ABSTRACT
Unbiased identification of individual immunogenic B-cell epitopes in major antigens of a pathogen remains a technology challenge for vaccine discovery. We therefore developed a platform for rapid phage display screening of deep recombinant libraries consisting of as few as one major pathogen antigen. Using the bicomponent pore-forming leukocidin (Luk) exotoxins of the major pathogen Staphylococcus aureus as a prototype, we randomly fragmented and separately ligated the hemolysin gamma A (HlgA) and LukS genes into a custom-built phage display system, termed pComb-Opti8. Deep sequence analysis of barcoded amplimers of the HlgA and LukS gene fragment libraries demonstrated that biopannng against a cross-reactive anti-Luk monoclonal antibody (MAb) recovered convergent molecular clones with short overlapping homologous sequences. We thereby identified an 11-amino-acid sequence that is highly conserved in four Luk toxin subunits and is ubiquitous in representation within S. aureus clinical isolates. The isolated 11-amino-acid peptide probe was predicted to retain the native three-dimensional (3D) conformation seen within the Luk holotoxin. Indeed, this peptide was recognized by the selecting anti-Luk MAb, and, using mutated peptides, we showed that a particular amino acid side chain was essential for these interactions. Furthermore, murine immunization with this peptide elicited IgG responses that were highly reactive with both the autologous synthetic peptide and the full-length Luk toxin homologues. Thus, using a gene fragment- and phage display-based pipeline, we have identified and validated immunogenic B-cell epitopes that are cross-reactive between members of the pore-forming leukocidin family. This approach could be harnessed to identify novel epitopes for a much-needed S. aureus-protective subunit vaccine.
This work is dedicated to the legacy of Carlos F. Barbas III.
INTRODUCTION
Directed evolution approaches (e.g., ribosome, phage, yeast, and mammalian cell display) offer robust high-throughput screening of minimally or broadly randomized libraries of peptides and proteins against a target. These approaches have been used extensively to discover antibodies (Abs) or fragments thereof with a desired specificity, including autoimmune and protective vaccine-elicited Abs that are then the starting points for biologic drug development. We reasoned that such an approach might also facilitate the subsequent characterization of the antigen target binding sites, and we posited that the thorough randomization of a focused library for a target antigen would instead be far better matched for the purpose of finding B-cell epitopes. We therefore developed a technology for this purpose and applied it to Staphylococcus aureus B-cell epitope discovery.
S. aureus is a significant human pathogen. Despite the ubiquity and the benign nature of S. aureus strains as residents of the microbiome communities of healthy individuals, S. aureus is responsible for a large variety of diseases in both community and hospital environments (1). In general, the past decade has witnessed alarming increases in the societal burden of invasive staphylococcal disease, particularly disease caused by antibiotic-resistant strains known as methicillin-resistant Staphylococcus aureus (MRSA) strains, which have severely restricted antimicrobial treatment options. Community-associated MRSA (CA-MRSA) accounts for much of this burden (2–6) and is a major public health threat; MRSA is on the World Health Organization list of bacteria that pose the greatest threats to human health (7). Indeed, MRSA strains kill an estimated 19,000 Americans per year, more than HIV, and more than Streptococcus pneumoniae, Neisseria meningitidis, Haemophilus influenzae, and Streptococcus pyogenes combined (3). In addition to skin and soft tissue infection (SSTI) in children and adults, S. aureus, particularly MRSA, causes invasive infections, such as pneumonia, septic arthritis and osteomyelitis. Indeed, S. aureus is one of the leading and most fatal causes of bacteremia, with an estimated mortality rate of 20% (8, 9). Patients with bacteremic disease are in danger not only of sepsis but also of metastatic infections that frequently persist and relapse even with optimal treatment (10, 11).
All clinical S. aureus strains carry genes for large arsenals of dozens of toxins, proteases, and immunoglobulin-binding proteins (12, 13), which have been implicated in facilitating colonization, host immune evasion, and invasion of host tissues, resulting in clinical infection syndromes. Posing a seeming paradox, antibody responses to S. aureus antigens are common in humans, whereas multiple mechanisms have been postulated by which this opportunistic pathogen may evade antibody-mediated host defenses (14). Hence, while on the one hand adaptive immunity is believed to be essential for an effective host defense, these responses appear to be inadequate in those who develop serious and often recurrent infections. However, all efforts to date to develop a vaccine against S. aureus have not yet been successful (reviewed in references 15 and 16).
In recent years, experimental systems have highlighted the importance in infection of several S. aureus toxins that may represent attractive vaccine candidates (17–19). We have therefore initially focused on the leukocidin (Luk) toxin family, the members of which serve as virulence factors in murine models of diseases and which are postulated to provide the pathogen with countermeasures for abrogation of host immune defenses (reviewed in reference 17). The Luk family includes nine distinct homologous toxin subunits that can be active during infection through specific heterodimeric pairing, which occurs during assembly on the membrane of a targeted host cell type. Formation of oligomeric pore-like structures by the different subunits on the plasma membrane causes toxicity and ultimately the death of neutrophils, lymphocytes, and other cell types essential for host defenses. In murine models, the deletion of a Luk gene from the infecting strain can greatly diminish the pathogenicity and seriousness of infection (20–27). Furthermore, pharmacologic blockade of a toxin binding to its receptor on host cells diminishes S. aureus in vitro toxin functional activity, as well as the associated morbidity and mortality, in small animal models (18, 28–33).
To identify primary sequence-dependent Luk B-cell epitopes that could eventually be incorporated into a vaccine, we hypothesized that phage display technology might provide uniquely well-suited opportunities. These systems have been widely used for the selection of monoclonal antibodies (MAbs), as well as engineered antibody and protein domain variants with improved specificity and affinity (34), and for selection from artificial peptide libraries (35). However, the potential utility of phage display for the molecular characterization of minimal B-cell/antibody epitopes targeted during the formation of effective host immune defenses has been largely overlooked (36).
Toward the goal of identifying Luk B-cell epitopes important for host defenses, we constructed the customized pComb-Opti8 vector to enable multivalent display of libraries of randomly digested fragments of the genes for several Luk toxin subunits. The pComb-Opti8 vector system enabled the identification of Luk antigenic sites that are directly and specifically recognized by murine MAbs generated via a conventional hybridoma approach. With this molecular tool, we then developed a pipeline for the identification of minimal B-cell epitope and optimization of synthetic peptide probes, in which each recapitulates the specificity and binding activity of the parent MAb to the target B-cell epitope in the intact native parental Luk toxin subunit protein.
RESULTS
Development of a vector system for identification of candidate B-cell epitopes from gene fragment libraries.For our discovery studies, we adapted a well-validated phagemid expression system by incorporating an optimized flexible linker and a truncated version of the filamentous phage major coat protein (adapted from reference 37), which we termed pComb-Opti8 (see Fig. S1 in the supplemental material) (see Materials and Methods). This phagemid vector was designed to facilitate multivalent surface display of copies of a fusion protein, mixed with a much greater number of wild-type gpVIII molecules required for structural integrity (38).
With the goal of identifying primary amino acid sequences containing minimal epitopes recognized by monoclonal antibodies, we developed and validated a protocol for generating libraries of gene fragments (see Fig. S2 in the supplemental material), which were then cloned into the pComb-Opti8 vector. Following helper phage rescue, each pComb-Opti8 gene fragment library represents a collection of infectious particles that each carry the encoding fusion gene, with the translated product of a Luk protein fragment fused with the major coat protein arrayed on the surface of that phage (illustrated in Fig. S3 in the supplemental material). We found that the unpanned hemolysin gamma A (HlgA) fragment library contains gene fragment inserts with a size distribution of all the double-stranded DNA (dsDNA) fragments, which had a maximum length of 159 nucleotides, representing 53 codons. These inserts varied in length with a mean of 48 nucleotides that were predicted to produce functional 16-codon peptide fragments, which indicated a wide range of potential small proteins that were large enough to mimic surface determinants, including to discontinuous epitopes.
Biopanning with the anti-LukE3 antibody selects for homologous HIgA fragments.For these proof-of-principle studies, we used a murine IgG2a antibody that was generated by immunization with S. aureus-isolated recombinant LukE recovered from a subcloned hybridoma, termed the anti-LukE3 MAb (see Fig. S4 in the supplemental material). This MAb was selected for further study, as it also recognizes the Luk subunit homologues, LukS, HlgA, and HlgC, with high reactivity (Fig. 1). Adapting previously described methods (39), the anti-LukE3 MAb was used in biopanning as the selecting “bait” from an HlgA fragment library, and the sequentially selected sublibraries were termed pan 1, pan 2, pan 3, and pan 4 libraries (Fig. 2). To gain insight into the representation and distribution of individual clones en masse, each (sub)library was barcoded and individually amplified, and each was sequenced and then analyzed using a gene assembly approach (see Materials and Methods).
Anti-LukE3 MAb uniquely cross-reacts with leukocidin homologues. (A) Mean fluorescence intensity (MFI) of Magplex beads coated with various S. aureus antigens (x axis) to anti-LukE3 MAb. The anti-LukE3 MAb was incubated with the beads at 500 ng/ml, 50 ng/ml, 5 ng/ml, and 0.5 ng/ml, with results plotted as sequential bars representing the mean with standard deviation (SD). (B) (Left) Superimposition of the three-dimensional (3D) crystallographic structures of LukS (PDB identifier [ID] 4iya), LukE (PDB ID 3roh), and HlgA (PDB ID 2qk7) illustrates the conserved tertiary structure of the LukS homologues. (Right) Rotated view. Image produced in PyMOL.
Sequentially biopanning against the anti-LukE3 MAb selected for a conserved HlgA gene subregion encoded within the HIgA leukocidin gene. (A) Deep sequencing of the sublibraries generated by biopanning of the HlgA fragment library by anti-LukE3 indicates that a gene subregion conserved among the leukocidins encodes a sequence that correlates with a loop in the stem domain, as identified in structural analyses. Phage fragment clones that distributed into one HlgA gene subregion were found to undergo preferential sequential enrichment with each round of anti-LukE3 MAb biopanning (see red box). (B) Monitoring the output of phage from sequential rounds of biopanning showed that from pan 1 through pan 4 of the HlgA phage display fragment library, there were high and modestly increasing output numbers with the anti-LukE3 antibody but not with the isotype control. (C) Testing of binding of the anti-LukE3 pan 4 recovered phage library demonstrates high reactivity in a dose-dependent manner for the anti-LukE3 antibody, as shown by enzyme-linked immunosorbent assay (ELISA). (D) This selected phage library does not react with wells coated with the irrelevant bovine serum albumin (BSA) protein.
The original HlgA library was comprised of gene fragment clones, which varied in size and were distributed throughout the parental gene in all three reading frames and in both orientations within the vector (Fig. 2A). Taken together, there was also a very high level of redundancy for fragment coverage over the entire gene. Most importantly, analyses of the HIgA sublibraries obtained by sequential biopanning against the anti-LukE3 MAb showed a massive decrease of clonal diversity within the resultant pan 1 sublibrary, with relative expansion in the representation of residual clones that encoded overlapping gene sequences within this subregion. These clones were heterogeneous, with an overall preferential representation of clones that included the HIgA 123 to HlgA 133 codons (Fig. 2A). Notably, with each subsequent selection round, this clonal preference was retained in the output sublibraries, and this pattern of clonal focusing was most pronounced following the fourth round of selection (i.e., pan 4).
After the third and fourth rounds of selection with the anti-LukE3 MAb, the ratio of output/input phage titers was well above that achieved with equivalent biopanning attempts with the IgG2a isotype control (Fig. 2B). Indeed, with this system we have generally found that biopanning experiments that succeeded in selecting with a MAb for binding phagemid clones were generally associated with increases in the output phage titers. Furthermore, immunoassay studies of phage form of the pan 4 sublibrary showed significantly stronger concentration-dependent binding interactions with the selecting anti-LukE3 MAb than with the isotype control (Fig. 2C). The phage form of this anti-LukE3 MAb-selected sublibrary was also devoid of detectable binding interactions with irrelevant protein-coated microtiter wells (Fig. 2D).
Individual colonies from the pan 4 sublibrary output plates were recovered, and DNA sequence analysis of 20 colonies revealed several cases of exact clone duplication. There were only 15 unique clones, which contained 45- to 126-nucleotide inserts that represented 15 to 42 in-frame codons per clone (Fig. 3A). Importantly, the sequences of these 15 unique clones all encoded overlapping portions of the same HIgA subregion that varied only in the retained residues from the parental sequence at their amino- and/or carboxy-terminal ends. These alignments showed that these 15 clones all encoded a conserved 11-amino-acid sequence, YLPKNKIDSAD, from the parental HIgA protein. In the solved HIgA structure, the peptide backbone of this candidate epitope sequence was localized to a solvent exposed surface of a loop subregion that connected β strands (Fig. 3B and C). As the anti-LukE3 MAb was also reactive with other leukocidins, we compared all of these primary sequences, which revealed a high level of sequence identity of the HIgA 11-amino-acid sequence with a homologous subregion in each of the related LukE, LukS, and HlgC toxins, with only minor variations that represented conservative amino acid substitutions (Fig. 3A).
Independently anti-LukE3 MAb-selected phage clones from a HlgA phage display fragment library overlap the sequences of the LukS clone selected by the same anti-LukE3 antibody. Homology comparisons are shown to the native HIgA deduced sequence. (A) The dominant HIgA phage clone fragment sequences align to a region conserved by the Panton-Valentine leukocidin (PVL) LukS homologues. YLPPNKIDSAD, underlined in red, which was conserved in MAb selected gene fragment clones, folded into a conserved loop representing the HlgA123-133 subregion. (B) Structural depictions of minimal conserved region in of HlgA (PDB ID 2qk7). The conserved, minimized structure of the alignment is highlighted in red, depicting the agreement of the clones and homologues. (C) Surface of the conserved aligned structure on the HlgA crystal structure.
Biopanning with the anti-LukE3 MAb selected for LukS fragment homologues.To directly test whether this system would also enable the identification of structurally conserved site(s) in different Luk family members, we repeated the anti-LukE3 MAb biopanning experiment with a library of gene fragments of the homologue toxin, LukS. Akin to the results with the HlgA library, four rounds of selection also yielded an increase of the out/in phage ratio that was well above the levels associated with biopanning using the isotype control (see Fig. S5a in the supplemental material). As expected, in a phage binding assay, the selected pan 4 sublibrary was highly reactive with the selecting anti-LukE3 MAb at levels well above that observed with the isotype control-selected phage (Fig. S5b).
Individual fragment clones from the LukS pan 4 were then recovered, and within the 20 examined colonies, we found many duplicated identical sequences that represented only two recurrent overlapping LukS gene fragment sequences, LukeS83 to LukS113 (LukS83–113) and LukS85–115 (Fig. 4A). The phage forms of these clones exhibited high binding reactivity with the selecting anti-LukE3 MAb (not shown). Notably, these clones expressed a LukS subregion that was homologous to the subregion recovered from the HIgA library (Fig. 3A). Furthermore, these sequences encoded a loop in LukS (Fig. 4B) that is highly conserved and superimposable with that in HIgA, and which represent nearly identical surface-exposed surfaces in the solved structures of these Luk subunits (Fig. 4C).
Comparisons of individual phage clones with binding activity from the LukS fragment library selected with anti-LukE3 MAb demonstrate sequence and structural relatedness. (A) Selected gene fragment clone primary sequences, LukS83 to LukS113 (LukS83–113) and LukS85–115, align to a region of the PVL LukS protein that is conserved in the homologues LukE, LukS, HlgA, HlgC, and LukE. Recurrent LukS fragment clones encoded a larger sequence (underlined in green), with other recurrent clones encoding a smaller 14-amino-acid homologous subregion (underlined in red). (B) LukS with LukS85–113 loop (PDB ID 4iya) identified (green). (C) The conserved stretch of amino acids recognized by the anti-LukE3 antibody represents a loop between β strands that is surface exposed in the solved crystallographic structure of the LukS (PDB ID 4iya) and homologues.
Epitope validation: synthetic peptides as epitope mimics.Based on the results of the above-described independent biopanning experiments, we identified candidate MAb-binding subregions that were conserved between the homologous Luk toxins. To assess the amino acid sequence requirements for binding by the anti-LukE3 MAb, we synthesized overlapping peptides derived from the HlgA fragment clones, namely, HlgA119–129, HlgA123–133, and HlgA124–129 (Fig. 5B). Results from immunoassays documented that the two larger peptides strongly reacted with anti-LukE3 MAb, with a somewhat stronger interaction with the HlgA123–133 peptide (Fig. 5B), that were each generated with biotins at their amino termini with an SGSG spacer. However, the smallest of these three peptides, HIgA124–129 (i.e., LPKNKI), was devoid of detectable anti-LukE3 MAb reactivity, indicating that these 6 core amino acids were insufficient in themselves to recapitulate the binding site (Fig. 5B). Taken together, these studies therefore, in part, documented that the LukE3 MAb-associated epitope retained functionality when reconstituted as small synthetic biotinylated peptides devoid of structural context of adjacent contributions from other components within the full-length native Luk protein or from the gpVIII fusion protein context in the pComb-Opti8 expression cloning vector.
Minimized LukE peptides derived from cross-reactive antigens require a conserved Lys residue for recognition by anti-LukE3. (A) N-terminally biotinylated LukE peptides derived from HlgA were designed to overlap and minimize the amino acid sequence of the PVL LukS homologues to which the anti-LukE3 antibody reacts. (B) The anti-LukE3 antibody specifically reacts with the HlgA119–129 and HlgA123–133 peptides but not with the Hlg124–129 peptide, indicating that the YLPKNKI amino acid sequence has residues essential for the minimal epitope. To assess the strength of binding interaction in replicate wells for each biotin-peptide, we included peptide at 10,000 ng/ml and further serial 5-fold dilutions. Biotinylated native full-length LukS was included as a positive control. (C) In parallel, studies with a similar design showed that the anti-LukS1 antibody only recognizes a distinct peptide, the synthetic peptide termed LukScyc188–213, and the native LukS holoprotein, but not the HIgA-derived peptides. (D) Anti-LukE3 reacts with LukE108–118 Y108A and HlgA123–133 Y123A at slightly lower optical density at 450 nm (OD450) values, whereas LukE108–118 K111A and HlgA123–133 K126A abolish all anti-LukE3 reactivity. Unmutated LukE108–118 and HlgA123–133 were used as controls here. Dashed lines represent 50% and 25% of the OD450 signal of the unmutated minimized peptides. (E) With control isotype mouse IgG2a, no binding signals were observed, indicating that the anti-LukE3 MAb binding interactions are specific and require a conserved Lys residue.
To further investigate the fine binding specificity of the anti-LukE3 antibody for structurally conserved cross-reactive epitope(s) common to other Luk subunits, synthetic peptides were also made with the homologous sequence in HIgC and in LukE that was used to generate the anti-LukE3 MAb. Briefly, sequences that align with the implicated subregions in the crystallographic structural solutions in the LukE, LukS, and HlgA structures were also examined for the minimal surface-exposed loop/peptide fragment (although there is currently no published HIgC monomer crystal structure). Immunoassays by enzyme-linked immunosorbent assay (ELISA) showed significantly raised dose-dependent levels of binding reactivity of the anti-LukE3 MAb with the minimized HlgA123–133, LukE108–118, LukS93–103, and HlgC96–106 peptides, which were akin to reactivity levels with the native HlgA, LukE, LukS, and HlgC holoproteins (Fig. 5; see also Fig. S6 in the supplemental material). Our analyses suggested that the parent MAb efficiently cross-reacted with the biotin-SGSG-YLPKNKIDSAD peptide.
Primary sequence alignment of the minimal epitope candidate LukE108–118 from LukE identified a high degree of conservation within the LukS, HlgA, and HlgC proteins, and modeling studies suggested that the residues Tyr108 and Lys111 were solvent exposed in the LukS and HlgA toxins and were potentially accessible for protein-protein interactions. To test the potential roles of these residues in binding interactions with the anti-LukE3 antibody, we generated the synthetic variant peptides LukE108–118 Y108A, LukE108–118 K111A, HlgA123–133 Y123A, and HlgA123–133 K126A, which were then tested side by side with epitope candidate-derived peptides LukE108–118 and HlgA123–133 (Fig. 5D). Compared to the reactivity of anti-LukE3 MAb with the epitope candidates with native sequences, reactivity to the LukE108–118 Y108A and HlgA123–133 Y123A peptides displayed significant measurable, but diminished, reactivity. In contrast, anti-LukE3 MAb had no detectable reactivity with either LukE108–118 K111A or HlgA123–133 K126A variant peptides. To assess the predicted structures of these subregions, computational ab initio folding studies were performed, which predicted that the three-dimensional (3D) loop-turn structure would not be significantly changed by the introduction of any of these sequence variations (Table 1). Hence, these studies agreed with the notion that the binding interaction with the MAb involved direct interactions with the side chains of these two specific amino acids within the candidate B-cell epitope sequence.
Predicted ab initio folding of anti-LukE3 MAb epitopes as synthetic peptidesa
To further document the specificity of anti-LukE3 MAb binding interactions, we compared the ability of the full-length LukE protein to inhibit the recognition of the immunizing LukE protein by the anti-LukE3 MAb with those of the LukE108–118 and LukE108–118 K111A peptides (see Fig. S7 in the supplemental material). As predicted, recognition of the full-length LukE protein by the anti-LukE3 MAb was inhibited by the full-length Luk protein (Fig. S7A). Remarkably, there was 50% inhibition of the recognition of the holoprotein with the anti-LukE3 antibody following preincubation with only about a 3-fold greater molar amount of the peptide homologue LukE108–118 (Fig. S7B). These findings further support the notions that the LukE108–118 is the true epitope for anti-LukE3 MAb and that the lysine residue at position 111 in LukE, and the analogous lysine residue at position 126 in HlgA, are required for binding to this epitope (Fig. SCc). In addition, we interpret the results as evidence that the LukE Y108 residue and the analogous HlgA Y123 residue may help to stabilize binding to the LukE3 epitope, although each of these may not be absolutely required for recognition.
Epitope validation: antigenicity demonstrated in human postinfection sera.Whereas the anti-LukE3 MAb was generated by experimental immunization of a naive mouse with a single purified recombinant leukocidin protein, human adult and adolescent anti-S. aureus antibody responses are believed to commonly arise following a multitude of episodes of past infection and colonization that occur with different strains of this opportunistic pathogen (discussed in reference 40). To assess whether anti-S. aureus immunity in humans can also involve recognition of the candidate minimized LukE3 epitope, we adapted a previously validated multiplex bead-based assay approach (40) for surveys of serum polyclonal IgG antibody responses. We found significant levels of IgG antibody recognition of the LukE3 epitope, represented by synthetic peptides (HlgA123–133, LukE108–118, LukS93–103, and HlgC96–106), in a substantial proportion of patients recovering from invasive skin and soft tissue infections, which were detected in parallel with responses to the full-length LukE toxin subunit holoprotein (Fig. 6). Furthermore, immune recognition of one of these LukE3-related epitope peptides by antibodies in a clinical serum sample from a patient strongly and significantly correlated with the recognition of the other three derived peptides from cross-reactive Luk (Table S1).
Anti-LukE3 epitope is recognized by serum IgG antibodies following clinical S. aureus infection. (A) Circulating IgG antibodies in patients recovering from S. aureus infection recognize the minimal LukE3 epitope peptide and cross-reactive epitopes generated as synthetic peptides in a multiplex bead-based assay. IgG reactivity with homologous sequence peptide from four members of the leukocidin family was also tested with these different sera. Reactivity with HlgA123–133, LukE108–118, LukS93–103, and HlgC96–106 was significantly higher than that with the control. Recognition by a patient’s IgG antibodies of an individual peptide was highly correlated with reactivity to the other LukE3 epitope-related peptides (Pearson correlation, r ≥ 0.98 and P < 0.0001; see Table S1 in the supplemental material). (B) All patients had high levels of IgG antibody reactivity to the full-length holoproteins HlgA, LukE, LukS, and HlgC. Assays were performed with sera at 1:100 dilution. A cutoff was assigned at a mean for the control analyte (i.e., human serum albumin [HSA] beads) plus 3 SD, which is shown as the dashed line. Tetanus toxoid reactivity was included as a positive control. Results were from 37 patients at 6 weeks after presentation with S. aureus skin and soft tissue infection (SSTI) who had each received a full course of antibiotics prior to evaluation at this time point. ***, P < 0.001; ****, P < 0.0001.
The genomes of S. aureus clinical isolates commonly include LukE3 epitope-linked sequences.Extending our recent studies of the S. aureus immunoproteome (41), sequence analyses were performed on the “closed” whole genomes of 250 S. aureus clinical isolates deposited in GenBank (see Fig. S8a in the supplemental material). Here, a custom BLAST-style alignment strategy queried for sequences that encoded the above-described 11-amino-acid sequences linked to the LukE3 epitope. As expected, these strains contained the core (i.e., highly conserved between different isolates) genome-associated sequences that potentially encoded the LukE epitope. Specifically, among these many S. aureus genomes, the hIgA gene was near-universally represented, with a somewhat lower but still very substantial level of representation of hIgC, then lukE, and then lukS (Fig. S8b and c). Hence, both for the infecting strains from patient sera studied in Fig. 6 and for a much larger set of genomes of S. aureus clinical isolates, the LukE3 epitope could be expressed in vivo during infection, although it remains uncertain whether the immune responses that we characterized were induced against the LukE3 epitope expressed during the index infections, prior immune exposures, or both.
Epitope validation: immunogenicity.Whereas the above-described studies investigated whether a murine monoclonal antibody, as well as human postinfection immune sera, recognized a primary-sequence-dependent minimal B-cell epitope when reconstituted as a small synthetic peptide, we sought to pose a much higher level of validation of the immunologic properties of this epitope.
We therefore synthesized the HlgA123–133 peptide as a conjugate with the keyhole limpet hemocyanin (KLH) carrier protein, which is known to be highly immunogenic and provides many T-cell epitopes. Indeed, KLH has been used extensively in prospective candidate vaccines evaluated in murine experiments and in human clinical trials (42). Using this conjugate, mice were primed, then boosted, and the resultant postimmunization sera were tested for binding against a panel of S. aureus toxins and peptides that included the immunizing peptide. As predicted, postimmunization mouse sera were highly reactive with the HlgA123–133 and LukE108–118 minimal LukE3 epitope, even at a 1:100,000 dilution (Fig. 7A; see also Fig. S9b in the supplemental material). Even more remarkably, these postimmunized mouse sera were also highly reactive with full-length HlgA, LukE, LukS, and HlgC S. aureus toxins (Fig. 7B to E and Fig. S9b), which in most cases was also detectable at serum dilutions of 1:100,000. Furthermore, there was no reactivity with the Luk HIgB subunit that shares only limited homology, the evolutionarily distant alpha-toxin (HLA), or the unrelated beta-toxin (Fig. 7F to H and Fig. S9b), which further documented the specificity of these induced responses. We found that immunization induced functional serum activity levels comparable to the activity of binding of the parental monoclonal antibody to peptide homologues (0.55 to 0.94 μg/ml MAb equivalents) (Table 2). Notably, these postimmunization serum IgG antibody responses reiterated the reactivity pattern with the anti-LukE3 MAb (Fig. S9c). Indeed, these peptide-induced in vivo responses reiterated the Lys (K) side chain requirements of the parental MAb, as well as the cross-reactivity with full-length Luk protein homologues (i.e., HIgA, HIgC, and LukS; 0.56 to 1.080 μg/ml MAb equivalent) (Table 2). Taken together, these in vivo immune challenge studies document the immunogenicity of the 11-mer primary sequence-defined peptide epitope and demonstrate that immune exposure to only one of the sequence variations of the LukE3 epitope will induce antibodies that also recognize the other cross-reactive full-length antigens.
Immunization with the candidate minimal epitope keyhole limpet hemocyanin (KLH)-HlgA123–133 conjugate induces in vivo IgG anti-LukE, anti-LukS, and anti-HlgA antibodies. A multiplex bead assay was used to assess the postboost induction of antibodies to PVL LukS homologues. Here, the KLH-HlgA123–133 conjugate was shown to induce high mean fluorescence intensity (MFI) values to the (A) HlgA123–133 epitope immunogen, (B) LukE full-length protein, (C) LukS full-length protein, (D) HlgA full-length protein, and (E) HlgC full-length antigen, while (F) HlgB, (G) alpha-toxin, and (H) beta-toxin, an unrelated S. aureus toxin, were negative controls. (I) KLH was used as a positive-control analyte. LukD, LukF, LukABCC8, and LukABCC30 were also shown to have no detectable MFI signal (data not shown).
Estimated levels of reactive IgG antibodies in pooled sera generated by murine immunizationa
DISCUSSION
To identify functional minimal B-cell epitopes on staphylococcal leukocidins, we developed a method for unbiased antibody-mediated selection from large libraries of toxin gene fragments displayed on filamentous phage (i.e., phage display). In our experiments, the clones selected from independent libraries of gene fragments made from two structurally related toxins demonstrated striking molecular convergence; the identical B-cell epitope was identified from the two different toxin libraries. We also proceeded to characterize individual selected clones for fusion protein expression and antibody binding analysis and to design synthetic peptides that recapitulated the tertiary B-cell epitopes as isolated entities, enabling precise mapping of the epitope by point mutants. Validation studies performed with these derived peptides confirmed both the antigenicity of this minimal epitope (including for clinical serologic surveys) and the in vivo functional immunogenicity of this epitope. Furthermore, to reveal the inner evolution occurring during sequential rounds of sublibrary selection, massive in-parallel sequence analysis of the selected sublibraries identified the kinetics of emergence of the sequence subregion shared by the selected gene fragment clones, which, in light of the resolving power of these in silico analyses, may in the future obviate individual fragment clone isolation. Although there are superficial features in common with approaches in which libraries of small synthetic overlap peptides of fixed size are probed (43), the gene fragment phage display libraries used in the current approach contain both small and larger gene fragments (Fig. S3), and these original libraries can be (re)used with other antibodies or other types of bait ligands to recover additional noncompeting unique leukocidin toxin epitopes for other antibodies. In addition, these gene fragments are not directly adhered to a solid phase and hence should have greater accessibility.
Our study illustrates that a minimized 11-mer peptide epitope (with an estimated 1.3-kDa mass), which represents about 4% of the 32-kDa full-length HlgA molecule, can be functionally equivalent to the same subregion when expressed in the full-length toxin holoprotein. Modeling studies support the notion that our LukE3-derived epitope peptides reiterate the fold and orientation represented by the B-cell epitope in the native parental protein. Furthermore, binding interactions in solution with these epitope peptides were shown to block the recognition by a monoclonal antibody of the native protein, while these small epitope peptides also retained immunogenicity properties that enabled induction of in vivo responses that recognized the toxin holoprotein. Such epitopes can therefore be described as examples of validated primary-sequence-defined conformational epitopes that retain key features shared with the much larger toxin protein.
With regard to the structure-function relationships of subregion(s) in the Luk molecule responsible for toxin activity, the LukE3 epitope localizes to what is predicted to be part of the stem domain in the Luk monomers. After assembly of the full oligomeric complex on the targeted host cell membrane, the stem domains then undergo conformational change to an extended conformation that is stabilized by intramolecular interactions with other stem domains in adjacent monomers, which together facilitates assembly and stabilization of the toxin pore responsible for cell killing (reviewed in references 17 and 44).
Among the panoply of exotoxins released during S. aureus infection, we chose the leukocidin family for our proof-of-principle study, in part because several of the members have been reported to be direct contributors to clinical syndromes of invasive and potentially life-threatening infections (40, 41, 45). In addition, our experimental design exploited the extensive literature on the shared structural features of Luk toxin subunits (46–49), which are integral to their properties as efficient molecular killing machines through formation of pores that cause host cell toxicity and death. Predictions based on our in silico modeling studies and confirmed by direct binding and competitive inhibition studies were consistent with the high level of conservation of the native conformation of the candidate epitope in different Luk members, which we showed was functionally emulated by an 11-mer synthetic peptide.
From a broader perspective, this report documents our success with a vector system and methodology tailored for the selection of minimal B-cell epitopes. Whereas others have shown that gpVIII-based systems can enable cloning of inserts of up to 2,000 bp (38, 50, 51), we used the pComb-Opti8 system in a cloning strategy incorporating enzymatic DNA digestion to generate small random fragments of S. aureus genes, with a mean length of about 50 bp but with great heterogeneity in size, ranging from 0 to 160 bp (Fig. S3). This cloning strategy used blunt-ended fragments, and due to the variations in potential reading frame and orientation, most of the library inserts would be predicted to be defective in their capacity to express in-frame fusion proteins on phage surfaces. In practice, this potential limitation was overcome by the high efficiency of transfection by electroporation, as each fragment library member was estimated to have greater than 1,000-fold redundancy in the original library (Fig. 2). Furthermore, we believe our success in the current studies was also in part linked to the compatibility of this bacterial-filamentous phage system with the expression, folding, and selection of S. aureus proteins. With the pComb-Opti8 system, we generated fragment fusion proteins with the truncated gpVIII domain that we postulate provided a compatible scaffold for the LukE3 epitope, which was localized to a loop between adjacent β-strands within the overall native Luk protein. We wonder whether the structural context of the resulting fusion proteins displayed on the phage surface is critical for fishing out of B-cell epitopes from libraries of fragments from the genes of Luk proteins with overall β barrel structures.
Based on the capacity to generate libraries with complete coverage of subregions of the parental gene (Fig. 2A), our fragment libraries should include sequences that encode subregions essential for toxin functionality, against which the formation of antibody complexes would be predicted to result in neutralization of this toxin activity. Furthermore, while an antibody may not be able to access the extended stem conformation that is only revealed in the fully assembled pore, the binding of an antibody to sites accessible in soluble monomers could stabilize the inactive conformation of these monomers and thereby interfere with sites essential for dimerization. Alternatively, an antibody could block subsequent intermolecular interactions essential to cell killing, or other mechanisms for neutralization could be relevant. In the next phase of our studies, we will focus on the discovery and validation of such epitopes.
In conclusion, we document here an approach that enabled the identification by a monoclonal antibody of a minimal epitope recapitulated in a small linear synthetic peptide that also displayed the immunogenicity of the native toxin. It is important to consider that immunodominant leukocidin epitopes recognized after murine immunization may be different in humans, who from an early age are commonly colonized and later suffer recurrent, although often minor, infections that induce circulating antibodies and memory B cells against many S. aureus antigens (40, 41). Indeed, although lasting postinfection enhancement of protection is uncommon or at best unpredictable (40), members of the Luk family are high on the hierarchy of protein antigens recognized by the immune responses of adults and children recovering from S. aureus infection (41, 45, 52). Thus, in future work, we will extend this optimized approach to identify epitopes targeted by human MAbs, which may shed light on whether there are fundamental differences in host responses between these two species against this opportunistic pathogen. Overall, our studies also illustrate the practical utility of the peptide epitope forms we were able to generate that facilitated our epitope immunoassay surveys. These synthetic peptide tools will also aid investigations of the dynamic nature of serologic and cellular immune responses and potentially help to elucidate how this opportunistic pathogen subverts host defenses. Taken together, these tools have the potential to accelerate investigations of the host-pathogen relationship, as well as to provide the epitope building blocks needed for the development of a practical protective vaccine.
MATERIALS AND METHODS
pComb-Opti8 vector construction.We modified the pComb3X phagemid system that was designed for oligomeric/monomeric display (53) by use of custom oligonucleotides that introduce a flexible linker followed by a truncated gpVIII gene optimized for expression of fusion proteins with this major coat protein (37). The resulting vector was termed pComb-Opti8 (Fig. S1).
To confirm the functionality of the system, we cloned a previously described high-affinity anti-tetanus toxoid (TT) human Fab (53) into this system, which was then electroporated into TG1 electrocompetent cells (Lucigen). Transfected bacteria were then grown overnight, with the addition of VCSM13 helper phage (Agilent) for phage rescue, and subsequent determination of the phage prep titer. An aliquot of these anti-TT Fab-expressing phage was then mixed in a ratio of 1:100,000 with pComb phagemid construct that instead contained nonfunctional stuffer fragments, with sequential rounds of biopanning on microtiter wells coated with TT antigen, resulting in the desired efficient selection and recovery of the anti-TT phagemid clone (data not shown).
Construction of gene fragment phagemid libraries.Using specific oligonucleotides designed to amplify genes of interest, the full-length genes of LukS (LACUSA300) and hemolysin gamma A (HIgA) (Newman) were each amplified without signal leader sequences or stop codons. The amplimer products of 4 PCRs for each gene of interest were separately pooled and precipitated, then evaluated by agarose electrophoresis to confirm predicted product size and homogeneity. Following the manufacturer’s protocol, the purified amplimer products were then individually fragmented with dsDNA Fragmentase (NEB) under conditions empirically shown to generate random gene fragments of about 50-bp average size. Products were visualized with the 4200 TapeStation system (Agilent), followed by polishing of the fragment ends with a Quick Blunting kit (NEB); products were then precipitated and quantified.
The pComb-Opti8 vector was prepared by digestion of the EcoRV site within the multiple cloning site (MCS). Following gel extraction and treatment with shrimp alkaline phosphatase (SAP; NEB), the vector was again gel extracted. Using the median size of the insert to calculate molarity, we used an insert to vector ligation ratio of 3:1 for ligations performed using T4 ligase (NEB), which was followed by electroporation into TG1 electrocompetent cells. Transfected bacteria were then grown overnight, with the addition of VCSM13 helper phage to enable rescue of the phage form of the library. Thereby, LukS and HigA gene fragment libraries were generated; each was estimated to have more than 107 independent members.
Generation of murine MAbs.Anti-LukE monoclonal antibodies were generated by splenocyte-myeloma cell fusion using a standard murine hybridoma protocol (Envigo). Briefly, primary subcutaneous immunization of BALB/c mice was performed with 50 to 100 μg of recombinant LukE (rLukE) emulsified in Freund’s complete adjuvant, followed by 3 boosts at monthly or bimonthly intervals using rLukE emulsified in Freund’s incomplete adjuvant. The immunized mouse was sacrificed, and splenocytes were isolated and fused with nonsecreting NS01 myeloma cells. Subcloned cell lines were then selected by screening supernatants for binding of the immunogen in an ELISA, and antibody gene sequences were determined as previously described (74).
Biopanning on microtiter wells.Adapting previously described methods from pComb antibody libraries (39), microtiter wells were coated overnight with anti-S. aureus mouse antibodies or an isotype control at 2 μg/ml. Plates were washed with 0.05% Tween/phosphate-buffered saline (PBS), then blocked with either 3% casein/PBS or 3% bovine serum albumin (BSA)/PBS for 2 h at 37°C. To each of four individual wells, we then added 50 μl of a phage library suspended in block solution with mouse isotype IgG at 1 μg/ml; the wells were coincubated at 37°C for 2 h, with each sequential round of biopanning (followed by elution of bound phage, bacterial infection, and phage rescue) enabling the next round of selection.
Deep sequencing of pComb-Opti8 gene fragment libraries selected by MAb biopanning.For sequence determinations of gene fragment inserts, DNA recovered from the original and sequentially selected sublibraries was separately amplified using universal primers (MiSeq; Illumina) designed to anneal at sites flanking the EcoRV cloning site in the vector backbone. In order to separate reads for each of the individual phage libraries, forward (AATGATACGGCGACCACCGAGATCTACACTATGGTAATTGTTGTGTGGAATTGTGAGCGGATAAC) and reverse (CAAGCAGAAGACGGCATACGAGATXXXXXXXXXXXXAGTCAGTCAGCCCAGCTCTTCTGTTATGTCTGTGAGG) primers, each with an unique barcode identifier (“XX”), were added along with the reverse primer to enable PCRs that also included the 5Prime HotMasterMix (Quantabio) and 20 ng of the miniprep of plasmid DNA from each library, following the manufacturer’s specifications. Using a T100 thermal cycler (Bio-Rad), the templates were denatured in the first cycle for 3 min at 94°C; this was followed by 35 cycles of a 45-s denaturation step at 94°C, a 1-min annealing step at 51°C, and extension for 1.5 min at 72°C. A final extension was performed for 10 min at 72°C before the samples were cooled to 4°C. PCR products were visualized on a 2% agarose gel run at 80 V for 1 h. Sequences were determined using a MiSeq instrument (Illumina) and visualized with SeqMan NGen software (DNAStar).
Phage binding ELISA.To assess the representation of phage particles in a prep, microtiter wells were coated with anti-M13 antibody (GE Healthcare) in PBS overnight at 4°C, then washed once with 0.05%Tween/PBS and blocked with 1% BSA/PBS for 1 h at 37°C. The blocking solution was removed and then wells were washed. In duplicate, the phage preps were then added at 1:5 serial dilutions into wells, then incubated at 37°C for 1 h. To remove unbound phage, wells were then washed four times with 0.05%Tween/PBS. Bound phage in each well were then detected by incubation with anti-M13-horseradish peroxidase (HRP) (GE Healthcare) for 1 h at 37°C, followed by 4-fold washing, then addition of the trimethylbenzidine (TMB) substrate (BioLegend), with the reaction stopped with 1 M phosphate buffer. Signal was assessed at 450-nm absorbance using a Synergy H1 Hybrid plate reader (Biotek). To assess the capacity of gene fragment phagemid particles to be bound by a MAb, microtiter wells were coated with anti-S. aureus MAb at 2 μg/ml overnight at 4°C and then plates were processed as described above.
Bead-based multiplex binding assay.Adapting a previously described protocol (40), we generated a custom protein array for the Magpix platform (Luminex) that included individually coupling purified recombinant S. aureus proteins, including bacterial antigens, biotinylated synthetic peptides, and control ligands, each coupled to individual microspheres (Table S2). For inhibition studies, the LukE108–118 peptide or full-length LukE in solution was separately incubated in with anti-LukE3 MAb for 1 h at room temperature, before addition and overnight incubation with analyte-coated beads.
For antigen-reactive IgG detection, 1,000 microspheres per analyte per well were premixed, sonicated, and then added to 100 μl of diluted serum or anti-S. aureus MAb. Bound IgG antibodies were detected with Fc-gamma-specific anti-human IgG R-phycoerythrin (R-PE) (eBioscience) or with anti-mouse IgG (Fc-specific) F(ab′)2 PE (Jackson ImmunoResearch). Data were acquired on a Magpix instrument (Luminex) and reported as mean fluorescence intensity (MFI) values (40).
Bioinformatics analysis of biopanning hits.The sequence of each clone isolated from a phage display library was aligned with the entire parent leukotoxin gene using ZEGA global sequence alignment software (54). From the MAb-selected sublibraries, most potential “hits” contained a continuous fragment identical in sequence to a subregion in the hololeukotoxin, enabling identification and visualization by PyMOL (Schrodinger) of the 3D structural location of the primary amino-acid subsequence hit within the holotoxin crystallographic structures available in the Protein Data Bank (PDB), as follows: LukE (PDB identifier [ID] 3roh), LukS (PDB ID 4iya), and HlgA (PDB ID 2qk7). The features of these candidate epitopes were then examined for surface exposure or for burial and inaccessibility, which allowed us to narrow down the exact continuous surface-exposed loop peptide fragment likely to be related to the molecular target of the selecting MAb (Fig. 4).
Design and synthesis of peptide probes.The solubility of Luk sequences that were predicted from the above analysis to be continuous surface exposed loop subregions was estimated using a solubility estimation tool (http://www.innovagen.com/proteomics-tools). If the solubility was acceptable, each sequence was subjected to biased probability Monte Carlo computational ab initio folding (55), which was used to predict the 3D conformation of the identified subsequences as free peptides in solution (56–67), with each searched conformation assessed for thermodynamic and physics-based energy components (68–71). Thereby, the lowest energy conformation that was biologically relevant was identified, along with the dynamic ensemble that enables assessment of the flexibility of the peptide structure. To assess the likelihood for each of the minimized gene fragment hits that an Ab/MAb will bind to a synthetic peptide form of a candidate B-cell epitope similarly as to the holodomain, this conformation was then compared to the in situ conformation of its parent subregion in the holodomain. These procedures were performed using algorithms in ICM-Pro software (MolSoft LLC, La Jolla, CA). Each of the primary amino acid sequences with acceptable estimated solubility that was predicted to retain a 3D conformation akin to the constrained structure within the full-length holotoxin was synthesized to enable further testing. To facilitate efficient arraying on neutravidin/streptavidin-coated beads or ELISA microtiter wells, these synthetic peptide probes were commercially generated with N-terminal biotin and an SGSG linker that was followed by the sequence of interest (Innovagen).
S. aureus isolate genomic analysis.The phylogenetic tree was constructed using AAF (72). Epitope sequences were extracted from the 210 gapless genomes, aligned using Muscle (73), and screened for polymorphism using a custom R script.
Murine immunization to assess immunogenicity.The HlgA123–133 peptide-keyhole limpet hemocyanin (KLH) conjugate was dissolved at 2 mg/ml into sterile PBS, which was then mixed in a 1:1 volume ratio with TiterMax Gold adjuvant (Sigma), then emulsified using mixing needles. Female ND4 Swiss Webster mice (9 to 10 weeks old; Envigo) were prebled, then given primary immunizations with 100 μg of peptide equivalent divided into two subcutaneous sites, followed by boosts at 14-day intervals. Before each boost, mice were bled and sera were prepared, then aliquoted and stored at −80°C until tested.
ACKNOWLEDGMENTS
This work was supported by National Institutes of Health–National Institute of Allergy and Infectious Diseases award HHSN272201400019C (G.J.S.). The Silverman lab is also supported by NIH-NIAMS P50 AR070591-01A1 and T32GM66704 (E.E.R.). The Torres lab was also supported by NIH-NIAID awards T32AI007180 (K.T.), R01AI105129 and R01AI099394 (V.J.T.), and R01AI37336 (B.S.). Flow cytometry and genomics support were provided by NYU Langone’s Cytometry and Cell Sorting Laboratory and the NYU Langone Health Genome Technology Center, which are supported in part by grant P30CA016087 from the National Institutes of Health–National Cancer Institute.
We thank Zhi Li for assistance in library analysis.
V.J.T. is an inventor on patents and patent applications filed by NYU that are currently under commercial license to Janssen Biotech, Inc., which provides research funding and other payments associated with licensing agreement.
FOOTNOTES
- Received 7 October 2019.
- Returned for modification 14 November 2019.
- Accepted 26 January 2020.
- Accepted manuscript posted online 3 February 2020.
Supplemental material is available online only.
- Copyright © 2020 American Society for Microbiology.
