Genome-wide screening and identification of antigens for rickettsial vaccine development

Abstract

The capacity to identify immunogens for vaccine development by genome-wide screening has been markedly enhanced by the availability of microbial genome sequences coupled to proteomic and bioinformatic analysis. Critical to this approach is in vivo testing in the context of a natural host-pathogen relationship, one that includes genetic diversity in the host as well as among pathogen strains. We aggregate the results of three independent genome-wide screens using in vivo immunization and protection against Anaplasma marginale as a model for discovery of vaccine antigens for rickettsial pathogens. In silico analysis identified 62 outer membrane proteins (Omp) from the 949 predicted proteins in the A. marginale genome. These 62 Omps were reduced to 10 vaccine candidates by two-independent genome-wide screens using IgG2 from vaccinates protected from challenge following vaccination with outer membranes (screen 1) or bacterial surface complexes (screen 2). Omps with broadly conserved epitopes were identified by immunization with a live heterologous vaccine, A. marginale ss. centrale (screen 3), reducing the candidates to three. The genome-wide screens identified Omps that have orthologues broadly conserved among rickettsial pathogens, highlighted the importance of identifying immunologically sub-dominant antigens, and supported the use of reverse vaccinology approaches in vaccine development for rickettsial diseases.

Introduction

Genome-wide screens have been both efficient and effective in identifying molecules required for invasion, replication, survival, persistence, and transmission of bacterial pathogens. In contrast, genome-wide identification of candidate vaccine antigens has been less widely applied: connecting immune effectors to expressed antigens requires integration of bioinformatics and proteomics with clear in vivo correlates of protection (Sette & Rappuoli, 2010). For rickettsial pathogens, defined broadly as those within the Order Rickettsiales, vaccine development has been constrained to the point that no vaccines to prevent human infection and/or disease are available. However, the availability of more than 30 complete rickettsial genomes has provided new resources to accelerate identification of the proteome and potential vaccine candidates. Linking the microbial proteome to host protective immunity now emerges as the major challenge in rickettsial vaccine development.

In vivo models are often selected based on their ability to recapitulate lesion pathogenesis and/or clinical disease. However, for vaccine discovery and development, recapitulation of protective immunity in the face of a relevant infectious challenge is also critical. In vivo vaccine discovery studies most frequently test protection in a murine model. Although advantageous from a study cost viewpoint, the predictive value of this approach is limited by the absence of a natural host-pathogen relationship as well as the lack of representative host genetic diversity in the major histocompatibility complex (MHC) loci. Multiple alternative approaches have been used to improve the predictive value of animal models for vaccine discovery, including the use of surrogate pathogens, usually closely related to the actual pathogen, in a natural animal host that recapitulates key features of infection and immunity. A cogent example is the use of Simian Immunodeficiency Virus (SIV) in non-human primates as a model for Human Immunodeficiency Virus (HIV) infection in humans (Morgan et al., 2008). This approach has been successful in identifying protection-inducing SIV antigens, which have analogs in HIV, as well as specific immunization strategies, including prime-boost and replicating viral vector protocols.

We propose that this approach, identification of candidate antigens and testing for capacity to induce protective immunity using natural pathogen-host models, is equally relevant for rickettsial vaccine development. In contrast to the SIV/HIV example, in which there are fewer than 20 lentiviral proteins, the much larger rickettsial proteome with closer to 1,000 proteins, provides a challenge to narrow the candidate pool to a number that can be tested in vivo. We have approached this challenge using Anaplasma marginale, a natural rickettsial pathogen of wild and domestic ruminants, as a model. Specific features critical to its applicability as a model for vaccine candidate discovery are: (i) protective immunity can be induced in a natural bovine host by vaccination (reviewed in Palmer et al., 1999); (ii) vaccination trials can incorporate MHC diversity (Brown et al., 2001); and iii) the natural mode of challenge (tick transmission) can be used to represent an appropriate and relevant challenge (Galletti et al., 2009). In this communication, we aggregate the results of three independent immunologic and proteomic screens to identify specific vaccine antigens and present the findings in the broader context of vaccine development for rickettsial pathogens.

Screen 1: Identification of immunogenic outer membrane proteins

Immunization with A. marginale outer membranes, isolated from intact bacteria using density gradient centrifugation, induces protection as measured by significant reduction in bacteremia following challenge with complete protection against infection in 20–30% of vaccinates (Brown et al., 1998; Noh et al., 2008; Tebele et al., 1991). Bioinformatic identification of outer membrane proteins (Omp) in the A. marginale genome using PSORT, PSORTB, and homology to experimentally identified surface proteins predicted 62 Omps within the 949 proteins predicted to represent the complete proteome (Brayton et al., 2005). To identify which of these predicted Omps were actually immunogenic within the context of protective immunity, calves representing multiple different MHC haplotypes were immunized with the outer membrane immunogen known to be protective against challenge. The post-immunization/pre-challenge sera were used to screen the A. marginale proteome using two-dimensional immunoblots; “spots” representing proteins that induced IgG2 were excised and subjected to mass spectroscopy (LC-MS/MS) to identify the protein by mapping to the in silico predicted proteome (Lopez et al., 2005). Of the 62 predicted Omps, 21 were identified as being present in the protective outer membrane immunogen and being capable of stimulating IgG2 (Table 1). Importantly, serum IgG2 represents the opsonizing bovine IgG subclass and reflects the requirement for MHC class II-restricted CD4⁺ T lymphocyte help for B cell isotype class switching, both of have been linked to protective immunity (Brown et al., 1998; Palmer et al., 1999).

Table 1.

Identification of rickettsial vaccine candidates through genome-wide screening and independent experimental in vivo immunization and protection trials

Protein1	Screen 1 (OMP)	Screen 2 (surface complex)	Screen 3 (live vaccine)	Ortholog2 with e value <e−20	Ortholog with e value <e−5
AM127	+	−		A E W N	R O
AM197	+	−		A E W	Rb
AM366	+	−		Ac
AM529	+	−		Ac
AM779	+3	+	+	A E W	R
AM854	+	+	+	A E W O	R N
AM956	+	−		A E W N	R O
AM1096	+	−		A E W N	R O
Msp1a	+3	+	−	Ac only
Msp2	+	+	−4	A E	W
Msp3	+	+	−4	A E	W
Msp4	+3	+	−	A E	W
Msp5	+	−		A E W N O R
Omp4	+	−		A	E
Omp7	+	+	+5	A	E
Omp8	+	+	+5	A	E
Omp9	+	+	+5	A	E
Omp10	+	−		A	E
Omp14	+	−		A	E
OpAG2	+	+	−	A E
virB9	+	−		A E W N O R

¹A. marginale proteins are designated by either the genome identifier (AM) or, for previously identified proteins by the designations Msp (Major surface protein), Omp (Outer membrane protein), OpAG (operon-associated antigen) or virB9 (Type IV secretion System protein).

²The presence of an orthologue with an e-value of ≤−20 or ≤ −5 is indicated for the following species: A = Anaplasma marginale ss. centrale, and Anaplasma phagocytophilum; E = Ehrlichia ruminantium, E. canis and E. chaffensis; W = Wolbachia species; N = Neorickettsia sennetsu and N. risticii; R = Rickettsia felis, R. bellii, R. africae, R. akari, R. conorii, R. rickettsii, R. typhi, R. prowazekii; O = Orientia tsutsugamushi. When Ac and Rb are indicated, A. marginale ss. centrale or R. bellii were the only species from the genus that contained an ortholog.

³Am779, Msp1a, and Msp4 were identified individually as inducing IgG2.

⁴Msp2 and Msp3 are related proteins; each consists of N- and C-terminal domains conserved among A. marginale strains and surface exposed central domains which are highly polymorphic among strains.

⁵Omps7–9 are collapsed into one orthologous Omp, designated Omp7/9, in the live vaccine strain.

Screen 2: Identification of surface exposed immunogenic Omps

The outer membrane immunogen used in the initial screen included all proteins in the isolated fraction (Lopez et al., 2005; Tebele et al., 1991). To narrow the candidate list to only those proteins with surface exposed domains, a macromolecular surface complex was isolated by cross-linking proteins on intact bacteria using 3,3′-dithiobis[sulfosuccinimidylpropionate], reactive with primary amines with a spacer arm of 1.2 nm, followed by electrophoretic separation (Noh et al., 2008). Cohorts of calves, matched for MHC diversity, were immunized with either the surface protein complex or the original outer membrane immunogen. Vaccinates in both groups were significantly protected against both bacteremia and disease (measured by anemia) as compared to identically challenged adjuvant-only controls with no significant difference between the outer membrane and surface complex immunogens (Noh et al., 2008). Pre-challenge serum IgG2 from the surface complex vaccinates bound only a subset of the complete outer membrane, consistent with the cross-linking restriction to Omps with surface domains, and mass spectrometry identified 10 individual Omps (Noh et al., 2008) (Table 1).

Screen 3: Identification of broadly conserved immunogenic Omps

The first two screens were based on immune protection against the homologous strain—useful for identifying initial candidates but limited in progressing to development of a vaccine effective against antigenically diverse strains (Dark et al., 2009; Kocan et al, 2003). To identify broadly conserved immunogenic Omps, further narrowing the vaccine candidate list, we took advantage of the ability of the live A. marginale ss. centrale vaccine strain to induce protective immunity against heterologous challenge. Originally isolated in South Africa (Theiler, 1911), this vaccine strain has been shown in experimental trials to provide protection against severe morbidity and mortality caused by A. marginale sensu stricto strains (Bock & de Vos, 2001; Pipano, 1995). While the difficulty in standardization, need for cryopreservation and thus a rigorous cold-chain, and risk of co-transmission of pathogens by the blood-based vaccine have impeded its use in much of the world, it has been used effectively in Australia and Israel for decades (Bock & de Vos, 2001; Kocan et al., 2003; Pipano, 1995). We immunized calves with the live Israel vaccine strain of A. marginale ss. centrale and used sera from the vaccinates, again representing diverse MHC haplotypes, to identify Omps of the heterologous St. Maries strain (A. marginale sensu stricto) with conserved epitopes. Vaccinates were then challenged by tick-transmission of the St. Maries strain and protection was confirmed (Galletti et al., 2009). This third screen eliminated 5 of the 10 remaining candidates (Table 1). Three Omps, all of which are orthologous between the vaccine strain and the challenge strain, were not consistently immunogenic among vaccinates of diverse MHC haplotypes. This result is not in itself totally unexpected as there are sequence differences between the vaccine strain and the sensu stricto strains for each of the orthologs (Dark et al., 2011; Herndon et al., 2010). Two additional Omps, Msp2 and Msp3, were identified as cross-reactive using IgG2 from the live strain vaccinates. However, the recognition was limited to the highly conserved N- and C-terminal domains with no binding to the relevant surface domains (Agnes et al., 2011), well-established to be highly antigenically variable among strains (Futse et al., 2008; Rodriguez et al., 2005). The remaining five Omps include Am779, Am854, and a closely related set of three proteins, Omps 7–9. Omps 7–9 are encoded by three genes in tandem within the St. Maries strain, with 70–75% identity among the three Omps (Brayton et al., 2005; Noh et al., 2006). This locus is collapsed in the vaccine strain to a single Omp, which retains 64–70% identity with the expanded Omp7–9. Thus the minimal number of identified vaccine candidates is three: Am779, Am854, and collapsed single Omp7/8/9.

Discussion

The ability to narrow the set of vaccine candidates from a genomic complement of 949 predicted proteins to three Omps with broadly conserved epitopes reflects the technical power of proteomic approaches coupled to immunization and challenge experiments. The biological independence of the three screens used, e.g., one screen did not predict the results of the subsequent screens, was critical in sharply narrowing the pool of candidate Omps. While bioinformatic analyses were an essential component of the approach, we emphasize that epitope conservation requires immunologic testing rather than in silico analysis. Bioinformatics analysis by Dark et al. (2011) classified all three Omps shown to have broadly conserved B cell epitopes (Am779, Am854, Omp7/9) as being “not conserved” based on only sequence comparisons. As epitope conservation, similar to active site conservation, can be “buried” by polymorphisms elsewhere in the protein, immunologic testing of conservation versus diversity is essential. All three of the vaccine candidates are classified as immunologically “sub-dominant” antigens. Technically, this explains why these candidates were not detected in previous approaches that lacked the sensitivity of the immunologic and proteomic assays used presently. Biologically, this identification is consistent with the hypothesis that protective antigens are more likely to be sub-dominant (Lopez et al., 2005, 2007), as shown experimentally by protective immunization against viral infections (Gallimore et al., 1998; Oukka et al., 1996). Consistent with the need for both class switching to IgG2 and affinity maturation associated with MHC class II-dependent CD4⁺ T lymphocyte help, specific CD4⁺ T cell responses have been identified for Am779, Am854, and Omp7/9 in vaccinates (Lopez et al., 2008).

The present approach has multiple outcomes relevant for vaccine development to protect against rickettsial infections. The first is the identification of three specific Omps as leading vaccine candidates. All three have orthologues in additional rickettsial pathogens (Table 1) and become candidates for testing experimentally in pathogen-specific animals models. The second is to highlight the importance of detecting subdominant antigens in a comprehensive identification strategy. The third is the efficiency of the strategy, the overall proteomic approach can be extended to high throughput screening of not only antigens inducing antibody responses but also T cell responses (Sette & Rappuoli, 2010; Lopez et al., 2008). Extending this reverse vaccinology approach based on in vivo protection in natural pathogen-host models to other rickettsial diseases will accelerate greatly needed vaccine development.