS

Bacteria (5 × 107–2 × 108 cfu) were harvested by centrifugation (5 min, 13 000 g at 4°C) and washed in 500 μl PBS. Bacteria were pelleted as described before and gently resuspended in 3% PFA in PBS and incubated for 5–10 min at room temperature (RT). Fixed bacteria were washed in PBS followed by incubation with biotinylated CGL2 at a 3 μg ml−1 final concentration for 1 h at RT. Lewis X antigen was detected by monoclonal anti-Lewis X antibody, at 1:100. Fucosylated structures were stained using RSL lectin coupled to FITC used at 4 μg ml−1, gift of Anne Imberty (Kostlanova et al., 2005) in the dark for 1 h at RT. Pelleted bacteria were washed with 500 μl PBS and biotinylated CGL2 was probed with Streptavidin-Alexa 647 (Vectorlabs) at 5 μg ml−1 and incubated for 1 h at RT in the dark. Anti-Lewis X was detected with anti-mouse IgM Alexa647 (Invitrogen) at 10 μg ml−1 for 1 h at RT (dark). Bacteria were washed in PBS and stored dark in 500 μl PBS at 4°C prior to FACS analysis. FACS acquisition was performed using FACS LSRII (BD Biosciences) using FACS Diva 5.0.3 and compensation controls were performed using single and unstained samples. Data were analysed with FlowJo V7.2.2 software.

Quantification of surface exposed carbohydrates by whole cell ELISA

Bacterial strains were grown, induced and harvested as described above and 2 × 108 cfu ml−1 were used per staining. Primary staining was essentially carried out as described for FACS but without fixation with PFA and the following modifications. The secondary antibody probing anti-Lewis X was anti-mouse IgM coupled to HRP (Santa Cruz) at 1:1000. Biotinylated lectins were detected by Streptavidin coupled to HRP (Vectorlabs) at 1 μg ml−1. Bacteria were washed twice in PBS and the final pellet was resuspended in 500 μl 70 mM citrate phosphate buffer, pH 4.2. Bacterial suspension was distributed in flat bottom 96 well plates (Nunc) for triplicate measurements. OD600 was measured with SpectraMax Plus 384 (Bucher Biotech) before adding 50 μl 4× ABTS buffer (4 mM ABTS, 70 mM citrate phosphate buffer pH 4.2 with 0.04% H2O2). Reaction velocities were monitored using ‘Kinetic ELISA with HRP and ABTS’ of the program SoftMax Pro 5.3 at 405 nm for 30 min.

Galectin ELISA

Bacteria were harvested after 16 h of induction, washed in PBS and fixed with final concentration of 7.4% formaldehyde and 1.5 × 108 cfu were allowed to adhere to 96 well plates (Nunc, MaxiSorp) for 30 min at RT. After three washes in PBS-Tween 0.05%, wells were blocked o/n at 4°C using 1% BSA in PBS. Wells were washed 3× with PBS-Tween 0.05% and Gal-1-GST or GST were added for 1 h at RT at 1 μg ml−1. For competitive blocking, galectins were pre-incubated for 15–30 min at 37°C in 1 mM LacNAc (Sigma) in PBS. Three washes were performed with PBS-Tween 0.05% before incubation with 0.5 μg ml−1 goat anti-GST (Rockland) for 1 h at RT. Wells were washed 3× as before and 0.2 μg ml−1 donkey anti-goat IgG-HRP (SantaCruz) in PBS was added for 45 min at RT. Wells were washed 4× and filled with 150 μl citrate phosphate substrate buffer. Reaction was started by adding 50 μl of 4× ABTS substrate and reaction kinetics were monitored using ‘Kinetic ELISA with HRP and ABTS’ of the program SoftMax Pro 5.3 at 405 nm for 10 min.

Localization of surface glycan epitopes by confocal microscopy

Strains were stained essentially as described for FACS quantification using either PFA fixed or live bacteria. After labelling, bacteria were immobilized onto poly-d-lysine (BD) coated glass slides and mounted using Vectashield hard set (VectorLabs). Images were recorded with a Zeiss Axiovert 200 microscope and an Ultraview confocal head (Perkin-Elmer) and analysed using Volocity software (Version 5.0.3., Improvision).

LOS isolation

Bacteria corresponding to OD600 units of 400–700 for small scale and 3500–4300 for NMR analysis were harvested by centrifugation at 2700 g for 10 min at 4°C. LOS was extracted based on a phenol–chloroform–petroleum ether method as described (Galanos et al., 1969; Ilg et al., 2010). Briefly, the pellet was washed with 50 ml 1× PBS and pelleted again by centrifugation. The cells were homogenized in PCP (phenol–chloroform–petroleum ether 1:2.5:4), extracted, precipitated and washed. The final LOS pellet was resuspended in ddH2O and lyophilized.

MALDI-TOF analysis

LOS was de-O-acylated prior to MALDI-TOF analysis by mild hydrazine treatment (Holst, 2000). Briefly, LOS was dissolved in 20 mg ml−1 hydrazine hydrate and incubated at 37°C for 2 h. LOS was precipitated after the cleavage of the O-linked acyl chains by drop-wise addition of 15 vols of ice-cold acetone and centrifugation at 16 000 g at RT for 15 min. The pellet was washed with acetone, spun for 15 min at RT at 10 000 g. Washing was repeated, and the pellet was air-dried. For de-O-acylated LOS profiling, samples were dissolved in water at a final concentration of 1 mg ml−1 and mixed 1:1 with the 6-aza-2-thio-thymine (ATT) matrix (20 mg ml−1 in 70% MeOH with 10 mM ammonium citrate). Data acquisition was performed on 4800 Proteomics Analyzer, (Applied Biosystems, Framingham, MA) using linear negative ion mode, with a total of 20 sub-spectra of 125 laser shots.

Polysaccharide purification for NMR

For NMR analysis, isolated LOS was delipidated by addition of 0.1 M sodium acetate pH 4.2 for 4 h at 99°C as described (Knirel et al., 1997). The precipitate was removed by centrifugation at 4000 g at 4°C for 90 min and supernatant was lyophilized.

The polysaccharide materials obtained after delipidation under acidic conditions were purified by size exclusion chromatography on a Superdex Peptide 10/300 GL (Tricorn) column (GE Healthcare) eluted with 1% 1-butanol in water at 1 ml min−1 with an ÄKTA purifier system (GE Healthcare, Sweden). RI and UV detection at 214 nm were used to monitor elution. The purified material, denoted LN and LX, were lyophilized and subsequently used in NMR analysis.

NMR experiments

1H and 13C NMR chemical shift assignments of the polysaccharides were performed in D2O (< 1 mg in 0.18 ml, 3 mm NMR tube) at pD 7–8 and 39°C on a Bruker Avance III 700 MHz spectrometer equipped with dual receivers and a 5 mm TCI Z-Gradient CryoProbe. 31P NMR chemical shifts were obtained at 39°C on a Bruker Avance II 500 MHz spectrometer equipped with a 5 mm BBO Z-Gradient probe. Chemical shifts are reported in ppm using external sodium 3-trimethylsilyl-(2,2,3,3-2H4)-propanoate (TSP, δH 0.00), 1,4-dioxane in D2O (δC 67.40) or 2% H3PO4 in D2O (δP 0.00) as references.

1H NMR spectra were recorded with 29 410 data points over a spectral width of 8.0 ppm, 600 scans and a repetition time of 12.6 s. Zero-filling to 128 k data points and an exponential weighting function using a line-broadening factor of 0.5 Hz were applied prior to Fourier transformation. The 31P NMR spectrum was recorded with 65 534 data points over a spectral width of 403 ppm and 18 688 scans. Zero-filling to 512 k points and an exponential weighting function using a line-broadening factor of 5 Hz were applied prior to Fourier transformation.

1H chemical shift assignments were performed using 1H,1H-TOCSY experiments (Bax and Davis, 1985) recorded over 6.0 ppm with 2048 × 256 data points and eight scans per t1-increment, using the States-TPPI method. An MLEV-17 spin-lock of 10 kHz and four different mixing times (10, 40, 70 and 100 ms) were used. Zero-filling was performed to 4096 × 1024 points. Prior to Fourier transformation 90° shifted squared sine-bell functions were applied in both dimensions.

13C chemical shifts were assigned using multiplicity-edited 1H,13C-HSQC experiments (Schleucher et al., 1994). The experiments were recorded with 1024 × 256 data points and 64 scans per t1-increment over a spectral region of 6.0 ppm for 1H and 60 ppm for 13C, employing the echo/antiecho method. Adiabatic pulses (Tannús and Garwood, 1997; Kupče, 2002) were used for 13C inversion (smoothed CHIRP, 20%, 80 kHz, 500 μs, Q = 5.0) and refocusing (composite smoothed CHIRP, 80 kHz, 2.0 ms). Prior to Fourier transformation forward linear prediction to 512 points in the F1-dimension and zero-filling to 4096 × 2048 points were performed; 90° shifted squared sine-bell functions were applied in both dimensions.

The 1H,1H-NOESY experiment was recorded with a spectral width of 6.0 ppm with 2048 × 256 data points and 32 scans per t1-increment (Kumar et al., 1980). A mixing time of 100 ms was used. Prior to Fourier transformation zero-filling was performed to 8192 × 1024 points and a 90° shifted squared sine-bell function was applied in both dimensions.

Translational diffusion measurements were performed at 298.1 K on a Bruker Avance III 600 MHz spectrometer equipped with an 5 mm inverse Z-gradient TXI probe (1H/13C/15N), using a pulsed field gradient spin-echo experiment (ledbpgp2s pulse sequence). The experiments were repeated five times and recorded with 16 k data points and 16 scans for each gradient step. The PFG strength (100% = 55.7 G cm−1) was increased linearly between 2% and 95% over 32 steps (Damberg et al., 2001). A fixed diffusion time (Δ) of 0.12 s and diffusion encoded gradient pulses (δ/2) of 2 ms were used. The pulsed field gradients were calibrated using a doped water sample (1% H2O in D2O + 1 mg ml−1 GdCl3) and a literature value of Dt = 1.90 × 10−9 m2 s−1 for the HDO diffusion coefficient in D2O at 25°C (Mills, 1973). The molecular mass was calculated from the following relationship (Viel et al., 2003): Dt = 8.2 × 10−9 MW−0.49. The measured diffusion coefficients were corrected with a factor of 1.06 for using 300.0 K in the calculations.

The chemical shifts were compared with those of the corresponding monosaccharides (Jansson et al., 1989).

Stimulation of bone marrow derived dendritic cells and detection of cytokines

C57Bl/6 mice were purchased from Janvier SAS (France) and tlr4−/− mice were bred at Harlan Laboratories (Füllinsdorf, Switzerland). Animal experiments were performed in accordance with institutional guidelines and were reviewed by the cantonal veterinary office (184/2009). BMDCs were generated as described elsewhere (Inaba et al., 1992) and 105 cells per well were cultured in 96 well round-bottomed plates in 200 μl culture medium (RPMI 1640 supplemented with glutamine, penicillin, streptomycin, 2-mercaptoethanol, all from Invitrogen) containing 10% (vol/vol) heat-inactivated FCS and GM-CSF. BMDCs were treated with 100 or 10 ng ml−1 of isolated LOS or formalin-inactivated bacteria at different ratios of bacteria to BMDCs for 24 h. To inactivate bacteria, strains were harvested after induction of glyco-epitopes and treated with 0.2% formalin at 4°C over night and washed 3× in PBS. TNF-α was measured in cell-free supernatants by sandwich ELISA using clones 6B8 and MP6-XT22. Nitric oxide production was estimated as the amount of nitrite released in the culture medium, by use of modified Griess reagent (Sigma).


This work was supported by an ETH postdoctoral fellowship to M. M., by grants from Swiss National Science Foundation (31003A_127098 to M. A. and PP00P3_123342 to S. L.-L.), the Swedish Research Council and The Knut and Alice Wallenberg Foundation. We thank the European Commission's Seventh Framework Programme (FP7/2007-2013 under grant agreement No. 215536) for funding the EuroGlycoArrays ITN, the Functional Genomics Center Zurich for instrument support, Silvio Hemmi and Viet Hung Trinh for providing purified Gal-1 and Elif Yavuz for biotinylated lectins.