Eliciting unnatural immune responses by activating cryptic epitopes in viral antigens

Abstract

Antigenic variation in viral surface antigens is a strategy for escaping the host’s adaptive immunity, whereas regions with pivotal functions for infection are less subject to antigenic variability. We hypothesized that genetically invariable and immunologically dormant regions of a viral surface antigen could be exposed to the host immune system and activated by rendering them susceptible to antigen-processing machinery in professional antigen-presenting cells (APCs). Considering the frequent antigen drift and shift in influenza viruses, we identified and used structural modeling to evaluate the conserved regions on the influenza hemagglutinin (HA) surface as potential epitopes. Mutant viruses containing the cleavage motifs of cathepsin S within HA were generated. Immunization of mice showed that the mutant, but not the wild-type virus, elicited specific antibodies against the cryptic epitope. Those antibodies were purified, and specific binding to HA was confirmed. These results suggest that an unnatural immune response can be elicited through the processing of target antigens in APCs, followed by presentation via the major histocompatibility complex, if not subjected to regulatory pathways. By harnessing the antigen-processing machinery, our study shows a proof-of-principle for designer vaccines with increased efficacy and safety by either activating cryptic, or inactivating naturally occurring, epitopes of viral antigens.—Lee, Y. J., Yu, J. E., Kim, P., Lee, J.-Y., Cheong, Y. C., Lee, Y. J., Chang, J., Seong, B. L. Eliciting unnatural immune responses by activating cryptic epitopes in viral antigens.

The body’s antiviral defense system has armed itself in various ways to successfully overcome infections during the long period of coexistence with pathogens (1). Such defense mechanisms mainly consist of innate and adaptive immune responses. Innate immune responses are immediately activated upon virus infection, and their main constituents, such as cytokines, complement components, and NK cells, induce an antiviral state that interferes with viral replication and spread (2, 3). Adaptive immune responses driven by T cells and antibodies provide highly specific and long-lasting protection against virus infections (4, 5). However, despite that elaborate immune machinery, viruses continue to infect humans by adopting a variety of strategies to circumvent or to inactivate host defense systems (6).

A well-established mechanism to escape the host’s adaptive immunity is antigenic variation, most prominently observed among RNA viruses (7). The error-prone nature of RNA polymerases causes sequence variability in their replication cycle, enabling frequent generation of “breakout” mutant species (8). Therefore, naturally occurring “epitopes” in viral antigens, usually targeted by neutralizing antibodies, are intrinsically subjected to antigenic variations. However, some sites on viral surface proteins are less susceptible to mutation (9–11). Those conserved regions are usually related to the functions essential for virus infection, such as receptor binding or membrane fusion (12) and, consequently, are much less affected by antigenic variability. For instance, despite frequent antigenic drift by influenza hemagglutinin (HA), its stalk domain remains relatively well conserved across viruses because of its essential role in viral-membrane fusion (13, 14). As such, redirecting the antibody responses from the variable HA1 domain, where most known, neutralizing epitopes are localized (15, 16), to the conserved HA2 domain, through engineered HA antigens, remains the basis for the “universal” vaccine approach (17–19). This probably means that most current approaches that use natural antigens, either from infectious viruses or from recombinant hosts, are intrinsically limited in inducing sufficient immunogenicity against the conserved and remaining cryptic sites. Thus, natural immune responses are compromised in eliciting effective protection from reinfection. To overcome this obstacle, new strategies are needed to activate or increase immunogenicity against the conserved regions.

After infection or vaccination, viral antigens are taken up by professional antigen-presenting cells (APCs), and the epitope peptides are subsequently loaded onto the major histocompatibility complex (MHC) on the surface of infected cells (20). We hypothesized that the genetically conserved domains are hidden from immune surveillance by a lack of processing in APCs, and subsequently fail to present on the MHC of infected cells. If conserved region–specific B cells take up and process viral antigens and then present that domain to helper CD4 T cells, the B cells could be selectively matured and elicit specific antibodies. The provision of a new proteolytic cleavage site adjacent to those invariant regions should permit scission of the antigen by directing the processing enzymes in APCs to the novel cleavage site, allowing the previously “cryptic” epitope to be presented in an MHC-dependent manner. A caveat for that assumption is that the epitope of interest is not cross-reactive with the self-proteome, such that the cryptic epitope (CE)-specific B cells are not selected for deletion during development.

Herein, this hypothesis was tested with influenza virus HA. Influenza is an ideal system for the present purpose because of its well-known antigenic drift and shift mechanisms (21–23). Because of its high propensity for genetic mutations of surface antigens, the virus can easily evade preexisting immunity systems acquired from previous infection or vaccination, resulting in annual outbreaks and occasional pandemics with enormous medical and socioeconomic burdens (24, 25). In this study, by structural modeling, we identified a conserved region on the surface of HA, and the amino acids flanking that site were modified into cathepsin S (Cat S) cleavage motifs. Cat S is one of the major proteases engaged in antigen processing (26–28). This enzyme is not only involved in the degradation of an invariant chain responsible for preventing premature loading of peptides in the antigen presentation of MHC class II (29, 30) but also has a direct role in antigen processing (27). Mutant viruses were generated in the genetic background of cold-adapted X-31 (X-31ca) with a reverse genetic (RG) system (31, 32). In a mouse model, the mutant virus elicited unnatural, and yet specific, antibody responses to the activated novel site, which otherwise remains cryptic in natural viruses. Implications of the present findings are discussed with respect to the rational design of vaccines.

MATERIALS AND METHODS

3D modeling of HA trimers and structural characteristic analysis

The 3D structures of HA trimers obtained from RCSB Protein Data Base (PDB; Research Collaboratory for Structural Bioinformatics, New Brunswick, NJ, USA; http://www.rcsb.org/) were used for homology modeling with SWISS-MODEL (Protein Structure Bioinformatics Group, Basel, Switzerland; http://swissmodel.expasy.org/) for visualization and structural analysis. The H1 structure of A/Puerto Rico/8/1934 (PR8; PDB ID 1RU7, 2YP2, and 2WR1) was used as the template for the homology modeling of H3 (A/Sydney/5/1997) and H5 (A/Indonesia/5/2005). University of California, San Francisco (UCSF) chimera (Resource for Biocomputing, Visualization, and Informatics, San Francisco, CA, USA; https://www.cgl.ucsf.edu;) was used for structural visualization. To calculate the properties of the amino acids in the predicted, conserved sites, all HETATMs (atomic coordinate record containing the X, Y, Z orthogonal Å coordinates for atoms in nonstandard residues of proteins in the PDB data file format), including water molecules, were removed. The Environment and virtual docking of atoms (ENVA) Program (Korea Institute of Science and Technology, Seoul, South Korea; http://www.biomapstore.com/), based on the Shrake–Rupley algorithm, was used to calculate the accessible surface area and polarity of the conserved sites (33).

Mutagenesis of the HA gene

PCR was used to insert the Cat S cleavage motifs into wild-type influenza HA protein (A/Indonesia/5/2005). Using a pHW2000 vector containing the wild-type influenza HA gene as a template (31), PCR was performed with primers containing the Cat S cleavage sites, in which the amino acid sequences at positions 303 and 325 (S303L and N325V) in the HA protein were replaced. This modified HA gene was rescued in infectious viruses by the RG system.

Generation of H5N1 cold-adapted vaccine containing Cat S cleavage motifs

The reassortant vaccine virus was constructed based on the previously established RG system (34). The mutant HA (S303L and N325V) and wild-type NA genes derived from A/Indonesia/5/2005 were cloned into a pHW2000 vector and cotransfected with the 6 pHW2000 plasmids, each encoding the internal gene of cold-adapted X-31 (X-31ca), into 293T cells. After 3 d of transfection, supernatants were collected and inoculated into 11-d embryonic chicken eggs for 3 d at 33°C. The mutant live virus, Indo M1/CA, was harvested from the allantoic fluid of infected eggs and was titrated by plaque assay with Madin–Darby canine kidney (MDCK; American Type Culture Collection, Manassas, VA, USA) cells. The mutations in the rescued transfectant virus were confirmed by sequencing and found to be stably maintained during 3 consecutive passages in eggs.

Growth test of H5N1 cold-adapted vaccine containing Cat S cleavage motifs

To analyze the temperature-dependent growth profiles of Indo M1/CA, MDCK cells were infected at a multiplicity of infection of 0.001 and were incubated at 30, 33, 37, or 39°C after infection. The supernatants were harvested every 24 h, and the infectious viral titers were determined by plaque assay in MDCK cells.

Protein expression

We used the N-terminal RNA interaction domain of human lysyl-tRNA synthetase (hRID) as a fusion partner to facilitate protein expression (35). The lysyl-tRNA synthetase (LysRS) gene of the plasmid pGE-LysRS (36) was replaced by hRID to generate the plasmid pGE-hRID. The expression plasmid pGE-hRID-CE-CE-6xhis (hRID-CEx2), containing the CE via Cat S cleavage, was obtained by PCR. The PCR amplicon was digested with BamHI/EcoRV and EcoRV/SalI, and then ligated into the plasmid pGE-hRID. By overlapping PCR using specific primers, the same Cat S cleavage motifs as those in Indo M1/CA were introduced to HA_247–406, and the final DNA fragments were digested at BamHI and SalI sites and inserted into the plasmid pGE-hRID to yield pGE-hRID-HA_247–406-6xhis (no mutation); pGE-hRID-HA_247–406 and S303L-6xhis (N-mutation); and pGE-hRID-HA_247–406 and N325V-6xhis (C-mutation). Using the expression plasmids, protein expression, SDS-PAGE analysis, and purification were conducted as previously described (37). The purified proteins were stored with 20% glycerol at −20°C.

Cat S cleavage test in vitro

Cleavage of each protein was conducted in a buffer, with conditions mimicking those of the antigen-processing compartment (38). Buffers of each protein solution were changed to citrate phosphate (pH 5.0), with centrifugal filters (Amicon Ultra; MilliporeSigma, Burlington, MA, USA). Cathepsin S (MilliporeSigma) was added to 6 mM l-cysteine (MilliporeSigma), and the mixtures were incubated at 37°C for 0, 1.5, and 3 h. The cleavage was monitored by SDS-PAGE followed by Coomassie blue R-250 staining. After SDS-PAGE, the sensitivity to Cat S was deduced by comparing the relative intensities of protein bands during densitometric scanning.

Animal test and ethics statement

Six-week-old female BALB/c mice were purchased from Orient Bio (Seongnam, South Korea) and maintained under specific-pathogen–free conditions. Mice were anesthetized by intramuscular injection of 400 μg alfaxalone (3-α-hydroxy-5-α-pregnane-11, 20-dione, Alfaxan; Jurox, Rutherford, Australia) before intranasal administration of 50 μl of virus suspension. Retro-orbital bleeding was performed to collect sera from immunized mice for IgG antibody titration. Mice were euthanized, and a lung single-cell suspension was obtained by passing lung tissue through a 70-μm cell strainer into serum-free Iscove’s modified Dulbecco’s medium (IMDM). The lung monoclonal cells were collected by centrifugation to investigate T cell responses. All animal studies were performed with 5 mice/group and in accordance with the Institutional Animal Care and Use Committee of Yonsei University (IACUC-A-201509-446-02).

Antibody analysis

ELISA was used to detect CE-specific antibodies in mouse sera. Plates were coated with 0.25 μg/well of hRID or hRID-CEx2. After blocking and washing, plates were incubated with diluted sera for 1 h at 20°C. After washing, the plates were incubated with horseradish peroxidase–conjugated secondary goat anti-mouse IgG antibody (Bethyl Laboratories, Montgomery, TX, USA) for 1 h at 20°C. After washing, the plates were incubated with 3,3′5,5′-tetramethylbenzidine (TMB) substrate solution (BD Biosciences, Franklin Lake, NJ, USA) for 30 min at 20°C in the dark. The colorimetric reaction was stopped by adding 2 N H₂SO₄ solution, and the absorbance was measured at 450 nm on the ELISA reader. An additional preincubation step was added to the ELISA protocol. Before the reaction between the third boosting sera and 10⁶ plaque-forming units (PFUs) of Indo M1/CA and hRID-CEx2, the mouse sera were preincubated with various concentrations of hRID or hRID-CEx2 (0, 10, 20, and 40 μg/ml) at 37°C for 1 h.

Binding assay

To evaluate the CE-specific antibodies for the binding of HA, a series of 2 tests were conducted: 1) purification of CE-specific antibodies from sera, and 2) ELISA of recombinant HA (rHA; Sino Biological, Beijing, China) or viruses using purified antibodies. Plates were coated with 0.25 μg/well hRID-CEx2 and incubated at 4°C overnight. To identify the background by hRID in hRID-CEx2, hRID was also used as a control under the same set of conditions. Then, blocking and reaction with diluted sera for 1 h at 20°C were conducted sequentially. After washing, glycine elution buffer (pH 2.8; 0.1 M glycine, 0.5 M NaCl, 0.05% Tween-20) was added and allowed to react for 3 min on ice for the elution of coating protein-specific antibodies, and then the eluate was immediately transferred to a 1.5-ml tube containing Tris buffer (1 M, pH 8.1) for neutralization. The elution and neutralization steps were repeated once, and then, the plate was washed, followed by a combination of the wash and the 2 eluates. Next, ELISAs were performed with the eluates containing hRID- or hRID-CE–specific antibodies and rHA or viruses as described in the antibody analysis section. Viruses were concentrated 10-fold by centrifugal filters (AmiconUltra-15) and incubated with 8 M urea at 50°C for 1 h. The solution was diluted to 2 M immediately before coating.

Flow cytometric analysis

To analyze cytokine-producing T cell populations, the lung lymphocytes were resuspended in IMDM supplemented with 10% heat-inactivated fetal bovine serum (10% IMDM) and stimulated with 10 μg of synthetic cryptic-epitope peptide (LSMPFHNIHPLTIGECPKYVKSVR) in 10% IMDM containing 10 ng of recombinant human IL-2 (BioLegend, San Diego, CA, USA) and Brefeldin A (1:1000; Thermo Fisher Scientific, Waltham, MA, USA) at 37°C for 5 h in dark. After stimulation, the cells were incubated with rat anti-mouse CD16/CD32 (BD Biosciences) blocking antibody for 10 min at room temperature, and then, the lung cells were stained with anti-CD4 (RM4-5; BioLegend) and anti-CD8 (53-6.7; BioLegend) antibodies for 30 min at 4°C in the dark. For intracellular cytokine staining, the stained cells were fixed with fluorescence-activated cell sorting lysing solution (BD Biosciences) for 20 min at room temperature, and permeabilized with fluorescence-activated cell sorting buffer (0.5% fetal bovine serum, 0.09% NaN₃ in PBS) containing 0.5% saponin (MilliporeSigma) for 15 min at room temperature. Then, those cells were stained with anti-IFN-γ (XMG 1.2; BioLegend) for 30 min at room temperature in the dark. The stained cells were analyzed by LSR Fortessa (BD Biosciences), and all flow cytometry data were analyzed by FlowJo software (Tree Star, Ashland, OR, USA).

Statistical analysis

All values are expressed as the means ± sd of each cohort. A Student’s t test was used when comparing a control and other groups. A value of P < 0.05 was considered statistically significant.

RESULTS

Identification and evaluation of conserved sites on HA and structure analysis

The present analysis identified 9 conserved sites in the HA1 subunit of influenza HA (10). Then, each site was further evaluated for potential accessibility to specific antibodies. The selection criteria included: 1) conservation among H1, H2, H3, and H5 subtypes; 2) sites containing ≥6 residues as a minimum length required for a peptide to elicit an antibody response; and 3) sites where >50% residues had a conservation score of 0.9–1.0 calculated from multiple sequence alignment. Previously, Sahini et al. (10) used the monomer conformation of H3 (PDB: 1HGJ), but here, we used the trimer structure as a more-relevant conformation for the present purpose, considering that influenza HA is a homotrimeric membrane glycoprotein, and some antigenic sites require HA trimerization (39). To determine whether the predicted sites are suitable for binding to their specific antibodies, structural characteristics, such as accessible surface area and polarity, were analyzed in HA trimers (Fig. 1 and Table 1). The H1 trimer structure (A/Puerto Rico/8/1934, 1RU7) was adopted from the PDB site, and trimers of H3 (A/Sydney/5/1997) and H5 (A/Indonesia/5/2005) were generated by homology modeling with SWISS-MODEL using PDB ID 2YP2 and 2WR1 as templates, respectively. Those structures were visualized with 9 conserved sites (sites 1–9) by UCSF, chimera (Fig. 1). Invariable regions were presented on the HA structures by position (top) and exposed surface area (bottom) (Fig. 1). Based on the trimers, the accessible surface area and polarity of each site were calculated by the ENVA program built on the Shrake–Rupley algorithm, and the data were expressed as percentages (Table 1). Despite excellent scores in 2 parameters, site 5 was excluded because it overlapped with the previously identified antigenic site D of H3 (40). Sites 1, 2, 4, 6, and 7 were of low priority because of overall poor scores. Site 3 represented relatively high exposed surface area but low polarity. Overall, sites 5, 8, and 9 were ranked relatively high on both indicators. Among those, site 8 carried the most conserved residues, was expected to maintain the conserved amino acids even after processing by endoproteases and exopeptidases in APCs, and was finally selected as a candidate novel epitope. A BLAST search (Basic Local Alignment Search, National Center for Biotechnology Information, Bethesda, MD, USA) of the human and mouse proteomes available at NCBI confirmed that there was little sequence homology (alignment scores <40) to the candidate sequence.

Figure 1.

Conserved regions mapped on the HA trimers. Trimeric structures of A/Puerto Rico/8/1934 (H1) (A), A/Sydney/5/1997 (H3) (B), and A/Indonesia/5/2005 (H5) (C) are shown with the conserved sites highlighted in different colors: site 1 in purple, site 2 in green, site 3 in blue, site 4 in yellow, site 5 in black, site 6 in cyan, site 7 in orange, site 8 in brown, and site 9 in pink. Sialic acid as the cell receptor is shown in red. The HA structures are represented with 90% transparency and ribbons to show the position of the invariable regions (top). The exposed surface areas of site 1–9 are visualized by changing the transparency of the HA structures to 0% (bottom).

TABLE 1.

Structural characteristics of predicted sites in HA

Site	Residues, n (conservation/total)	Average of surface (%)	Average of polarity (%)
1	4/7	4.58 (6)	46.67 (4)
2	6/8	5.5 (5)	41.06 (5)
3	6/13	14.89 (2)	41.06 (5)
4	6/9	0.67 (8)	35.56 (8)
5	4/9	11.58 (3)	57.83 (1)
6	5/6	0.13 (9)	27.6 (9)
7	5/9	1.67 (7)	40.93 (7)
8	8/18	8.04 (4)	46.79 (3)
9	4/8	16.92 (1)	56.58 (2)

The average values for each site are represented. Numbers in parentheses indicate ranks.

Generation and properties of a mutant virus carrying Cat S cleavage motifs

Cat S is one of the lysosome-resident cysteine proteases involved in antigen processing. It is not only essential for the late stages of Ii degradation but also has a crucial role in the generation of particular epitopes (27, 41). Although the substrate specificity for Cat S is not yet established, a previous study (42) suggested that the specificity was mainly determined by the P-2, P-1′, and the P-3′ substrate positions. Guided by that study, we attempted to introduce the cleavage sites on both sides of site 8 by PCR-based site-directed mutagenesis using the HA genes of A/Puerto Rico/8/1934, A/Sydney/5/1997, and A/Indonesia/5/2005 as templates. Two types of mutants were made separately: one that minimized genetic alterations, and another that increased Cat S specificity (Supplemental Fig. S1). However, we failed to generate most of the attempted mutant viruses, except for mutation 1 in the A/Indonesia/5/2005 strain (Fig. 2A), probably because of the loss of viral fitness by structural perturbation within HA induced by mutations. Mutation 1 replaced 2 residues from the original HA (S303L, N325V), which turned out to involve the fewest changes tested among different mutants. The mutations were supposed to generate potential CE, given that the region was expected to be processed by Cat S, which was flanking site 8 (A/Indonesia/5/2005, HA_303–326).

Figure 2.

Growth, attenuation, genetic stability, and Cat S cleavage of Indo M1/CA. A) The amino acid sequence of A/Indonesia/5/2005 HA Site 8 (HA_301–330) is shown in <Original>. Mutation 1 involved a change in 2 residues from the Original (S303L, N325V). Modified amino acids are highlighted in red for S303L and N325V mutations in <Mutation 1>, showing the CE in brackets. B) MDCK cells were infected with 0.001 multiplicity of infection (MOI) of Indo M1/CA and were incubated at different temperatures. Every 24 h after infection, supernatants were collected for viral titration by plaque assay on MDCK cells. C) Mice were vaccinated with 10⁵, 5 × 10⁵, and 10⁶ PFU of Indo M1/CA or received PBS as a control. Weight changes were observed daily for 2 wk. D) Chromatograms show the nucleotide sequences of purified Indo M1/CA plaques after 3 passages in eggs. Modified nucleotides for introduction of Cat S cleavage motifs (S303L and N325V) are shown in blue shading. E) Model proteins for Cat S cleavage test (No mutation, N-mutation, and C-mutation) were incubated with Cat S in the cleavage buffer at 37°C for 0, 1.5, or 3 h and were analyzed by SDS-PAGE. Compared with those before the reaction with Cat S (0 h), the relative intensities of bands after the reaction (1.5 and 3 h) were estimated by densitometric scanning.

Using an RG system, we generated a reassortant virus, Indo M1/CA, carrying HA (mutation 1) and NA from A/Indonesia/05/2005 in the genetic background of X-31ca (31, 32). To determine whether Indo M1/CA maintained the cold-adapted (ca) and temperature-sensitive phenotypes, the growth properties at various temperatures in MDCK cells were examined (Fig. 2B). The viral titers at lower temperatures of 30 and 33°C were ∼10⁵ PFU/ml, ∼10-fold higher than those at 37°C. Indo M1/CA was highly restrictive in replication at 39°C. These results confirmed the intrinsic ca and temperature-sensitive phenotypes associated with X-31ca (32). However, the peak titer of Indo M1/CA was ∼10-fold lower than that of Indo WT/CA, containing the HA and NA genes from A/Indonesia/05/2005 and the other 6 genes from X-31ca (43). Indo M1/CA was then evaluated for attenuated phenotypes in mice. Female BALB/c mice were intranasally inoculated with 10⁵, 5 × 10⁵, or 10⁶ PFU of Indo M1/CA or PBS as a control, and weight changes were monitored daily for 2 wk (Fig. 2C). Although mice infected with 10⁵ PFU did not show any symptoms of virulence or weight loss, similar to those in the control group, mice infected with 5 × 10⁵ PFU exhibited ∼10% weight loss at 3 d after infection, but began to steadily recover at 4 d after infection. In the case of the infection with 10⁶ PFU, body weight was reduced by 20% at 4 d after infection but recovered thereafter. Thus, all mice survived inoculation with various doses of Indo M1/CA, albeit with weight loss at the higher doses. Cat S cleavage motifs in Indo M1/CA were genetically stable over 3 consecutive passages in embryonated eggs (Fig. 2D).

Aided by hRID as a solubility and folding enhancer (35, 36), all 3 model proteins carrying Cat S cleavage motifs were expressed and purified as soluble forms. The no-mutation construct represented a part of A/Indonesia/5/2005 HA that included site 8 (HA_247–406) without mutation, whereas the N- and C-mutation constructs carried the S303L and N325V replacements flanking the epitope, respectively. The proteins were treated with Cat S, and cleavage was monitored for 3 h by SDS-PAGE analysis (Fig. 2E). Evidently, the band intensity for the N- and C-mutation constructs decreased by 54 and 27%, respectively, reflecting the sensitivity to Cat S. This was in contrast to the no-mutation construct, which remained stable over prolonged incubation (92%, after 3 h reaction). The results showed that the mutations introduced into Indo M1/CA increased the sensitivity to Cat S in vitro.

Antibody responses against the conserved site in HA

To determine whether Indo M1/CA could induce specific antibody responses against the conserved epitope region, mice were vaccinated with various doses of Indo M1/CA (5 × 10⁵ or 10⁶ PFU) or Indo WT/CA (10⁵ or 5 × 10⁵ PFU) 4 times at 2-wk intervals. Mouse sera were collected for each booster dose (first, second, and third booster). Compared with that of Indo WT/CA (5 × 10⁵ PFU), a higher dose of Indo M1/CA (10⁶ PFU) was used, taking into account the increase in attenuation introduced to HA (Cat S cleavage mutation) in addition to cold adaptation in internal genes (44). The potential reactivity of mouse sera to the CE was assessed by ELISAs using hRID-CEx2 or hRID as the coating antigen (Fig. 3). The model protein hRID-CEx2 was designed to express repeated CE units (target site of Cat S in mutation 1) for efficient recognition by specific antibodies. CE was composed of 24 aa containing the CE sequence flanked by Cat S cleavage motifs (Fig. 2A). Sera from mice vaccinated with Indo M1/CA developed a low seroreactivity against hRID-CEx2 after the first booster, but that reactivity was increased by the second and third boosters (Fig. 3A). The higher-dosed group (10⁶ PFU) showed stronger reactivity than the lower-dosed group (5 × 10⁵ PFU) under the same conditions. Control sera from mice vaccinated with Indo WT/CA reacted only weakly against hRID-CEx2, when compared with those from mice vaccinated with Indo M1/CA, and that reactivity failed to increase even after repeated boosting (Fig. 3A). As an additional control, hRID without the epitope peptides showed very low reactivity via ELISA in all sera tested (Fig. 3B). The specific ELISA response of hRID-CEx2 calculated by subtraction of the control value of hRID cargo is summarized in Fig. 3C. Evidently, a specific-antibody response was directed to the activated epitope, and the magnitude of the response was augmented by boost immunization. Further, a depletion experiment was also performed. The sera from the mice vaccinated 4 times with 10⁶ PFU of Indo M1/CA were preincubated with various concentrations of hRID-CEx2 or hRID (0, 10, 20, and 40 μg/ml), and then used as primary antibodies in ELISAs against hRID-CEx2 (Fig. 3D). Preincubation with the control hRID failed to reduce the ELISA reactivity, irrespective of the concentration of hRID. However, the seroreactivity of the sera fell sharply after preincubation with hRID-CEx2 in a concentration-dependent manner. Of note, Cat S cleavage mutations in CE units may create novel epitopes because of the CE–CE repeat, and that may contribute to non–CE-specific binding. However, hRID-CEx2 showed the similar reactivity against CE-specific antibodies, regardless of the mutations (Supplemental Fig. S2). These results demonstrated that Indo M1/CA, carrying Cat S cleavage motifs, activated the epitope in HA, which otherwise would have remained cryptic and induced unnatural antibody responses to a novel epitope.

Figure 3.

Confirmation of CE-specific antibody production from the mice vaccinated with Indo M1/CA. Mice were vaccinated with Indo M1/CA (5 × 10⁵ or 10⁶ PFU), Indo WT/CA (10⁵ or 5 × 10⁵ PFU), or PBS as a control and were then boosted 3 times at 2-wk intervals with the same dose (priming, first booster, second booster, and third booster). Sera were collected from vaccinated mice at 4, 6, and 8 wk (first booster, second booster, and third booster) and then used as primary antibodies in ELISA. A, B) ELISAs were performed with 0.25 μg/well hRID-CEx2 (A) or hRID (B), and the sera were diluted to 1:100. C) In the same sample, the difference in absorbance between the 2 proteins was calculated. D) The third booster sera with 10⁶ PFU of Indo M1/CA were incubated with various titers of hRID-CEx2 or hRID (0, 10, 20, and 40 μg/ml) at 37°C for 1 h, and then ELISAs were conducted using 0.25 μg/well hRID-CEx2. The sera were preincubated and diluted to 1:100. OD, optical density. *P < 0.05, **P < 0.01 when compared with the PBS control or with 0 μg/ml from the preincubated group.

Binding assay between purified CE-specific antibodies and HAs

To confirm whether CE-specific antibodies induced by Indo M1/CA vaccination could directly bind to homologous and heterologous HAs, mice were inoculated with Indo M1/CA (10⁶ PFU), Indo WT/CA (5 × 10⁵ PFU), or PBS (as a control) 4 times at 2-wk intervals and then sacrificed to collect sera. To purify CE-specific antibodies, sera were bound to hRID-CEx2 coated on the ELISA plate. After rinsing, the antibodies were eluted off the plate by treating with glycine elution buffer (pH 2.8). hRID as a coating protein was included as a control. ELISAs were then performed with eluates as purified antibodies and rHAs from 3 different subtypes—PR8 (A/Puerto Rico/8/1934, H1), Sydney (A/Sydney/5/1997, H3), and Indonesia (A/Indonesia/5/2005, H5)—as coating antigens (Fig. 4A, B). Specific binding was observed between eluates purified from Indo M1/CA by hRID-CEx2 and the homologous Indonesia HA, whereas antibodies from Indo WT/CA failed to bind (Fig. 4A). None of the purified antibodies bound to the control cargo protein hRID (Fig. 4B). To confirm whether the CE-specific antibodies binding to homologous rHA could also bind to urea-denatured viruses, viruses (mutant, Indo M1/CA; wild type, Indo WT/CA) were concentrated 10-fold and incubated with urea before coating on the plate. The final concentration of the mutant coated on the plate was ∼10-fold less than that of the wild-type virus because of the attenuation of growth induced by the mutation (Indo M1/CA, 10⁷ PFU/ml; Indo WT/CA, 10⁸ PFU/ml) (43). The viruses were pretreated with 8 M urea before binding to the plate, and ELISA was performed with eluates as purified antibodies (Fig. 4C, D). The eluates purified from Indo M1/CA by hRID-CEx2 showed specific binding to mutant and wild-type viruses. The higher signal in response to wild-type Indo WT/CA could be ascribed to the relatively high concentration of mutant virus (Indo M1/CA, 10⁸ PFU/ml; Indo WT/CA, 10⁹ PFU/ml) (Fig. 4C). At the same virus concentration (10⁸ PFU/ml), the ELISA response was similar (Supplemental Fig. S3A, B). The purified antibodies failed to bind to the control cargo protein hRID, confirming that the ELISA response is epitope specific (Fig. 4D). These results further corroborate a specific immune response targeted to the activated epitope by design.

Figure 4.

Binding of purified antibodies against rHAs and urea-denatured viruses. A, B) Mice were vaccinated with Indo M1/CA (10⁶ PFU), Indo WT/CA (5 × 10⁵ PFU), or PBS as a control and were boosted with the same dose 3 times at 2-wk intervals. Sera were collected at the eighth week. Plates were coated with 0.25 μg/well of hRID-CEx2 or hRID, and sera diluted to 1:100 were added. After washing, a low-pH buffer was added to elute antibodies bound to the coated protein. Then, ELISAs were performed with the eluates and 0.25 μg/well of hRID-CEx2 (A) or rHA (B) of the 3 strains: PR8 (A/Puerto Rico/8/1934, H1), Sydney (A/Sydney/5/1997, H3), and Indonesia (A/Indonesia/5/2005, H5). C, D) As with rHA, ELISAs were performed with the eluates and 10⁷ PFU/well of Indonesia mutant virus (Indo M1/CA) or 10⁸ PFU/well of wild-type virus (Indo WT/CA). The viruses were incubated with 8 M urea at 50°C for 1 h, and then urea was diluted to 2 M when coating on the plate. OD, optical density. *P < 0.05, **P < 0.01 when compared with the PBS control.

Plaque-reduction neutralization tests were conducted to determine whether CE-specific antibodies could neutralize the infectivity of homologous and heterologous viruses, but the antibodies were unable to reduce infections (data not shown). Similar results were obtained, even when CE-specific antibodies were concentrated by 10-fold, suggesting that the antibodies could bind homologous influenza HA but could not block virus infection. These data indicate that the CE that is not exposed on live viruses becomes accessible to antibodies only after urea denaturation of the virus, and therefore, the antibodies against this epitope do not neutralize the virus.

DISCUSSION

During long periods of coexistence with their hosts, pathogens have developed various strategies to establish successful infection by either avoiding or subverting the host immune system (6, 7). Based on the hypothesis that viruses hide the conserved regions of their surface antigens from immune detection, the present study purported to identify potential epitopes that remained cryptic but could be activated to be immunogenic. For that purpose, we screened for potential candidates that are conserved among multiple subtypes of influenza HA and used structural modeling. Considering the substrate specificity of Cat S, a major endopeptidase involved in antigen processing in APCs, a Cat S–dependent cleavage site was introduced in HA. After confirming the in vitro cleavage of the model peptide (Fig. 2E) (42), viruses carrying the mutations were generated by an RG system with X-31ca as the backbone (31, 32). Immunization of mice showed that mutant Indo M1/CA induced production of specific antibodies against the conserved sites, and the antibody level was elevated by boosting; in contrast, the control Indo wild type failed to elicit specific antibodies (Fig. 3). The antibodies specific to the epitope were purified, and ELISA finally confirmed the specific binding to viral antigen (Fig. 4), indicating that an unnatural immune response was indeed elicited against the epitope. Previously, an autoimmune T cell response could be elicited against a self-protein (mouse lysozyme) by inserting mutations increasing the sensitivity to proteases (45). The present results provide a proof-of-principle that a specific antibody response can be directed to epitopes of infectious viral origin by design.

Structural parameters in antigenic determinants, e.g., surface exposure and hydrophilicity, are known to influence the cross-reactions between proteins and antibodies (46, 47), although the mechanism for that remains unclear. We analyzed the structural characteristics of the naturally occurring antigenic sites of PR8 HA (Ca, Cb, Sa, and Sb), and the average exposed surface and polarities of the sites were 40.32 and 70.03%, respectively (Supplemental Table S1). Those values corresponding to sites 1–9 were consistently lower, 0.13–16.92% and 27.60–57.83%, respectively (Table 1), consistent with the immunologic inactivity of those sites. Site 5 partially overlapped with the known epitope D of H3, which explains its high score in those criteria. Despite that, site 8 ranked high [exposed surface (8.04%) and polarity (46.79%)] (Fig. 1 and Table 1) under the present evaluation criteria. Indeed, it was confirmed by immunization of mice that specific antibodies were induced to that site by inserting Cat S cleavage motifs into flanking regions (Fig. 3) and that the antibodies could bind to HA (Fig. 4). However, >3 inoculations were required to induce detectable levels of antibodies. Moreover, although the antibodies could bind recombinant HA, the binding was not detected with live viruses (Supplemental Fig. S3C, D) but became apparent only after prior exposure to a high concentration (8 M) of urea (Fig. 4C). Low immunogenicity could be ascribed to the relatively low exposed surface area and polarity of the target epitope (Table 1). Exposure of the epitope via conformational change was, therefore, required for binding, probably suggesting that the epitope was structurally inaccessible to antibodies in the native structure of HA in live viruses. These results, however, suggest that it is indeed possible to elicit specific antibodies even if structural indicators associated with the epitope are low for antigen recognition and processing.

Cat S is one of the major proteases directly involved in antigen processing (26–28). The choice of cleavage sites for Cat S belonging to the papain superfamily is difficult to predict because that family has relatively broad substrate specificity (48–50). Therefore, we were guided by previous analyses on Cat S, where its substrate specificity was mainly determined by the P-2, P-1′, and P-3′ positions (42). Those mutations rendered the model protein susceptible to digestion by Cat S in vitro (Fig. 2E), and the motif inserted into the N-terminal side (N-mutation) was less effective than that inserted into the C-terminal side (C-mutation). The results are consistent with those in a previous study (42), which reported that the cleavage efficiency is less for peptides with isoleucine at P-2 position. Our attempt to replace isoleucine into other amino acids with better cleavage, such as valine and methionine, was unsuccessful because of a failure in rescuing the viruses carrying the desired mutations (I301V or I301M). This is probably due to local structural perturbation in HA that would affect viral fitness (Supplemental Fig. S1).

Indo M1/CA was greatly attenuated by >10⁵-fold when compared with highly pathogenic H5N1 influenza viruses (50% minimum lethal dose <10 PFU) (Fig. 2C) (51, 52). Previously, inoculation of mice with 10⁵ PFU Indo WT/CA resulted in ∼10% loss in body weight (43); however, mice inoculated with the same dose of Indo M1/CA did not show any symptoms of virulence or weight loss. Indo M1/CA showed virulence at 10⁶ PFU (weight loss up to 20%), which is a very high titer, but the inoculated mice recovered rapidly after 5 d. Therefore, the dose-dependent morbidity in Fig. 2C suggests that the introduction of Cat S cleavage mutations further contributes to attenuation established by cold adaptation.

Immunization with Indo M1/CA induced specific antibodies in mice against the conserved site, CE (Fig. 3), and the antibodies could directly bind the HA of the Indonesia strain (Fig. 4). In contrast, the response to Indo WT/CA without Cat S cleavage motifs was as low as that of the PBS control. Of note, the antisera were not able to neutralize the infectivity of the virus (data not shown). The activated epitope was located far from the receptor binding site (RBS) (Fig. 1), and therefore, the antibodies may not have physically interfered with the receptor binding and infectivity of the virus. The antisera were also specific to Indo M1/CA and failed to bind other influenza viruses carrying similar CEs (Fig. 4A). Harnessing better in silico tools for predicting 3-dimensional structures, judicious choosing of physically overlapping CEs with RBS, and establishing experimental protocols for a more-efficient RG system would allow us to test a repertoire of candidate CEs for potential neutralizing activities.

In addition to antibody responses, T cell immune responses against CEs were measured. To determine whether Indo M1/CA could induce CE-specific T cell responses, antigen-specific T cell responses were analyzed (53). Mice were vaccinated 3 times with Indo M1/CA, Indo WT/CA, or PBS as a control, and lungs were obtained 6 d after the last vaccination. Cells obtained from lungs were stimulated with synthetic CE peptide or not stimulated as a control and, then, were stained with fluorescent antibodies (anti-CD4 and anti-CD8). The numbers of CD4 and CD8 T cells in mice vaccinated with Indo M1/CA after stimulation with the CE peptide were not significantly different from those in unstimulated controls (Supplemental Fig. S4A, B). The results suggest a relatively inefficient presentation of the epitope either by endosomal or cytosolic pathways, although the epitope-specific antibody was present for both recombinant and influenza virus HA (Fig. 4). After staining with anti-CD4 and anti-CD8 (surface staining), the cells were permeabilized and stained with anti–IFN-γ (intracellular cytokine staining) to analyze IFN-γ–positive T cells. The numbers of IFN-γ–positive CD4 and CD8 T cells in mice vaccinated with Indo M1/CA after CE peptide stimulation almost doubled when compared with those in unstimulated controls (Supplemental Fig. S4C, D). The IFN-γ–positive CD8 T cells were more sensitive to CE peptide stimulation in mice vaccinated with Indo M1/CA than in mice vaccinated with Indo WT/CA. However, the level of stimulation was not considered significant when compared with that in the control. Certainly, further studies are necessary to delineate the role of T cells in activating the immune network for eliciting unnatural immune responses.

Finally, what is the presentation mechanism of the CE? The present X-31ca–based, live, attenuated vaccine mimics natural infection in delivery (nasal route) and subsequent antigen processing events after vaccination. Processing of foreign antigens is executed by 2 major routes. The endogenous pathway involves cytosolic degradation of endogenous antigen in the endoplasmic reticulum to generate peptides for MHC I display, whereas the exogenous pathway operates via endolysosomal degradation of exogenously acquired antigens to generate MHC II peptides (54). Although endogenous antigens accessed by viral infection are usually processed to generate mainly MHC class I peptides, recent studies show that antigens processed through 1 pathway can also be presented by the other pathway (cross-presentation) (55, 56). Influenza HA, a viral glycoprotein, can be cannibalized by lysosomes by autophagy (57), and the proteolytically cleaved peptides can be presented by both MHC class I and class II molecules. In that scenario, lysosomal enzymes, including cathepsins, might have a central role in processing of viral antigens. A previous study found that cathepsins are responsible for processing the HIV envelope into peptides of potential MHC class I and II epitopes that contribute to eliciting CD4⁺ T cell responses for developing adaptive immune responses toward protection (58, 59). Proper processing in APCs and loading on MHC does not, however, warrant successful elicitation of antibody responses. It is possible that CE-specific B cells are selected against, e.g., by deletion during development because of cross-reactivity with self or because of control by regulatory T cells (60). The CE sequence was not homologous to the mouse or human proteomes (data not shown) and, therefore, was probably not subjected to negative selection during the maturation process. Because of multilayered immune mechanisms (especially pronounced synergistic innate and adaptive responses and/or mucosal immune responses, which are usually weak or absent in other vaccine types) (61), the live attenuated vaccine approach employed herein could prevail and overshadow the protective effect conferred by humoral antibodies. Parallel studies involving a recombinant vaccine approach may further elucidate the strength of neutralizing antibodies in protecting against lethal challenge. Targeting the amino acid sequences in the vicinity of and, thereby, physically overlapping with the RBS in the surface antigen would be beneficial for eliciting neutralizing antibodies. Certainly, a detailed mechanism of the activation of CEs warrants further study.

In summary, a viral antigen was made amenable to antigen processing, eliciting unnatural antibody responses to CEs. Further understanding of the antigen-processing machineries, better in silico mining tools, and screening of potentially neutralizing epitopes in vitro and in vivo are required to translate the present approach into the design of vaccines, e.g., for conferring broad protection against multiple subtypes (62–64) or developing novel vaccines against viruses for which effective vaccines are not yet available (65, 66). Alternatively, preexisting, known epitopes could be made dormant by mutations for eliminating nonneutralizing antibodies responsible for the enhancement of infection (67, 68). It remains to be further explored whether this proof-of-principle, involving harnessing the antigen-processing pathway, could be translated into designer vaccines with increased efficacy and safety.

Supplementary Material

This article includes supplemental data. Please visit http://www.fasebj.org to obtain this information.

Glossary

APC

antigen-presenting cell

Cat S

cathepsin S

CE

cryptic epitope

HA

hemagglutinin

hRID

N-terminal RNA interaction domain of human lysyl-tRNA synthetase

Indonesia

A/Indonesia/5/2005

IMDM

Iscove’s modified Dulbecco’s medium

MDCK

Madin–Darby canine kidney

MHC

major histocompatibility complex

PFU

plaque-forming unit, PR8, A/Puerto Rico/8/1934

RBS

receptor binding site

RG

reverse genetic

rHA

recombinant hemagglutinin

Sydney

A/Sydney/5/1997

X-31ca

cold-adapted X-31

AUTHOR CONTRIBUTIONS

Young J. Lee and B. L. Seong designed the research study and wrote the manuscript; Young J. Lee conducted the research study and analyzed the data; J. E. Yu designed and expressed all proteins in the experiments; P. Kim conducted the 3-dimensional modeling and analyzed the structural characteristics; J. Y. Lee and J. Chang analyzed T-cell responses; Y. C. Cheong and Yoon J. Lee provided technical assistance; and B. L. Seong supervised all the experiments and prepared the manuscript.