Self Antigens Displayed on Liposomal Nanoparticles above a Threshold of Epitope Density Elicit Class-switched Autoreactive Antibodies Independent of T-cell Help

Abstract

Epitope density has a profound impact on B cell responses to particulate antigens, the molecular mechanisms of which remain to be explored. To dissect the role of epitope density in this process, we have synthesized a series of liposomal particles, similar to the size of viruses, that display a model self antigen peptide at defined surface densities. Immunization of C57BL/6J mice using these particles elicited both IgM and class-switched IgG1, IgG2b and IgG3 autoreactive antibodies that depended on the epitope density. In C57BL/6 gene knockout mice lacking either functional T-cell receptors or MHC class II molecules on B cells, the liposomal particles also elicited IgM, IgG1, IgG2b and IgG3 responses that were comparable in magnitudes to wild-type mice, suggesting that this B cell response was independent of cognate T cell help. Notably, the titer of the IgG in wild-type animals could be increased by more than 200-fold upon replacement of liposomes with bacteriophage Qβ virus-like particles that displayed the same self antigen peptide at comparable epitope densities. This enhancement was lost almost completely in gene knockout mice lacking either T-cell receptors or MHC class II molecules on B cells. In conclusion, epitope density above a threshold on particulate antigens can serve as a stand-alone signal to trigger secretion of autoreactive and class-switched IgG in vivo in the absence of cognate T cell help or any adjuvants. The extraordinary immunogenicity of Qβ viral-like particles relies in large part on their ability to effectively recruit T cell help after B cell activation.

Introduction

Often, foreign particulate antigens such as viral particles can effectively prime the immune system for elicitation of protective antibody responses, with a few exceptions. This B cell response forms the basis for the majority of licensed antiviral vaccines (1). However, how a particulate antigen such as a viral particle activates the immune system to bring about the protective antibody responses, especially at quantitative and mechanistic level, remains largely uncharacterized. A breakthrough in our understanding of this process was put forward by Bachmann et al. (2), who showed that antigen organization had a profound impact on B cell responses to antigens. Remarkably, the envelope glycoprotein displayed on the surface of vesicular stomatitis virus broke the tolerance of B cells in transgenic mice that expressed the same glycoprotein under the control of H-2K promoter (2). Along a similar vein, studies led by Schiller showed that potent and long-lasting autoreactive IgG antibodies were elicited in mice upon immunization with self antigens incorporated or conjugated to papillomavirus-like particles (3–5). These discoveries have led to potentially exciting applications in efforts to elicit therapeutic antibodies through vaccination approach (6, 7). However, at a fundamental mechanistic level, there remain questions unanswered regarding the extraordinary immunogenicity of these viruses or viral-like particles. Specifically, what are the molecular components or quantitative features in these particles that are required for the potent B cell activation? If collaboration with other immune cells is also required to yield the antibody response, to what extent does the production of these antibodies rely on those cells? Answers to these questions will help the design and engineering of vaccines, and also assist in understanding the spectra of host immune responses to viral pathogens.

Among various features shared by the particles listed above, epitope density appears to be a dominant parameter in the quality of the antibody response. Ensemble estimations based on RNA and proteins present in viral particles yielded approximately 500 glycoproteins per vesicular stomatitis viral particle (8), while papillomavirus displayed 360 copies of the major capsid protein L1 per particle based on a cryoelectron microscopy study (9). Thus both viruses display high number of viral-specific epitopes per particle, which is in fact a feature shared by most of the human viral pathogens with licensed vaccines (1). Furthermore, Chackerian et al. showed that reduction of epitope density on papillomavirus-like particles significantly decreased the IgG autoantibody production (3), highlighting the critical role of high epitope density in breaking the B cell tolerance. Similarly, studies led by Bachmann also showed that epitope density on bacteriophage Qβ viral-like particles strongly influenced the resulting IgG antibody response (10). However, high epitope density alone does not seem to entail the full story. It was showed early on that very high density of antigens could induce B cell tolerance both in vitro and in vivo (11, 12). A recent study of mouse B cells also suggested that too high epitope density in the absence of immediate T-cell help might trigger B-cell death instead of activation (13).

In order to understand quantitatively and mechanistically B cell responses to particulate antigens and define the role of epitope density in this process, we have taken a ‘deconstructive’ approach: chemically synthesizing particulate antigens similar to the size of viruses and using these particles for immunization in mice. We chose liposomes as carriers for the epitope of interest for two reasons: (i) the nonimmunogenic nature of liposomes by themselves, and (ii) the versatility of liposomes in epitope conjugation and engineering (14). In particular, as we have shown recently (15), epitopes of interest can be conjugated onto the surface of these particles in a controlled fashion that yields particles with varied epitope densities, an important tool to unravel the role of epitope density in this process. As we show, these liposomal particles, in the absence of any adjuvants, can elicit IgG autoreactive antibodies in mice that depends on epitope density. To the best of our knowledge, this is the first time that a self antigen, upon conjugation to liposomes above a threshold of epitope density, is shown to induce IgG antibody responses in the absence of T-cell help. Furthermore, replacement of liposomes with bacteriophage Qβ viral-like particles at comparable epitope densities dramatically enhanced the titer of the IgG response. Studies using gene knockout mice revealed that the superior immunogenicity of Qβ viral-like particles originated, in large part, from their extraordinary ability to recruit MHC Class II dependent T-cell help after B-cell activation. Our study has thus uncovered a fundamental aspect in B-cell activation by particulate antigens and offered valuable insights to future vaccine design targeting self antigens.

Materials and methods

Synthesis of liposomes and antigen conjugation

Liposomes were prepared using oil-in-water emulsion precursor followed by membrane extrusion as described ^1–3. Three different lipids of designated molar ratios were used in the synthesis of liposomes: 1,2-dimyristoyl-sn-glycero-3-phosphocholine (DMPC), 1,2-distearoyl-sn-glycero-3-phosphoethanolamine-N- [maleimide (polyethylene glycol)-2000] ammonium salt (DSPE-PEG (2000) maleimide) and cholesterol (Avanti lipids). Briefly, lipid mixture (7.5 μmol in total) in chloroform were added to a round-bottom flask, blown dry with purified argon and desiccated by vacuum to form a thin film at the bottom of the flask. 1mL HEPES buffer (50mM HEPES, 150mM NaCl, pH6.9) (16) was added to hydrate the lipid film through vortex and short bursts of sonication in a water bath. After hydration, the lipid film resuspension was extruded using polycarbonate membrane with pore sizes of 1000 nm and 100 nm sequentially for ten times each. Synthesized liposomes were incubated with a synthetic peptide of the following sequence, CSSQNSSDKPVAHVVANHQVE (TNF-α peptide, >95% purity, Biomatik Scientific) at a chosen molar ratio between the maleimide functional group and the peptide. The sequence of the peptide was derived from mouse tumor necrosis factor α. The N-terminal cysteine of the peptide was engineered for the purpose of maleimide conjugation. After overnight incubation at 22°C, the liposome peptide mixture was applied to Sepharose CL-4B gel filtration column, as we described previously (15), to separate TNF-α conjugated liposomes from unconjugated free peptides. To prepare peptide-conjugated liposomes of smaller diameters, the lipid film resuspension was extruded using polycarbonate membrane with pore sizes of 1000 nm and 50 nm sequentially for ten times each and the rest procedures were the same as described above. To prepare dual-functional liposomes for flow cytometry analysis of mouse splenocytes, 6% DSPE-PEG (2000) maleimide was included in the lipid mixture. The lipid thin film was hydrated using a solution of 0.6 mM Alexa-594 NHS ester freshly dissolved in PBS. After extrusion, the liposomes encapsulating Alexa-594 dye molecules were loaded onto Sepharose CL-4B gel filtration column to purify liposomes away from excess fluorescent dyes. The eluted liposomal fractions were concentrated by centrifuge through Amicon Ultra-4 centrifugation unit (100 kD cut-off), and then mixed with TNF-α peptide for conjugation overnight at 22°C. The free peptide was removed on the second morning by running liposomes through a Sepharose CL-4B gel filtration column a second time. The peptide-conjugated liposomes were stored at 4°C in PBS.

Quantitation of average peptide density on liposomes

The average density of TNF-α peptide covalently conjugated on liposomes was quantitated using the method that we established previously (15). Briefly, we estimate the concentrations of liposome and peptide respectively for the liposomes purified through gel filtration column. The ratio between the two concentrations represents the average number of TNF-α peptide per liposome (15). To estimate the concentration of liposomes, we used the established Stewart assay (17) to determine the phospholipid content in the purified liposomes, based on which the molarity of the liposomes can be further estimated as described (15). Briefly, for each liposomal sample, 20 μL purified liposomes were added to 500 μL of chloroform in a clean glass tube, and then 500 μL ferrothiocynate solution containing 0.1 M ferric chloride hexahydrate and 0.4 M ammonium thiocyanate was added. The mixture was vigorously vortexed for 20 seconds and then centrifuged for 10 minutes at 1000 g at 22°C. The lower chloroform layer was then taken for absorption measurement at 470 nm. The phospholipid concentration was calculated based on standard curves constructed from known quantities of DMPC and DSPE-PEG (2000) maleimide, respectively, taking into account the compositions of the two lipids in the liposome samples. To measure the concentration of TNF-α peptide conjugated on liposomes, we used bicinchoninic acid (BCA) protein assay (18). Briefly, 100μL of the liposome sample was incubated with 2 mL BCA solution at 60 °C for 30 minutes before absorption measurement at 562 nm. 8% SDS was also included in this mixture to minimize the interference of lipids to BCA assay as reported (19). The TNF-α peptide concentration was calculated based on the comparison with a standard curve constructed from known quantities of the TNF-α peptide under the same BCA assay conditions.

Stability of epitope density in serum-containing media

Peptide-conjugated liposomes containing 20% of maleimide-lipid were chosen for stability studies as described below. An equal volume of fetal bovine serum was added to peptide-conjugated liposomes to reach 50% serum condition and incubated at 37°C. After 1, 3, 5, and 7 days of incubation respectively, a fraction of liposome-serum mixture was taken and loaded onto Sepharose CL-4B gel filtration column for purification of liposomes away from serum components. After purification, the liposomes were evaluated using Stewart Assay for quantitation of lipid concentration. Incubation with serum may lead to the binding of serum proteins on liposomal surface. Thus the protein-based BCA assay was no longer reliable in quantitating the amount of peptide conjugated on liposomes. We therefore used an alternative approach based on silver staining of a denaturing polyacrylamide gel to quantitate TNF-α peptides conjugated on liposomes. Liposomes from different time points of serum incubation were normalized based on lipid content and loaded onto a Tricine-SDS polyacrylamide gel (20) for electrophoresis. For optimal resolution, we used 16% acrylamide separating gel with 1cm of 10% spacer gel and 4% stacking gel. The gel was run at 120 V for the initial 20 minutes followed by 180 V for the remainder of the time. Silver staining was then carried out for quantitation of peptides on the gel based on the intensity of staining as we reported previously (15).

Conjugation of TNF-α peptide to bacteriophage Qβ viral-like particles

The bacteriophage Qβ viral-like particles were prepared as described (21). The TNF-α peptide, which contained a single cysteine at its N-terminus, was conjugated to the surface of the viral-like particles using the heterobifunctional crosslinker Succinimidyl-6-[(β-maleimidopropionamido)hexanoate] (SMPH). Briefly, the viral-like particles were first derivatized with SMPH at a 10-fold molar excess of SMPH over Qβ coat protein. The mixture was incubated at 22°C for two hours and the excess crosslinker was removed by centrifugation at 4°C in an Amicon Ultra-4 centrifugal unit with a 100 kD cutoff. The TNF-α peptide was then added to the derivatized viral-like particles at a molar ratio of 10:1 over Qβ coat protein. The mixture was incubated at 4°C overnight and again the excess peptide was removed by centrifugation at 4°C in an Amicon Ultra-4 centrifugal unit with a 100 kD cutoff. The quantity of conjugated peptide was assessed using a denaturing polyacrylamide gel based on the intensity from silver staining in comparison with a standard curve obtained from the same gel.

Mice immunizations

All animal procedures were approved by the University of Michigan Animal Care and Use Committee. Female C57BL/6 mice (8 weeks, Jackson Laboratory) were used for immunizations. Prior to inoculation, all injection samples were filtered through 0.45 μm pore size membrane to eliminate potential microbial contamination. 100 μL samples were injected to each mouse subcutaneously, 50 μL on each flank. The immunization was boosted on Day 14 and 28 using the same materials. Mouse blood was collected submentally using Microvette serum collection tube (Sarstedt) three days before the first injection, and 11 days after each inoculation. The serum was harvested by centrifugation at 10,000 g for 10 minutes and immediately frozen and stored at −20°C.

Two strains of gene knockout mice were purchased from Jackson Laboratory: Tcrb^tm1Mom/ Tcrd^tm1Mom (#002122) and Ciita^tm1Ccum (#003239). Tcrb^tm1Mom/ Tcrd^tm1Mom mice (TCR⁻/⁻ herein) were deficient in both alpha beta and gamma delta T-cell receptors (22), while Ciita^tm1Ccum mice (MCII⁻/⁻herein) did not have MHC class II molecules on the surface of splenic B cells or dendritic cells (23). Gene knockout mice were housed in germ-free environment and immunization protocol were carried out as described above for C57BL/6 mice.

Enzyme-Linked Immunosorbent Assay (ELISA)

Blood serum was tested for ELISA in order to quantitate B cell antibody responses to various immunizations. 96-well plates (Nunc MaxiSorp, Invitrogen) were coated overnight at 4°C with 5 μg of TNF-α peptide in PBS. After blocking with 1% Bovine Serum Albumin (BSA, Fisher) in PBS, mouse sera of different dilution factors (1: 100, 1: 400 and 1:1600) were added to each well for incubation at 22°C for 2 hours. After three washes using PBS with 0.05% Tween-20, secondary anti-mouse-IgG-HRP antibody (# 616520, ThermoFisher Scientific) or anti-mouse-IgM-HRP antibody (# 626820, ThermoFisher Scientific) was added in blocking buffer at 1:5000 dilution and incubated for 1 hour at 22°C. Following three washes, 100 μL substrate 3,3’,5,5’-Tetramethylbenzidine (Thermal Scientific) was added to each well and incubated in dark for 10 minutes. The reaction was stopped by addition of 100 μL 2M sulfuric acid in each well. The optical density of each well at 450nm was measured using a microplate reader (Bio-Tek Synergy HT). All the OD values reported were background subtracted by comparison between two wells that were coated with TNF-α peptide and PBS, respectively. Similarly, for ELISA using TNF-α protein as the bait, 200 ng of recombinant murine TNF-α protein (≥ 98% purity, #315–01A, PeproTech) was coated in each well and the rest procedures were carried out as described above.

To determine the titers of anti-TNF-α antibodies, we used serial dilution of serum (from 1:100 dilution to 1:1638400 by a dilution factor of 4) for ELISA with either TNF-α peptide or the recombinant TNF-α protein. Cutoff values were calculated using the following equation as reported (24): Cutoff = $\overline{\mathrm{X}}+SD\cdot f$, where $\overline{\mathrm{X}}$ and SD are the mean and standard deviation of control well OD reading values, f is the standard deviation multiplier corresponding to different confidence levels. Specifically in our assays, f = 2.631 when the number of control wells was 4 and confidence level was 95%. The titer value was determined as the highest dilution factor of the serum that still yielded OD450 value higher than the above cutoff value in ELISA.

To determine the subclasses of IgG antibodies elicited in mice upon immunization by various agents, we used the following secondary antibodies during ELISA: goat anti-mouse IgG1 HRP antibody (#1071–05, SouthernBiotech), goat anti-mouse IgG2b HRP antibody (#1091–05, SouthernBiotech), goat anti-mouse IgG2c HRP antibody (#1078–05, SouthernBiotech), and goat anti-mouse IgG3 HRP antibody (#1101–05, SouthernBiotech).

To determine the avidity of anti-TNF-α antibodies against recombinant mouse TNF-α proteins, the ELISA assay was conducted as described above except that after incubation of the diluted sera in the wells for 2 hours, the plate was washed once using PBS with 0.05% Tween-20. 8 M urea was then added to the designated wells for incubation for exactly 5 min (5), after which the wells were washed twice and the regular ELISA procedure was continued. The avidity index was calculated as the ratio of the mean OD values of urea-treated wells to PBS-treated control wells multiplied by 100 (25).

Flow cytometry analysis of mouse splenocytes

At specific time points following mouse inoculation with various agents, mice were euthanized by CO2 inhalation followed by rapid dissection of the spleen. Mouse spleens were minced on ice and a cellular suspension was separated from tissue using a 40 μm cell strainer. Red blood cells were removed by rapid exposure of mouse splenocytes to ACK lysis buffer followed by washing with FACS buffer (PBS with 2% fetal bovine serum and 2 mM EDTA). Single cell suspensions of mouse splenocytes were first incubated with rat anti-mouse CD16/CD32 (BD Pharmingen #553142) on ice for 15 min, followed by incubation with fluorophore-conjugated antibodies and/or TNF-α conjugated liposomes encapsulating Alexa-594 for 10 min on ice covered by aluminum foil. The cells were washed twice with 200 μl FACS buffer, resuspended in FACS buffer, and fixed in the presence of 2% PFA at 22°C for 1 hour before data acquisition on a Miltenyi MACSQuant VYB flow cytometer equipped with 3 spatially-separated lasers (405, 488 and 561 nm) and 10 detection channels. A minimum of 100,000 events were collected for all experiments. The fluorophore-conjugated antibodies were pacific blue rat anti-mouse CD19 (Biolegend #115526), FITC rat anti-mouse GL7 antigen (Biolegend #144603). The data were analyzed using FlowJo (BD).

Statistical Analysis

Statistical analysis was carried out using the Statistics toolbox in MATLAB (Mathworks, MA). Data sets were analyzed using one- or two-way analysis of variance as described in figure legends. p-values less than 0.05 were considered statistically significant. All values were reported as mean ± standard error unless otherwise noted.

Results

Liposome-based particulate antigens that carry epitopes of varying density

To obtain a quantitative and mechanistic understanding of B cell responses to particulate antigens such as viruses, we decided to choose liposomes as our platform for antigen design. The unique advantages of using engineered liposomal particles as opposed to a specific virus are threefold: (i) the use of liposomes by themselves will not elicit an immune response and therefore provide a baseline for quantification of immunogenicity; (ii) it allows us to use defined components to build a particle that mimic certain features of a natural virus particle and therefore there is no ambiguity on the constituents of the input antigen; and (iii) it allows us to use model antigens, such as the TNF-α peptide used in this study, to systematically investigate the individual contribution of various virus features on B-cell responses, including epitope density, the presence or absence of T-cell epitope and Toll-like receptor ligands, since these different features can be built into a liposomal nanoparticle in a controllable fashion.

In current study, we have selected DMPC (1,2-dimyristoyl-sn-glycero-3-phosphocholine) as the major lipid for construction of our liposomes, which has a phase transition temperature of 23°C (26) and therefore offers ease in extrusion through polycarbonate membranes for formation of unilamellar liposomes. For conjugation of B-cell epitopes onto the surface of liposomes, we chose a sequence of 20 amino acids from mouse tumor necrosis factor α (TNF-α) as our model self antigen. This peptide corresponds to amino acids 83–102 of mouse TNF-α. Based on crystal structures of the highly homologous human TNF (27, 28), the C-terminal half of the above peptide is predicted to be solvent exposed and contains critical residues that are directly involved in binding with TNF receptors (27). Therefore, antibodies directed against this peptide are expected to bind the TNF-α protein and potentially block TNF-α and receptor interactions. Indeed, Chackerian et al. were able to elicit potent and long-lasting autoreactive TNF-α antibodies in mice using this peptide fused to a truncated form of streptavidin that was conjugated to biotinylated bovine papillomavirus-like particles (5). Similarly and interestingly, Spohn et al. were able to elicit high titers of antibodies in mice that were autoreactive towards soluble but not membrane-bound TNF-α protein by conjugation of a similar peptide to viral-like particles of bacteriophage Qβ (29), suggesting the presence of a mouse B-cell epitope in this peptide. The lack of cysteine in the above peptide sequence allowed us to engineer a single cysteine at the N-terminus of the peptide for conjugation of this peptide to the surface of liposomes using maleimide chemistry in an orientation that exposed its C-terminal residues. For this maleimide chemistry, we have specifically chosen DSPE-PEG (2000) maleimide as the phospholipid for incorporation into liposomes. The presence of polyethylene glycol chain between the lipid head group and the maleimide moiety can increase the residence time of liposomes in the bodily fluid as demonstrated for many liposomal systems (30–34).

To check the covalent conjugation of the TNF-α peptide onto the liposomes, we ran a 16% denaturing polyacrylamide gel for the purified liposomes and subjected the gel to silver staining. The peptide itself had a molecular weight of 2.2 kD (indicated by the arrow in lane 4), and the peptide conjugated DSPE-PEG (2000) maleimide lipid was expected to have a molecular weight of ~ 5.1 kD, as indicated by the band migrating between 3.5 and 6.5 kD in lane 2 in Fig. 1A. Thus, this was exactly as expected for a 1:1 conjugation of the peptide onto the lipid. We also characterized these liposomal particles using dynamic light scattering (DLS). As shown in Fig. 1B, the DLS data showed a well-defined peak around 153 nm for the peptide conjugated particles, which compared with 137 nm for bare liposomes, suggesting that the peptide may form a dense layer on liposomal surface under this set of conditions. Consistent with this notion, quantitation of peptide content in the purified liposomes (Materials and Methods) revealed 6,030 ± 428 (mean ± standard deviation) molecules of TNF-α peptide per liposome on average, which yielded a density of 8.2×10⁴ peptides/μm² that was more than twofold higher than the epitope density on papillomavirus-like particles (3.2×10⁴ molecules/μm²) (1).

Figure 1. Preparation and characterization of p-liposomes of varied epitope densities.

(A) Tricine-SDS-PAGE and silver staining result of samples: lane 1: peptide marker (Bio-Rad, # 1610326); lane 2: 0.12 pmol peptide-conjugated liposomes containing 20% maleimide lipids (p-liposome_20%-m); lane 3: 0.12 pmol liposome_20%-m without peptide conjugation; lane 4: 1 μg TNF- α- peptide. (B) Dynamic light scattering size analysis of liposome_20%-m and p-liposome_20%-m. (C) Estimated number of TNF-α peptide per liposome as a function of the maleimide percentage incorporated in liposomes. The error bars represent standard deviations measured from three independent batches of liposomes prepared with various maleimide percentages listed as follows: 1.1%, 2.2%, 4.4%, 6.6% and 20%. (D) Average number of TNF-α peptide molecules per liposome for four batches of p-liposome_20%-m independently prepared. Error bars represent the standard deviation from two independent estimations of the epitope density. (E) Average diameter of p-liposome_20%-m upon incubation in 50% serum-containing media as a function of incubation time. The error bars represent standard deviations from three independent repeats of the same experiment. (F) Estimated number of TNF-α peptide per liposome for p-liposome_20%-m upon incubation in 50% serum for various amount of time. Day 0 is before incubation. The error bars represent standard deviations from three independent repeats of the same experiment.

In order to test the effect of epitope density on B cell responses, it is necessary to prepare liposome particles with varied densities of the antigenic peptides. To this end, we systematically changed the molar percentage of the DSPE-PEG (2000) maleimide included during liposome synthesis from 1.1 to 20%. After conjugation of the TNF-α peptide, these liposomes were purified through gel-filtration column and the epitope density was measured as described (Materials and Methods). As shown in Fig. 1C, the epitope density on these liposomes displayed a monotonic increase with increasing percentage of the DSPE-PEG (2000) maleimide lipids. This dependence was well described with a linear relationship (the black line in Fig. 1C), yielding R-squared value of 0.9922. Specifically, the epitope density decreased to 360 ± 64 for liposomes containing 1.1% maleimide lipids, which was 6.6-fold lower than the epitope density on papillomavirus-like particles. This procedure has a good reproducibility, as shown in Fig. 1D using liposomes containing 20% maleimide lipids as an example. Four independent repeats of the same synthesis and conjugation procedure all yielded an epitope density around 6,000 TNF-α peptides per liposome on average.

Lastly, because these peptide-conjugated liposomal particles (p-liposome) were to be used in vivo for mouse immunizations, it is important that the epitope density is stable in biological milieu so that the data can be interpreted using epitope density with confidence. To this end, we incubated p-liposomes in 50% fetal bovine serum at 37°C for varied time and then purified liposomes away from serum components using a gel filtration column. The resulting particles were then characterized in terms of size and epitope density. As shown in Fig. 1E for p-liposomes prepared with 20% maleimide lipids (p-liposome_20%-m), the particles remained relatively constant in their size after exposure to serum-containing medium. Also, the density of TNF-α peptide conjugated on liposomes remained at a similar level over time (Fig. 1F). Thus, we can interpret our results in terms of epitope density with confidence.

TNF-α peptides conjugated on liposomal surface at high density elicit autoreactive IgG antibody in wild-type mice

To examine whether p-liposomes can elicit autoreactive antibodies, we first inoculated mice with p-liposome_20%-m, which displayed 6030 ± 430 TNF-α-peptides per liposome on average. Each group of mice received three doses of p-liposome_20%-m, each dose containing 4.5 μg TNF-α-peptide. No substance other than p-liposome_20%-m was included in the injections. Antibody responses were measured by ELISA using serum collected 11 days after each inoculation. As shown in Fig. 2A–B, p-liposome_20%-m elicited both IgM and IgG responses towards the TNF-α-peptides after just one dose. Both responses were significantly higher than either the PBS or bare liposome controls, and also significantly higher than those elicited by soluble TNF-α-peptides at the same dose (4.5 μg). In fact, the levels of IgM and IgG responses from the soluble TNF-α-peptides were statistically indistinguishable from those of PBS or bare liposome controls, consistent with the fact that this peptide by itself is a bona fide self antigen and no apparent antibody responses were elicited upon immunization. The results shown in Fig. 2A–B also suggest that upon conjugation of the peptide to a liposomal nanoparticle, it became immunogenic and could induce class-switched IgG antibody responses, even though the liposomal nanoparticles by themselves were not immunogenic.

Figure 2. Anti-TNF-α-peptide antibody response in wild-type C57BL/6 mice.

(A-B) ELISA OD values for IgM (A) and IgG (B) antibody from 1:100 diluted mouse sera after 1^st inoculation using PBS, liposome_20%-m, TNF-α peptide (4.5 μg), and p-liposome_20%-m containing 4.5 μg peptide. (C-D) Time courses of ELISA OD values for IgM (C) and IgG (D) from 1:100 diluted mouse sera collected from animals immunized with p-liposome_20%-m (upper triangle) and the soluble peptide (hollow circle), respectively. The upward arrows denote the time of inoculations. Blood sera were collected at week 2, week 4 and week 6, 11 days after each inoculation, and at week 10, week 14 and week 18 after the first inoculation. Throughout the four panels, each data point in the figure represents the mean of OD450 values obtained from four mice of each group. Error bars represent the standard errors. Statistical difference between soluble TNF-α peptide and p-liposome_20%-m at each time point was determined by Student’s T-test (***: p-value < 0.001; **: p-value < 0.01; *: p-value < 0.05; NS (Not Significant): p-value > 0.05).

To further examine the antibody responses elicited by p-liposome_20%-m, we continued to measure both IgM and IgG responses at different time points along the process, up until 18 weeks after the first inoculation. As shown in Fig. 2C, the IgM response from p-liposome_20%-m declined after successive injections, and it continued to decline until the levels of responses became indistinguishable from those induced using soluble peptides by Week 10. Similarly, the IgG response from p-liposome_20%-m also declined after successive injections, and became indistinguishable from that of the soluble peptide by Week 14 (Fig. 2D). This trend of IgG responses after successive inoculations suggest the lack of apparent antibody affinity maturation or B cell clonal expansion, which was in sharp contrast to the boosting effect typically observed for foreign antigens (35).

Elicitation of autoreactive IgG antibody requires a threshold of epitope density

The above results revealed that a self antigen, once conjugated onto the surface of a liposomal nanoparticle at high density, could elicit both IgM and IgG antibody responses that were much higher than those from soluble peptide controls at the same antigen dose in the absence of any adjuvants. To further examine the dependence of this immunogenicity on epitope density, we inoculated C57BL/6 mice using p-liposomes conjugated with varied densities of TNF-α peptide. They were p-liposomes prepared with 6.6% maleimides (p-liposome_6.6%-m), p-liposomes prepared with 2.2% maleimides (p-liposome_2.2%-m), and p-liposomes prepared with 1.1% maleimides (p-liposome_1.1%-m), which displayed on average 2480 ± 322, 730 ± 92, and 360 ± 64 (mean ± standard deviation) TNF-α-peptides per liposome, respectively (Fig. 1C). Together with p-liposome_20%-m and the soluble peptide control, a total of five groups of mice were inoculated. Throughout these experiments, the peptide dose was kept the same regardless of the antigen carriers (4.5 μg TNF-α peptide per injection). Three inoculations were performed for each group, and the antibody responses were measured starting 11 days after the first inoculation. As shown in Fig. 3A, the IgM response from these various liposomes showed a dependence on epitope density and declined by half at p-liposome_1.1%-m. The IgG response from these liposomes also showed a clear dependence on epitope density, and declined with decreasing epitope density (Fig. 3B). Interestingly, the IgG response from p-liposome_1.1%-m was indistinguishable from background, which was in contrast to the IgM response induced by this immunogen. These results revealed that the IgG response towards the TNF-α peptide was more sensitive to the change in epitope density than IgM responses. In particular, low valence liposomal display (360 ± 64 TNF-α peptides) elicited higher IgM response than the soluble peptide, but was incapable of eliciting an autoreactive IgG response, clearly showing the dependence of IgG response on epitope density. Qualitatively, a similar phenomenon was also reported previously by Jegerlehner et al. for foreign epitopes conjugated to viral-like particles (10).

Figure 3. Anti-TNF-α-peptide antibody response in wild-type C57BL/6 mice measured 11 days after first inoculation with p-liposomes of varied epitope densities.

Percentages of maleimide lipid in total lipids are indicated. (A-B) ELISA OD values for IgM (A) and IgG (B) antibody from 1:100 diluted mouse sera. The dose of TNF-α peptide per injection was the same for all immunizations, 4.5 μg, including the soluble TNF-α peptide. (C-D) ELISA OD values for IgM (C) and IgG (D) antibody from 1:100 diluted mouse sera. For each inoculation, 20 μg CpG (ODN1826 from Invivogen) was mixed with various antigens before inoculation. The dose for TNF-α peptide remained 4.5 μg per injection. For all four panels, p-liposome_1.1%-m was used as a reference to determine statistical difference between data points by Student’s T-test. The statistical difference was denoted using the same set of symbols as Figure 2. Error bars represent the standard errors (N=4).

All the immunization experiments reported above were conducted in the absence of other immunostimulatory agents. To examine the effect of Toll-like receptor activation on these antibody responses, we next performed the above experiments in the presence of CpG, a potent Toll-like receptor 9 agonist. For each inoculation, 20 μg CpG was mixed with either p-liposomes or the soluble peptide control and then administered into animals. The immunization followed the same schedule as before and blood was collected at designated time for ELISA. As shown in Fig. 3C, all four p-liposomes induced similar levels of IgM response. The difference of the IgM response between p-liposome_1.1%-m and other p-liposomes was diminished, suggesting that the presence of CpG weakens the dependence of IgM responses on epitope density. For IgG, three groups of p-liposomes induced similar levels of IgG response (p-liposome_20%-m, p-liposome_6.6%-m, and p-liposome_2.2%-m) (Fig. 3D). The IgG response elicited by p-liposome_1.1%-m remained the lowest among the four groups of liposomes, but in the presence of CpG, its IgG response was clearly detectable above the background and the soluble peptide control. This was in contrast to Fig. 3B, where p-liposome_1.1%-m was incapable of eliciting an IgG response above background. These results thus suggest that the presence of CpG likely modulates the sensitivity of B cells to epitope density. Specifically, the presence of CpG may sensitize B cells towards a lower threshold of epitope density that is required to elicit an IgG response. We continued to monitor these animals at later time points. As shown in Supplementary Fig. 1, IgG response did not show any boosting effect upon successive immunizations, which was true for all these liposomal preparations in the absence or presence of CpG, consistent with the patterns we observed in Fig. 2C–D.

TNF-α peptides conjugated on Qβ viral-like particles elicit potent autoreactive IgG antibody in wild-type mice

Even in the presence of CpG, it did not escape our attention that the titers of IgG response from these liposomes were only on the order of 10³, which was much lower than the IgG titer reported for the same peptide but noncovalently conjugated to papillomavirus-like particles (5). To examine the mechanisms behind this apparent difference, we conjugated the TNF-α peptide to bacteriophage Qβ viral-like particles (VLP_Qβ) using the heterobifunctional crosslinker SMPH. Bacteriophage Qβ has been used as an antigen presentation platform for elicitation of antibodies both in mice and in human clinical trials (21, 36, 37). The single available thiol group present in our synthetic TNF-α peptide allowed us to conjugate this peptide to the surface of SMPH-derivatized VLP_Qβ. The excess free peptide was removed by filtration of the reaction mixture through Amicon filtration units. As shown in Fig. 4A, using a silver stained gel to monitor this conjugation reaction, the SMPH-derivatized VLP_Qβ showed a major band around 15 kD before conjugation with TNF-α peptide (lane 7), which corresponded to the molecular weight of Qβ coat protein. Upon crosslinking with TNF-α peptide, a series of bands were seen on the gel (lane 8), which matched with the expected molecular weights for the coat protein conjugated with one TNF-α peptide, two TNF-α peptides, and so on. Based on silver staining intensity, we estimated that on average there were 260 TNF-α peptides per VLP_Qβ particle. We also characterized these VLP_Qβ particles using DLS. As shown in Fig. 4B, the DLS data showed a well-defined peak around 42 nm for the peptide conjugated particles (p-VLP_Qβ), which compared with 36 nm for the unconjugated VLP_Qβ particles, suggesting that the peptide formed a dense layer on VLP_Qβ surface that was consistent with our estimation for the average number of peptides per VLP_Qβ particle. These measurements yielded an epitope density of 4.7×10⁴ peptides/μm² on these p-VLP_Qβ particles, which was comparable to the epitope density on papillomavirus-like particles (3.2×10⁴ molecules/μm²) (1).

Figure 4. Characterization of Qβ particles conjugated with TNF-α peptides and the antibody responses in wild-type C57BL/6 mice immunized with Qβ particles conjugated with TNF-α peptides.

(A) Tricine-SDS-PAGE and silver staining result of samples listed as follows: lane 1: protein marker (Thermal Scientific, Cat. 26614); lane 2–6: hen egg lysozyme of various quantities (800ng, 400ng, 200ng, 100ng, 50ng); lane 7: VLP_Qβ (2 μg Qβ); lane 8: p-VLP_Qβ (4 μg Qβ). (B) Dynamic light scattering size analysis of VLP_Qβ and p-VLP_Qβ. (C-D) Time courses of ELISA OD values for IgM (C) and IgG (D) antibody from 1:100 diluted mouse sera upon three successive inoculations using antigens as indicated in the figure. For p-VLP_Qβ, the dose of TNF-α peptide was 0.5 μg per injection; for p-liposome_20%-m the dose of TNF-α peptide was 4.5 μg per injection. The upward arrows denote the time of inoculations. Student’s T-test was done to compare p-VLP_Qβ with p-liposome_20%-m. The statistical difference was denoted using the same set of symbols as Figure 2. Error bars represent the standard errors (N=4). (E) IgG titer as a function of time after the first inoculation of the respective antigens as indicated. Each data point represents the IgG titer result from each individual mouse in an immunization group. Lines represent the mean of titers.

We then used the p-VLP_Qβ to immunize C57BL/6 mice. In these experiments, we followed the same immunization schedules as we did for p-liposomes. Also, in order to compare the immune responses elicited by p-VLP_Qβ to those previously published in literature using papillomavirus-like particles, we used p-VLP_Qβ containing 0.5 μg TNF-α peptides for each inoculation, which was the peptide dose published previously using papillomavirus-like particles. As shown in Fig. 4C, the p-VLP_Qβ elicited twofold higher IgM responses than p-liposome_20%-m containing 4.5 μg TNF-α peptides. Notably, the p-VLP_Qβ elicited IgG responses that saturated the OD reading at 450 nm under the same serum dilution (1:100; Fig. 4D). Control experiments using a simple admixture of VLP_Qβ and the soluble TNF-α peptide at the same particle and peptide doses did not elicit any IgM or IgG response above the background (hollow squares), clearly indicating that it was the conjugation of the TNF-α peptide onto the VLP_Qβ that mattered.

The saturation in OD450 required further dilution of the sera in order to obtain meaningful titer values for the IgG antibody responses. By performing a serial dilution of the collected sera and repeating the ELISA measurements (Materials and Methods), we determined that the IgG response elicited by p-VLP_Qβ had a titer between 10⁵ and 10⁶ (Fig. 4E, circles) 11 days after the first inoculation, and these titers were stable after the second and third inoculations. These titer values compared favorably with the titer values previously reported for the same peptide but conjugated to papillomavirus-like particles at the same peptide dose (5). In contrast, the IgG titer elicited by p-liposome_20%-m was 10³ (Fig. 4E, triangles) 11 days after the first inoculation, and the titer remained at 10³ after the second and third inoculations. No boosting effect was observed for p-liposomes despite the fact that these antibody levels were more than 100-fold lower than those elicited by p-VLP_Qβ. These data clearly indicated that p-VLP_Qβ particles were much more potent than liposomal nanoparticles in eliciting IgG responses.

Previous studies by Chackerian et al. showed that the TNF-α peptide conjugated to papillomavirus-like particles could elicit antibodies that were capable of binding to the recombinant TNF-α protein. To test whether our construction of the p-VLP_Qβ particles could elicit the same effect, we thus conducted ELISA using a recombinant TNF-α protein coated on the ELISA plate. As shown in Fig. 5A, we could clearly detect a positive ELISA signal above background for the sera from p-VLP_Qβ immunized animals at a dilution of 1:25600. In contrast, a dilution of 1:400 for the serum was required to observe similar OD values for p-liposome_20%-m immunized animals. Thus for both p-liposome and p-VLP_Qβ particles, the IgG antibodies were able to bind the TNF-α protein, although at different efficiencies. To determine the avidity of the IgG antibody response against the mouse TNF- α protein, we used 8M urea to treat the antigen-antibody complexes for 5 min after 2 hours of incubation, followed by washes and regular ELISA procedures. As shown in Fig. 5B, both IgG responses measured at Day 11 after first injection fell significantly after this treatment. Specifically, for sera from p-liposome_20%-m immunized mice, even at 1:100 dilution, no significant IgG response could be detected above background; for sera from p-VLP_Qβ immunized mice, IgG response could be detected at 1:1600 dilution, but decreased to background levels upon further dilution. These results thus suggest that at Day 11 after first immunization, both IgG responses were still of low avidity, with an avidity index < 30%, although both could bind to the recombinant mouse TNF-α proteins.

Figure 5. Anti-TNF-α-protein antibody response in wild-type C57BL/6 mice measured 11 days after first inoculation with either p-liposome_20%-m or p-VLP_Qβ.

Sera from mice immunized with p-liposome_20%-m (4.5 μg TNF-α peptide, grey columns) and p-VLP_Qβ (0.5 μg TNF-α peptide, white columns) were serial diluted as indicated. Each column in the figure represents the mean of OD450 values obtained from four mice of each group (N=4). Error bars represent the standard errors. After two hours of incubation in ELISA plates, the antigen-antibody complexes were treated with either PBS for 5 min (Panel A) or 8 M urea for 5 min (Panel B), followed by washes and regular ELISA procedures in order to estimate antibody avidity.

Why were p-VLP_Qβ particles much more potent than the aforementioned liposomal nanoparticles in eliciting IgG responses? Among other possibilities, one difference was the size of the particles. The liposomal particles that we so far prepared had diameters around 150 nm, whereas the p-VLP_Qβ particles were approximately threefold smaller. To examine the effect of particle size on liposomal immunogenicity, we prepared liposomal particles that were about 75 nm in diameter. Again, using 20% maleimide lipids, we conjugated TNF-α peptides to these liposomes at a density of 9.4×10⁴ peptides/μm² (1720 molecules per liposome on average), similar to our previous p-liposome_20%-m with peptide density of 8.2 ×10⁴ peptides/μm² (6030 ± 428 molecules per liposome). Following the same immunization protocol and using the same peptide dose, we immunized C57BL/6 mice. As shown in Fig. 6, these smaller particles elicited IgM and IgG responses that were very much comparable to those of larger liposomes. Therefore, the size of these particles did not apparently explain the huge difference in the immunogenicity between p-VLP_Qβ and p-liposomes.

Figure 6. Anti-TNF-α-peptide antibody response in wild-type C57BL/6 mice as a function of liposomal particle size.

Time courses of ELISA OD values for IgM (a) and IgG (b) from 1:100 diluted mouse sera collected at week 2, week 4 and week 6, 11 days after each inoculation. Each data point in the figures represents the mean of OD450 values obtained from four mice of each group (N=4). Error bars represent the standard errors. Data from soluble TNF-α peptide (4.5 μg) were also plotted for comparison (cross).

Antibody responses in gene knockout mice lacking cognate T cell help

In general, B cells must receive T help in order to efficiently proliferate, class switch, and differentiate into long-lived plasma cells. T cells that recognize self antigens are usually eliminated or tolerized during their maturation (38). Because the TNF-α peptide is a bona fide self antigen and is the only protein antigen in the liposome-based vaccines, cognate T help may not be available to sustain B-cell proliferation. Search of the immune epitope database (www.iedb.org) (39) did not yield any known T-cell epitope for this peptide. The lack of boosting effect upon successive inoculations, in spite of the low titer values we observed upon a series of p-liposome immunizations (Fig. 2 and Supplementary Fig. 1), suggest the lack of T cell help in these IgG responses. In contrast, both the bacteriophage Qβ coat protein and major capsid protein L1 in papillomavirus-like particles are foreign to mice, the inclusion of these proteins in the antigen presentation platforms may have thus served this role of foreign antigens for efficient recruitment of T cell help to sustain B cell activation for more potent IgG response.

To test this hypothesis, we utilized two different strains of gene knockout mice that were both in the C57BL/6 background (Materials and Methods). The first line was deficient in both alpha/beta and gamma/delta T-cell receptors (22) (denoted as TCR⁻/⁻), while the second line did not express MHC class II molecules on the surface of splenic B cells or dendritic cells (23) (denoted as MCII⁻/⁻), and thus both lines were deficient in T-dependent B cell activation. We first inoculated these two lines of gene knockout mice using p-liposome_20%-m, following the same immunization schedule and doses as we used for wild-type C57BL/6 mice. As shown in Fig. 7A, these two lines of gene knockout mice produced IgM responses that were identical (within error) to the wild-type C57BL/6 mice. Remarkably, these two lines of gene knockout mice also produced class-switched IgG responses that were similar in magnitudes to the wild-type C57BL/6 mice (Fig. 7B). One-way ANOVA test conducted for these data yielded p-values >0.05 for all three groups at Week 2, 4, 6, 10, 14 and 18, demonstrating that neither IgM nor IgG responses elicited in these animals were statistically different. Furthermore, no boosting effect of IgG response was observed in either of the gene knockout animals, consistent with the result seen in wild-type mice (Fig. 2). These data thus confirmed that the IgG response elicited by p-liposomes in wild-type mice was a result of T-independent B cell activation. These data also showed that class-switched IgG antibody responses against a self antigen could be elicited by these p-liposomes in the absence of cognate T cell help.

Figure 7: Comparison of anti-TNF-α-peptide antibody response in gene knockout mice with wild-type C57BL/6.

(A-B) Three groups of mice were inoculated with the same dose of p-liposome_20%-m (4.5 μg TNF-α peptide) following the same schedule: wild-type mice (C57BL/6J), TCR deficient mice (TCR^−/−) and MHC Class II deficient mice (MCII^−/−). The upward arrows denote the time of inoculations. Time courses of ELISA OD values for IgM (A) and IgG (B) measured from 1:100 diluted mouse sera, collected at week 2, week 4 and week 6, 11 days after each immunization, and week 10, week 14 and week 18 after the first inoculation. Each data point in the figures represents the mean of OD450 values obtained from four mice of each group (N=4). Error bars represent the standard errors. One-way ANOVA was conducted for sera from three mice groups collected at the same time point and p-values are >0.05 for all six groups at week 2, 4, 6, 10, 14, and 18. (C-D) Three groups of mice were inoculated with the same dose of p-VLP_Qβ (0.5 μg TNF-α peptide) following the same schedule: wild-type mice (C57BL/6J, circles), TCR deficient mice (TCR^−/−, diamond) and MHC Class II deficient mice (MCII^−/−, down triangle). The upward arrows denote the time of inoculations. Time courses of ELISA OD values for IgM (C) and IgG (D) measured from 1:100 diluted mouse sera, collected at week 2, week 4 and week 6, 11 days after each immunization. Each data point in the figures represents the mean of OD450 values obtained from four mice of each group (N=4). Error bars represent the standard errors. Data from p-liposome_20%-m (4.5 μg TNF-α peptide) were also plotted in panels C and D for comparison purpose (upper triangle). For panels C and D, p-liposome_20%-m was used as a reference to determine statistical difference between data points by Student’s T-test. The statistical difference was denoted using the same set of symbols as Figure 2.

We then used p-VLP_Qβ to immunize the two lines of gene knockout mice, following the schedules and doses that we used for p-liposomes and collected sera at designated time points to measure antibody responses using ELISA. As shown in Fig. 7C, the IgM response from these animals was on the same order of magnitudes in comparison to p-liposome_20%-m. Remarkably, the IgG response from these animals had dropped substantially to a level that could be compared to that elicited by p-liposome_20%-m (Fig. 7D). By Week 6 after the third inoculation, the differences in IgG responses had become insignificant between wild-type mice immunized with p-liposome_20%-m and gene knockout mice immunized with p-VLP_Qβ. This result thus demonstrates that the extraordinary immunogenicity of p-VLP_Qβ is determined in large part, by the ability of these particles to recruit T-cell help after B cell activation.

IgG subclasses of antibodies induced upon immunization

In order to obtain more mechanistic information about the antibody class switching, we determined the subclasses of the IgG antibodies induced in mice upon immunization under various conditions. These results are shown in Fig. 8 A through D. There are two major observations from these results. First, the TNF-α peptide conjugated to the surface of liposomes (p-liposome_20%-m) elicited IgG1 antibodies in wild-type C57BL/6 (condition 4), together with IgG2b and IgG3. This result was also confirmed in the TCR−/− (condition 5) and MCII−/− (condition 6) gene knockout mice, suggesting that this elicitation of IgG1 can occur in the absence of T cells. Previously it was shown that T-independent type II antigens such as NP-Ficoll could induce IgM antibodies together with lower amounts of IgG3 and IgG2b, but not IgG1 in C57BL/6 mice (40). Our results here thus suggest that the peptide-conjugated liposomal antigens use mechanisms that are distinct from those used by T-independent type II antigens in the induction of class-switched antibodies. Second, the peptide-conjugated liposomal antigens did not elicit IgG2c (Fig. 8C). On the other hand, the elicitation of IgG2c seems to be well correlated with the presence of nucleic acids in the immunization agents. The TNF-α peptide conjugated to Qβ was administered to wild-type C57BL/6 (condition 8), TCR−/− (condition 9) and MCII−/− (condition 10) mice respectively. In all cases, IgG2c was strongly elicited (Fig. 8C). Interestingly, when we added CpG DNA oligos to p-liposome_20%-m in the form of an admixture (condition 11), this agent also induced IgG2c, in sharp contrast to p-liposome_20%-m alone (condition 4). Throughout, neither liposome only (condition 2) nor the soluble TNF-α peptides (condition 3) elicited any statistically significant IgG subclass responses as compared to PBS control (condition 1), which is consistent with the results in Fig. 2B. In summary, this study on IgG subclasses clearly revealed that the form of antigens could bias the distributions of IgG subclasses elicited upon immunization, and p-liposome_20%-m alone could trigger the production of IgG1 in the absence of T cell help.

Figure 8. Anti-TNF-α-peptide IgG subclasses in mice measured 11 days after first inoculation with various agents.

Panels A-D show ELISA OD values for IgG1 (A), IgG2b (B), IgG2c (C) and IgG3 (D), respectively. The conditions 1 through 11 are PBS injection of C57BL/6 wild-type mice (1), liposome only injection of C57BL/6 wild-type mice (2), soluble peptides injection of C57BL/6 wild-type mice (3), p-liposome_20%-m injection of C57BL/6 wild-type mice (4), p-liposome_20%-m injection of TCR−/− mice (5), p-liposome_20%-m injection of MCII−/− mice (6), VLP_Qβ and soluble peptide injection of C57BL/6 wild-type mice (7), p-VLP_Qβ injection of C57BL/6 wild-type mice (8), p-VLP_Qβ injection of TCR−/− mice (9), p-VLP_Qβ injection of MHCII−/− (10), and p-liposome_20%-m and CpG injection of C57BL/6 wild-type mice (11). Throughout, the antibody responses were from 1:100 diluted mouse sera except condition 8, where 1:1600 diluted sera were used to avoid the saturation of ELISA readings. The dose of TNF-α peptide per injection was 4.5 μg for conditions 3, 4, 5, 6 and 11, while the dose of TNF-α peptide per injection was 0.5 μg for conditions 7, 8, 9 and 10. For condition 11, 20 μg CpG was mixed with p-liposome_20%-m before inoculation. For all four panels, soluble peptides (condition 3) was used as a reference to determine statistical difference between data points by Student’s T-test. The statistical difference was denoted using the same set of symbols as Figure 2. Error bars represent the standard errors (N=4). The IgG1 response in gene knockout mice (conditions 5, 6, 9 and 10) were low in magnitudes. However, these responses were reproduced in an independent repeat of the experiments using a second set of animals and statistically significant compared to those from soluble peptides.

Antigen-specific B cells revealed by flow cytometry

To further examine the mechanisms involved in this production of class-switched antibodies, we have developed a dual-functional liposome for tracking of antigen-specific B cells using flow cytometry (Materials and Methods). This dual-functional liposome (fp-liposome_6%-m) encapsulated the fluorescent dye Alexa-594 in the interior of the liposome, whose surface was also covalently conjugated with the aforementioned TNF-α peptide. Therefore, we expect that this dual-functional liposome can bind to TNF-α specific B cells and serve as a fluorescent indicator to reveal these antigen-specific B cells via flow cytometry. To this end, we inoculated C57BL/6 wild-type mice with PBS, p-liposome_20%-m, or p-VLP_Qβ, respectively. At Day 4 after inoculation, we prepared single-cell suspensions of splenocytes, incubated them with fp-liposome_6%-m, together with markers for mouse CD19, and subjected the cells to flow cytometry. As shown in Fig. 9, control mice inoculated with PBS showed negligible amounts of B cell populations that were specific for the fluorescent liposome (Q2 in Panel A), while mice inoculated with either p-liposome_20%-m (Panel B) or p-VLP_Qβ (Panel C) showed reproducible B cell populations that were specific for fp-liposome_6%-m. The percentages of these populations were significantly higher than those of control mice inoculated with PBS (Panel D). This difference suggests that the TNF-α specific B cells were triggered to expand after relevant antigen exposure, which allowed them to be detected above background. Furthermore, the fraction of TNF-α specific B cells from mice inoculated with p-VLP_Qβ was about threefold of that from mice inoculated with p-liposome_20%-m, which was qualitatively consistent with ELISA results. This study thus supports that antigen-specific B cells can expand upon exposure to the liposomal self-antigen in vivo and these B cells can be detected as early as Day 4 after immunization.

Figure 9. Flow cytometry analysis of TNF-α specific B cells from splenocytes at Day 4 after inoculation with various reagents.

Panels A-C show CD19⁺ B cells from mouse splenocytes harvested at Day 4 after inoculation with PBS (A, 100 μl), p-liposome_20%-m (B), or p-VLP_Qβ (C). The number of events shown in A-C are 101997, 172028 and 124628, respectively. The dose of TNF-α peptide was 4.5 μg for (B) and 0.5 μg for (C), respectively. The cytograms were gated for TNF-α peptide specific populations, indicated by Q2 in Panels A-C. The percentages of TNF-α specific B cells from Q2 are further shown and compared in Panel D. PBS-inoculated mice was used as a reference to determine statistical difference between data points by Student’s T-test in Panel D. The statistical difference was denoted using the same set of symbols as Figure 2. Error bars represent the standard errors (N=4).

No germinal center formation above background upon p-liposome_20%-m immunization

Induction of germinal centers is a hallmark of humoral immune responses to foreign antigens. We asked whether germinal centers were formed in response to vaccination with self antigen displaying liposomes. To address this question, we inoculated C57BL/6 wild-type mice with PBS control, p-liposome_20%-m, or p-VLP_Qβ, respectively. At Day 12 after inoculation, we prepared single-cell suspensions of splenocytes, and incubated them with markers for mouse CD19 and GL7 antigen, and subjected the cells to flow cytometry. As shown in Fig. 10, control mice inoculated with PBS did show a low percentage of GL7⁺ B cells, which might correspond to background level activation of B cells (Panel A). However, mice inoculated with p-liposome_20%-m also showed comparable level of GL7⁺ B cells (Panel B). This level of activation was statistically indistinguishable from that of PBS control (Panel D), suggesting that inoculation of p-liposome_20%-m alone did not bring up the level of B cell activation above background at Day 12 after inoculation. In contrast, GL7⁺ B cell population was evident in mice immunized with p-VLP_Qβ (Panel C), which was statistically higher than either PBS control or the mice immunized with p-liposome_20%-m (Panel D). These results indicate that there was no formation of stable germinal centers that could be distinguished from background at Day 12 after inoculation of the liposomal self-antigen. These results were also qualitatively consistent with the fact that no apparent increase of antibody titer was observed for the IgG antibodies elicited upon repeated inoculation of the liposomal antigen (Fig. 2).

Figure 10. Flow cytometry analysis of GL7 antigen positive B cells from splenocytes at Day 12 after inoculation with various reagents.

Panels A-C show CD19⁺ B cells from mouse splenocytes harvested at Day 12 after inoculation with PBS (A, 100 μl), p-liposome_20%-m (B), or p-VLP_Qβ (C). The number of events shown in A-C are 126549, 128190 and 114264, respectively. The dose of TNF-α peptide was 4.5 μg for (B) and 0.5 μg for (C), respectively. The cytograms were further gated for GL7⁺ populations, indicated by Q2 in Panels A-C. The percentages of GL7⁺ B cells from Q2 are further shown and compared in Panel D. PBS-inoculated mice was used as a reference to determine statistical difference between data points by Student’s T-test in Panel D. The statistical difference was denoted using the same set of symbols as Figure 2. Error bars represent the standard errors (N=4).

Discussion

The response of B cells to particulate antigens is highly relevant to both our understanding of host immune responses towards viruses and rationale design of vaccines. In this study, we have used carefully engineered liposomal nanoparticles that display a model self antigen peptide at tailored densities in order to investigate the factors that lead to antibody responses. In our particle design, we have chosen maleimide chemistry for conjugation of the self antigen onto liposomal surface. This covalent conjugation ensures the density of antigen is stable with time (Fig. 1F), which is in contrast to the alternative noncovalent nickel-nitrilotriacetic acid (Ni-NTA) technology, where the epitope density will decrease rapidly upon dilution into biological milieu (15). The stability of epitope density on these particles therefore allows us to interpret our results based on varied epitope densities with confidence.

When the epitope density exceeded certain threshold, these liposomal particles could elicit both IgM and IgG autoreactive antibodies in the absence of any adjuvants. In particular, we identified a threshold condition where IgG response was no longer significant despite the fact that these liposomal particles carried 360 ± 64 molecules of the antigenic peptides per particle on average (Fig. 3B). The IgG response was also elicited in two lines of gene knockout mice that were defective in either T cell receptors or MHC class II molecules on B cells, which further uncovered that this class switching is independent of cognate T cells. These conclusions were further supported by flow cytometry experiments where we could detect TNF-α specific B cells in mice as early as Day 4 after immunization with the liposomal antigen (Fig. 9B). The absence of these B cells in control mice suggest that these cells arose as a result of expansion in response to the specific antigen exposure. Altogether, these results provided an experimental validation for the hypothesis that epitope density could serve as a stand-alone signal (1) to trigger B cell secretion of class-switched IgG against a self antigen in vivo, independent of cognate T cell help and in the absence of any other adjuvants. The profiles of IgG subclasses induced by the liposomal self antigen include IgG1, IgG2b and Ig3 (Fig. 8), which are different from the classic Type II T-independent antigens such as NP-Ficoll, and suggest a different mechanism involved in this IgG elicitation.

Class switch recombination (CSR) provides the immune system with antibodies of different effector functions (41). IgG, in particular, offers benefits over IgM since its smaller size can easily gain access to extravascular space and offers protection at those sites (42). In addition, class-switched B cells may also gain advantage in signaling and survival over unswitched B cells due to the difference in Ig cytoplasmic tail (43). Although CSR can be conveniently induced in cell culture using Toll-like receptor ligands, the signal(s) to trigger CSR in vivo has been less clear (44). It was reported that gut B cells can undergo CSR to produce IgA in the absence of CD40 signaling or germinal center formation (45). Additionally, evidence suggesting that B-1 B cells and marginal zone B cells could produce class-switched IgG and IgA antibodies through T-cell independent pathways has been reported (46), although the order of molecular events that led to these T-cell independent CSR was not clear. In a recent study, it was shown that bacteriophage Qβ viral-like particles could induce class-switched and somatically mutated memory B cells in the absence of T cell help and also Bcl-6 expression in pregerminal center B cells (47). However, because Qβ viral-like particles possess both high epitope density and single-stranded RNA packaged inside the particles, which could serve as a potent Toll-like receptor 7 ligand, it remains unclear if epitope density alone could trigger CSR in the absence of T cells in vivo.

The results from current study show that epitope density is a distinct signal that can trigger B cell activation and secretion of class-switched IgG in vivo, independent of the cognate interactions with helper T cells and other adjuvants such as Toll-like receptor ligands. Whether this phenomenon is extensible to other antigens in general remains to be determined in the future. However, if it were true for other antigens, it could provide benefits to the immune system for fending various viral agents that typically display a dense array of epitopes on particle surface (1). Different viral strains carry surface antigens of different structures. While it may take time for affinity maturation to develop a perfect antibody against the different structures of viral surface antigens, to recognize and respond to the epitope density, a common feature of most viral agents, might offer a good strategy. Upon initial recognition of the epitopes on particulate antigens, even though the affinity towards individual epitope is low, the epitope density can drive B cell response quickly in the absence of T cells to secret both IgM and class-switched IgG antibodies. The IgG response, even in the absence of affinity maturation, may play important roles in B cell functionality and also the control of pathogen proliferation during early phases of infection. For example, it was observed that influenza-specific IgG responses could be mounted in influenza-infected mice that were defective in cognate T cell help (48). Although this CD4 T-cell independent IgG antibody response was low in its titer, it could promote resolution of primary influenza virus infection and also help prevent reinfection in mice (48). The signal(s) to trigger the production of the influenza-specific IgG in the absence of T cell help was unclear. However, based on current studies, it is tempting to speculate that this CD4 T-cell independent IgG antibody may arise from the mechanism that we hypothesized above, i.e., a threshold density of influenza-specific surface antigens such as the hemagglutinin or the neuraminidase, which is a subject for future research.

The results described in this work also open up several important questions that are worth studying in the future. First, are there particular subsets of B cells required for the T-cell independent IgG response observed in current study? Marginal zone B cells play critical roles in early response to T-independent particulate antigens (49, 50). The ability to track antigen-specific B cells using flow cytometry (Fig. 9) offers a potential way to identify those cells early on during this activation process and examine their status of B cell activation upon antigen exposure. Second, are other cells, such as dendritic cells, required for this T-cell independent IgG response? This question is relevant since dendritic cells have been implicated in the T-independent class switching of B1 and marginal zone B cells (44). On the other hand, recent studies by Hong et al. have clearly demonstrated that antigen-specific B cells themselves are essential and sufficient to present antigens to naïve CD4⁺ T cells and initiate antiviral response (51), in lieu of dendritic cells that are thought by many as required for the initiation of CD4⁺ T cell activation. These studies thus uncovered novel aspects of B cell function in humoral immunity and in the future it will be relevant to study the impact of epitope density on B cell secretion of cytokines, because differences in cytokine secretion by B cells may well signal other immune cells for collaborative efforts to contain the invading pathogens in a timely manner.

Our current study also suggests that high epitope density alone may break B cell tolerance in the absence of cognate T cell help or any other adjuvants in vivo. The model antigen that we employed in this study was a peptide derived from mouse TNF-α protein, a proinflammatory cytokine that is important in the pathogenesis of a number of chronic inflammatory diseases (52). Although the in vivo concentration of this protein that is required to develop B cell tolerance towards this protein has not been determined, our results indicate that potentially self-reactive B cells are available that can be activated by particulate antigens that display relevant self epitopes above a threshold of epitope density, in the absence of any other adjuvants. In fact, the ability of the IgG to bind to the recombinant TNF-α protein in our studies further supports this hypothesis. The implication of this result on the maintenance of B cell tolerance also warrants future studies.

Compared to the IgG response elicited by viral-like particles in wild-type mice, the IgG titer from the above T-independent B cell activation is lower by two to three orders of magnitudes (Fig. 4E). Consistent with this phenomenon, we could not detect germinal center formation above background in mice at Day 12 after inoculation of the liposomal self-antigen (Fig. 10). Strikingly, when the antigen-conjugated viral-like particles were administered into gene knockout mice that were deficient in T cell help, the IgG response fell to a level that could be compared to that of liposomal nanoparticles within the same order of magnitudes. The results clearly illustrate that T cell help dominates the IgG response to viral-like particles in wild-type mice and the major difference between antigen-conjugated viral-like particles and liposomal particles is their differential ability to recruit T cell help and maintain germinal centers after the initial B cell activation.

It is worth noting, however, that our current results do not exclude the possibility of transient formation of germinal centers upon antigen exposure. Germinal center formation in response to T-independent antigens has been reported (53, 54). In particular, a study led by MacLennan showed that large germinal centers could form in the absence of T cells in response to NP-Ficoll, although these germinal centers aborted abruptly at Day 5 after immunization (55). The formation of these transient germinal centers required the multivalent nature of the antigen, a threshold of the antigen dose, and also a threshold frequency of antigen-specific B cells. The antigen dose that we have used in current studies is more than six fold lower than the antigen dose used in those studies. Studies are underway to examine if germinal center structures may have formed even transiently in response to the liposomal self-antigens and if there is any effect of antigen dose on this process.

The molecular components or features of viral-like particles that are critical for this T cell recruitment remains to be determined in the future. Among other things, two potential factors that may come into play are: (i) the rigidity of epitope display, and (ii) the presence of nucleic acids in viral-like particles that may synergize B cell activation through the engagement of Toll-like receptors. In particular, substantial studies have supported the role of Toll-like receptor 9 ligands in augmenting B cell antiviral responses (56).

In summary, we showed that a peptide derived from a self antigen, upon conjugation to liposomes above a threshold of epitope density, could induce class-switched antibody responses in the absence of cognate T-cell help or other adjuvants. To the best of our knowledge, this is the first time that the epitope density was shown to be a stand-alone signal to trigger B cell secretion of class-switched antibodies in vivo in the absence of T cells. Our study has thus uncovered a fundamental aspect regarding B-cell activation in vivo and offered valuable insights to future vaccine design targeting self antigens, which could be useful in treating diseases where self antigens are associated, such as rheumatoid arthritis and Alzheimer’s disease.

Key points.

1. A threshold of epitope density for a self-antigen elicits IgG1, IgG2b and IgG3

2. The class-switched autoreactive antibodies were induced in the absence of T cell help

Abbreviations used in this article

NP-Ficoll

(4-hydroxy-3-nitrophenyl) acetyl-Ficoll

p-liposome

peptide-conjugated liposomes

p-liposome_20%-m

peptide-conjugated liposomes prepared with 20% molar ratio of maleimide lipids

CpG

A 20-mer DNA oligo of the following sequence, 5’-tccatgacgttcctgacgtt-3’ with a full phosphorothioate backbone

VLP_Qβ

bacteriophage Qβ viral-like particles

SMPH

Succinimidyl-6-[(β-maleimidopropionamido)hexanoate]

p-VLP_Qβ

peptide-conjugated bacteriophage Qβ viral-like particles

TCR^−/−

a gene knockout mouse deficient in both alpha beta and gamma delta T-cell receptors

MCII^−/−

a gene knockout mouse deficient in MHC class II molecules on splenic B cells or dendritic cells