INTRODUCTION
The shift from an empirical to a rational method for vaccine development associated with an enhanced understanding of host–pathogen interactions is synergistic with advances in genomics and bioinformatics. Since the first complete sequencing of a DNA genome over 20 years ago, that of the phage Φ-X174, with only 5368 base pairs (Sanger et al. Reference Sanger, Nicklen and Coulson1977), the number of sequenced genomes, proteomes and transcriptomes of different pathogens has risen exponentially. There are now over 6500 complete genomes published from over 59 000 organisms, representing bacteria, viruses, parasites and eukaryotes, as available on the GOLD database (https://gold.jgi-psf.org/index). This includes the genomes, proteomes and/or transcriptomes of human, nonhuman primate, and rodent species of Plasmodium. This wealth of information is a direct consequence of technological advances focused at the molecular level and allows for multidisciplinary approaches to identify target antigens for the development of vaccines, drugs and diagnostic tests, and for the development and application of methods to identify immune correlates of protection (Doolan et al. Reference Doolan, Apte and Proietti2014). Complementary methods, such as next-generation sequencing of lymphocyte epertoires (Mehr, Reference Mehr2014), complete T- and B-cell phenotype analysis (Zarnitsyna et al. Reference Zarnitsyna, Evavold, Schoettle, Blattman and Antia2013), highly-sensitive gene expression measurement using Fluidigm (Spurgeon et al. Reference Spurgeon, Jones and Ramakrishnan2008) or Nanostring (Geiss et al. Reference Geiss, Bumgarner, Birditt, Dahl, Dowidar, Dunaway, Fell, Ferree, George, Grogan, James, Maysuria, Mitton, Oliveri, Osborn, Peng, Ratcliffe, Webster, Davidson, Hood and Dimitrov2008), high-throughput profiling technologies using CyTOF mass cytometer (Kidd et al. Reference Kidd, Peters, Schadt and Dudley2014; Hansmann et al. Reference Hansmann, Blum, Hsin-Ju, Liedtke, Robinson and Davis2015), and biophotonic imaging for visualizing the infectious disease process (Andreu et al. Reference Andreu, Zelmer and Wiles2011), among other advances, have the potential to enhance understanding of the interaction between host and pathogens at the molecular level. The availability and refinement of large-scale bioinformatic databases containing information on both host and pathogen can further advance the acquisition, analysis and application of research data to yield more clinically relevant outcomes, ideally leading to the development of vaccines that provide sterile life-long protective immunity without the need for boosting, or sensitive and specific biomarkers of pathogens exposure or protective immunity. Such applications are highly desirable in the malaria community.
The term ‘immunomics’ was coined in 2001 by Klysik (Klysik, Reference Klysik2001), who suggested that ongoing advances in technology should serve to address the correlations between genes and the functional properties of their protein products. Today, the term immunomics refers to an integration of molecular immunology, genomics, proteomics, transcriptomics and bioinformatics, effectively providing a much-needed link between these fields (Doolan, Reference Doolan2011) and enabling an effective correlation between immunology research and clinical application. Immunomics is the study of the immunome, which can be defined as the set of antigens or epitopes that interface with the host immune system (Sette et al. Reference Sette, Fleri, Peters, Sathiamurthy, Bui and Wilson2005). However, it is necessary to consider that the definition of immunomics might be subject to different interpretations, as usually happens with the fast-changing fields of study like the ones comprised in the general domain of –omics, and is likely to expand and embrace concepts that might still be under investigation. Thus, the evolution of methods and techniques in the fields that are currently under the concept of immunomics will likely influence the definition and context of this term. In this paper, we are using the Sette et al. definition for immunomics.
Immunomics is distinct from reverse vaccinology, systems immunology and vaccinomics. Reverse vaccinology aims to identify the complete repertoire of antigens that an organism is capable of secreting or expressing on its surface (Rinaudo et al. Reference Rinaudo, Telford, Rappuoli and Seib2009). Systems immunology is a sub-discipline of systems biology (Schubert, Reference Schubert2011), and deals with molecular mechanisms of how the components of the immune system work together as a whole (Narang et al. Reference Narang, Decraene, Wong, Aiswarya, Wasem, Leong and Gouaillard2012). Vaccinomics, on the other hand, integrates immunogenetics and immunogenomics with systems biology and immune responses (Poland et al. Reference Poland, Kennedy and Ovsyannikova2011), Aimed at creating vaccines that are personalized/individualized.
Immunomics is as dependent on the host as it is on the pathogen, since the immune system and the infecting pathogens have been co-evolving for thousands of years. Furthermore individual characteristics of the parasite (e.g. species, strain, virulence, etc) interact with individual characteristics of the host immune system (e.g. age, immune status, epigenetic traits) (Tournier and Quesnel-Hellmann, Reference Tournier and Quesnel-Hellmann2006; Stilling et al. Reference Stilling, Bordenstein, Dinan and Cryan2014). Each of the components of the immune response is extremely complex on its own, and the interactions between them create an even more complex network of reactions. This complexity creates a challenge for in-depth, comprehensive analyses and increases the cost of experimental verification. An immunomics-based approach offers a solution to this challenge since high-throughput screening is performed (at least in part) in silico, prior to in vitro and in vivo experimental verification. High-throughput screening is used intensely (and successfully) for lead and drug discovery (Balls et al. Reference Balls, Bennet, Kendall, Kayser and Warzecha2012; Annang et al. Reference Annang, Perez-Moreno, Garcia-Hernandez, Cordon-Obras, Martin, Tormo, Rodriguez, de Pedro, Gomez-Perez, Valente, Reyes, Genilloud, Vicente, Castanys, Ruiz-Perez, Navarro, Gamarro and Gonzalez-Pacanowska2015). Translation of this technique to immunomics in order to address the challenge created by the complexity of the immune response to a pathogen includes the use of powerful computational analysis of next-generation, high-density peptide microarrays for rapid discovery and mapping of antigenic determinants (Hecker et al. Reference Hecker, Lorenz, Steinbeck, Hong, Riemekasten, Li, Zettl and Thiesen2012; Carmona et al. Reference Carmona, Nielsen, Schafer-Nielsen, Mucci, Altcheh, Balouz, Tekiel, Frasch, Campetella, Buscaglia and Aguero2015). This allows for a more thorough, unbiased and rational approach. In this fashion, data-driven studies of the immunome facilitate identification and characterization of key antigens and epitopes.
TRADITIONAL VACCINOLOGY
Pathogens with complex life cycles, complex genomes, proteomes and transcriptomes, and correspondingly complex immunomes, represent a great challenge for the development of effective vaccines. The Plasmodium parasite which causes malaria exemplifies this challenge. Such pathogens express a broad repertoire of antigens and epitopes that could be available for recognition by the host immune system. In some cases, effective immune responses directed against only a subset of these antigens and epitopes are sufficient for competent protection. This is demonstrated by the effectiveness of subunit vaccines such as the recombinant HBsAg antigen based vaccine for hepatitis B (Arnon, Reference Arnon, Rappuoli and Bagnoli2011; Schetgen, Reference Schetgen2014). However, for many pathogens, subunit vaccines based on only one or a few antigens have proved poorly effective (Meeusen et al. Reference Meeusen, Walker, Peters, Pastoret and Jungersen2007; Foged et al. Reference Foged, Rades, Perrie and Hook2014). For many vaccines that are currently in use worldwide, for example the highly effective whole-organism based Bacillus Calmette-Guérin (BCG) vaccine, the mechanism of protective immunity remains unknown (Kaufmann et al. Reference Kaufmann, Juliana McElrath, Lewis and Del Giudice2014). This gap in knowledge highlights the difficulties in assessing clear interactions between host and pathogens. Due to the poor efficacy of most subunit vaccines, until very recently almost all licensed vaccines were based on the whole organism, typically either live-attenuated pathogens or inactivated/killed pathogens (Grimm and Ackerman, Reference Grimm and Ackerman2013). In the case of malaria where extensive efforts directed towards subunit vaccines have thus far failed (Schwartz et al. Reference Schwartz, Brown, Genton and Moorthy2012), a resurgence of effort towards development of a whole-organism vaccine has occurred (Hoffman et al. Reference Hoffman, Billingsley, James, Richman, Loyevsky, Li, Chakravarty, Gunasekera, Chattopadhyay, Li, Stafford, Ahumada, Epstein, Sedegah, Reyes, Richie, Lyke, Edelman, Laurens, Plowe and Sim2010; McCarthy and Good, Reference McCarthy and Good2010; Good, Reference Good2011; Mikolajczak et al. Reference Mikolajczak, Lakshmanan, Fishbaugher, Camargo, Harupa, Kaushansky, Douglass, Baldwin, Healer, O'Neill, Phuong, Cowman and Kappe2014). A genetically attenuated Trypanosoma cruzi parasite is also being considered for vaccination against Chagas disease (Sanchez-Valdez et al. Reference Sanchez-Valdez, Perez Brandan, Ferreira and Basombrio2015). However, these whole-organism vaccines are associated with number of problems. In some cases, they have been associated with reversion to virulence, causing a proportion of vacinees to develop some form of the disease they had been vaccinated against (Bonanni and Santos, Reference Bonanni and Santos2011). Furthermore, whole-organism vaccines are almost exclusively restricted to pathogens that can be cultured in vitro and have relatively low antigen variability, which can be difficult for pathogens like Plasmodium (malaria), Mycobacterium tuberculosis (tuberculosis) or Schistosoma (schistosomiasis) (Doolan et al. Reference Doolan, Apte and Proietti2014).
Reverse vaccinology for rational vaccine design
An alternative strategy to culture-based in vitro approaches for the development of an effective vaccine against complex pathogens is that of reverse vaccinology, pioneered by Rappuoli in 2001 (Rappuoli, Reference Rappuoli2001). This approach analyses the entire genome of a pathogen to rapidly identify putative protective antigens and predict potential vaccine candidates (Rinaudo et al. Reference Rinaudo, Telford, Rappuoli and Seib2009; Heinson et al. Reference Heinson, Woelk and Newell2015). Reverse vaccinology studies were among the first to harness the wealth of information generated by genome sequencing for vaccine development. The proof of concept for this approach was established by the screening of surface-exposed proteins in the Neisseria meningitides genome, a causative agent of meningococcal disease, and resulted in a phase III vaccine trial after 40 years of setbacks using conventional methods (Pizza et al. Reference Pizza, Scarlato, Masignani, Giuliani, Arico, Comanducci, Jennings, Baldi, Bartolini, Capecchi, Galeotti, Luzzi, Manetti, Marchetti, Mora, Nuti, Ratti, Santini, Savino, Scarselli, Storni, Zuo, Broeker, Hundt, Knapp, Blair, Mason, Tettelin, Hood and Jeffries2000; Serruto et al. Reference Serruto, Bottomley, Ram, Giuliani and Rappuoli2012). Briefly, the sequence of the N. meningitides virulent strain MC58 was analysed using bioinformatics algorithms for surface-exposed proteins, which were then recombinantly expressed in E. coli, purified and tested in mice for their potential to induce bactericidal antibodies. Humoral response was analysed by Western blot and surface localization of the target protein was confirmed by enzyme-linked immunosorbent assay (ELISA) and flow cytometry. Of the 91 proteins found to be positive in the bioinformatic screening, 28 were able to induce antibodies with bactericidal activity and were prioritized based on their ability to induce broad protection. Ultimately five proteins were combined in a multicomponent vaccine named 4CMenB (Pizza et al. Reference Pizza, Scarlato, Masignani, Giuliani, Arico, Comanducci, Jennings, Baldi, Bartolini, Capecchi, Galeotti, Luzzi, Manetti, Marchetti, Mora, Nuti, Ratti, Santini, Savino, Scarselli, Storni, Zuo, Broeker, Hundt, Knapp, Blair, Mason, Tettelin, Hood and Jeffries2000; Serruto et al. Reference Serruto, Bottomley, Ram, Giuliani and Rappuoli2012). Very recently, two multicomponent vaccine formulations based on sequencing of the whole meningococcal genome to identify surface antigens of the meningococcal strains, called Bexsero® (Novartis) and Trumemba™ (Pfizer), were made available for clinical immunization against invasive group B meningococcal disease, although still subject to additional monitoring; however, unfavourable cost-effectiveness ratios for application of these vaccines have been reported (Christensen et al. Reference Christensen, Hickman, Edmunds and Trotter2013; Pouwels et al. Reference Pouwels, Hak, van der Ende, Christensen, van den Dobbelsteen and Postma2013; Leca et al. Reference Leca, Bornet, Montana, Curti and Vanelle2015; Tirani et al. Reference Tirani, Meregaglia and Melegaro2015).
Reverse vaccinology has limitations; in particular, it cannot predict polysaccharides, lipids or glycolipids which may be active compounds for a vaccine (Kanampalliwar et al. Reference Kanampalliwar, Rajkumar, Girdhar and Archana2013; Bertholet et al. Reference Bertholet, Reed, Rappuoli, Nor, Acosta and Sarmiento2014). It is also unlikely that a reverse vaccinology approach by itself to be able to predict a good correlate of protective immunity (Bertholet et al. Reference Bertholet, Reed, Rappuoli, Nor, Acosta and Sarmiento2014).
IMMUNOMICS-BASED VACCINOLOGY
In a field of study as complex as vaccinology, where the intricacy of the human immune system is evident from the cohorts of non-homogenous groups where disease phenotype and molecular profile (among other elements) have immense variation (Falus, Reference Falus2008), elegant approaches are required to decipher the host immune response to pathogens. Immunomics facilitates such a rational, systematic and comprehensive approach to antigen selection and prioritization for vaccine development. The wealth of information that immunomics draws on, namely large-scale genomic, proteomic and transcriptomic datasets, can be accessed via large public-access databanks such as GenBank (http://www.ncbi.nlm.nih.gov/genbank) or UniProt (http://www.uniprot.org/); pathogen-specific databases such as PlasmoDB (www.PlasmoDB.org) or TriTrypDB (http://tritrypdb.org/tritrypdb); or immunology based databases such as the Immune Epitope Database (http://www.iedb.org/) or the Innate Immune Database (http://www.innatedb.com). The available information in these databases is used as primary input for epitope prediction and therefore it is crucial that sequences are verified, annotated and curated. Moreover, immunomics also faces the challenges associated with computational predictive algorithms, which are subject to a series of conditional instructions for weightings and outputs dependent on ‘learned’ behaviour or characteristics. If those conditional instructions are flawed, the error is magnified exponentially, resulting in inaccurate data.
However, the multifactorial nature of protective immunity (Pulendran and Ahmed, Reference Pulendran and Ahmed2006), the large number of antigenic determinants or epitopes that can be recognized by the cells of the immune system as immunodominant or subdominant epitopes (Sette and Sundaram, Reference Sette, Sundaram and Frelinger2006), the broad range of putative epitopes restricted by multiple human leukocyte antigen (HLA) alleles prevalent in the human population (del Guercio et al. Reference del Guercio, Sidney, Hermanson, Perez, Grey, Kubo and Sette1995) combined with the high variation in the frequencies of different major histocompatibility complexes (MHC) alleles in different ethnicities (Sidney et al. Reference Sidney, Steen, Moore, Ngo, Chung, Peters and Sette2010), precludes analysing such large and complex datasets without the use of computation. The need for accurate prediction of biologically relevant epitopes for rational vaccine design is therefore crucial.
Computational predictive methods
The prediction of peptide epitopes from primary protein sequences is not a modern achievement. In fact, reports from as early as 1987 show that peptide epitopes could be predicted from protein sequence by consideration of hydrophobicity and amphipathic helices, and these sequences could be synthesized to enable further study (Cease et al. Reference Cease, Margalit, Cornette, Putney, Robey, Ouyang, Streicher, Fischinger, Gallo and DeLisi1987; Gotch et al. Reference Gotch, Rothbard, Howland, Townsend and McMichael1987; Margalit et al. Reference Margalit, Spouge, Cornette, Cease, Delisi and Berzofsky1987). Subsequently, it was established that T cell epitopes are generally linear and continuous. MHC class I molecules typically bind peptides that are 8–15 amino acids long (Rammensee et al. Reference Rammensee, Bachmann, Emmerich, Bachor and Stevanovic1999), while MHC class II molecules usually bind longer peptides of 12–25 amino acids in length (Jardetzky et al. Reference Jardetzky, Brown, Gorga, Stern, Urban, Strominger and Wiley1996). These features provided the foundation for improved prediction algorithms that considered the affinity of binding of a specific peptide sequence to a given MHC molecule. Using experimental affinity data deposited in public databases as training data, researchers developed statistical methods to take the early prediction algorithms to a new level. Widely used algorithms include those in the Immune Epitope DataBase (Kim et al. Reference Kim, Ponomarenko, Zhu, Tamang, Wang, Greenbaum, Lundegaard, Sette, Lund, Bourne, Nielsen and Peters2012; Vita et al. Reference Vita, Overton, Greenbaum, Ponomarenko, Clark, Cantrell, Wheeler, Gabbard, Hix, Sette and Peters2015) such as average relative binding (ARB) (Bui et al. Reference Bui, Sidney, Peters, Sathiamurthy, Sinichi, Purton, Mothe, Chisari, Watkins and Sette2005), or alternate algorithms such as support vector machine for prediction of MHC-binding peptides (SVMHC) (Donnes and Kohlbacher, Reference Donnes and Kohlbacher2006) or NetMHCII-2·2 (Nielsen and Lund, Reference Nielsen and Lund2009). More recent prediction algorithms consider additional features, such as proteosomal cleavage sites and transporter associated with antigen processing (TAP)-binding patterns, further enhancing accuracy (Tenzer et al. Reference Tenzer, Peters, Bulik, Schoor, Lemmel, Schatz, Kloetzel, Rammensee, Schild and Holzhutter2005; Antonets and Bazhan, Reference Antonets and Bazhan2013). Current computational models consider quantitative matrices, artificial neural networks, hidden Markov models, support vector machines (SVMs), quantitative structure activity relationship and molecular docking simulations (Brusic et al. Reference Brusic, Bajic and Petrovsky2004; Desai and Kulkarni-Kale, Reference Desai and Kulkarni-Kale2014). These improvements are crucial given the high degree of MHC polymorphism and complexity of generation and presentation of T cell epitopes (Desai and Kulkarni-Kale, Reference Desai and Kulkarni-Kale2014). Overall, in silico epitope predictions represent a more targeted, cost- and time-effective strategy as compared with more traditional approaches such as screening pools of overlapping peptide pools by enzyme-linked immunospot (ELIspot) or intracellular cytokine staining (ICS), or other epitope identification methods like X-ray crystallography and nuclear magnetic ressonance (NMR) techniques (Sun et al. Reference Sun, Ju, Liu, Ning, Zhang, Zhao, Huang, Ma and Li2013).
Physicochemical properties associated with the T cell receptor also critically influence effective cell mediated immune responses (Osuna et al. Reference Osuna, Gonzalez, Chang, Hung, Ehlinger, Anasti, Alam and Letvin2014, Madura et al. Reference Madura, Rizkallah, Holland, Fuller, Bulek, Godkin, Schauenburg, Cole and Sewell2015). Very recent studies have shown that the T cell receptor (TCR) undergoes conformational changes upon engagement with a peptide, allowing for discrimination between peptides. This conformational change causes the peptide to diverge its amino terminus partly away from the MHC peptide binding groove, forming a higher affinity interface with the TCR than is formed with the MHC groove (Dyson, Reference Dyson2015; Madura et al. Reference Madura, Rizkallah, Holland, Fuller, Bulek, Godkin, Schauenburg, Cole and Sewell2015). This discovery suggests that future epitope predictions strategies that consider details such as antigen TCR-MHC affinity may result in improved accuracy of epitope predictions and consequently reduced time for validation studies.
With regard to B-cell epitope predictions, most of the computational methods and databases currently available focus on continuous or linear B cell epitopes (Ansari and Raghava, Reference Ansari and Raghava2010). However, many B-cell epitopes are conformational and discontinuous (Braga-Neto and Marques, Reference Braga-Neto and Marques2006; Ansari and Raghava, Reference Ansari and Raghava2010), corresponding to the tridimensional features on the surface of the antigen where recognition by the immune system occurs (Braga-Neto and Marques, Reference Braga-Neto and Marques2006). This creates difficulties for the bioinformatic prediction of B-cell epitopes. Ideally, predictive algorithms would use tridimensional surface models of the protein antigens and measure surface energy interactions of variable regions of the immunoglobulins that correlate with B-cell activation. However, due to the computational complexity associated with analysing tridimensional interactions and the limited number of known antibody-antigen complex structures, only a limited number of prediction methods exist for discontinuous epitopes [reviewed in (Yao et al. Reference Yao, Zheng, Liang and Zhang2013)]; most are considered to perform poorly (Sun et al. Reference Sun, Ju, Liu, Ning, Zhang, Zhao, Huang, Ma and Li2013). It has been suggested that combining multiple classifiers for B-cell epitope definition, as an ensemble, could improve the performance of computational B-cell epitope prediction tools (El-Manzalawy and Honavar Reference El-Manzalawy and Honavar2014), but further multidisciplinary understanding of conformational epitopes may also contribute to the improving performance of these predictive algorithms.
Enabling technologies
Immunomics-based approaches can enhance our understanding of key features of the immune system in health and disease. Classical methods of assessing immune responses have typically focused only on the frequency and magnitude of a single immune parameter, e.g. antibody titre. Immunomics, on the other hand, allows for a multifactorial view of the response and considers the relevant biological outcome, e.g. protective immunity, by effectively taking into account pertinent and appropriate elements, such as particular epitope combinations, the cytokine response measured, and the T cell population used as the target. Multiple experimental methods, with different strengths, must also be used, and consequent results integrated. For example, using an immunomics approach to a disease model, Quintana and his team have shown that immunomics can be used to predict future disease (Quintana et al. Reference Quintana, Hagedorn, Elizur, Merbl, Domany and Cohen2004). Moreover, it has been computationally shown how the immune response to an epitope modulates the behaviour of an immune network, providing evidence for immunomic regulatory networks from immunomic microarray data (Braga-Neto and Marques, Reference Braga-Neto and Marques2006).
The quality of the immune response is at least as important as the quantity, as quality can be a key determinant of protection (Zepp, Reference Zepp2010; Doolan et al. Reference Doolan, Apte and Proietti2014). Some of the more conventional methods for assessment of immune responses as a measure of T- or B-cell reactivity include immunoassays such as the Jerne plaque assay (Jerne and Nordin Reference Jerne and Nordin1963), the splenic focus assay (Klinman and Aschinazi Reference Klinman and Aschinazi1971), ELISA (Engvall and Perlmann Reference Engvall and Perlmann1971), interferon-γ (IFN-γ) ELIspot (Czerkinsky et al. Reference Czerkinsky, Nilsson, Nygren, Ouchterlony and Tarkowski1983), fluorescence-activated cell sorting (Hayakawa et al. Reference Hayakawa, Ishii, Yamasaki, Kishimoto and Hardy1987; McHeyzer-Williams et al. Reference McHeyzer-Williams, McLean, Lalor and Nossal1993), peptide-induced ICS (Ozen et al. Reference Ozen, Tucker and Miller1998), and tetramer staining (Altman et al. Reference Altman, Moss, Goulder, Barouch, McHeyzer-Williams, Bell, McMichael and Davis1996; Skinner et al. Reference Skinner, Daniels, Schmidt, Jameson and Haase2000). More recently, technological and conceptual advances have resulted in the development and application of novel methods to comprehensively assess immune responses, including approaches to simultaneously examine a large number of cell functions and phenotypic markers, including at the single cell level. Such methods include, but are not limited to, the analysis of gene expression using Nanostring (Geiss et al. Reference Geiss, Bumgarner, Birditt, Dahl, Dowidar, Dunaway, Fell, Ferree, George, Grogan, James, Maysuria, Mitton, Oliveri, Osborn, Peng, Ratcliffe, Webster, Davidson, Hood and Dimitrov2008) or Fluidigm (Spurgeon et al. Reference Spurgeon, Jones and Ramakrishnan2008), as well as mass spectrometry based methods such as CyTOF technologies (Cheung and Utz, Reference Cheung and Utz2011). The quality of those responses can be assessed via phenotypic markers, differentiation state, profile of secreted cytokines, avidity, affinity and repertoire diversity (Siegrist, Reference Siegrist and Offit2013). If used singularly, none of the methods cited above can describe the complete set of characteristics that define an antigen-specific response. Immunomics can address such problems by integrating different fields of study, thus providing a platform for combining the strengths and compensating the flaws of different methods and approaches to vaccine development.
Particularly transformative are multidisciplinary computational and mathematical methods developed to cope with multiplex data, including host–pathogen interactions (Raman et al. Reference Raman, Bhat and Chandra2010), for which experimental analysis may be costly and laborious if the pathogen is complex. These mathematical and bioinformatic platforms may also support other aspects of vaccinology such as adjuvant discovery (Schellhammer and Rarey, Reference Schellhammer and Rarey2004; Sollner et al. Reference Sollner, Heinzel, Summer, Fechete, Stipkovits, Szathmary and Mayer2010).
Immunomics for rational vaccine design
The use of an immunomics based approach to vaccinology is a promising alternative for efficacious vaccine design. Unlike reverse vaccinology, immunomics also considers the immune system. It provides a means to systematically identify the antigens and epitopes that interact with the host immune system (Sette et al. Reference Sette, Fleri, Peters, Sathiamurthy, Bui and Wilson2005; De Groot, Reference De Groot2006). The premise of an immunomics-based, rational vaccine design is a consistent induction of the desired immune response against the key pathogen antigen(s) or epitopes which are targeted by protective immune responses (Barbosa and Barral-Netto Reference Barbosa and Barral-Netto2013; Doolan et al. Reference Doolan, Apte and Proietti2014; Slifka and Amanna, Reference Slifka and Amanna2014).
Immunomics of viruses
A number of immunomics approaches have been applied to viruses. In the case of HIV, an effective vaccine might require the design or discovery of immunogens which elicit good neutralizing antibodies against circulating strains of the virus (Kwong et al. Reference Kwong, Mascola and Nabel2011). For many years, only a few neutralizing monoclonal antibodies against the virus were known; the most intensely studied have been the antibodies directed against the glycoproteins gp120 and gp41 (Burton et al. Reference Burton, Poignard, Stanfield and Wilson2012). With the advent of immunomics, more neutralizing monoclonal antibodies, some with unexpected epitopes, have been identified (see Burton et al. Reference Burton, Poignard, Stanfield and Wilson2012). Most recently, Gallerano et al. (Reference Gallerano, Ndlovu, Makupe, Focke-Tejkl, Fauland, Wollmann, Puchhammer-Stöckl, Keller, Sibanda and Valenta2015) have analysed polyclonal antibody responses of HIV-infected persons to overlapping peptides covering the complete amino acid sequences of the gp120 and gp41 proteins, and identified major epitopes that can be recognized by antibodies.
Another immunomics study has prospected the influenza A (H1N1) virus immunome and transcriptome to show a complex host response pathway to the virus, and unravelled interactions between virus and host (Dimitrakopoulou et al. Reference Dimitrakopoulou, Dimitrakopoulos, Wilk, Tsimpouris, Sgarbas, Schughart and Bezerianos2014). A different study has used immunomics to better understand the repertoire of T cell specificities for H1N1, with the objective of developing an universal vaccine for influenza virus to combat the continuous antigenic drift of the virus (Assarsson et al. Reference Assarsson, Bui, Sidney, Zhang, Glenn, Oseroff, Mbawuike, Alexander, Newman, Grey and Sette2008). This study used over 4000 peptides from a panel of 23 influenza A virus strains based on predicted high-affinity binding to HLA class I or class II and high conservancy levels. Peripheral blood mononuclear cells (PBMCs) from healthy human donors were tested for reactivity against HLA-matched peptides by using IFN-γ ELIspot. One epitope, called PB1, was found to be the major target for both CD4+ and CD8+ T cell responses; 54 other non-redundant epitopes (38 class I and 16 class II) were also identified, and provide a potential base for the development of a universal influenza vaccine (Assarsson et al. Reference Assarsson, Bui, Sidney, Zhang, Glenn, Oseroff, Mbawuike, Alexander, Newman, Grey and Sette2008).
Immunomics of bacteria
Immunomics has been also successfully applied to bacteria. In one example, a Francisella tularensis protein microarray was generated and probed with serum from experimentally immunized mice to identify 11 of the 12 antigens previously discovered using traditional methods plus an additional 31 new antigens (Eyles et al. Reference Eyles, Unal, Hartley, Newstead, Flick-Smith, Prior, Oyston, Randall, Mu, Hirst, Molina, Davies, Milne, Griffin, Baldi, Titball and Felgner2007); this study further demonstrated an IgG subclass bias towards IgG2a in protected animals.
T cell based approaches have also been pursued to advance bacterial vaccine development. Moise et al. (Moise et al. Reference Moise, Moss and De Groot2012) evaluated the Helicobacter pylori genome for CD4+ T cell epitopes by using the predictor algorithm EpiMatrix; resulting epitopes were experimentally validated for MHC binding and T cell reactivity in p27 knockout mice infected with the mouse-adapted H. pylori strain. The immunoreactive epitopes were assembled into a multi-epitope vaccine that induced a broad immune response as determined by IFN-γ production in ELIspot assays (Moise et al. Reference Moise, Moss and De Groot2012).
For Mycobacterium tuberculosis, CD4+ T cells are crucial for controlling the infection (Woodworth et al. Reference Woodworth, Aagaard, Hansen, Cassidy, Agger and Andersen2014). Recently, a genome-wide screening for CD4+ T cell reactivity against M. tuberculosis (Arlehamn et al. Reference Arlehamn, Gerasimova, Mele, Henderson, Swann, Greenbaum, Kim, Sidney, James, Taplitz, McKinney, Kwok, Grey, Sallusto, Peters and Sette2013) identified a number of novel epitopes and antigens that may represent a potential vaccine candidates. This immunomics-based study screened the sequences from five complete M. tuberculosis genomes available from the National Center for Biotechnology Information (NCBI) database and protein sequences were parsed into 15-mer peptides, which were then ranked by consensus percentile of HLA binding for the 22 alleles most commonly present in the general population. For each protein, no less than two of the best-predicted binders were selected for synthesis, creating a synthetic library of over twenty thousand peptides that were tested by IFN-γ ELIspot against T cells from latent TB-infected donors. The reactive epitopes were ranked on the basis of magnitude of response and mapped to individual bacterial antigens using a reference genome. The results revealed a very heterogeneous response to infection: 82 antigens were recognized by more than 10% of donors. Hundreds of novel epitopes recognized by the human immune system have been identified, attesting to the potential of genome-wide screening strategy (Arlehamn et al. Reference Arlehamn, Gerasimova, Mele, Henderson, Swann, Greenbaum, Kim, Sidney, James, Taplitz, McKinney, Kwok, Grey, Sallusto, Peters and Sette2013; Arlehamn and Sette, Reference Arlehamn and Sette2014).
Immunomics of parasites
We and others have had a particular interest in the immunomics of Plasmodium spp. (Doolan, Reference Doolan2011). In early studies which served as proof of concept for the protein microarray platform, 250 genes representing putative proteins from P. falciparum were selected from a genomic sequence database according to the pattern of stage-specific gene or protein expression, subcellular localization, secondary structure, and known immunogenicity or antigenicity in human and animal models (Doolan et al. Reference Doolan, Mu, Unal, Sundaresh, Hirst, Valdez, Randall, Molina, Liang, Freilich, Oloo, Blair, Aguiar, Baldi, Davies and Felgner2008). Each of these 250 sequences was then printed onto a protein microarray which was probed with human sera from individuals differing in immune status. This study showed that the protein microarray platform could be successfully applied to identify antigens recognised as serodominant by individuals naturally or experimentally exposed to malaria. A more comprehensive array of 2320 protein fragments representing 23% of the P. falciparum proteome was then fabricated; the protein selection considered stage-specific transcription or protein expression, subcellular localization, secondary protein structure, and documented immunogenicity in humans or animal models at the time of antigen selection, as indicated by multidimensional protein identification technology. Subsequent studies with this array identified a signature of 16 proteins that were associated with the sterile immunity induced by experimental immunization with radiation attenuated sporozoites (Trieu et al. Reference Trieu, Kayala, Burk, Molina, Freilich, Richie, Baldi, Felgner and Doolan2011), and a signature of 49 antigens associated with the anti-disease immunity induced by natural exposure to malaria (Crompton et al. Reference Crompton, Kayala, Traore, Kayentao, Ongoiba, Weiss, Molina, Burk, Waisberg, Jasinskas, Tan, Doumbo, Doumtabe, Kone, Narum, Liang, Doumbo, Miller, Doolan, Baldi, Felgner and Pierce2010). Those data provide experimental support for a multivalent vaccine. These and other studies identified a number of antigens that had not been previously described as immunologically reactive (reviewed in (Davies et al. Reference Davies, Duffy, Bodmer, Felgner and Doolan2015)).
Our group has also shown that immunomics-based approaches can be applied to the study of cellular responses against Plasmodium. In the proof of concept demonstration of an epitope-based T cell screening approach in the P. falciparum model, the parasite genomic sequence was scanned to identify and prioritize a set of genes representing antigens potentially expressed in the sporozoite and intrahepatic stage of the parasite life cycle. A total of 27 proteins putatively expressed in the sporozoite proteome were selected according to their level of expression in the sporozoite proteome as determined by MudPIT, as well as stage specificity. This panel included 10 antigens expressed only in sporozoites, and 17 antigens common to other stages of the parasite life cycle. Evaluation of these proteins has shown that 16 of them were reproducibly recognised by peripheral blood mononuclear cell (PBMC)s from irradiated sporozoite immunized volunteers but not by naive controls, and nine of these antigens were more antigenic than other well-characterized antigens considered as leading vaccine candidates (Doolan et al. Reference Doolan, Southwood, Freilich, Sidney, Graber, Shatney, Bebris, Florens, Dobano, Witney, Appella, Hoffman, Yates, Carucci and Sette2003). Subsequently, we have applied this strategy to the complete pre-erythrocytic stage proteome and shown that only approximately 30% of the proteome is recognized, and identified the set of antigens that are highly reactive for T cell responses (Proietti & Doolan, in preparation).
With regard to immunomics for other parasites, Schistosoma proteome and transcriptome have been mined to identify surface-derived proteins, a subset of which were then expressed and printed on a protein microarray (Driguez et al. Reference Driguez, Doolan, Loukas, Felgner and McManus2010; Loukas et al. Reference Loukas, Gaze, Mulvenna, Gasser, Brindley, Doolan, Bethony, Jones, Gobert, Driguez, McManus and Hotez2011). These arrays have been probed with specimens from humans representing distinct clinical categories as well as experimentally immunized rodents. Several proteins which are predicted to be good potential vaccine targets have been identified (Loukas et al. Reference Loukas, Gaze, Mulvenna, Gasser, Brindley, Doolan, Bethony, Jones, Gobert, Driguez, McManus and Hotez2011; McWilliam et al. Reference McWilliam, Driguez, Piedrafita, McManus and Meeusen2012). These include the tetraspanin SmTSP-2, the tegumental antigen Sm29, and the very low-density lipoprotein-binding protein SjSVLBP as well as other novel proteins (reviewed in (McWilliam et al. Reference McWilliam, Driguez, Piedrafita, McManus and Meeusen2012). More recently, these protein microarrays have been screened for IgE and multiple IgG subclasses responses using sera from resistant or susceptible individuals (Gaze et al. Reference Gaze, Driguez, Pearson, Mendes, Doolan, Trieu, McManus, Gobert, Periago, Correa Oliveira, Cardoso, Oliveira, Nakajima, Jasinskas, Hung, Liang, Pablo, Bethony, Felgner and Loukas2014). The resultant antibody profiles could distinguish between protected vs non-protected cohorts and allowed identification of antigens that might represent excellent vaccine candidates (Gaze et al. Reference Gaze, Driguez, Pearson, Mendes, Doolan, Trieu, McManus, Gobert, Periago, Correa Oliveira, Cardoso, Oliveira, Nakajima, Jasinskas, Hung, Liang, Pablo, Bethony, Felgner and Loukas2014). The same study also identified protein associated with potentially deleterious hypersensitivity responses if used as subunit vaccines in endemic populations.
CONCLUSION
Infectious diseases continue to pose a major threat to public health worldwide and the need for prophylactic or therapeutic vaccines is urgent. Many of the diseases with high indexes of mortality or morbidity are caused by pathogens with large complex genomes and multistage life cycles, which present substantial challenges for the development of an effective vaccine. Immunomics, by focusing on the key components of host–pathogen interactions, provides a sound foundation to systematically search for critical determinants of immunity, namely key target antigens and epitopes, which could form the base of rationally designed new generation vaccines. Provided that immunomics continues to exploit state-of-the-art techniques and technologies and is able to respond to the inherent challenges associated with large datasets, this approach offers, in our opinion, an enormous potential as a 21st century solution to the challenge of rationally designing vaccines which are highly effective against complex pathogens such as the causative agent of malaria.
ACKNOWLEDGEMENTS
KdS is supported by the International Postgraduate Research Scholarship from the University of Queensland. DLD is supported by a National Health and Medical Research Council (NHMRC) Principal Research Fellowship; support by a Pfizer Australia Senior Research Fellowship is also gratefully acknowledged.
FINANCIAL SUPPORT
This research received no specific grant from any funding agency, commercial or not-for-profit sectors.