A Structure-Based Strategy for Epitope Discovery in ... - Cell Press

47 downloads 2625 Views 876KB Size Report
Nov 15, 2012 - tial epitopes were synthesized as free peptides and tested for ... host immune system with B. pseudomallei used a protein micro- array, covering .... whereas domain C is formed by an internal contiguous segment. (residues ...
Structure

Article A Structure-Based Strategy for Epitope Discovery in Burkholderia pseudomallei OppA Antigen Patricia Lassaux,1,7 Claudio Peri,3,7 Mario Ferrer-Navarro,4 Louise J. Gourlay,1 Alessandro Gori,3 Oscar Conchillo-Sole´,4 Darawan Rinchai,5 Ganjana Lertmemongkolchai,5 Renato Longhi,3 Xavier Daura,4,6 Giorgio Colombo,3,* and Martino Bolognesi1,2,* 1Department

of Biosciences Nazionale delle Ricerche, Institute of Biophysics University of Milan, Milan 20133, Italy 3Consiglio Nazionale delle Ricerche, Institute for Chemistry of Molecular Recognition, Department of Computational Biology, Milan 20131, Italy 4Institute of Biotechnology and Biomedicine, Universitat Auto ` noma de Barcelona, Bellaterra 08193, Spain 5Center for Research and Development of Medical Diagnostic Laboratories, Faculty of Associated Medical Sciences, Khon Kaen University, Khon Kaen 40002, Thailand 6Catalan Institution for Research and Advanced Studies, Barcelona 08010, Spain 7These authors contributed equally to this work *Correspondence: [email protected] (G.C.), [email protected] (M.B.) http://dx.doi.org/10.1016/j.str.2012.10.005 2Consiglio

SUMMARY

We present an approach integrating structural and computational biology with immunological tests to identify epitopes in the OppA antigen from the Gram-negative pathogen Burkholderia pseudomallei, the etiological agent of melioidosis. The crystal structure of OppABp, reported here at 2.1 A˚ resolution, was the basis for a computational analysis that identified three potential epitopes. In parallel, antigen proteolysis and immunocapturing allowed us to identify three additional peptides. All six potential epitopes were synthesized as free peptides and tested for their immunoreactivity against sera from healthy seronegative, healthy seropositive, and recovered melioidosis patients. Three synthetic peptides allowed the different patient groups to be distinguished, underlining the potential of this approach. Extension of the computational analysis, including energy-based decomposition methods, allowed rationalizing results of the predictive analyses and the immunocapture epitope mapping. Our results illustrate a structure-based epitope discovery process, whose application may expand our perspectives in the diagnostic and vaccine design fields. INTRODUCTION Structure-based antigen design is emerging as a strategy for next-generation vaccine development (Dormitzer et al., 2008). This approach enables antigens to be discovered and engineered for improved biochemical and immunological properties, to enhance vaccine efficacy, possibly generating cross-reactive vaccines when microbial antigenic variation is a factor (Dormitzer

et al., 2008; Schneewind and Missiakas, 2011). Its prerequisite is a detailed knowledge of the three-dimensional (3D) structure of an antigenic protein, which provides atomic-level information on the overall antigen fold and epitope location/formation. Complemented by computational analyses, the conformational properties of the antigen and the molecular determinants of antibody recognition may be further probed to facilitate the engineering process. An elegant use of computer-aided structure-based design is illustrated by the grafting of the typically transient HIV-1 gp41 neutralizing epitope onto a more stable protein backbone scaffold (Ofek et al., 2010). Nevertheless, despite some successful examples (Nuccitelli et al., 2011), the whole field is still largely in its early development stages. Melioidosis vaccine research has been intensely pursued over the last decade in the therapeutic and also biodefense areas (inhalation is a principal infection route of B. pseudomallei). This intracellular Gram-negative bacterium is endemic in tropical areas of Southeast Asia and Northern Australia (Currie et al., 2008). Because of the polymorphic nature of B. pseudomallei infections, translating into symptoms mimicking different diseases, and due to its multidrug resistance, B. pseudomallei proves itself difficult to diagnose and treat (Cheng and Currie, 2005). Early diagnosis, coupled to vaccination, would accelerate and improve melioidosis treatment, with a vaccine being a cost-effective alternative to antibiotic administration, preventing both relapse and reinfection; however, all potential melioidosis vaccines tested to date have proven unsuccessful (Sarkar-Tyson and Titball, 2010). A recent study launched to investigate the interaction of the host immune system with B. pseudomallei used a protein microarray, covering over 1,000 proteins, to probe for immunoreactivity against melioidosis patient sera, producing a short list of 49 antigens that were significantly reactive in comparison with sera from healthy individuals (Felgner et al., 2009). Several antigens were further studied for their immunogenic properties, including the oligopeptide-binding protein A (OppABp), which is the focus of this communication. OppABp is part of an ATPbinding cassette (ABC) transport system, a group of proteins

Structure 21, 167–175, January 8, 2013 ª2013 Elsevier Ltd All rights reserved 167

Structure Epitope Discovery for Structural Vaccinology

regarded as potential targets for the development of therapeutic interventions against bacterial infections, given their key role in bacterial survival, virulence, and pathogenicity. Components of several ABC transporter systems, from both Gram-negative and Gram-positive bacteria, have been proposed as candidates for vaccine development because they were shown to react with convalescent patient sera (Garmory and Titball, 2004; Tanabe et al., 2006). OppA is part of the oligopeptide transport system OppABCDF involved in nutrient uptake and recycling of cell-wall peptides (Monnet, 2003). The transport system consists of five subunits: two integral membrane proteins forming the pore, two proteins responsible for ATP hydrolysis, and a substrate-binding protein (SBP) (Monnet, 2003). OppA is a receptor, or SBP, and determines the recognition properties of the system, delivering the substrates to its cognate-binding partners. Precedent studies on OppA from different pathogens revealed that OppA from Listeria monocytogenes is a virulence factor important for intracellular survival (Borezee et al., 2000) and that OppA from Yersinia pestis is a protective antigen (Tanabe et al., 2006). With regard to OppABp, it is recognized by T cells primed by B. pseudomallei, triggers IFN-g production, and stimulates both humoral and cell-mediated responses. However, sera raised against OppABp failed to offer protection in a mouse infection model (Harland et al., 2007). Therefore, OppABp is seen as a suitable target for structure-based antigen analysis and improvement of its antigenic properties. In this framework, we studied the crystal structure of OppABp (at 2.1 A˚ resolution), which was used as a template for computational epitope design/predictions using matrix of local coupling energies (MLCE) (Scarabelli et al., 2010) and electrostatic desolvation profiles (EDP) (Fiorucci and Zacharias, 2010) methods, resulting in the identification of three consensus peptides (COMP1–COMP3). In parallel, we performed experimental epitope mapping, based on proteolytic digestion of OppABp and immunocapture of peptide fragments containing the epitopes recognized by murine anti-OppABp polyclonal antibodies, which provided three potential epitope sequences (EXP4–EXP6). The COMP and EXP peptides were synthesized and tested for their immunoreactivity against sera from uninfected, healthy-infected, and melioidosis-recovered patients. Plasma antibodies from patients who had recovered from B. pseudomallei infection significantly recognized all COMP peptides and EXP4, relative to plasma antibodies from the seronegative group. Human plasma antibodies recognizing all peptides but COMP2 were significantly higher in the healthy seropositive group than in the seronegative one. Interestingly, antibodies recognizing EXP5 and EXP6 were also significantly higher in the healthy seropositive group than in the recovered melioidosis groups. It is also remarkable that response to COMP3 was significantly different in all three groups. Comparison of experimental and computational results led us to further develop our in silico epitope prediction methods, by incorporating an energy-based decomposition approach to divide the antigenic protein into fragments prior to MLCE and EDP analyses. Such an approach was instrumental in improving the agreement between structure-based epitope predictions and immunocapture epitope mapping. Our results highlight the potential of a structure-to-epitope prototypic pipeline that

combines structural and computational biology, to identify antigenic substructures/epitopes in view of their application in diagnostic tools or vaccine development. RESULTS 3D Structure of OppABp The OppABp gene (GeneDB code BPSS2141) encoding for protein residues 39–554, devoid of its predicted signal peptide (residues 1–38), was amplified, cloned in pET14b, and expressed in BL21 Star (DE3) Escherichia coli cells; the protein construct was purified as a N-terminal His-tag fusion protein and crystallized. The crystal structure of OppABp was solved at a resolution of 2.1 A˚ (Rgen and Rfree values of 15.8% and 20.9%, respectively), as described in the Experimental Procedures (statistics for the data collection and model refinement are shown in Table 1). Electron density was visible for residues 48–553 but absent for the first 30 N-terminal residues (21 corresponding to the His-tag region and 9 to the protein) and for the C-terminal residue 554. The 3D structures of several oligopeptide-binding proteins have been previously reported by Berntsson et al. (2009), Dunten and Mowbray (1995), Levdikov et al. (2005), Sleigh et al. (1999), Tanabe et al. (2007). Despite low sequence similarity, superimposition of the C-a backbones (http://bioinfo3d.cs.tau.ac.il/c_alpha_match/) highlighted significant structure conservation between OppABp and the six OppA homologs, with rmsd values in the 1.1–1.9 A˚ range. All OppAs consist of two main domains, with the polypeptide chain crossing over between the two domains, two or three times, resulting in the formation of a hinge that opens and closes upon ligand binding and release. Indeed, OppABp is composed of two lobes (domains AB and C) comprising a/b folds, connected by two loops (residues 299–307 and 521–528), which form the hinge (Figure 1). Domain AB is formed by the N and C termini regions (residues 48–303 and 524–553, respectively), whereas domain C is formed by an internal contiguous segment (residues 304–523). Domain C has an a/b topology, in which a central mixed b sheet (b17-b12-b16) is flanked by eight a helices and a b sheet (b14–b15) (Figure 1). Domain AB can be further divided into two subdomains, with subdomain A hosting the N- and C-terminal parts of the protein (residues 48–264 and 541–553) and presenting an a/b fold. Subdomain B (residues 265–303 and 524–540), linked to domain C, consists mostly of meandering loops with small b strands (b18, b19, and b20) and two parallel a helices (a6 and a7). We observed the presence of a disulphide bond (between C469 and C479) in domain C of OppABp, a variable feature among OppA family members that contributes to formation of the substrate-binding site. OppAs bind small peptides with high affinity without any sequence specificity, which explains why they are often copurified with their bound peptide, as seen for OppABp. Subdomain A and domain C form the ligand-binding pocket. Based on electron density fitting, we assigned the sequence D-V-A to the tripeptide, which binds as reported for OppA from Salmonella typhimurium (Sleigh et al., 1999). Epitope Prediction and Design of Antigenic Peptides Two different epitope prediction computational methods were combined and applied to representative structures obtained

168 Structure 21, 167–175, January 8, 2013 ª2013 Elsevier Ltd All rights reserved

Structure Epitope Discovery for Structural Vaccinology

Table 1. Data Collection and Refinement Statistics for OppABp OppABp (Residues 48–553) Data Collection Space group

P21

Cell dimensions a, b, c (A˚)

46.9, 81.9, 74.2

a, b, g ( )

90, 104.3, 90

Resolution (A˚)

40.97–2.1 (2.21–2.1)

Rmergea

0.089 (0.287)

I/sIa

13.2 (5.6)

Completeness (%)a Redundancy

a

Experimental Epitope Mapping Epitope mapping experiments were carried out using recombinant OppABp and cognate polyclonal sera. To this aim, we adopted and extended an immunocapturing approach successfully used previously with monoclonal antibodies (Koehler et al., 2011; Soriani et al., 2010). The approach involves proteolytic digestion (using diverse proteases) of the target antigen prior to immunocapturing and subsequent analysis of antibodybound peptides (containing epitopes or fragments thereof) by mass spectrometry. The sera collected from three immunized mice were independently analyzed and produced identical results. Using trypsin for the partial digestion of OppABP, four different peptides were captured by the polyclonal IgGs, with masses of 1768.909, 1991.121, 2096.056, and 2638.095 Da. MS/MS analysis of the 1768.909 Da peptide showed that it corresponded to the N-terminal His-tag segment 2-GSSHHH HHHSSGLVPR-17 (1767.841 Da). The other three peptides corresponded to the sequences 121-WSNGQPVTAADFVYAWQR138 (peptide EXP4, 2096.056 Da), 299-ELRPGLQLATYYYYLK314 (EXP5, 1991.121 Da), and 336-EILTSKITQAGEVPM(oxy) YGLM(oxy)PKGVK-359 (EXP6, 2638.095 Da; M(oxy) standing for oxidized methionine).

99.96 (99.87) 4.5 (4.5)

Refinement Resolution (A˚)

38.5–2.1

No. of reflections

31,796

Rgen/Rfree

15.8/20.9

No. of atoms Protein

4,060

Peptide

22

Water

221

Glycerol

56

Chloride ion B factors (A˚2)

1

Protein

22.9

Ligand

22.8

Water

26.2

Glycerol

61.1

Chloride ion

27.8

Rmsds Bond lengths (A˚)

0.006

Bond angles ( )

0.909

Ramachandran plot (%) Favored regions

98.4

Allowed regions

1.6

COMP1 is a conformational epitope, composed of two loops (residues A107–D109 and T185–P188), COMP2 is a polypeptide segment devoid of secondary structure that extends from R363 to P367, and COMP3 corresponds to a highly solvent-accessible area on the protein surface, comprising the most exposed residues of the N480–L509 stretch. The conformational properties of the predicted epitopes were further characterized through MD simulations (see Supplemental Experimental Procedures; Figure S1), with COMP1 and COMP2 being flexible and COMP3 maintaining its conformation with minor fluctuations.

A single data collection was allowed to solve the structure. Rmerge = P P rI(I)r I 3 100, where I is the intensity of a reflection, and (I) is the average intensity; Rfree was calculated from 5% of randomly selected P P data for cross-validation; R factor = rFoFcr/ rFor 3 100. a Values in parentheses are for the highest-resolution shell.

from molecular dynamics (MD) simulations run on the OppABp crystal structure (see Experimental Procedures) (Fiorucci and Zacharias, 2010; Scarabelli et al., 2010). MLCE was specifically developed to pick antigenic epitopes, whereas the scope of EDP is broader, having been developed to identify general protein-protein interaction interfaces and also being validated for antigen-antibody interactions. The prediction results individually produced by MLCE and EDP, with regard to the protein sequence, are presented in Figure S2, which is available online. Consensus analysis of the results from the two methods led to the identification of three different minimal epitopes, labeled COMP1–COMP3, which are displayed on the OppABp 3D structure (Figure 2).

Synthesis of Epitope Peptides The results of the structure-based predictions were used to design and synthesize a set of polypeptides corresponding to the selected epitope regions, with different supporting tags (using two polyethylene glycol [PEG] moieties as spacers and human serum albumin [HSA] as a carrier protein) (see Experimental Procedures). The final peptide sequences were HSA-Cys-PEGPEG-KAPDTggKTEVPVSY (COMP1); HSA-Cys-PEG-PEG-GVK GVQRPFTPDWA (COMP2); HSA-Cys-PEG-PEG-AEANQKLDD GARAALLTQAHDLA (COMP3); HSA-Cys-PEG-PEG- WSNGQPV TAADFVYAWQR (EXP4); HSA-Cys-PEG-PEG- ELRPGLQLATYY YYLK (EXP5); and HSA-Cys-PEG-PEG-EILTSKITQAGEVPMYGL MPKGVK (EXP6). COMP1 is a conformational epitope: the introduction of two glycine residues (lowercase residues in the sequence) was estimated to be sufficient to mimic the distance between the two termini and the orientation of the two constituting stretches observed in the crystal structure. The computationally predicted epitope residues are underlined. Peptide Detection by Human Plasma Antibodies To evaluate their immunogenicity, the synthetic peptides were tested for antibody recognition in plasma samples from 19 healthy donors and 20 recovered melioidosis cases (from Khon Kaen University and Srinakarin Hospital, Thailand), by indirect ELISA. Healthy donors were divided into two groups,

Structure 21, 167–175, January 8, 2013 ª2013 Elsevier Ltd All rights reserved 169

Structure Epitope Discovery for Structural Vaccinology

Figure 1. Tertiary Structure of OppABp Secondary structure ribbon representation of OppABp bound to its tripeptide ligand (purple ball and stick). Domains AB and C are illustrated in yellow-red and green, respectively. b Strands, a helices, and 310 helices are labeled S1–S17, H1–H15, and h1–h7, respectively. The N and C termini (N term and C term, respectively) of OppABp are labeled. This figure was generated using PyMOLWin.

that the peptides were recognized by the humoral-mediated immune response to B. pseudomallei infection in humans.

seronegative and seropositive subjects, based on indirect hemagglutinin assay (IHA) antibody titers (see Supplemental Experimental Procedures). As expected, control experiments showed that plasma antibodies to a crude B. pseudomallei antigen extract (Figure 3A) and to recombinant OppABp (Figure 3B) in seropositive and recovered groups were significantly higher than those in the seronegative group (Mann-Whitney U test; p < 0.001 and p < 0.05, respectively). Interestingly, the average response of recovered plasma antibodies recognizing our recombinant OppABp was much higher than previously reported by Suwannasaen et al. (2011). Parallel experiments with the COMP1–COMP3 peptides revealed that the recovered group generally showed higher reactivity against all three peptides, as compared to the seronegative group (Mann-Whitney U test; p < 0.05, p < 0.01, and p < 0.01, for COMP1, COMP2, and COMP3, respectively; Figures 3C–3E). In addition, COMP1 and COMP3 peptides were significantly recognized by plasma from seropositive subjects (MannWhitney U test; p < 0.05) that were classified as asymptomatic healthy control subjects (subclinical melioidosis), compared to seronegative plasma samples (Figures 3C and 3E). Interestingly, the reactivity of COMP3 was significantly diverse among the three groups (Figure 3E), highlighting its potential use as a diagnostic tool. The EXP4–EXP6 peptides showed a distinct reactivity pattern toward plasma of all three groups (Figures 3F–3H). In fact, the EXP peptides reacted strongly with the plasma from asymptomatic healthy patients relative to the healthy individuals. Particularly, the EXP5 and EXP6 were not significantly recognized by the plasma from recovered subjects, hinting at the potential application of both peptides for discriminating between asymptomatic versus clinical melioidosis in endemic areas (Figures 3G and 3H). Thus, the results obtained confirm

In Silico Fragment Subdivision Coupled to Epitope Prediction The identification of distinct sets of epitopes by the computational approach (COMP1–COMP3) and by the polyclonal antibodies induced in young mice (used for immunocapturing; EXP4–EXP6) may depend on different factors. One of these concerns a rapid processing of OppABp into smaller fragments after intraperitoneal injection in mice, which may facilitate recognition of sequences that are inaccessible in the fully folded protein (used by the surface-oriented computational approaches). Based on such considerations, we set out to extend the reach of our computational prediction methods to include regions of low accessibility in the folded protein. Both MLCE and EDP methods are based on the analysis of the surface properties of the protein, where accessibility and optimal antibody recognition of certain epitopes may be limited, e.g., by steric hindrance. To overcome this limitation, we combined the MLCE-EDP epitope prediction strategy with a recently developed domain decomposition approach that allows in silico dissection of a folded protein into smaller fragments (Genoni et al., 2012) (see Experimental Procedures). The underlying hypothesis is that such fragments expose sequence stretches that may be targeted by antibodies under conditions of partial unfolding or degradation of the antigen protein. Application of the domain decomposition algorithm allowed us to identify six boundaries and cluster the results into three different fragments: A0 (residues 83–227 and 428–507), B0 (residues 48–82, 228–302, and 528–553), and C0 (residues 303–422 and 508–527), which partially overlap with the OppABp structural domains (Figure 4). EXP4 and EXP6 peptides are entirely contained in fragments A0 and C0 , respectively, whereas EXP5 extends across fragments B0 and C0 . MLCE and EDP predictions were then applied to the isolated A0 , B0 , and C0 fragments, identifying potential epitopes that very satisfactory overlap with the peptides identified by immunocapture. In fact, in accordance with experimental mapping, part of the helical segment of EXP4 (consensus residues A130–R138) was identified as a potential epitope region upon MLCE-EDP analysis of fragment A0 . Computational analysis of fragment B0 identified as a potential epitope an exposed part (consensus

170 Structure 21, 167–175, January 8, 2013 ª2013 Elsevier Ltd All rights reserved

Structure Epitope Discovery for Structural Vaccinology

Figure 2. OppABp Computational Epitope Predictions Consensus MLCE-EDP predictions. The left panel is a 3D surface representation of OppABp: MLCE (red), EDP (green), and consensus (yellow) predicted epitopes are highlighted on the surface. Right panel is a secondary structure representation of the consensus epitope region containing the three epitopes: COMP1, blue; COMP2, green; and COMP3, red. These figures were produced using VMD ver. 1.9. Figure S1 presents the results of the MD relaxation run on OppABp crystal structure previous to the computational epitope predictions. Figure S2 is a visual representation showing the MLCE and EDP predictions.

residues Y309–K314) of EXP5, whereas analysis of the C0 fragment retrieved a large portion of EXP6 and COMP2 (consensus residues L353–P364 and T366–A370). The extended algorithm, however, did not identify the EXP5 portion located on fragment C0 as a possible epitope. DISCUSSION Structural vaccinology is a developing field that first aims to devise viable strategies for vaccine design, based on the identification of immunogenic determinants (epitopes), through computational and 3D structure analyses. In this context, the crystal structure of the OppABp antigen was used as a test case in a computational search strategy to identify epitopes, in view of a future broader implementation in the diagnostic and vaccine fields but also with methodological aims. Available data regarding the immunogenic potential of OppABp confirm that the protein is recognized by the humoral and cell-mediated branches of the human immune response, both known to be necessary for protection against B. pseudomallei infection (Healey et al., 2005). OppABp is recognized by antibodies present in the plasma of melioidosis-infected and recovered subjects, although it did not offer protection in preliminary studies on mice challenged with B. pseudomallei (Felgner et al., 2009; Harland et al., 2007; Suwannasaen et al., 2011). With regard to cell-mediated responses, OppABp is recognized by T cells in plasma from seropositive individuals (Tippayawat et al., 2009). The inability of recombinant OppABp to be used in its native form in a vaccine renders it an ideal target for structure-based engineering to foster its antigenicity. The conformational dynamics of an antigen is a key property in determining its immunogenic potential (Ofek et al., 2010; Westhof et al., 1984); its characterization helps locate solventaccessible regions of the protein that are most likely to house epitope regions. In this respect, MD simulations were carried out on the crystal structure of OppABp, as a prerequisite for subsequent computational epitope predictions. Two computational prediction methods (MLCE and EDP) were then combined, resulting in the identification of three consensus epitopes (COMP1–COMP3). In parallel to the computational predictions, experimental epitope mapping was carried out on recombinant OppABp using proteolysis, polyclonal sera from mice, and MS

analysis; three epitope-carrying peptides were thus identified (EXP4–EXP6). To assess antigenicity, all six epitopes were synthesized as HSA-conjugated peptides and tested for antibody response in human plasma samples. COMP1–COMP3 and EXP4 peptides were selectively recognized by antibodies present in plasma samples from patients that had recovered from B. pseudomallei infection, in comparison with healthy seronegative controls (Figure 3); notably, EXP4 and COMP1 presented very similar reactivity patterns. Human plasma antibodies reactive versus all peptides but COMP2 were also significantly higher in the healthy seropositive group than in the seronegative one. In addition, plasma antibodies recognizing EXP5–EXP6 were significantly higher in the seropositive group than in the recovered melioidosis group, suggesting that these peptides might find application in discriminating between asymptomatic and symptomatic melioidosis. A particular feature was the distinct immunoreactivity displayed by the synthetic COMP3 peptide, which produced characteristic profiles in seronegative, healthy seropositive, and melioidosis-recovered groups, suggesting its potential use in diagnostics. Developments along these lines appear relevant, considering that it is still very difficult to diagnose B. pseudomallei infections due to the lack of specific symptoms. Indeed, melioidosis is often misdiagnosed as tuberculosis, delaying correct antibiotic treatment and thus impairing the patients’ ability to overcome infection. As an additional point and unexpectedly, the control recombinant OppABp was found to be much more immunoreactive in recovered patient sera relative to what had been reported earlier by Suwannasaen et al. (2011). This could imply that different protein preparations may affect the results of these studies and confirms the importance of further investigation of OppABp. The application of MLCE-EDP methods to the OppABp crystal structure allowed us to identify peptides COMP1–COMP3 as potential antigens; these, however, did not match in their sequence location the EXP4–EXP6 peptides. Because all six peptides synthesized were recognized by human antibodies, we conclude that they all are epitopes (or contain fragments of OppABp epitopes). Besides the obvious consideration that every protein hosts several distinct epitopes, the aforementioned observation suggests that different prediction methods may preferentially lead to the identification of different epitope

Structure 21, 167–175, January 8, 2013 ª2013 Elsevier Ltd All rights reserved 171

Structure Epitope Discovery for Structural Vaccinology

Figure 3. Antibody Response to B. pseudomallei OppABp Epitopes in Plasma of Healthy and Recovery Melioidosis Subjects Crude B. pseudomallei antigen (A), recombinant OppABp (B), and synthetic peptides COMP1 (C), COMP2 (D), COMP3 (E), EXP4 (F), EXP5 (G), and EXP6 (H) were coated onto ELISA plates and probed with diluted plasma samples of healthy seropositive individuals (S+; n = 12) tested by means IHA (titer >40), healthy seronegative individuals (S; n = 7), and recovery melioidosis individuals (R; n = 20), and quantified by indirect ELISA. Data represent the absorbance index (AI) of individual samples = (OD of tested  OD of uncoated)/OD of uncoated. Experiments were performed in duplicate, and results represent mean (AI) ± SE, Mann-Whitney U test; *p < 0.05, **p < 0.01, and ***p < 0.001 values. ns, not significant.

regions. Additionally, the multiepitope nature of any antigen may explain the different antibody responses depending on the progression of the disease. The aforementioned considerations prompted us to search for an extension of the computational methods that could help improve the agreement between in silico predictions and the experimental epitope mapping results. Because parts of the experimentally identified peptides (EXP4–EXP6) were internally located in the OppABp fold, we considered that such sequence stretches were not probed by the MLCE-EDP algorithms, which exclusively focus on solvent-accessible residues. To account for the potential exposure of internal sequence segments under conditions of partial unfolding or degradation, our computational approach was integrated with an energybased ab initio analysis designed to dissect the protein 3D structure into smaller fragments. In such a way, sequence stretches that are partially buried, or not optimally accessible due to steric reasons, can be identified as possible antibody-

binding sites. In agreement with such theoretical considerations, consensus epitope predictions (domain decomposition coupled to MLCE-EDP) run on the resulting three distinct OppABp fragments resulted in improved identification of protein regions matching the experimentally mapped epitopes (EXP4–EXP6) (see Figure 4). As a methodological speculation, we consider that the MLCE-EDP consensus prediction method applied to intact antigen structure proved successful in the design of peptides that house immunogenic determinants recognized by B cell antibodies. In a complementary way, incorporating in silico domain fragmentation of the antigenic protein, based on rational physicochemical principles, improved the match to epitopes experimentally mapped by means of murine polyclonal antibodies, proteolysis, and immunocapturing. The combination of all approaches reported here depicts a successful methodological pipeline with potential in a structural vaccinology context. Among different applications, the computational part could be implemented for screening libraries of known antigens, to generate a large, yet focused, collection of B cell epitope predictions in a minimal time. One possible limitation at this stage might be represented by the necessity of a high-quality structure of the protein antigen. However, this problem may be alleviated by the constant increase in the number of experimentally solved protein structures as well as the steady improvement in structure prediction methods. There are several applications for antigenic peptides. For example, with regard to melioidosis, immunization with lipopolysaccharide and capsular polysaccharide resulted in a delayed mortality in mice challenged with B. pseudomallei; conjugation to antigenic peptides from OppABp may result in increased

172 Structure 21, 167–175, January 8, 2013 ª2013 Elsevier Ltd All rights reserved

Structure Epitope Discovery for Structural Vaccinology

Figure 4. Energy-Based Domain Decomposition of OppABp and Epitope Prediction The 3D structure of OppABp is shown with the defined fragments A0 , B0 , and C0 highlighted in yellow, red, and green, respectively. The fragments predicted through the decomposition algorithm are also shown individually in the shaded panels. The epitope regions mapped by immunocapture (EXP4, EXP5, and EXP6) are highlighted in light blue in both the full structure and in the individual fragments. Regions of the EXP epitopes matching those predicted by MLCE plus EDP on the isolated fragments are shown in dark blue and marked by arrows.

Epitope Prediction and Design The crystal structure of OppABp was used as a starting point for a 30 ns all-atom MD simulation in explicit water at 300 K. The simulations and the analysis of the trajectories were performed using the GROMACS 4.5.1 software package (Hess et al., 2008), GROMOS96 force field (van Gunsteren et al., 2006), and the SPC water model (Berendsen et al., 1987). The procedure employed is fully described in Supplemental Experimental Procedures.

protection (Nelson et al., 2004). Because the majority of presentday vaccines contain more than one antigen, such peptides could also be conjugated to, or used in conjunction with, other antigens (Purcell et al., 2007). As a final note, as for the OppABp peptides that were recognized differently by antibodies from healthy seropositive (EXP5–EXP6) and recovered patients (COMP1–COMP3 and EXP4), but not by healthy seronegative plasma, the identified antigenic peptides may provide a direct route for the development of diagnostic tools, which in the case of B. pseudomallei infections, remains a major unsolved challenge. EXPERIMENTAL PROCEDURES Crystallization, Data Collection, Model Building, and Refinement OppABp crystals were grown in a sitting drop setup at 20 C, in a 300 nl drop containing 30% protein solution (5 mg/ml) and reservoir solution (0.1 M HEPES [pH 7.5] and 20% PEG 8000). Crystals were cryoprotected in a solution containing 0.1 M HEPES (pH 7.5), 25% PEG 8000, and 30% glycerol. One crystal was used to collect X-ray diffraction data at the ID23-2 beamline at the European Synchrotron Radiation Facility (Grenoble, France). Data were processed using programs available from the CCP4 suite (Collaborative Computational Project, Number 4, 1994; Evans, 2006; Leslie, 2006), and the structure was solved by molecular replacement using Phaser (McCoy, 2007) and the structure of OppA from Salmonella typhimurium (PDB code 1QKA) as the search model (Sleigh et al., 1999). The structure was refined to satisfactory crystallographic and geometric parameters with PHENIX.refine, under the PHENIX platform (Adams et al., 2010; Afonine et al., 2010; Chen et al., 2010; Emsley and Cowtan, 2004; Murshudov et al., 1997) (Table 1). Additional details are provided in the Supplemental Experimental Procedures. Atomic coordinates and structure factors have been deposited in the Protein Data Bank (http://www.rcsb.org/pdb), under accession code 3ZS6 (Berman et al., 2000).

MLCE Method Epitope predictions were carried out on the representative structure of the most populated structural cluster obtained using the method developed by Daura et al. (1999). The MLCE method (Scarabelli et al., 2010) is based on the eigenvalue decomposition of the matrix of residue-residue energy couplings (Colacino et al., 2006a, 2006b; Morra and Colombo, 2008; Ragona et al., 2005; Tiana et al., 2004) (see Supplemental Experimental Procedures), available as the free web tool BEPPE (http:// bioinf.uab.es/BEPPE). EDP Method In order to increase the consistency of the epitope predictions, the MLCE output was crossed with the EDP output for consensus surfaces. The EDP method calculates the free energy penalty for desolvation placing a neutral probe at various protein surface positions. Surface regions with a small free energy penalty for water removal may correspond to preferred interaction sites. Evidence suggests that it is easier for an antibody to bind to an epitope when properties required for high-affinity binding like low desolvation penalty are met (Fiorucci and Zacharias, 2010). Epitope Identification Based on MLCE-EDP Consensus The final predicted epitopes, which represent the sequences chosen for peptide synthesis, were obtained from the consensus results of the two predictions. From sequence alignments, we selected the corresponding overlapping regions, which were defined as candidate epitopes (COMP1–COMP3). The candidate epitopes were produced via solid-phase synthesis as reported below. Epitope Mapping with Murine Sera Peptide mixtures were obtained by trypsin digestion in 50 mM ammonium bicarbonate buffer at a ratio of 10:1, at 37 C for 3 hr. To capture the epitope-containing peptide, a 25 ml suspension of Dynabeads Pan Mouse IgG (uniform, superparamagnetic polystyrene beads of 4.5 mm diameter coated with monoclonal human anti-mouse IgG antibodies) was used. The beads were washed twice with PBS using a magnet and resuspended in the initial volume. A total of 50 ml of the murine serum, prepared as described in Supplemental Experimental Procedures, was added and incubated for 30 min at room temperature (RT), after which the beads were washed five

Structure 21, 167–175, January 8, 2013 ª2013 Elsevier Ltd All rights reserved 173

Structure Epitope Discovery for Structural Vaccinology

times with PBS to remove serum debris. A total of 0.5 ml of Protease Inhibitor Mix (GE Healthcare) was added before the peptide mixture to avoid potential antibody degradation. The sample was then incubated for 2 hr at RT with gentle tilting and rotation. After incubation, beads were washed three times with 1 ml PBS, and the bound peptides were eluted in 50 ml of 0.2% TFA. The elute fraction was concentrated and washed with C18 ZipTips (Millipore) and eluted in 2 ml of 50% ACN and 0.1% TFA. Subsequent MALDI-MS analysis of the eluted fractions was carried out as described in the Supplemental Experimental Procedures. Synthesis of Epitope Peptides In view of their use as immunogens to generate antibodies, the original COMP peptides were elongated by including two flanking residues at the N and C termini, based on the consideration that longer sequences induce better immunogenic responses (Purcell et al., 2007). PEG moieties (MW 308.16) were introduced at the N terminus of each peptide (COMP1–COMP3 and EXP4–EXP6) as spacers for subsequent immunological tests. A cysteine residue preceded the PEG group to facilitate conjugation to HSA, the carrier protein. All peptides were prepared in free- and HSA-conjugated forms. COMP1 is a discontinuous conformational epitope composed of two short peptides that are brought close together in the tertiary structure. Measurement of the average distance between the termini of COMP1 from MD simulations suggested that insertion of two glycine residues would be sufficient to bridge the two short sequences. Indirect ELISA to Detect Antibodies to B. pseudomallei Antibody recognition of the peptides was detected using indirect ELISA using 96-well microtiter plates (Nunc; MaxiSorp) that were uncoated or coated with 50 ml/well of 3 mg/ml crude B. pseudomallei-extracted antigens (Crude Bps), or 10 mg/ml of recombinant OppABp as controls, or 3 mg/ml peptides COMP1, COMP2, COMP3, EXP4, EXP5, and EXP6 in 0.1 M carbonate-bicarbonate buffer (pH 9.6), incubating at 37 C for 3 hr. Immunoreactivity was revealed, probing the plates with 1:300 diluted plasma samples from healthy and recovery melioidosis subjects. See Supplemental Experimental Procedures for details of the procedure and human plasma sample preparation. Domain Mapping and Cleavage Site Prediction Domain mapping was performed using a newly developed ab initio computational method by Genoni et al. (2012) (see Supplemental Experimental Procedures). This technique is an expansion of the energy decomposition method (Tiana et al., 2004), which diagonalizes the protein-nonbonded interaction energy matrix Enb (namely, van der Waals and electrostatics) to identify the amino acids necessary for the stabilization of the protein fold from the eigenvector associated with the lowest eigenvalue.

SUPPLEMENTAL INFORMATION Supplemental Information includes two figures and Supplemental Experimental Procedures and can be found with this article online at http://dx.doi. org/10.1016/j.str.2012.10.005. ACKNOWLEDGMENTS This work was supported by CARIPLO Foundation Project ‘‘From Genome to Antigen: a Multidisciplinary Approach towards the Development of an Effective Vaccine against Burkholderia pseudomallei, the Etiological Agent of Melioidosis’’ (contract number 2009-3577). Support was also received from MIUR PRIN 2008 (Grant No.2008K37RHP) and MIVR PMN 2008 (Grant No. 2008K37RHP). P.L. and C.P. were fellowship recipients of the CARIPLO grant, and L.J.G. is a recipient of Assegno di Ricerca (2012) from the University of Milano, all of which are gratefully acknowledged. Received: June 21, 2012 Revised: September 10, 2012 Accepted: October 4, 2012 Published online: November 15, 2012

REFERENCES Adams, P.D., Afonine, P.V., Bunko´czi, G., Chen, V.B., Davis, I.W., Echols, N., Headd, J.J., Hung, L.W., Kapral, G.J., Grosse-Kunstleve, R.W., et al. (2010). PHENIX: a comprehensive Python-based system for macromolecular structure solution. Acta Crystallogr. D Biol. Crystallogr. 66, 213–221. Afonine, P.V., Grosse-Kunstleve, R.W., Chen, V.B., Headd, J.J., Moriarty, N.W., Richardson, J.S., Richardson, D.C., Urzhumtsev, A., Zwart, P.H., and Adams, P.D. (2010). phenix.model_vs_data: a high-level tool for the calculation of crystallographic model and data statistics. J. Appl. Cryst. 43, 669–676. Berendsen, H.J.C., Grigera, J.R., and Straatsma, T.P. (1987). The missing term in effective pair potentials. J. Phys. Chem. 91, 6269–6271. Berman, H.M., Westbrook, J., Feng, Z., Gilliland, G., Bhat, T.N., Weissig, H., Shindyalov, I.N., and Bourne, P.E. (2000). The Protein Data Bank. Nucleic Acids Res. 28, 235–242. Berntsson, R.P., Alia Oktaviani, N., Fusetti, F., Thunnissen, A.M., Poolman, B., and Slotboom, D.J. (2009). Selenomethionine incorporation in proteins expressed in Lactococcus lactis. Protein Sci. 18, 1121–1127. Borezee, E., Pellegrini, E., and Berche, P. (2000). OppA of Listeria monocytogenes, an oligopeptide-binding protein required for bacterial growth at low temperature and involved in intracellular survival. Infect. Immun. 68, 7069– 7077. Chen, V.B., Arendall, W.B., 3rd, Headd, J.J., Keedy, D.A., Immormino, R.M., Kapral, G.J., Murray, L.W., Richardson, J.S., and Richardson, D.C. (2010). MolProbity: all-atom structure validation for macromolecular crystallography. Acta Crystallogr. D Biol. Crystallogr. 66, 12–21. Cheng, A.C., and Currie, B.J. (2005). Melioidosis: epidemiology, pathophysiology, and management. Clin. Microbiol. Rev. 18, 383–416. Colacino, S., Tiana, G., Broglia, R.A., and Colombo, G. (2006a). The determinants of stability in the human prion protein: insights into folding and misfolding from the analysis of the change in the stabilization energy distribution in different conditions. Proteins 62, 698–707. Colacino, S., Tiana, G., and Colombo, G. (2006b). Similar folds with different stabilization mechanisms: the cases of Prion and Doppel proteins. BMC Struct. Biol. 6, 17. Collaborative Computational Project, Number 4. (1994). The CCP4 suite: programs for protein crystallography. Acta Crystallogr. D Biol. Crystallogr. 50, 760–763. Currie, B.J., Dance, D.A., and Cheng, A.C. (2008). The global distribution of Burkholderia pseudomallei and melioidosis: an update. Trans. R. Soc. Trop. Med. Hyg. 102 (Suppl 1 ), S1–S4. Daura, X., Gademann, K., Jaun, B., Seebach, D., van Gunsteren, W.F., and Mark, A.E. (1999). Peptide folding: when simulation meets experiment. Angew. Chem. Int. Ed. Engl. 38, 236–240. Dormitzer, P.R., Ulmer, J.B., and Rappuoli, R. (2008). Structure-based antigen design: a strategy for next generation vaccines. Trends Biotechnol. 26, 659–667. Dunten, P., and Mowbray, S.L. (1995). Crystal structure of the dipeptide binding protein from Escherichia coli involved in active transport and chemotaxis. Protein Sci. 4, 2327–2334. Emsley, P., and Cowtan, K. (2004). Coot: model-building tools for molecular graphics. Acta Crystallogr. D Biol. Crystallogr. 60, 2126–2132. Evans, P. (2006). Scaling and assessment of data quality. Acta Crystallogr. D Biol. Crystallogr. 62, 72–82. Felgner, P.L., Kayala, M.A., Vigil, A., Burk, C., Nakajima-Sasaki, R., Pablo, J., Molina, D.M., Hirst, S., Chew, J.S., Wang, D., et al. (2009). A Burkholderia pseudomallei protein microarray reveals serodiagnostic and cross-reactive antigens. Proc. Natl. Acad. Sci. USA 106, 13499–13504. Fiorucci, S., and Zacharias, M. (2010). Prediction of protein-protein interaction sites using electrostatic desolvation profiles. Biophys. J. 98, 1921–1930. Garmory, H.S., and Titball, R.W. (2004). ATP-binding cassette transporters are targets for the development of antibacterial vaccines and therapies. Infect. Immun. 72, 6757–6763.

174 Structure 21, 167–175, January 8, 2013 ª2013 Elsevier Ltd All rights reserved

Structure Epitope Discovery for Structural Vaccinology

Genoni, A., Morra, G., and Colombo, G. (2012). Identification of domains in protein structures from the analysis of intramolecular interactions. J. Phys. Chem. B 116, 3331–3343.

Purcell, A.W., McCluskey, J., and Rossjohn, J. (2007). More than one reason to rethink the use of peptides in vaccine design. Nat. Rev. Drug Discov. 6, 404–414.

Harland, D.N., Chu, K., Haque, A., Nelson, M., Walker, N.J., Sarkar-Tyson, M., Atkins, T.P., Moore, B., Brown, K.A., Bancroft, G., et al. (2007). Identification of a LolC homologue in Burkholderia pseudomallei, a novel protective antigen for melioidosis. Infect. Immun. 75, 4173–4180.

Ragona, L., Colombo, G., Catalano, M., and Molinari, H. (2005). Determinants of protein stability and folding: comparative analysis of beta-lactoglobulins and liver basic fatty acid binding protein. Proteins 61, 366–376.

Healey, G.D., Elvin, S.J., Morton, M., and Williamson, E.D. (2005). Humoral and cell-mediated adaptive immune responses are required for protection against Burkholderia pseudomallei challenge and bacterial clearance postinfection. Infect. Immun. 73, 5945–5951. Hess, B., Kutzner, C., van der Spoel, D., and Lindahl, E. (2008). GROMACS 4: algorithms for highly efficient, load-balanced, and scalable molecular simulation. J. Chem. Theory Comput. 4, 435–447. Koehler, C., Carlier, L., Veggi, D., Balducci, E., Di Marcello, F., Ferrer-Navarro, M., Pizza, M., Daura, X., Soriani, M., Boelens, R., and Bonvin, A.M. (2011). Structural and biochemical characterization of NarE, an iron-containing ADP-ribosyltransferase from Neisseria meningitidis. J. Biol. Chem. 286, 14842–14851. Leslie, A.G. (2006). The integration of macromolecular diffraction data. Acta Crystallogr. D Biol. Crystallogr. 62, 48–57. Levdikov, V.M., Blagova, E.V., Brannigan, J.A., Wright, L., Vagin, A.A., and Wilkinson, A.J. (2005). The structure of the oligopeptide-binding protein, AppA, from Bacillus subtilis in complex with a nonapeptide. J. Mol. Biol. 345, 879–892. McCoy, A.J. (2007). Solving structures of protein complexes by molecular replacement with Phaser. Acta Crystallogr. D Biol. Crystallogr. 63, 32–41. Monnet, V. (2003). Bacterial oligopeptide-binding proteins. Cell. Mol. Life Sci. 60, 2100–2114. Morra, G., and Colombo, G. (2008). Relationship between energy distribution and fold stability: insights from molecular dynamics simulations of native and mutant proteins. Proteins 72, 660–672. Murshudov, G.N., Vagin, A.A., and Dodson, E.J. (1997). Refinement of macromolecular structures by the maximum-likelihood method. Acta Crystallogr. D Biol. Crystallogr. 53, 240–255. Nelson, M., Prior, J.L., Lever, M.S., Jones, H.E., Atkins, T.P., and Titball, R.W. (2004). Evaluation of lipopolysaccharide and capsular polysaccharide as subunit vaccines against experimental melioidosis. J. Med. Microbiol. 53, 1177–1182. Nuccitelli, A., Cozzi, R., Gourlay, L.J., Donnarumma, D., Necchi, F., Norais, N., Telford, J.L., Rappuoli, R., Bolognesi, M., Maione, D., et al. (2011). Structurebased approach to rationally design a chimeric protein for an effective vaccine against Group B Streptococcus infections. Proc. Natl. Acad. Sci. USA 108, 10278–10283. Ofek, G., Guenaga, F.J., Schief, W.R., Skinner, J., Baker, D., Wyatt, R., and Kwong, P.D. (2010). Elicitation of structure-specific antibodies by epitope scaffolds. Proc. Natl. Acad. Sci. USA 107, 17880–17887.

Sarkar-Tyson, M., and Titball, R.W. (2010). Progress toward development of vaccines against melioidosis: a review. Clin. Ther. 32, 1437–1445. Scarabelli, G., Morra, G., and Colombo, G. (2010). Predicting interaction sites from the energetics of isolated proteins: a new approach to epitope mapping. Biophys. J. 98, 1966–1975. Schneewind, O., and Missiakas, D. (2011). Structural vaccinology to thwart antigenic variation in microbial pathogens. Proc. Natl. Acad. Sci. USA 108, 10029–10030. Sleigh, S.H., Seavers, P.R., Wilkinson, A.J., Ladbury, J.E., and Tame, J.R. (1999). Crystallographic and calorimetric analysis of peptide binding to OppA protein. J. Mol. Biol. 291, 393–415. Soriani, M., Petit, P., Grifantini, R., Petracca, R., Gancitano, G., Frigimelica, E., Nardelli, F., Garcia, C., Spinelli, S., Scarabelli, G., et al. (2010). Exploiting antigenic diversity for vaccine design: the chlamydia ArtJ paradigm. J. Biol. Chem. 285, 30126–30138. Suwannasaen, D., Mahawantung, J., Chaowagul, W., Limmathurotsakul, D., Felgner, P.L., Davies, H., Bancroft, G.J., Titball, R.W., and Lertmemongkolchai, G. (2011). Human immune responses to Burkholderia pseudomallei characterized by protein microarray analysis. J. Infect. Dis. 203, 1002–1011. Tanabe, M., Atkins, H.S., Harland, D.N., Elvin, S.J., Stagg, A.J., Mirza, O., Titball, R.W., Byrne, B., and Brown, K.A. (2006). The ABC transporter protein OppA provides protection against experimental Yersinia pestis infection. Infect. Immun. 74, 3687–3691. Tanabe, M., Mirza, O., Bertrand, T., Atkins, H.S., Titball, R.W., Iwata, S., Brown, K.A., and Byrne, B. (2007). Structures of OppA and PstS from Yersinia pestis indicate variability of interactions with transmembrane domains. Acta Crystallogr. D Biol. Crystallogr. 63, 1185–1193. Tiana, G., Simona, F., De Mori, G.M., Broglia, R.A., and Colombo, G. (2004). Understanding the determinants of stability and folding of small globular proteins from their energetics. Protein Sci. 13, 113–124. Tippayawat, P., Saenwongsa, W., Mahawantung, J., Suwannasaen, D., Chetchotisakd, P., Limmathurotsakul, D., Peacock, S.J., Felgner, P.L., Atkins, H.S., Titball, R.W., et al. (2009). Phenotypic and functional characterization of human memory T cell responses to Burkholderia pseudomallei. PLoS Negl. Trop. Dis. 3, e407. van Gunsteren, W.F., Bakowies, D., Baron, R., Chandrasekhar, I., Christen, M., Daura, X., Gee, P., Geerke, D.P., Gla¨ttli, A., Hu¨nenberger, P.H., et al. (2006). Biomolecular modeling: goals, problems, perspectives. Angew. Chem. Int. Ed. Engl. 45, 4064–4092. Westhof, E., Altschuh, D., Moras, D., Bloomer, A.C., Mondragon, A., Klug, A., and Van Regenmortel, M.H. (1984). Correlation between segmental mobility and the location of antigenic determinants in proteins. Nature 311, 123–126.

Structure 21, 167–175, January 8, 2013 ª2013 Elsevier Ltd All rights reserved 175