Immunization with outer membrane vesicles displaying designer glycotopes yields class-switched, glycan-specific antibodies

Summary

The development of antibodies against specific glycan epitopes poses a significant challenge due to difficulties obtaining desired glycans at sufficient quantity and purity, and the fact that glycans are usually weakly immunogenic. To address this challenge, we leveraged the potent immunostimulatory activity of bacterial outer membrane vesicles (OMVs) to deliver designer glycan epitopes to the immune system. This approach involved heterologous expression of two clinically important glycans, namely polysialic acid (PSA) and Thomsen-Friedenreich antigen (T antigen) in hypervesiculating strains of non-pathogenic Escherichia coli. The resulting glycOMVs displayed structural mimics of PSA or T antigen on their surfaces, and induced high titers of glycan-specific IgG antibodies following immunization in mice. In the case of PSA glycOMVs, serum antibodies potently killed Neisseria meningitidis serogroup B (MenB), whose outer capsule is PSA, in a serum bactericidal assay. These findings demonstrate the potential of glycOMVs for inducing class-switched, humoral immune responses against glycan antigens.

Introduction

Complex carbohydrates, or glycans, are a ubiquitous feature on the surface of cells from all three domains of life. For example, capsular polysaccharides (CPS) or lipid-linked lipopolysaccharides (LPS) present on the surface of pathogenic bacteria are well known to mediate host-pathogen interactions (Comstock and Kasper, 2006). Alternatively, by displaying glycans that are structurally similar to those of their host, certain pathogens are able to avoid immune recognition (Comstock and Kasper, 2006). In eukaryotes, surface glycans participate in a variety of key biological processes including adhesion, cell-cell recognition, differentiation, and immune recognition (Varki et al., 2009), and are also known to feature prominently in disease (Ohtsubo and Marth, 2006). Indeed, glycans on the surfaces of tumor cells are commonly expressed at atypical levels or with altered structural attributes, and these aberrant structures serve as unambiguous markers of malignancy for a number of cancers (Pinho and Reis, 2015).

At present, the study of glycans and their myriad roles remains a daunting task due in large part to their inherent structural complexity and the relative lack of tools for their biosynthesis, analysis, and recognition. Antibodies (Abs) specific for glycan epitopes (glycotopes) are particularly useful clarifying the functions of glycans. Glycan-targeting Abs can be elicited by immunization with carbohydrate antigens, and the resulting Abs can be used to probe the structure and function of glycans (Calarese et al., 2005; Nonaka et al., 2014) or target glycans therapeutically (Luo et al., 2010; Zhang et al., 2010). Nonetheless, the creation of glycan-specific Abs by immunization poses a significant challenge for several reasons. First, it is very difficult to isolate glycan-based immunogens from cells and tissues at purities and quantities that are sufficient for mAb isolation. Glycans and glycoconjugates are almost always a heterogeneous mixture of structures when isolated from natural sources (Raman et al., 2005), which dilutes any potential antigenic response. Total chemical synthesis and chemoenzymatic synthesis can often yield more uniform glycotopes (Wang and Lomino, 2012), however, these techniques are labor intensive, difficult to scale, and exist predominantly in the laboratories of a handful of experts. Second, glycans alone usually elicit weaker T-cell independent immune responses, which are short-lived and lack IgM-to-IgG class switching (Avci and Kasper, 2010).

A common strategy for enhancing the immunogenicity of carbohydrates is to covalently couple a glycan to a T-cell dependent antigen. For example, conjugates composed of bacterial CPS or LPS chemically bound to an immunogenic carrier protein induce high-affinity, class-switched mAbs (Astronomo and Burton, 2010; Avci and Kasper, 2010). Unfortunately, production of traditional conjugate vaccines is a complex, multistep process that is expensive, time consuming, and low yielding (Frasch, 2009). A simplified alternative for generating glycoconjugates known as protein glycan coupling technology (PGCT) has been described recently (Cuccui and Wren, 2014; Terra et al., 2012). This approach leverages laboratory strains of Escherichia coli for the expression of recombinant bacterial polysaccharides (e.g., O-polysaccharide antigens), which are conjugated in vivo to a co-expressed carrier protein by the Campylobacter jejuni oligosaccharyltransferase PglB. However, while PGCT has been used to make several novel protein/glycan combinations, it is limited by variable glycan conjugation efficiency as observed for certain heterologous polysaccharide substrates (Cuccui et al., 2013; Ihssen et al., 2015; Ihssen et al., 2010) and a challenging purification of the product antigen. This is particularly pertinent in the context of producing glycoconjugates carrying mammalian-like glycans (Cuccui and Wren, 2014).

Here, we sought to develop an efficient method for generating class-switched, anti-glycan Abs that overcomes many of the challenges discussed above. To this end, our approach combined custom glycan biosynthesis with outer membrane vesicle (OMV) formation in laboratory strains of E. coli. OMVs are naturally occurring nanospherical structures (~20–250 nm) produced constitutively by all Gram-negative bacteria. They are composed of proteins, lipids, and glycans derived from the outer membrane and periplasm, and have natural adjuvant properties that strongly stimulate the innate, and more importantly, the adaptive immune response (Alaniz et al., 2007; Baker et al., 2014; Ellis et al., 2010). To expand the immunostimulatory potential of OMVs, genetic engineering techniques have been used to load OMVs with foreign protein antigens by targeting expression to the outer membrane or to the periplasm of an OMV-producing host strain (Chen et al., 2010; Muralinath et al., 2011). These OMV-associated recombinant proteins elicited strong and specific antibody responses following immunization in mice. Building on these earlier observations, we engineered hypervesiculating strains of E. coli (Bernadac et al., 1998) to produce OMVs that displayed foreign glycans on their exteriors. This involved creation of two heterologous pathways for biosynthesis of structural mimics of clinically important carbohydrates, namely poly-α2,8-N-acetyl neuraminic acid (polysialic acid or PSA) and Galβ1-3GalNAcα1 (Thomsen-Friedenreich antigen or T antigen). The resulting glycosylated OMVs (glycOMVs), whose surfaces were remodeled with the custom-designed PSA or T antigen epitopes, induced strong glycan-specific IgG antibody titers following immunization in BALB/c mice. Taken together, our results show that engineered glycOMVs represent an effective strategy for generating functional Abs against structurally defined glycotopes of biomedical importance.

Results

A bottom-up engineered pathway for biosynthesis of T antigen on the surface of OMVs

T antigen is one of many ‘self’ antigens expressed on a variety of malignancies including breast, colon, prostate, and stomach cancer, and Abs recognizing T antigen could have clinical benefit (Astronomo and Burton, 2010; Heimburg-Molinaro et al., 2011). However, the low intrinsic immunogenicity of T antigen poses a barrier to vaccination even after conjugation to a carrier protein (Adluri et al., 1995). Here, we hypothesized that the adjuvanticity of OMVs could be leveraged to overcome the weak immunogenicity of the T antigen epitope. Since LPS is a major component of released OMVs (Baker et al., 2014) and can be engineered to display foreign glycans (Ilg et al., 2010; Valderrama-Rincon et al., 2012), we attempted to remodel the carbohydrate component of LPS with T antigen-containing glycans. Such remodeling involved the lipid carrier undecaprenylpyrophosphate (Und-PP) as an acceptor of engineered glycans, which are flipped to the periplasmic side of the inner membrane and subsequently transferred to lipid A-core by the O-polysaccharide antigen ligase WaaL (Fig. 1a). In most laboratory strains of E. coli, Und-PP is primed with N-acetylglucosamine (GlcNAc) by the enzyme WecA. To elaborate the native Und-PP-GlcNAc with Galβ1-3GalNAc, we expressed two heterologous glycosyltransferases (GTases): the α1,3-GalNAc-transferase (PglA) from Campylobacter jejuni for transfer of GalNAc to Und-PP-GlcNAc (Glover et al., 2005); and the β1,3-galactosyltransferase (WbnJ) from E. coli O86 for stereospecific addition of the terminal galactose residue (Yi et al., 2005). Additionally, the UDP-GlcNAc 4 epimerase (Gne) from the same locus as C. jejuni PglA was added to supply the requisite UDP-GalNAc (Bernatchez et al., 2005). Formation of the T antigen epitope on inner membrane lipids was assessed by introducing plasmid pTF in E. coli K12 strain MC4100 lacking the waaL gene, which causes accumulation of UndPP-linked glycans in the cytoplasmic membrane. Lipid-linked oligosaccharides (LLOs) were extracted from these cells, and the glycan portion was released and purified as previously described (Valderrama-Rincon et al., 2012). MALDI-TOF mass spectrometry (MS) analysis of these glycans identified a major peak consistent with the expected Gal-terminal T antigen structure (m/z = 609) (Fig. 1b). Following treatment of the isolated glycans with β1,3-galactosidase, MS analysis revealed a major peak (m/z = 447) consistent with the removal of a single hexose residue (Fig. 1b), thereby corroborating the linkage of a terminal β1,3 Gal residue.

Figure 1. Expression of recombinant T antigen glycotope on OMVs.

(a) Schematic of recombinant T antigen biosynthesis, which begins with glycan assembly on endogenous Und-PP-GlcNAc structure in the inner membrane and involves the coordinated action of heterologously expressed GalE epimerase and the glycosyltransferases PglA and WbnJ (shown in red). Trisaccharide glycan is flipped into the periplasmic space by Wzx where it is subsequently transferred onto lipid A core by WaaL and translocated to the outer membrane. (b) MALDI-MS profile of glycans released from LLOs by acid hydrolysis. LLOs were extracted from E. coli MC4100 ΔwaaL::kan cells carrying plasmid pTF. Glyans released from LLOs were dried and resuspended in dH₂O. The major signal at m/z 609 corresponded to HexHexNAc2. Following digestion of glycans with β-1,3-galactosidase at 37°C, the major signal at m/z 447 corresponded to HexNAc2. (c) Dot blot analysis of OMV fractions, generated from plasmid-free JC8031 cells (empty), JC8031 cells carrying pTF, and JC8032 cells, which lacked waaL, carrying pTF. Blots were probed with peanut agglutinin to confirm presence or absence of T antigen on exterior of OMVs. Fetuin and asialofetuin served as negative and positive controls, respectively.

Next, to determine whether the recombinant T antigen could be displayed on the exterior of OMVs, the hypervesiculating E. coli K12 strain JC8031, which overproduces OMVs due to deletion of tolRA (Bernadac et al., 1998), was transformed with the pTF plasmid. OMVs isolated from these cells were subjected to dot blot analysis whereby intact OMVs were spotted directly onto nitrocellulose membranes without any denaturation steps, and membranes were probed with peanut agglutinin (PNA), a lectin that binds the Galβ1-3GalNAc structure of the T antigen (Lotan et al., 1975). Consistent with the observation that outer membrane glycolipids are a major component of OMVs (Baker et al., 2014), we observed a strong signal from the non-denatured OMV fraction derived from JC8031 cells carrying pTF (Fig. 1c).

To confirm that this signal was due to incorporation of the recombinant T antigen into LPS structures, we analyzed the OMV fraction from a knockout mutant of JC8031 that lacked waaL (hereafter JC8032), which encodes the O-polysaccharide antigen ligase responsible for transferring engineered Und-PP-linked glycans to lipid A-core (Feldman et al., 2005). As expected, the formation of T antigen on OMVs was blocked in JC8032 cells (Fig. 1c), confirming that display of engineered glycotopes on lipid A-core involved WaaL-dependent assembly.

The incorporation of foreign glycotopes into E. coli LPS structures had no visible effect on vesicle nanostructure. For example, the spherical bilayered shape of T antigen-containing OMVs was indistinguishable from control OMVs as evidenced by transmission electron microscopy (TEM) microscopy (Supplementary Fig. 1a). Likewise, analysis by dynamic light scattering (DLS) revealed that the majority of the purified vesicles had a diameter of 20–60 nm (Supplementary Fig. 1b), consistent with the size of E. coli-derived OMVs that were characterized previously (Park et al., 2010). To determine whether recombinant T antigen detected in the pelleted supernatant was associated with intact vesicles, rather than with released outer membrane fragments or other cellular debris, the OMV-containing fraction isolated from JC8031 cells carrying pTF was separated by density gradient ultracentrifugation. Coomassie staining and Western blotting of the resulting fractions revealed that total OMV proteins, the outer membrane protein OmpA, and recombinant T antigen all co-migrated to denser fractions (Supplementary Fig. 2a–c), reminiscent of the gradient profiles seen previously for intact OMVs and OMV-associated proteins (Kim et al., 2008).

A bottom-up engineered pathway for biosynthesis of PSA on the surface of OMVs

PSA is a CPS that coats the surface of MenB and E. coli K1, and is also expressed in human tissues, most notably on neural cell adhesion molecule (NCAM) (Moe et al., 2009). As a result of this latter point, PSA has proven to be a particularly difficult target for antibody generation. Indeed, PSA is poorly immunogenic even when conjugated to a carrier protein (Krug et al., 2004), possibly because of the similarity to self-antigens. To overcome this barrier, we engineered a hypervesiculating E. coli K12 strain to produce OMVs decorated with recombinant PSA glycans. Since E. coli K12 strains do not produce PSA naturally, this first required the creation of an artificial pathway for PSA biosynthesis. To create the core onto which PSA could polymerize, we generated plasmid pPSA, which enabled heterologous expression of the GTases LgtB from Neisseria gonorrhoeae and CstII from Campylobacter jejuni. These GTases were predicted to catalyze the successive transfer of galactose and sialic acid (N-acetyl neuraminic acid; NeuNAc), respectively, to the Und-PP acceptor. Plasmid pPSA also encoded the neuBACS genes from E. coli K1, which collectively coordinate the formation of precursor CMP-NeuNAc and polymerization of NeuNAc (Fig. 2a). Finally, the neuD gene from E. coli K1, which promotes efficient sialic acid synthesis by enhancing the activity of other proteins (e.g., NeuBAC) in the sialic acid pathway (Daines et al., 2000), was cloned on a separate plasmid named pNeuD. These two plasmids were introduced into a ΔnanA derivative of the hypervesiculating strain JC8031 (hereafter JC8033), which is unable to catabolize free NeuNAc due to absence of the N-acetylneuraminate lyase enzyme, NanA (Priem et al., 2002).

Figure 2. Expression of recombinant PSA glycotope on OMVs.

(a) Schematic of PSA biosynthesis pathway, which involves NeuABCD for the formation of CMP-NeuNAc and NeuS for its polymerization into PSA. (b) Positive-ion MALDI-TOF spectrum of permethylated PSA. Glycolipids were extracted from JC8033 cells carrying plasmids pPSA and pNeuD. The major signal at m/z 791.4, with additional significant peaks at m/z 1152.6, 1513.7, and 1874.9 corresponding to tri, tetra, and penta NeuNAc structures. The predominance of the dimer seen here is consistent with preliminary NMR analysis of the same material. (c) Dot blot analysis of non-denatured OMV fractions generated from plasmid-free JC8033 cells (−) or JC8033 cells carrying either pNeuD, pPSA, or pPSA/pNeuD together. Also shown are OMV fractions from JC8033 cells carrying PSA/pNeuD where the pPSA plasmid lacked cstII (pPΔC), lgtB (pPΔL), or both (pPΔCL). Strain EV36 served as a positive control. (d) Dot blot analysis of non-denatured OMV fractions generated from plasmid-free JC8033 cells (−) or the following strains each carrying pPSA/pNeuD: JC8033; JC8034, which lacked waaL; and JC8035, which lacked wecA. Also shown are OMV fractions from hypervesiculating versions of MG1655 and ClearColi (MG1655-ves and ClearColi-ves, respectively) carrying pPSA/pNeuD. Blots were probed with SEAM 12 antibody to confirm the presence or absence of PSA on exterior of OMVs.

LLOs were extracted from JC8033 cells carrying pPSA and pNeuD, and the glycan portion was purified and permethylated. Positive-ion mode MALDI-TOF MS of the permethylated glycans identified a major peak (m/z = 791.4) corresponding to a NeuNAc disaccharide, as well as minor peaks corresponding to tri-, tetra-, and pentasaccharides of NeuNAc (Fig. 2b). A minor peak consistent with a NeuNAc-NeuGc structure was also identified. To determine whether PSA was incorporated in OMVs derived from these cells, membrane vesicles were isolated and subjected to dot blot analysis using SEAM 12, a murine Ab that is cross-reactive with PSA and exhibits potent complement-mediated bactericidal activity against MenB (Granoff et al., 1998). As with the engineered T antigen, we observed a strong signal from the non-denatured OMV fraction derived from JC8033 cells carrying both pPSA and pNeuD (Fig. 2c). This signal was comparable to that obtained by similarly probing intact EV36 cells, a K-12/K1 hybrid E. coli strain that natively expresses PSA on its surface (Fig. 2c) (Vimr et al., 1989). In contrast, no detectable signal was observed from OMV fractions derived from JC8033 cells without any plasmids or carrying either the pPSA and pNeuD plasmid individually (Fig. 2c), indicating that NeuD was required for engineered PSA biosynthesis. Likewise, the absence of LgtB or CstII, or both, resulted in a similar lack of signal in dot blots probed with the SEAM 12 antibody (Fig. 2c). It is noteworthy that nearly identical signals were observed using a commercial anti-polysialic acid-NCAM antibody (Millipore) that recognizes α2,8-linked PSA.

Density gradient ultracentrifugation of the OMV fraction from JC8033 carrying pPSA and pNeuD revealed that PSA co-migrated with total OMV proteins and OmpA (Supplementary Fig. 2d–f), and thus appeared to be associated with intact vesicles. It is also noteworthy that incorporation of foreign PSA glycan into E. coli LPS structures had no visible effect on vesicle nanostructure (Supplementary Fig. 1a), with vesicle size again ranging from 20–60 nm in diameter (Supplementary Fig. 1b).

To confirm whether PSA was produced on the Und-PP acceptor, we analyzed the OMV fraction from the waaL knockout mutant of JC8033 (hereafter JC8034) carrying the pPSA and pNeuD plasmids. Unlike the case of T antigen biosynthesis above, the display of PSA on the exterior of OMVs was not dependent on WaaL (Fig. 2d). Likewise, PSA display was still observed in JC8033 cells lacking wecA (hereafter JC8035), which transfers GlcNAc to Und-PP and forms the hypothesized Und-PP-GlcNAc acceptor for LgtB (Fig. 2d). In light of these results, we hypothesized an alternative mechanism for incorporation of recombinant PSA into LPS structures involving direct conjugation of foreign saccharides to lipid A-core structures (Ilg et al., 2010). The basis for this hypothesis stems from the following: in E. coli K-12 strains, lipid A-core contains glucose residues that might serve as substrates for heterologously expressed LgtB, a promiscuous biocatalyst that can transfer galactose to a variety of different glucose- and glucosamine-containing acceptors (Blixt et al., 2001). To test this hypothesis, we generated a hypervesiculating derivative of E. coli ClearColi, a K-12 strain that produces truncated LPS structures, called lipid IV_A, that lack saccharide acceptors for our heterologously expressed GTases (e.g., LgtB) (Mamat et al., 2008). This was accomplished by deleting the nlpI gene that is known to increase vesiculation on par with tolRA mutants (Kim et al., 2008; McBroom et al., 2006), resulting in strain ClearColi-ves. OMVs produced from ClearColi-ves cells carrying pPSA and pNeuD were indeed blocked for PSA display (Fig. 2d). In contrast, OMVs derived from the parental strain MG1655, also lacking nlpI (MG1655-ves) and carrying the pPSA and pNeuD plasmids, were decorated with PSA at a level that rivaled JC8033 carrying the same PSA pathway plasmids (Fig. 2d). These results support the notion that PSA was incorporated in LPS structures by direct conjugation to saccharides in lipid A-core.

Immunization with glycOMVs elicits glycan-specific antibodies

We next sought to assess the immunological potential of glycOMVs displaying the T antigen and PSA epitopes. Specifically, BALB/c mice were immunized via subcutaneous (s.c.) injection with either T antigen- or PSA-containing glycOMVs, after which blood was collected at regular intervals. Controls included ‘empty’ OMVs from plasmid-free JC8031 or JC8033 cells, LOS extracted from MenB strain S3446 (NmBLOS), and phosphate buffered saline (PBS). To determine whether glycOMVs generated glycan-specific Abs, the total T antigen- and PSA-specific IgG titers at the endpoint were measured by ELISA using the model glycoprotein carrier protein scFv13-R4 with a C-terminal glycosylation motif (Valderrama-Rincon et al., 2012) bearing the T antigen or native LOS from MenB, respectively, as immobilized antigen. In the case of the T antigen epitope, glycOMVs elicited a significantly higher (p < 0.01) level of glycan-specific IgGs compared to both the empty OMV and PBS control groups (Fig. 3a). Similarly, the total PSA-specific IgG titers were significantly increased (p < 0.01) for the group immunized with PSA glycOMVs compared to all other immunized groups (Fig. 3b). It is particularly noteworthy that the IgG titers for the group immunized with native MenB LOS were not significantly different (p > 0.2) than those measured in the PBS group, consistent with the weak immunogenicity of glycans alone. Hence, the immunogenicity of engineered carbohydrates was boosted by display on the exterior of OMVs. IgG titers were further broken down by analysis of IgG1 and IgG2a titers, wherein mean IgG1 to IgG2a antibody ratios served as an indicator of a Th1- or Th2-biased immune response. Mice immunized with T antigen and PSA glycOMVs showed a significant (p < 0.05) increase in mean titers of glycan-specific IgG1 and IgG2a in comparison to all other groups (Supplementary Fig. 3a and b). The similar levels observed for IgG1 versus IgG2a titers suggested no measurable Th1/Th2 bias.

Figure 3. GlycOMVs boost production of glycan-specific IgG antibodies.

Median antigen-specific IgG titers of individual mice immunized with (a) T antigen glycOMVs or (b) PSA glycOMVs in endpoint (day 54 and 84, respectively) serum. For the T antigen epitope, three groups of BALB/c mice were immunized s.c. with: 10 μg OMVs from plasmid-free JC8031 cells (empty), 10 μg OMVs from JC8031 cells carrying pTF (T antigen glycOMVs); or PBS. Glycosylated scFv13-R4 bearing T antigen was used as immobilized antigen. For the PSA epitope, four groups of BALB/c mice immunized s.c. with: 10 μg OMVs from plasmid-free JC8031 ΔnanA cells (empty); 10 μg OMVs from JC8031 ΔnanA cells carrying pPSA and pNeuD (PSA glycOMVs); 2 μg MenB LOS; or PBS. NmBLOS was used as immobilized antigen. Mice were boosted on day 28 and 56 with same doses. Asterisk (*) represents statistical significance (p < 0.01; Tukey-Kramer Post-Hoc HSD) versus all other groups.

T antigen-specific antibodies detect target antigen in Western blot format

To determine the diagnostic potential of these glycan-specific Abs, we performed Western blot analysis using sera generated through glycOMV immunization. As expected, Abs generated by immunization with T antigen glycOMVs cross-reacted exclusively with the model carrier protein scFv13-R4 bearing the T antigen, generating a signal that was on par with that obtained using PNA lectin (Fig. 4a). In contrast, when the membrane was probed with Abs generated by immunization with empty OMVs, there was no visible binding to either glycosylated or aglycosylated scFv13-R4 (Fig 4a).

Figure 4. Diagnostic and therapeutic potential of glycan-specific antibodies.

(a) Western blot analysis of scFv13-R4 glycosylated with T antigen (+) or aglycosylated scFv13-R4 (−). Blots were probed with PNA, polyclonal sera from groups immunized with T antigen glycOMVs or empty OMVs, or anti-His-HRP antibody. (b) Representative killing activity of antibodies in the serum of mice immunized with: PBS, empty OMVs, and PSA glycOMVs. Survival data is derived from standard serum bactericidal assay (SBA) where dilutions of serum from immunized mice were tested against MenB strain H44/76 in the presence of human complement. Murine antibodies against MenB (SEAM 12) and MenC (anti-MenC) served as positive and negative controls, respectively. The SBA curves for PBS and empty OMVs are representative of all mice in the group (n = 6) and three of six PSA glycOMV that had the highest response to the vaccine.

PSA-specific antibodies exhibit complement-mediated bactericidal activity

We next investigated whether the serum Abs produced by glycOMV immunization were immunologically relevant. For this, we performed a complement-mediated serum bactericidal activity (SBA) assay using the sera collected from mice immunized with PSA glycOMVs. SBA is an established method by which the activity of Abs against N. meningitidis is measured, and it correlates with protection for all serogroups of the pathogen (Martin et al., 2005). Here, we hypothesized that PSA-specific Abs generated by glycOMV immunization would bind to capsular PSA on the surface of MenB and, in the presence of components of the human complement system, would mediate bacteriolysis of the pathogen. In the group immunized with PSA glycOMVs, 50% SBA was observed at over 100-fold dilutions of the serum, a level that was on par with the anti-MenB antibody, SEAM 12 (Fig. 4b). In contrast, no killing was observed for sera collected from any of the control groups, or for the control anti-MenC antibody, over the dilutions tested (Fig. 4b). Complete killing was observed in immunized groups at dilutions as high as 10-fold, indicating that the Abs present in serum from glycOMV-immunized mice were immunologically functional.

Discussion

We have developed a new approach for generating class-switched, anti-glycan Abs that leverages the immunostimulatory properties of OMVs (Alaniz et al., 2007; Chen et al., 2010; Ellis et al., 2010; Sanders and Feavers, 2011) to boost the immune response to glycan epitopes, which are notoriously weak antigens (Astronomo and Burton, 2010; Avci and Kasper, 2010). An important first step involved converting laboratory strains of E. coli into factories for glycosylated OMV production by combining bacterial vesiculation with engineered pathways for designer glycan biosynthesis. We anticipate that this strategy could be generalized to create many other structurally diverse and biomedically relevant glycotopes on the exterior of OMVs for both diagnostic and therapeutic applications.

It is worth noting that compared to the expression of protein antigens in OMVs, expression of glycans is a more elaborate undertaking. For protein antigens, expression in OMVs simply requires targeting the antigen of interest either to the periplasmic space by genetic fusion of an N-terminal export signal or to the cell surface by genetic fusion to an outer membrane carrier protein (Baker et al., 2014). Following vesiculation, the periplasmic- or outer membrane-targeted proteins become constituents of the OMV lumen or exterior, respectively. In contrast, display of carbohydrate antigens on OMVs requires the coordinated expression of multiple heterologous glycosyltransferases for directing the synthesis of desired glycans onto bacterial lipid carriers that subsequently localize to the outer membrane and become constituents of released OMVs. Despite the challenges, several groups including ours have used glycoengineering as a tool to remodel the bacterial outer membrane with mammalian glycotopes of interest including ganglioside GM3 (Ilg et al., 2010), Lewis Y (Le^Y) antigen (Yavuz et al., 2011), and trimannosyl core N-glycan (Valderrama-Rincon et al., 2012). Presumably, expression of these different cell surface glycans in a hypervesiculating host strain would yield uniquely glycosylated OMVs, although this remains to be shown.

Importantly, once new glycan structures are created, however, production of glycOMV immunogens is significantly less complicated, less time consuming, less expensive, and more scalable than conventional approaches for producing glycoconjugates. It requires only one cultivation step to generate the final product, which can be easily and economically isolated by a single ultracentrifugation step (Chen et al., 2010). Moreover, the clinical translation of OMVs has also been established recently by Bexsero, a four component vaccine against N. meningitidis serogroup B that has been approved in the U.S. for preventing infection caused by this serogroup. The active components of this vaccine are three recombinant proteins identified by reverse vaccinology combined with detergent extracted OMVs prepared from a naturally occurring epidemic strain (Gorringe and Pajon, 2012). While engineered OMVs such as the ones we described herein have not yet found their way into the clinic, Bexsero represents a first important step in that direction.

The ability of T antigen- and PSA-modified glycOMVs to elicit class-switched, glycan-specific IgGs in mice is significant in light of the low intrinsic immunogenicity that has been observed for these glycans, even when conjugated to a carrier protein (Adluri et al., 1995; Krug et al., 2004). The poor immunogenicity of these glycans has been attributed to immunologic tolerance that arises due to their resemblance with structures present in human and murine hosts (i.e., self antigens). Hence, our observation that glycOMVs triggered high titers of class-switched IgGs represents a significant advance in the pursuit of Abs against glycotopes of interest. While a molecular-level understanding of how OMVs boost the immune response to these glycans remains to be determined, we suspect that it stems from the potent adjuvanticity afforded by OMVs which: (i) are readily phagocytosed by professional antigen-presenting cells; (ii) carry pathogen-associated molecular patterns (PAMPs) within their structure that can stimulate both innate and adaptive immunity; and (iii) possess strong proinflammatory properties (Alaniz et al., 2007; Ellis et al., 2010; Sanders and Feavers, 2011).

The fact that PSA and T antigen are self antigens of humans and a potential cause of immunopathology has hindered their development as vaccines. However, there are reasons to believe that many tumor-specific carbohydrate antigens including T antigen and PSA have a number of attributes that make them viable targets for vaccine development, most notably their widespread and high expression in several different cancers and their low or cryptic expression on normal cells (Astronomo and Burton, 2010; Heimburg-Molinaro et al., 2011). Interestingly, in the case of PSA, Miller and colleagues recently published an essay in which they reviewed the data on PSA as a self antigen and concluded that (2→8)-α-Neu5Ac conjugates will be as safe and effective as the polysaccharide protein conjugate vaccines for the other four meningococcal serogroups (Robbins et al., 2011). Aside from the potential immunotherapeutic applications enabled by glycOMVs, their ability to elicit glycan-specific Abs can be exploited to produce high-affinity reagents for glycobiology and glycomedicine. Currently, carbohydrate-binding lectins are the primary means to detect glycans in numerous analytical assays; however, lectins are limited by their poor sensitivity and binding affinity as well as lack of specificity towards less common glycan structures (Haab, 2012). Our demonstration that polyclonal serum from immunization with T antigen glycOMVs can be used for immunodetection of glycans illuminates the diagnostic potential of serum Abs elicited by glycOMVs and ensures that these Abs will find use even in cases where vaccine-induced autoimmunity proves to be an insurmountable obstacle.

The potent bactericidal activity of the PSA glycOMV-stimulated serum Abs against N. meningitidis serogroup B in the presence of human complement further confirmed the authenticity of the engineered glycotope mimics as well as the full functionality of the Abs they elicited. To our surprise, vaccinations using empty OMV controls elicited measurably higher IgG titers in mice as compared to titers measured for the PBS control mice; however, none of these serum Abs were bactericidal. It should also be noted that the PSA-specific IgG titers for the empty OMV-immunized groups were still significantly (p < 0.01) less than those generated from the PSA glycOMV-immunized groups. Nonetheless, we attribute these unexpected ELISA signals to the presence of serum Abs against other components in the OMVs, which cross-reacted with similar components present in the NmBLOS. Indeed, when ELISA plates were instead coated with a synthetic PSA-ADH derivative (Granoff et al., 1998), the signal from the empty OMV group was notably lower.

Overall, the results of this study reveal that glycOMVs are a reliable and robust method for generating class-switched Ab responses to glycan epitopes of interest. Compared to current approaches for achieving the same goal, glycOMVs represent a solution that is considerably less complicated and significantly more scalable. By rewiring the glycan biosynthetic pathway, it should be possible to generate glycOMVs displaying a wide array of biomedically relevant glycotopes found on the surfaces of bacteria and human cells, as we demonstrated here with T antigen and PSA. Moreover, the ability of OMVs to overcome immunologic tolerance by eliciting strong immune responses to glycans characterized as self antigens should further expand the palette of glycans that can be targeted by this approach. There also exist opportunities to couple glycOMVs with emerging techniques for Ab discovery such as immune repertoire mining (Lavinder et al., 2015), which could provide unprecedented access to a renewable source of high-quality, glycan-binding affinity reagents for interrogating the glycomes of living organisms or treating human disease. And since immunization with PSA glycOMVs yielded Abs that were fully functional (i.e., bactericidal), it stands to reason that glycOMVs themselves might eventually find use in a therapeutic context.

Experimental Procedures

Bacterial strains and plasmids

A description of all bacterial strains and plasmids used in this study, including those that were constructed herein, is provided in the Supplementary Methods, along with a complete list in Supplementary Table 1. Briefly, unless otherwise stated, most strains used herein are based on E. coli strain JC8031, a tolRA mutant strain that is known to hypervesiculate (kindly provided by Roland Lloubes, Centre national de la Recherche Scientifique) (Bernadac et al., 1998).

Cell growth and OMV preparation

OMVs were prepared as described previously (Chen et al., 2010). Briefly, cells were freshly transformed with plasmids for T antigen or PSA biosynthesis and selected on medium supplemented with the appropriate antibiotic. An overnight culture of a single colony was subcultured into 100–200 mL of Luria-Bertani (LB) medium. The culture was grown to mid-log phase, at which time protein expression was induced with L-arabinose (0.2%) and/or IPTG (0.1 mM), if necessary. Cell-free culture supernatants were collected 16–20 h post-induction and filtered through a 0.2 μm filter. Vesicles were isolated by ultracentrifugation (Beckman-Coulter; TiSW28 rotor; 141,000xg; 3 h; 4°C) and resuspended in PBS. OMVs were quantified by the bicinchoninic-acid assay (BCA Protein Assay; Pierce) using BSA as the protein standard.

OMVs were further separated by density-gradient ultracentrifugation as previously described (Kim et al., 2008). Briefly, OMVs were prepared as described above but resuspended in a 50 mM HEPES (pH 6.8) solution. This solution was adjusted to 45% (v/v) Optiprep (Sigma) in 1.5 mL. All other Optiprep layers were prepared using the same 50-mM HEPES (pH 6.8) solution. Optiprep/HEPES gradient layers were added to a 12-mL ultracentrifuge tube as follows: 0.33 mL of 10%, 0.33 mL of 15%, 0.66 mL of 20%, 0.66 mL of 25%, 0.9 mL of 30%, 0.9 mL of 35%, 1.5 mL of 45% containing the prepared OMVs, and enough 60% to nearly fill the tube. Gradients were centrifuged (Beckman-Coulter; TiSW41 rotor; 180,000xg; 3 h; 4°C), then, a total of ten fractions of 0.5 mL each were removed sequentially from the top of the gradient. These fractions were analyzed by Western blot and dot blot analyses as described below.

Glycoprotein expression and purification

E. coli MC4100 ΔwaaL::kan was co-transformed with plasmid pTrc99A-ssDsbA-scFv13-R4^DQNAT (Valderrama-Rincon et al., 2012) and either pTF or empty vector control pMW07. An overnight culture was used to inoculate 100 mL of LB medium containing ampicillin and chloramphenicol. Cultures were grown to an OD₆₀₀ of ~ 2.0, and induced overnight with 0.2% arabinose and 0.1 mM IPTG. Cells were harvested after which glycosylated or aglycosylated scFv13-R4 proteins were purified using Ni-NTA spin columns (Qiagen) according to manufacturer’s instructions. Proteins were buffer exchanged to PBS and the concentration adjusted to 1 mg/mL.

Western blot and dot blot analysis

OMV and LPS samples were prepared for SDS-PAGE analysis by boiling for 15 min and cooling to room temperature in the presence of loading buffer containing β-mercaptoethanol. OMV samples were normalized by total protein concentration, which was quantified using the BCA method as detailed above. Samples were run on Any kD polyacrylamide gels (BioRad, Mini-PROTEAN® TGX) and transferred to a PVDF membrane. After blocking with Carbo-Free blocking solution (Vector Labs), membranes with T antigen samples were probed first with biotinylated peanut agglutinin (Vector Labs) and then with streptavidin-HRP (Abcam). Similarly, after blocking with a 5% milk solution, membranes with PSA samples were probed first with SEAM 12 primary antibody specific against N. meningitidis B CPS (Granoff et al., 1998) and then with the corresponding anti-mouse HRP-conjugated secondary antibody (Promega). Membranes from density-gradient samples were also probed with an OmpA-specific antibody (kindly provided by Wilfred Chen, University of Delaware) and then with anti-mouse HRP-conjugated secondary antibody (Promega). Signal was visualized using HRP substrate and either an X-ray film developer or a ChemiDoc Imaging System (BioRad).

Western blot analysis of glycosylated and aglycosylated scFv13-R4 was performed according to standard protocols. Briefly, protein-containing samples were separated by SDS-PAGE and transferred to PVDF membranes. Membranes were then probed with either: biotinylated PNA (Vector labs) and then with streptavidin-HRP (Vector Labs); immunized sera (from either empty OMV or T antigen glycOMV groups) at a concentration of 1:1000, followed by anti-mouse IgG HRP (VWR); or anti-His-HRP (Sigma). Signal was visualized using HRP substrate and imaged using a ChemiDoc Imaging System (BioRad).

For dot blot analysis, OMV samples were normalized by total protein concentration, which was quantified using the BCA method as detailed above, and spotted directly onto a nitrocellulose membrane. Alternatively, OMVs were boiled for 10 min and cooled to room temperature prior to spotting on the membrane. After blocking with a 5% milk solution, membranes with PSA samples were probed first with SEAM 12 and then with HRP-conjugated anti-mouse IgG. For the T antigen, membranes were blocked with Carbo-Free blocking solution (Vector Labs), and then probed with biotinylated peanut agglutinin (Vector Labs) and subsequently streptavidin-HRP (Abcam). Signal was visualized using HRP substrate and either an X-ray film developer or a ChemiDoc Imaging System (BioRad).

Electron microscopy

Structural analysis of vesicles was performed via transmission electron microscopy as previously described (Chen et al., 2010). Briefly, vesicles were negatively stained with 2% uranyl acetate and deposited on 400-mesh Formvar carbon-coated copper grids. Imaging was performed using a FEI Tecnai F20 transmission electron microscope.

MALDI-TOF MS analysis

For structural characterization of T antigen, LLOs were extracted from MC4100 ΔwaaL::kan cells carrying pTF. An overnight culture of a single colony was subcultured into LB medium. Cultures were grown at 30°C and induced when optical density at 600 nm (OD₆₀₀) reached ~2.0. Cultures were then harvested after 20 h. Cell pellets were collected, resuspended in methanol and lysed via sonication. Material was then dried at 60°C and subsequently resuspended in 2:1 chloroform:methanol solution (v/v, CM) via sonication and washed two times with the CM solution. The pellet was then washed in water. Lipids were extracted with 10:10:3 chloroform:methanol:water (v/v/v, CMW) followed by methanol. Extracts were then loaded into a DEAE cellulose column and eluted with 300 mM NH₄OAc in CMW. The LLOs were extracted with chloroform and dried. Glycans released from LLOs were dried and resuspended in dH₂O. 10-μL β-1,3-galactosidase (NEB) reactions were prepared using supplied buffer and incubated at 37°C. Reaction products were monitored by MALDI/TOF-MS. For structural characterization of PSA, the carbohydrate was isolated from the cells following a published procedure (Willis et al., 2013), except that the cells were disrupted by French press and that the gel-filtration step was omitted. The yield of carbohydrate was about 1 mg. Next, the samples were permethylated as described (Anumula and Taylor, 1992) and analyzed by MALDI/TOF-MS in reflector positive-ion mode on a ABISciex 5800 MALDI/TOF-TOF using α-dihyroxybenzoic acid (DHBA, 20 mg/mL solution in 50% methanol:water) as matrix.

Mouse immunizations

Three groups of four or five BALB/c female mice aged six- to eight-weeks old (The Jackson Laboratory) were each immunized s.c. with either PBS alone (control) or 100 μL of PBS containing: 10 μg of OMVs from JC8031 cells carrying no plasmid (empty OMVs) or 10 μg of OMVs from JC8031 cells harboring pTF (T antigen glycOMVs). Separately, four groups of five or six BALB/c female mice aged six- to eight-weeks old (The Jackson Laboratory) were each immunized s.c. with either PBS alone (control) or 100 μL of PBS containing: 2 μg of native LOS from N. meningitidis serogroup B (NmBLOS), 10 μg of OMVs from JC8033 cells carrying no plasmid (empty OMVs), or 10 μg of OMVs from JC8033 cells harboring pPSA and pNeuD (PSA glycOMVs). NmBLOS was prepared from MenB strain S3446 identically as described previously (Apicella et al., 1997). PSA content of the PSA glycOMV and NmBLOS doses were similar and determined via reactivity to the SEAM 12 antibody (Granoff et al., 1998). All PBS used was at pH 7.4. Each group of mice was boosted with an identical dosage of antigen 28 days and 56 days after the priming dose. Blood was collected from each mouse from the mandibular sinus immediately before and 14 days after the first immunization, immediately before and 14 days after the first boosting dose, immediately before the second boosting dose, and at 14 days and 28 days after the second boosting dose. The protocol number for the animal studies (2009-0096) was approved by the Institutional Animal Care and Use Committee at Cornell University. Statistical analysis was performed using one-way analysis of variance (ANOVA) followed by the post-hoc Tukey-Kramer test for multiple comparisons.

Enzyme-linked immunosorbant assay (ELISA)

Glycan-specific Abs produced in immunized mice were measured via indirect ELISA using a modification of a previously described protocol (Chen et al., 2010). Briefly, sera were isolated from the collected blood draws after centrifugation at 2,200xg for 10 min. 96-well plates (Maxisorp; Nunc Nalgene) were coated with E. coli-derived LPS containing T antigen or NmBLOS (2 μg/mL in PBS pH 7.4) and incubated overnight at 4°C. For comparison, ELISAs were also performed using a synthetic PSA-adipic acid dihydrazide (ADH) derivative, which was prepared as described (Granoff et al., 1998) and used to coat microtiter plates at a concentration of 20 μg/ml in PBS (pH 7.4). The next day, plates were washed 3 times with PBST (PBS, 0.05% Tween-20, 0.3% BSA) and blocked overnight at 4°C with 5% nonfat dry milk (Carnation) in PBS. Samples were serially diluted, in triplicate, between 1:100-1:12,800,000 in blocking buffer and added to the plate for 2 h at 37°C. Plates were washed 3 times with PBST and incubated for 1 h at 37°C in the presence of one of the following horseradish peroxidase-conjugated Abs: goat anti-mouse IgG (1:5000; Abcam), anti-mouse IgG1 (1:5000; Abcam), or anti-mouse IgG2a (1:5000; Abcam). After 3 additional washes with PBST, 3,3′-5,5′-tetramethylbenzidine substrate (1-Step Ultra TMB-ELISA; Thermo Scientific) was added and the plate was incubated at room temperature for 30 min. The reaction was halted with 2M H₂SO₄. Absorbance was quantified via microplate spectrophotometer (Molecular Devices) at a wavelength of 450 nm. Serum antibody titers were determined by measuring the lowest dilution that resulted in signal three standard deviations above background. Statistical significance was determined using Tukey-Kramer post hoc honest significant difference test and compared against the PBS control case.

Complement-mediated bactericidal assay

Bactericidal assays were conducted similar to a previously published protocol (Moe et al., 1999). Briefly, MenB strain H44/76 was grown overnight on chocolate agar plates (Remel). Single colonies were used to inoculate a culture in Franz media starting at an OD₆₂₀ ~0.15 and grown at 37°C, 5% CO₂ to OD₆₂₀ ~0.6. The bacteria were diluted in Dulbeccos PBS with Ca²⁺ and Mg²⁺ (DPBS), containing 1% human serum albumin (Sigma-Aldrich). Approximately 300–400 CFU meningococci were incubated with 20% human serum (from a healthy adult with no detectable anticapsular antibody to group B polysaccharide) that had been depleted of IgG with a Protein G column and serum collected from mouse immunizations. Percent survival was calculated as the CFU/mL after 60 min incubation of bacteria compared to baseline CFU/ml at time zero determined by average bacterial growth in buffer alone, with heat-inactivated complement, with active complement, or a mixture of heat-inactivated and active complement. Murine antibodies SEAM 12 and anti-MenC were used as positive and negative controls, respectively.