Recognition of sialylated meningococcal lipopolysaccharide by siglecs expressed on myeloid cells leads to enhanced bacterial uptake

Authors


E-mail p.r.crocker@dundee.ac.uk and m.virji@bristol.ac.uk; Tel. (+44) 1382 345781 and (+44) 117 928 7527; Fax (+44) 1382 345873 and (+44) 117 928 7527.

Summary

Sialic acid-binding immunoglobulin-like lectins (siglecs) are expressed predominantly in the haemopoietic and immune systems and exhibit specificities for both the linkage and the nature of sialic acids in N-glycans, O-glycans and glycolipids. Several siglecs, including sialoadhesin (Sn, siglec-1) and siglec-5, bind to NeuAcα2,3Gal, a terminal capping structure that can also be displayed on the lipopolysaccharide (LPS) of Neisseria meningitidis (Nm). In the present study, we examined the potential of siglecs expressed on cells of the immune system to function as receptors for sialylated Nm. We used sialylated and non-sialylated LPS derivatives of two serogroups (A and B) of Nm in this study. Using recombinant chimeric soluble receptors, siglec-transfected cell lines and macrophages from wild-type and Sn-deficient mice, we observed that sialylated but not non-sialylated variants of either genetic background were specifically recognized by Sn and siglec-5, whereas other siglecs examined were ineffective. In addition, macrophages expressing Sn, as well as transfectants expressing Sn or siglec-5, bound and phagocytosed sialylated bacteria in a siglec- and sialic acid-dependent manner. This study demonstrates that Nm LPS sialylation can lead to increased bacterial susceptibility to phagocytic uptake, a phenomenon in direct contrast to previously reported protective effects of LPS sialylation.

Introduction

Neisseria meningitidis (Nm) is a human specific commensal and pathogen that resides in the nasopharynx of its host, and further sequestration occurs from this primary and specific site of colonization. The complete complement of factors that is necessary for colonization or that lead to serious pathogenic conditions remains to be described. Of the recognized virulence factors of meningococci, the surface polysaccharides, capsule and lipopolysaccharides (LPS or lipooligosaccharide, LOS) seem to play a critical role in bacterial survival in vivo. Capsule and sialylated LPS are expressed in disseminated isolates and are believed to protect the organism against antibody/complement and phagocytosis. They are also expressed by a number of carrier isolates and may have functions that allow the organism to exist in the nasopharynx (allow avoidance of mucosal immunity) or are physically protective against extra-host environment (antidesiccation property of capsular polysaccharide). However, acapsulate meningococci are frequently isolated from the nasopharynx (Cartwright, 1995). In vitro studies show that adhesion and particularly invasion of epithelial cells are enhanced (aided by Opa and Opc outer membrane adhesins) in the absence of capsule (Virji et al., 1992; 1993). In addition, meningococci from the nasopharynx often express the LPS type that resists sialylation due to the absence of lacto-N-neotetraose (LNnT) structure, an acceptor for sialic acid (Cartwright, 1995; Jennings et al., 1995; McLeod Griffiss et al., 2000). Sialylation of LPS has functional consequences similar to capsulation, and imparts resistance to immune mechanisms of the host (Hammerschmidt et al., 1994; Kahler and Stephens, 1998; Vogel and Frosch, 1999) and, in doing so, masks the functions of many outer membrane adhesins (Virji et al., 1993; 1995). To date, the interplay between surface polysaccharides and various adhesins and invasins has been investigated with an emphasis on the inhibitory function of sialic acids on cellular interactions (Virji et al., 1995; de Vries et al., 1996).

Meningococcal LPS is an amphipathic glycolipid of Mr≈ 4.8 kDa. Twelve immunotypes (ITs) of Nm LPS have been described based on their antigenic differences (Kogan et al., 1997; Kahler and Stephens, 1998; Tsai et al., 2002). Carbohydrate portions of Nm LPS mimic human glycosphingolipid blood group antigens and may be important in bacterial host interactions (Mandrell and Apicella, 1993). Two core heptose molecules bear the side-chains (α and β or γ) of the terminal carbohydrates (Fig. 1). The variable α chain, of, for example Nm strain MC58 LPS, bears the terminal sialo or asialo lacto-N-neotetraose {[R-Galβ(1,4)GlcNAcβ(1,3)Galβ(1,4)Glc] where R = α(2,3)-linked sialic acid (Neu5Ac)}. This immunotype L3 may be converted to L8 that lacks the terminal Lac(NAc) (Fig. 1) by phase variation of the enzyme lgtA (Jennings et al., 1995). The resulting disaccharide of L8 IT cannot be sialylated. Thus in Nm serogroups B and C strains, that intrinsically sialylate their LPS, sia+/sia− phase variation may occur via lgtA. In serogroup A strains, intrinsic sialylation of the LPS does not occur as they do not synthesize CMP-N-acetyl neuraminic acid (CMP-NeuAc), the substrate for LPS sialyltransferase. However, provision of the substrate in vivo from host secretions or in vitro in supplemented growth media results in sialylation when the strain contains the terminal acceptor structure such as LNnT. The sialylation of LPS, besides increasing serum resistance, may also provide ligands for haemopoietic cell receptors such as sialic acid-binding immunoglobulin-like lectins (siglecs).

Figure 1.

Schematic presentation of N. meningitidis LPS (adapted from Tsai et al., 2002). The α side-chains attached to Hep I of the immunotypes L2–5, L7 and L9 contain LNnT. An alternative side-chain occurs in immunotype L1 that contains a trisaccharide as shown. The L6 structure lacks the terminal galactose of LNnT, and L8 immunotype lacks Gal-GlcNAc; neither can be sialylated by Nm α2,3 sialyltransferase. L1 and LNnT containing LPSs can be intrinsically sialylated in serogroup B and C strains. Two very short side-chains occur at Hep II. Phosphoethanolamine (PEA) may replace the glucose moiety in the β chain in some immunotypes.

Siglecs are transmembrane proteins of the immunoglobulin (Ig) superfamily that have an amino-terminal V-set domain and varying numbers of C2 domains, ranging from 16 in sialoadhesin/siglec-1 (Sn) to one in CD33/siglec-3. In general, these proteins are involved in both adhesive and signalling functions (reviewed by Crocker, 2002). There are 11 known bona fide siglecs in humans that can be divided into two subgroups based on sequence similarity in the extracellular region. Sn, CD22 and myelin-associated glycoprotein/siglec-4 (MAG) form one subgroup and share around 25–30% sequence identity, whereas CD33 and siglecs 5–11 form a second subgroup and share around 50–85% identity. Each siglec exhibits distinct carbohydrate-binding properties with specificity for both the sialic acid linkage and the nature of the sialic acid itself (Table 1). The sialic acid binding site is localized to the V-set Ig domain, and structures of this domain have been determined by X-ray crystallography for both Sn and siglec-7 (May et al., 1998; Alphey et al., 2003). With the exception of MAG (siglec-4), all siglecs characterized to date are expressed on distinct subsets of haemopoietic cells, such as macrophages (Sn/siglec-1), neutrophils (siglecs-5, -9), monocytes (siglecs-3, -5, -7, -9), B cells (siglecs-5, -6, -9, -10), natural killer (NK) cells (siglec-7) and eosinophils (siglec-8, -10) (reviewed by Crocker and Varki, 2001).

Table 1.  Properties of siglecs and anti-siglec monoclonal antibodies used in this study.
SiglecAbbreviations
used in the
text/figures
NeuAcα2,3
Lac(NAc)
bindinga
Expressed onBlocking mAbs used
Mouse mAbsbReference
  • a. 

    Other specificities are described by Crocker and Varki (2001).

  • b. 

    Mouse hybridoma culture supernatants containing antibodies were used at the dilutions shown.

Sialoadhesin
(Siglec-1)
Sn+++Tissue macrophagesSER-4 (1:5)
+1C2 1:20
Crocker and Gordon (1989);
Crocker et al.
(1991)
CD33
(Siglec-3)
S3+Myeloid progenitors,
monocytes
  
Siglec-5S5++Monocytes, neutrophils
macrophage subsets
1A5 1:5Cornish et al. (1998)
Siglec-7S7++Monocytes, NK cells  
Siglec-8S8+Eosinophils7C9 (1:5)Floyd et al. (2000)
Siglec-9 Siglec-10S9 S10++++Monocytes, neutrophils
Monocytes (weak),
eosinophils (weak), B cells
  

The differences in expression and ligand-binding properties suggest that each siglec mediates a specific, non-redundant function in haemopoietic cell biology. Apart from Sn and MAG, all human siglecs contain two or more tyrosine-based motifs in their cytoplasmic tails that can become phosphorylated, recruit SHP-1 and mediate inhibitory signalling functions (reviewed by Crocker and Varki, 2001). Several siglecs recognize α2,3-linked sialic acids that occur not only on mammalian cells but also on human pathogens, notably on Nm LPS. In order to investigate whether any of these siglecs also function as receptors for sialylated bacteria, we have investigated their potential to interact with sialic acids expressed on Nm LPSs. In this study, we used sialylated (L3) and non-sialylated (L8) LPS variants from the intrinsically sialylated serogroup B strain MC58 as well as variants of a serogroup A strain C751 that were grown in media with exogenous CMP-NeuAc. These studies have demonstrated the potential of two distinct siglecs to recognize Nm LPS sialic acids.

Results

The investigations used glutaraldhehyde-fixed, heat-killed or viable Nm derivatives of strains MC58 and C751. Several clonal isolates of C751 expressing or lacking opacity proteins were included. The phenotypic characteristics of the meningococcal derivatives used are shown in Table 2.

Table 2. Neisseria meningitidis strains and variants used in this study.
Strain/derivativeSerogroupPhenotypePrevious
nomenclature
Reference
CapaPiliOpaOpcLPS
  • a. 

    cap, capsule expression. Strain MC58 variant with L3 LPS immunotype becomes intrinsically sialylated. Strain C751 variants express the L9 immunotype that can be sialylated when an extrinsic source of sialic acid (CMP-NeuAc) is provided in the growth medium.

MC58B      Virji et al. (1995)
 Sia– L8⊄11 
 Sia+ L3⊄12 
C751A      Virji et al. (1993; 1992)
 OpaA+ AL9A7i20 
 OpaB+ BL9A7i16 
 OpaD+ DL9A7i5 
 Opc+ +L9A7b 
 Op– L9A7i 

Siglec Fc-chimera binding of N. meningitidis derivatives

To investigate whether siglecs can recognize the sialylated LPS structures present on Nm, we selected a panel of siglec-Fc chimeras, corresponding to siglecs known to be expressed on leucocytes and able to interact with α2,3-linked sialic acids (Table 1). A suspension binding assay was developed using glutaraldehyde-fixed Nm, in which siglec-Fc chimeras were complexed with anti-Fc-alkaline phosphatase in order to allow multivalent, high-avidity interactions with sialylated glycans. Intrinsically sialylated L3 immunotype of MC58 (sia+) and an L8 phenotype lacking LNnT (sia–) as well as a range of sialylated and non-sialylated variants of serogroups A strain C751 that had been grown in the presence or absence of CMP-NeuAc were used. These assays were carried out in parallel with suspension binding assays of human red blood cells (RBCs) as an internal comparison, as the siglecs under study have been shown to bind to these cells in a sialic acid-dependent manner (Crocker et al., 1994; Cornish et al., 1998; Nicoll et al., 1999; Floyd et al., 2000; Zhang et al., 2000; Hartnell et al., 2001; Munday et al., 2001).

Surprisingly, only Sn and siglec-5 were found to mediate sialic acid-dependent binding to the Nm derivatives (Figs 2A and 3A). In these experiments, binding of sialylated Nm to Sn-Fc was consistently higher than to siglec-5-Fc. As the binding assay depends on multivalent clustering, relatively small differences in affinity of Sn and siglec-5 for Nm LPS could result in large differences in binding. The level of Nm binding to mouse and human Sn recombinant proteins was very similar and is consistent with previous data showing indistinguishable binding properties to sialoglycoconjugates and mammalian cell populations (Hartnell et al., 2001). This validates the use of mouse Sn-expressing cells as a model for examining interactions with Nm, as described below.

Figure 2.

Binding of siglec-Fc chimeras to N. meningitidis variants.
A. Siglec-Fc chimeras at 1 µg ml−1 were precomplexed with anti-human Fc–alkaline phosphatase conjugate and used in suspension binding assays with glutaraldehyde-fixed Nm. All were human proteins except for mSn, which was the murine form. Strain MC58 variants (sia+ and sia–; see Table 1) and strain C751 Op Nm variants grown with or without CMP-NeuAc were used. Sialylated phenotypes are represented by filled bars, and non-sialylated variants are represented by open bars. Data are expressed as means ± SD of triplicate samples representative of three independent experiments.
B (assay control). Binding of precomplexed siglec-Fc proteins to sialylated (filled bars) and desialylated (open bars) glutaraldehyde-fixed human RBCs. Data are expressed as means ± SD of triplicate samples representative of three independent experiments.

Figure 3.

Binding of sialylated and non-sialylated N. meningitidis opacity variants to mouse sialoadhesin-Fc (Sn-Fc) and siglec-5-Fc (S5-Fc).
A. Chimeric receptor binding to distinct phenotypes of meningococci (glutaraldehyde-fixed) was carried out as described in the legend to Fig. 2. In each case, sialylated bacteria (filled bars) bound in larger numbers than non-sialylated bacteria (open bars) independently of the expression of opacity proteins. Data are expressed as means ± SD of triplicate samples representative of two independent experiments.
B. Binding of native Nm (viable at the start of the experiment) to soluble chimeric receptors in dot-blot overlay experiments. Receptor binding was detected using anti-human Fc–alkaline phosphatase conjugate and appropriate substrates. Relative binding was determined by densitometric analysis using the NIH scion image program. The paired filled and paired blank bars represent mean values of duplicates from two independent experiments (errors < 25% of the means). The C751 isolate shown here expressed Opc. However, similar results were obtained with Opa-expressing and Op bacteria.

No sialic acid-dependent binding to any Nm derivative was seen with siglecs-7, -8, -9 and -10 or a mutant R97A form of Sn (May et al., 1998) carrying an inactivating mutation in the sialic acid binding site (data not shown). In comparison, siglec-7 and siglec-8 were shown to bind variably to Nm strains in a sialic acid-independent manner (Fig. 2A). This is likely to reflect non-specific binding of these recombinant proteins as CHO cells transfected with siglec-7 and -8 did not bind Nm at greater levels than non-transfected controls (Table 3). With the exceptions of siglecs-5 and -8, all siglec-Fc proteins bound RBCs in a sialic acid-dependent manner in parallel suspension binding assays (Fig. 2B). As siglecs-5 and -8 were found previously to bind RBCs in solid phase adhesion assays (Cornish et al., 1998; Floyd et al., 2000), this suggests that their interaction with RBCs in solution was too weak to be detected under the conditions used. In conclusion, these results indicate that the sialylated LPS structures on the surface of Nm in a variety of phenotypes (with or without opacity proteins) and in strains of distinct serogroups have the potential to act as ligands for Sn and siglec-5.

Table 3.  Adhesion of N. meningitidis to CHO-siglec transfectants: inhibition by anti-siglec mAbs.
Nm strain/variantCHO transfectants
NeomSnS3S5S7S8S9S10
  1. CHO transfectants expressing the indicated siglecs at the cell surface were infected with meningococci as described in the text, and bacterial binding was assessed. Approximate numbers of bacteria adherent per cell as determined by immunofluorescence microscopy are shown. mAbs used and their concentrations are shown in Table 1. Inhibition of bacterial binding was assessed by immunofluorescence as well as in a viable count assay in the case of CHO-S5. From viable count assays, percentage inhibition of bacterial adhesion in the presence of mAb compared with that in its absence is shown in parenthesis.

  2. Neo, sham-transfected CHO cells. mSn, CHO cells transfected with mouse Sn; the rest were human siglec transfectants.

MC58 sia–<12–5<1<1<1<1<1<1
MC58 sia+<1>50<120–50<1<1<1<1
MC58 sia+ + anti-Sn 2–5      
MC58 sia+ + anti-S5 >50  5–10 (60%)    
MC58 sia+ + anti-S8   20–50 (2%)    

Receptor overlay dot blots of live N. meningitidis

To confirm the siglec binding properties with live bacteria, suspensions of sia+ and sia– phenotypes of strain MC58 as well as C751 variants (Op–, Opc+, OpaB+) were dotted onto nitrocellulose and overlaid with chimeric receptor molecules as described in Experimental procedures. Densitometric analysis demonstrated the marked difference between the sialylated and non-sialylated bacterial binding to human siglec-5-Fc and mouse Sn-Fc independently of opacity protein expression (Fig. 3B). No significant binding was observed to any of the other siglec-Fc molecules. Quantitative differences observed in different experiments are most likely to result from heterogeneity of LPS and its sialylation levels between different cultures.

Cellular interaction of live meningococci via siglecs

To investigate whether cell-expressed siglecs can interact with Nm in a similar way to recombinant soluble siglecs, we used stably transfected Chinese hamster ovary (CHO) cells expressing several siglecs and determined the interactions of live sialylated and non-sialylated MC58 isolates. This included CD33/siglec-3, which was not represented in the panel of siglec-Fc proteins described above. Immunofluorescence assays indicated that sialylated bacteria adhered to transfected cells that expressed either human siglec-5 or mouse Sn but not CHO cells expressing other siglecs or untransfected CHO cells (Table 3, Fig. 4). The interactions of sialylated bacteria could be inhibited in a specific manner with appropriate monoclonal antibodies (mAbs) or with F(ab)2 fragments of sheep polyclonal antibodies (Table 3, Fig. 4). Viable count experiments confirmed the observations of the immunofluorescence adhesion assays. Approximately 30- and 10-fold differences in binding of sialylated bacteria versus non-sialylated bacteria to CHO siglec-5 transfectants were observed for strain MC58 and C751 Op derivatives. Gentamicin protection assays showed that more Nm were internalized by CHO transfectants if they were sialylated compared with their non-sialylated counterparts, albeit the numbers of gentamicin survivors were relatively low. Approximately 0.7% of associated MC58 sia+ bacteria were apparently internalized or survived after uptake by viable count assay, and a six- to 10-fold difference in internalization of sia+ and sia– bacteria was observed.

Figure 4.

Adherence of sialylated and non-sialylated N. meningitidis to cell-expressed receptors assessed by immunofluorescence assays. Monolayers of CHO cells expressing mouse Sn (CHO-Sn) or human siglec-5 (CHO-S5) were incubated with sia+ or sia– variants of strain MC58 for 2 h at 37°C (infection ratio bacteria:target cells of 200:1). Non-adherent bacteria were washed away, and adherent bacteria were detected with rabbit antiserum against Nm and secondary anti-rabbit antibody conjugated to rhodamine. For blocking with specific antibodies, monolayers of CHO-Sn were preincubated for 15–20 min with monoclonal antibodies (SER-4 at 1:5 and 1C2 at 1:20 dilutions) and CHO-S5 with F(ab)2 fragments (20 µg ml−1) derived from affinity-purified sheep polyclonal antibodies against siglecs-5 and -8 before incubation with bacteria.

Flow cytometry analysis of uptake of sialylated phenotypes by CHO transfectants expressing siglecs

We further confirmed the internalization of sia+ phenotype by flow cytometry and measured quantitatively the uptake of heat-killed fluorescein isothiocyanate (FITC)-labelled bacteria by CHO cells expressing siglec-5 or mouse Sn. In this assay, non-internalized bacteria are excluded from the analysis by the addition of trypan blue, which quenches extracellular fluorescence. Compared with non-sialylated bacteria, significantly larger numbers of sialylated bacteria were taken up into siglec-5-expressing CHO cells than wild-type CHO cells, and this could be inhibited by pretreatment of the target cells with blocking antibodies against siglec-5 (Fig. 5A and B). In the presence of the blocking antibody, the overall degree of uptake was similar to or less than that of the control samples. Comparable results were also obtained with CHO cells stably transfected with Sn (data not shown).

Figure 5.

Flow cytometric analysis of uptake of fixed FITC-labelled N. meningitidis by CHO-siglec-5. Siglec-5-expressing CHO cells were incubated at 37°C for 60 min with FITC-labelled sialylated and non-sialylated Nm variants as shown at a ratio of 1:20. After trypan blue quenching of extracellular bound bacteria, uptake into cells was evaluated by flow cytometry (solid histograms) as described in the text. Empty histograms show background fluorescence of cells incubated in the absence of bacteria. The values in each histogram show the median fluorescence intensities of cell-associated bacteria within the indicated gates. Sia+ bacteria were taken up to a greater extent than Sia– bacteria by CHO-S5 (A). Preincubation of CHO transfectants with sheep anti-siglec-5 polyclonal IgG selectively reduced the uptake of sialylated variants (B). Unmasking of siglec-5 at the cell surface by sialidase pretreatment of CHO cells enhanced uptake, especially in the case of the sialylated Nm variants (C). The enhanced uptake of sialylated Nm variants by sialidase-treated CHO cells was selectively reduced after pretreatment of CHO cells with anti-siglec-5 IgG (D). Data are from one experiment, representative of two independent experiments.

As siglecs can be masked at the cell surface as a result of cis interactions with sialic acids expressed on the same cell (Freeman et al., 1995; Cornish et al., 1998; Razi and Varki, 1999; Floyd et al., 2000; Zhang et al., 2000), we used Vibrio cholerae sialidase to unmask or at least reduce the masking of siglec-5. Pretreatment of siglec-5 CHO cells with sialidase resulted in a marked increase in the overall uptake of bacteria (Fig. 5C). This increase occurred to a much greater extent with the sialylated bacteria than with the non-sialylated bacteria, suggesting that it was not purely a non-specific effect of removing negatively charged molecules from the CHO cell surface. This was confirmed by using anti-siglec-5 blocking antibodies that restored uptake of sialylated bacteria to background levels (Fig. 5D).

Macrophage phagocytosis of sialylated N. meningitidis

Sialoadhesin is naturally expressed at high levels on subsets of tissue macrophages. To test whether Sn–Nm interactions may lead to altered uptake of the bacteria, we carried out phagocytosis assays using bone marrow-derived macrophages from wild type (+/+) and Sn-deficient (–/–) mice. Compared with Sn-deficient macrophages, wild-type macrophages showed an enhanced uptake of sialylated Nm over a 30 min time course, the influence of Sn on bacterial uptake being seen as early as 10 min (Fig. 6A). All macrophage populations took up non-sialylated bacteria poorly, and uptake of sialylated bacteria could be inhibited either by pretreating macrophages with blocking Sn antibodies or by sialidase pretreatment of Nm (Fig. 6B). At later time points, up to 60 min, the differences were less obvious (data not shown), suggesting that additional macrophage receptors such as the class A scavenger receptor (Peiser et al., 2002) bind and phagocytose both sialylated and non-sialylated bacteria. The data indicate that Sn, despite being regarded as a non-phagocytic receptor, can lead to enhanced phagocytic clearance of sialylated Nm variants, an observation of clear biological importance.

Figure 6.

Influence of Sn on macrophage phagocytosis of N. meningitidis.
A. Adherent bone marrow-derived macrophages from wild-type (solid lines, open triangles) and Sn-deficient (dotted lines, filled circles) mice were incubated with FITC-labelled sialylated (left) or non-sialylated (right) MC58 Nm variants for up to 30 min at 37°C at a ratio of 1:10 (macrophage: bacteria). After fixation and ethidium bromide quenching of non-phagocytosed bacteria, the percentages of macrophages with two or more phagocytosed bacteria were determined by fluorescence microscopy. Sialylated bacteria were phagocytosed more rapidly by wild-type macrophages than by Sn-deficient macrophages, but no differences were seen for non-sialylated bacteria. Data are expressed as means of duplicate wells and are representative of three independent experiments.
B. Adherent bone marrow-derived macrophages from wild-type (+/+) and Sn-deficient (–/–) mice were incubated with FITC-labelled sialylated (filled bars) or non-sialylated (open bars) MC58 Nm variants for 30 min. The enhanced phagocytic activity of wild-type macrophages towards sialylated Nm variants was abolished upon macrophage pretreatment with blocking anti-Sn mAbs and sialidase pretreatment of bacterial variants. Data are expressed as mean percentage ± SD of quadruplicates and are representative of two independent experiments.

Discussion

In general, sialylation of meningococcal LPS results in the inhibition of molecular interactions at the bacterial surface including complement and antibody binding to outer membrane components (Hammerschmidt et al., 1994; Vogel and Frosch, 1999) and receptor–ligand interactions mediated by opacity proteins (Rest and Frangipane, 1992; Virji et al., 1995; 1996; McNeil and Virji, 1997; Dehio et al., 2000). However, in one study, sialic acid-bearing LPS was found to cause a greater increase in proinflammatory cytokine release from macrophages than asialo-LPS, suggesting direct interactions of Nm LPS sialic acids with the macrophages (Quakyi et al., 1997; Tsai et al., 1998). In the current investigations, we have provided evidence that sialic acids of Nm LPS can interact with phagocytic cell receptors. In the current investigations, we have provided evidence that sialic acids of Nm LPS can interact with phagocytic cell receptors. Using meningococcal strains of serogroups A and B, we have shown that two distinct human siglecs, namely sialoadhesin and siglec-5, recognize the α2,3-linked sialic acids on Nm LPS. This was observed with Nm LPS of either L3 or L9 immunotypes and was demonstrated in a variety of binding and internalization assays which used recombinant soluble and membrane-expressed forms of the proteins. We also investigated whether the presence of the two basic outer membrane proteins Opa and Opc, which are present in clinical isolates, had any effect on the interactions. The studies clearly show no interference by these ligands in the assays used.

The composition of the oligosaccharide side-chains of Nm can vary between isolates, and this is the basis for antigenic variability that defines the 12 immunotypes (Fig. 1). Sialylation usually occurs on the α side-chain, catalysed by an endogenous α2,3 sialyltransferase. This gives rise to a terminal capping structure corresponding to 3′ sialyllactosamine (NeuAcα2,3Galβ1,4GlcNAc) (Gilbert et al., 1997; Tsai et al., 2002). The requirements for sialylation include at least a trisaccharide linked to Hep 1 in addition to a terminal galactose as the disaccharide lactose with terminal galactose of L8 immunotype cannot be sialylated (Tsai et al., 2002). Apparently, the length of this side-chain determines whether LPS may be sialylated or not, and it has been suggested that steric hindrance imposed by the neighbouring Hep II may explain why the shorter lactose moiety could not act as an acceptor for sialic acid (Tsai et al., 2002) (Fig. 1). However, lactoneotetraose occurs frequently on the α chains of several ITs and is present in immunotypes most commonly isolated from disseminated infections including L3 and L9.

Based on the presence of the 3′ sialyllactosamine moiety on LPS, it was expected that all siglecs tested would interact with the sialylated Nm as, in previous studies, Sn, siglec-5, -7, -8, -9 and -10 showed robust binding to synthetic polyacrylamide probes coupled to 3′ sialyllactose or 3′ sialyllactosamine (Cornish et al., 1998; Nicoll et al., 1999; Angata and Varki, 2000; Floyd et al., 2000; Zhang et al., 2000; Hartnell et al., 2001; Munday et al., 2001). The failure of siglecs-7, -8, -9 and -10 to bind Nm LPS in a sialic acid-dependent manner was surprising and suggests that the underlying glycans (e.g. attached to Hep II; Fig. 1) may negatively affect recognition by these siglecs but do not influence recognition by Sn and siglec-5.

The crystal structure of Sn V-set domain complexed to 3′ sialyllactose revealed that most of the molecular contacts were made with sialic acid; there were limited interactions with galactose and none with glucose (May et al., 1998). Both Sn and siglec-5 can bind α2,3-, α2,6- and α2,8-linked sialic acids, suggesting a relaxed specificity for sialylated glycans that may be important in their ability to interact with sialylated LPS molecules (Crocker et al., 1991; Brinkman-Van der Linden and Varki, 2000). In contrast, recognition of α2,3-linked sialic acids by siglec-9 can be dramatically affected by underlying glycans (Angata and Varki, 2000). In addition, recent structural and biochemical analyses of siglec-7 have shown that this siglec has a more extensive and complex binding site than Sn, interacting strongly with α2,8-linked disialic acids on complex gangliosides as well as an internally branched α2,6-sialylated glycan, LSTb (Yamaji et al., 2002; Alphey et al., 2003). Further studies are needed to probe the fine specificity of different siglecs to determine the molecular basis for the differential recognition of sialylated Nm LPS described here.

When expressed naturally at the cell surface, the binding site of most siglecs is ‘masked’ by cis interactions with sialic acids co-expressed on the same plasma membrane. Sn is unusual in this respect, as it is naturally unmasked on macrophages where it functions as a sialic acid-dependent adhesion molecule. With 17 Ig-like domains, Sn is thought to extend its amino-terminal binding site away from the plasma membrane, thereby reducing cis interactions with sialic acids. In contrast, siglec-5 has only four Ig domains and, by using soluble glycoprobes, it has been shown to be largely masked at the cell surface, on COS cells (Cornish et al., 1998), CHO cells (C. Jones and P. R. Crocker, unpublished observations) or neutrophils (Razi and Varki, 1999). Therefore, it was of interest that, in the present study, siglec-5 expressed on CHO cells could bind and internalize Nm in a sialic acid-dependent manner even without prior unmasking. It is possible that the sialylated LPS outcompetes the cis-interacting sialic acids on CHO cells, resulting in the significantly high levels of siglec-5-dependent binding in trans. These findings raise the possibility that siglec-5 expressed naturally on monocytes, neutrophils and macrophages could interact with sialylated LPS on Nm and influence their recognition and uptake. However, the degree to which this occurs may depend on the sialylation status of the myeloid cells, which could affect both masking and non-specific electrostatic repulsive effects. It is of interest to note that, in previous studies, sialylated LPS reduced non-opsonic interactions of acapsulate Nm with phagocytes mediated by opacity proteins (McNeil and Virji, 1997). This suggests that the level of masking of siglecs on resting neutrophils may be such that (at least under in vitro conditions) interactions in trans with LPS sialic acids do not occur. Bacterial sialic acids in this case may become protective by increasing electrostatic repulsive effects. However, experiments using polyacrylamide-based glycoprobes indicated that siglec-5 can become unmasked on neutrophils after activation (Razi and Varki, 1999), but whether this occurs in vivo and leads to trans interactions with LPS sialic acids remains to be shown.

To demonstrate further the biological potential of Nm sialic acids in recognition by phagocytes, we used macrophages from wild-type and Sn-deficient mice. The use of this model was validated by the observation that both human and mouse Sn behaved similarly in soluble chimeric receptor binding assays. As in the case of CHO transfectants, Sn-replete macrophages engulfed sialylated bacteria more efficiently than Sn-deficient macrophages. In previous studies, Sn expressed on macrophages was unable to mediate phagocytosis of bound RBCs, even over several days of culture (Crocker and Gordon, 1986). It is possible that Sn may be capable of directly mediating the internalization of small particles such as bacteria, but not larger particles such as RBCs. Alternatively, the enhanced Sn-dependent internalization of sialylated Nm by macrophages and CHO cells could result from synergy between Sn and another (phagocytic) receptor such as the macrophage class A scavenger receptor that ligates as yet unidentified bacterial component(s) (Peiser et al., 2002). A similar type of synergistic effect could, in principle, also operate in the case of the siglec-5-dependent increased uptake of Nm observed in transfected CHO cells. However, one important difference between Sn and siglec-5 is the presence in the latter of two tyrosine-based motifs (YASL and YSEI), which have the potential to function as signals for clathrin-dependent endocytosis. Generally, tyrosine-based signals are of the form NPXY or YXXØ (where X is any amino acid and Ø is an amino acid with a bulky hydrophobic side-chain) (Royle et al., 2002), which interact with the µ2 subunit of adaptor protein-2 associated with clathrin in coated pits. Further studies are required to investigate these possibilities.

Importantly, the tyrosine-based motifs in siglec-5 are also potential substrates for src-like tyrosine kinases and are strongly implicated in signalling functions. The membrane-proximal motif conforms to the canonical immune receptor tyrosine-based inhibitory motif (ITIM) which, on tyrosine phosphorylation, can recruit the inhibitory tyrosine phosphatase, SHP-1, leading to inhibition of activatory signals (reviewed by Ravetch and Lanier, 2000). Indeed, several CD33-related siglecs have been shown to be capable of delivering inhibitory signals when co-cross-linked with activating receptors (reviewed by Crocker and Varki, 2001), and similar findings have been made in the case of siglec-5 (T. Avril and P. R. Crocker, unpublished). A recent study has also demonstrated that ligation of siglec-8 expressed on eosinophils can trigger a caspase-dependent apoptotic response (Nutku et al., 2003). Based on their sialic acid-binding activity and their potential to modulate leucocyte activities, it has been proposed that CD33-related siglecs may function to inhibit ‘self’-reactivity of leucocytes via interaction with sialic acids on host cells (Crocker and Varki, 2001). Thus, it is conceivable that the acquisition of sialic acid by pathogens may be important for modulating leucocyte functions via interaction with siglecs. Further studies are required to determine whether ligation of siglec-5 on neutrophils and macrophages can deliver inhibitory signals that might favour the survival of sialylated versus non-sialylated bacteria.

In the case of Sn, which lacks tyrosine-based signalling motifs, enhanced macrophage uptake could result in enhanced killing of bacteria and increased production of proinflammatory cytokines, as proposed by others (Quakyi et al., 1997; Tsai et al., 1998). Interestingly, however, a recent study has shown that the macrophage class A scavenger receptor can promote macrophage uptake of Nm, but that uptake is uncoupled from LPS/Toll-like receptor-dependent triggering of cytokine release (Peiser et al., 2002). If this were also the case with Sn, it would suggest that Sn could be important for enhancing macrophage clearance of sialylated bacteria in tissues and at sites of inflammation, without triggering the release of potentially deleterious proinflammatory cytokines.

In conclusion, these studies demonstrate that siglecs can interact with sialylated N. meningitidis and play a potentially protective role during infections. Sn is expressed on many resident and inflammatory macrophage populations in humans that would probably be exposed to bacteria in the circulation and at sites of inflammation and would therefore be expected to contribute to macrophage clearance and destruction of these microorganisms. In the case of siglec-5, expression is seen predominantly on circulating neutrophils, but also on inflammatory macrophages, especially those in reactive lymph nodes (Connolly et al., 2002). Whether recognition via siglecs leads to efficient bacterial elimination or whether, after internalization via this route, bacteria may survive and translocate in a ‘Trojan horse’ manner within phagocytes across epithelial or endothelial barriers remains to be shown. Regardless, the observations described here demonstrate the biological potential of siglecs in controlling meningococcal survival in vivo via interactions with otherwise protective sialic acids on Nm LPSs.

Experimental procedures

Bacteria and growth conditions

The meningococcal serogroup A strain C751 and serogroup B strain MC58 and intrastrain variants have been described previously. Briefly, clonal populations of acapsulate, Opc– mutants of strain MC58 were selected further to isolate Opa–, pili– phase-variant phenotypes with either L8 or L3 LPS immunotype expression (Virji et al., 1995). Clonal populations of opacity protein phase variants of strain C751 were derived from single colony isolates of a spontaneous acapsulate, non-piliated phenotype and were frozen in liquid nitrogen after a single subculture (Virji et al., 1993; 1992). All experiments used 16–18 h cultures grown on agar, inoculated directly from liquid nitrogen stocks. This ensured that the inoculum was> 90% pure with regard to the expression of the ligands of interest. Bacteria were grown on brain–heart infusion (BHI) agar supplemented with 10% heated horse blood in 5% CO2 in air at 37°C. Bacterial culture conditions and enumeration have been described previously (Virji et al., 1991).

Phenotypic characterization and properties

The expression of major surface proteins as well as LPS was investigated by SDS-PAGE and immunoblotting using monoclonal antibodies. The detailed characterization of the mutants/variants has been reported previously (Virji et al., 1993; 1992; 1995). By these criteria, the MC58 derivatives used in the current studies were acapsulate, Opc–, Opa–, pili– with L8 (sia–) or L3 (sia+) LPS immunotype expression (Table 2). Intrinsic sialylation occurs in this strain but only when L3 immunotype is present. The presence of sialic acid on the LNnT of the L3 immunotype has been demonstrated (Jennings et al., 1995). Growth of this variant on cytidine monophospate N-acetyl neuraminic acid (CMP-NeuAc) was undertaken in some experiments, but no significant difference in cellular adhesion over L3 immunotype grown in the absence of CMP-NeuAc was seen. The strain C751 derivatives differed from each other only in the presence or absence of Opa or Opc on their surface. None of the variants expressed pili, but pilin subunits were detected in the outer membrane using the monoclonal antibody SM1 against class I pili. All variants were capsule deficient, and their LPS (immunotype L9) produced a low Mr component on SDS-PAGE, which was replaced by one of higher Mr when bacteria were grown on CMP-NeuAc. This LPS also incorporated radiolabel from neuraminic acid-labelled CMP-NeuAc provided in the culture medium (Virji et al., 1993; 1994). Thus, intrinsic sialylation of LPS did not take place in this strain.

Preparation of heat-killed bacteria

Bacteria were harvested after 16–18 h growth and suspended in Dulbecco's phosphate-buffered saline (PBS) containing Ca and Mg salts (PBSB). Numbers of bacteria were enumerated as described previously by measurement of their DNA content, and standard suspensions were prepared. The suspensions were immediately heated at 65°C for 30 min. Heat-killed bacteria were stored at −20°C in 10% glycerol and thawed immediately before use.

Transfected cell lines and soluble chimeric receptors

CHO cells expressing full-length forms of human siglecs-3, -5, -7, -8, -9 and -10 were prepared as described previously (Crocker et al., 1994; Cornish et al., 1998; Nicoll et al., 1999; Floyd et al., 2000; Zhang et al., 2000; Munday et al., 2001). Owing to the unavailability of CHO cells expressing the full-length human Sn, CHO transfectants expressing mouse Sn were used (Crocker et al., 1994). Previous studies have shown indistinguishable ligand binding by both mouse and human sialoadhesins (Hartnell et al., 2001). CHO cells stably secreting chimeric proteins comprising the extracellular regions of siglec proteins fused to the Fc portion of human IgG1 were prepared as described previously (Crocker et al., 1994; Cornish et al., 1998; Nicoll et al., 1999; Floyd et al., 2000; Zhang et al., 2000; Munday et al., 2001). All contained the entire extracellular region of each siglec with the exception of mouse and human Sn, which had the first three and four N-terminal Ig domains respectively. In the case of mouse Sn-Fc, the binding properties of this construct are equivalent to those of the full-length form, composed of 17 Ig domains (Crocker et al., 1994). Human Sn-Fc, which behaves indistinguishably from mouse Sn-Fc in binding assays, was prepared as described from transiently transfected COS-1 cells (Hartnell et al., 2001). All Fc proteins were used in the form of serum-free tissue culture supernatants, containing 1–10 µg ml−1 Fc-protein, as determined by enzyme-linked immunosorbent assay (ELISA).

Antibodies to siglecs and bacteria

Monoclonal antibodies (mAbs) to siglecs were used as tissue culture supernatants as described in Table 1. Polyclonal sheep antisera (Diagnostics Scotland) were raised to the purified extracellular regions of siglecs-5 and -8, cleaved from Fc-proteins by 3C protease cleavage (Zhang et al., 2000) (Diagnostics Scotland). IgG was purified from immune sera using protein G Sepharose (Amersham Pharmacia Biotech) and affinity purified using the relevant siglec-Fc protein coupled to cyanogen bromide-preactivated Sepharose CL-4B (Amersham Pharmacia Biotech). F(ab)2 fragments were generated by pepsin digestion followed by purification on an AKTA purifier using a HiPrep 26/60 Sephacryl S200 HR column (Amersham Pharmacia Biotech). Purity was determined to be> 90% by SDS-PAGE. Specific binding was demonstrated by fluorescence-activated cell sorting (FACS) analysis against a panel of CHO cell transfectants expressing human siglecs-3 and 5–10. Antibodies to Nm were raised in rabbits immunized with mixtures of lysed Nm strains of distinct serogroups.

Preparation of human RBCs

RBCs were isolated from blood of healthy adult volunteers by Ficoll-Paque density gradient centrifugation (Amersham Pharmacia Biotech). RBCs were washed extensively in PBS and stored in Alsever's solution at 4°C for up to 2 weeks.

Sialidase treatment of cells and bacteria

RBCs and bacteria were washed three times in PBS, resuspended at 1% (v/v) in sialidase buffer (154 mM NaCl, 10 mM NaAc, 9 mM CaCl2, pH 5.5) and sialidase treated using 0.05 U ml−1Vibrio cholerae (Calbiochem) sialidase for 2 h at 37°C with rotation. Cells were washed with PBS and then fixed in PBS + 0.125% glutaraldehyde for 10 min at 4°C. After washing, aldehyde groups were blocked with 10% fetal calf serum (FCS) for 10 min. The cells were then washed twice and resuspended in appropriate assay buffer. Sialidase treatment of CHO cells was carried out in HAM's F10 containing 0.1 U ml−1 sialidase for 1 h at 37°C with occasional agitation. Cells were then washed in HAM's F10 before being used in the flow cytometry assays.

Siglec-Fc suspension binding assay

Siglec-Fc chimeras (1 µg ml−1) were complexed with an alkaline phosphatase (1:3000)-conjugated anti-human IgG (Fc specific; Sigma) in PBS + 0.1% bovine serum albumin (BSA) + 10 mM sodium azide for 60 min at room temperature with rotation. The siglec-Fc–alkaline phosphatase complexes were then mixed with 108 glutaraldehyde-fixed Nm variants or 0.5 µl of glutaraldehyde-fixed RBCs in a total volume of 50 µl per well. After incubation at 4°C for 60 min, bacteria and RBCs were washed with ice-cold Tris-buffered saline (TBS) containing 1% BSA (16 250 g, 5 min, 4°C). Bacterial pellets were resuspended in 100 µl of 10 µM fluorescein diphosphate (Molecular Probes) incubated at room temperature for 15–30 min. Bacteria were then pelleted by centrifugation, supernatants transferred to 96-well plastic plates and fluorescence readings measured using a fluorescent plate reader (Cytofluor, PerSeptive Biosystems).

Receptor/ligand overlay dot blotting

For soluble chimeric receptor binding to native bacterial suspensions, freshly prepared live bacterial suspensions were applied to nitrocellulose (≈ 107 per dot), air dried, blocked in 3% BSA in PBS (BSA-PBS) and overlaid with siglec-Fc-soluble proteins (1–2 µg ml−1). Binding was detected using anti-human Fc–alkaline phosphatase conjugate and nitroblue tetrazolium and 5-bromo-4-chloro-3-indoylphosphate as substrates (Sigma). The levels of reactivity were determined by densitometric analysis using the NIH scion image program.

Relative adhesion of live bacteria to target cells expressing siglecs

Monolayers of CHO cells or transfectants expressing distinct siglecs were infected with bacteria in the presence or absence of antibodies at concentrations shown in Table 1 or figure legends at an infection ratio of ≈ 200 bacteria per cell and incubated for 2 h at 37°C. Non-adherent bacteria were removed by repeated washing, and adherence was measured by immunofluorescence or viable count assays.

Immunofluorescence detection of adherent bacteria

Monolayers with adherent bacteria were fixed with absolute methanol for 10 min, washed and non-specific sites were blocked for 1 h with 3% BSA-PBS containing 0.05% Tween 20. Bacteria were stained using polyclonal rabbit antiserum against Nm and secondary antibody conjugated to rhodamine (Sigma).

Viable count assays for cell association

Cell-associated bacteria were released by the addition of 1% saponin and cultured for determination of colony-forming units (Virji et al., 1991). All bacteria used in such experiments were equally resistant to treatment with 1% saponin. All estimations were done at least in triplicate, and each experiment was repeated three or more times.

Gentamicin protection assay for internalization of viable bacteria

The internalization of bacteria was determined using gentamicin to eliminate extracellular adherent bacteria (Virji et al., 1992). The internalized population is resistant to the action of gentamicin, which is excluded from intact eukaryotic cells. Internalized bacteria were released using saponin as above and enumerated by the viable count method.

FITC labelling of N. meningitidis variants

Frozen, heat-inactivated glycerol stocks of Nm were washed by centrifugation in PBSB at 16 250 g for 10 min. Bacteria were then labelled with 1 mg ml−1 FITC in 0.05 M NaHCO3, pH 9.2, 0.1% glycerol for 15 min at 25°C (McNeil et al., 1994). After extensive washing in PBSB to remove unbound FITC, the bacterial pellets were resuspended in the appropriate assay buffer.

Flow cytometric assay for internalization of heat-killed bacteria by CHO-siglec-5

CHO cells stably expressing siglec-5 were detached from tissue culture flasks with PBS + 0.5 mM EDTA, washed with HAM's F10 + 20 mM Hepes + 0.1% BSA and transferred to a 1 ml Eppendorf tube and mixed with FITC-labelled Nm variants at a ratio of 1:20 in a final volume of 400 µl. In some experiments, CHO cells were sialidase treated before the assay as described above. Tubes were sealed and incubated for 60 min at 37°C with gentle tumbling. At the end of incubation, cells were washed in PBS and fixed in 2% paraformaldehyde for 10 min at 4°C. After further washing, the fluorescence of extracellular bacteria was quenched by the addition of an equal volume of 0.4% trypan blue for 10 min at room temperature. Excess trypan blue was removed by washing, and CHO cells were analysed by flow cytometry using a FACSort (Becton Dickinson). Side and forward scatter properties were used to distinguish CHO cells from free bacteria, and 10 000 events from the CHO cell gate were collected. The effectiveness of trypan blue quenching was routinely checked in each experiment by treating fixed bacterial suspensions in the absence of CHO cells, followed by flow cytometric analysis. In some experiments, CHO cells were pretreated for 1 h with 2 µg ml−1 affinity-purified sheep anti-siglec-5 IgG before the addition of bacteria.

Phagocytosis assay with wild-type and Sn-deficient macrophages

Full details of the production and characterization of Sn-deficient mice will be reported elsewhere. Bone marrow cells from Sn-deficient (–/–) and wild-type (+/+) mice on a C57Bl/6 genetic background were incubated at 106 ml−1 in 10-cm-diameter bacterial Petri dishes at 37°C in Dulbecco's modified Eagle medium (DMEM) + 10% FCS + 20% L929 conditioned medium + tumour growth factor (TGF)-β (1 ng ml−1) for ≈ 12 days. The bone marrow-derived macrophages were detached from the Petri dishes with PBS + 0.5 mM EDTA, washed and then seeded at a density of 2 × 104 cells per well in culture medium in plastic 8-well chamber slides (Nunc) for 2–3 h at 37°C to allow for adhesion. Macrophages were washed with prewarmed assay medium (DMEM + 0.1% BSA) and then incubated for up to 60 min at 37°C with FITC-labelled untreated or sialidase-treated Nm variants (ratio 1:10 macrophage:bacteria). In some experiments, macrophages were pretreated for 1 h with a mixture of SER-4 and 1C2 anti-Sn mAbs (Table 1) before the addition of bacteria. After incubation, unbound bacteria were removed by washing three times with ice-cold Hanks’ balanced salt solution, and cells were fixed in PBS + 0.125% glutaraldehyde for 10 min at room temperature. Fluorescence of the extracellular bound bacteria was quenched with 50 µg ml−1 ethidium bromide for 10 min at room temperature as described previously (Drevets and Campbell, 1991). Slides were mounted in Vectashield (Vector Laboratories), and phagocytosis was assessed by fluorescence microscopy as the percentage of macrophages ingesting two or more bacteria (Axioskop; Zeiss). A minimum of 300 cells was scored for each treatment from randomly selected fields. Examination of fixed bacteria alone was a routine control to check the effectiveness of fluorescence quenching. For each preparation of macrophages, surface expression of Sn on wild-type macrophages was confirmed by flow cytometry.

Acknowledgements

These studies were supported by a Wellcome Trust Senior Research Fellowship (to P.R.C.), the Human Frontiers Science Programme, the Medical Research Council, the Meningitis Research Foundation and the Spencer Dayman Meningitis Laboratories. We are grateful to Ms Debbie Evans for technical assistance.

Ancillary