Design of a Streptococcus pyogenes M protein immunogen to elicit M type cross-reactivity

Coiled coil-forming M proteins of the widespread and potentially deadly bacterial pathogen Streptococcus pyogenes (Strep A) are immunodominant targets of opsonizing antibodies. However, antigenic sequence variability into >220 M types, as defined by the M protein hypervariable region (HVR), has been considered to limit its utility as a vaccine immunogen due to type-specificity in the antibody response. Surprisingly, a multi-HVR immunogen in clinical vaccine trials elicited M type cross-reactivity. The basis for this cross-reactivity is unknown but may be due in part to antibody recognition of a three-dimensional (3D) pattern conserved in many M protein HVRs that confers binding to human C4b-binding protein (C4BP). To test this hypothesis, we asked whether a single M protein immunogen carrying the 3D pattern would elicit cross-reactivity against other M types carrying the 3D pattern. We found that a 34-amino acid sequence of M2 protein bearing the 3D pattern retained full C4BP-binding capacity when fused to a coiled coil-stabilizing sequence from GCN4. This immunogen, called M2G, elicited cross-reactive antibodies against a number of M types that carry the 3D pattern but not against those that lack the 3D pattern. The M2G antiserum recognized M proteins as displayed natively on the Strep A surface, and promoted the opsonophagocytic killing of Strep A strains expressing these M proteins. As C4BP-binding is a conserved virulence trait of Strep A, targeting the 3D pattern may prove advantageous in vaccine design.


Streptococcus pyogenes (Group A Streptococcus or Strep A) is a globally widespread
gram-positive bacterial pathogen that causes a variety of diseases, ranging from mild and selflimiting (e.g., pharyngitis and impetigo) to invasive and deadly (e.g., necrotizing fasciitis and streptococcal toxic shock syndrome) (1). Strep A infection can also lead to autoimmune diseases (e.g., acute rheumatic fever and rheumatic heart disease), which remain serious causes of morbidity and mortality in the developing world (2)(3)(4). Approximately 500,000 deaths occur annually due to diseases caused by Strep A (5). At present there is no vaccine against Strep A (6), with one of the major impediments being the sequence variability of its immunodominant surface antigen, the bacterial cell wall-anchored M protein (7)(8)(9)(10)(11).
More than 220 M protein types have been identified (12). The primary sequence of M proteins in general have heptad repeats, which are diagnostic of α-helical coiled coils (13), and structural studies have directly confirmed that M proteins do indeed form parallel, dimeric αhelical coiled coils (14)(15)(16), albeit with functionally significant sequence and structural irregularities (17,18). The sequence of the N-terminal 50 amino acids of the mature form of M proteins (with their signal sequences removed) is hypervariable and defines the M type. These N-terminal hypervariable regions (HVRs) elicit protective, opsonizing antibodies (7)(8)(9)(10)(11). In contrast, other portions of M proteins are often not immunogenic or do not elicit opsonizing antibodies (7,19). In addition, M protein HVRs do not elicit autoreactive antibodies (20), which other portions of M proteins do, presenting a concern for initiating autoimmune diseases (21).
While M protein HVRs have highly favorable features as vaccine immunogens, antibody reactivity tends to be type-specific and therefore limited to a single M type strain (22)(23)(24)(25).
Surprisingly, a Strep A vaccine immunogen composed of multiple M protein HVRs elicited an M type cross-reactive response. This vaccine immunogen, StreptAnova TM , consists of 30 different M protein HVRs fused into four separate polyproteins (~45-50 kDa per polyprotein), and upon immunization of rabbits elicited reactivity against these 30 M types as well as cross-reactivity against ~50 M types not included in the immunogen (20,26,27). This reactivity promoted opsonophagocytic killing of Strep A (26,27). The human immune system also appears to be capable of generating an M type cross-reactive response (28,29). The basis for M type cross-reactivity of StreptAnova TM is not known. However, our own work in understanding how human C4b-binding protein (C4BP) binds multiple M protein HVRs provides a plausible model (16,30).
C4BP limits the generation of the major opsonin C3b and thereby functions as a downregulator of the complement system (classical and lectin pathways). Recruitment of C4BP by M protein to the Strep A surface is an essential virulence trait, preventing opsonization, phagocytic uptake, and consequent killing (16, [31][32][33][34][35][36][37][38]. A large-scale study found that 90 of 100 Strep A strains of differing M types bound C4BP (33). Because C4BP-binding has been attributed only to M protein HVRs (33), these results suggested that C4BP is cross-reactive for M protein HVRs. To understand the basis of M type cross-reactivity of C4BP, we determined X-ray crystal structures of four M protein HVRs (M2, M22, M28, and M49) each bound to C4BPα1-2, a fragment of C4BP that is necessary and sufficient to bind M protein HVRs (16).
These structures revealed these M protein HVRs display a similar spatial or three-dimensional (3D) pattern of amino acids that contact a common site in C4BP. The amino acids of this shared 3D pattern are surrounded in space and in primary sequence by a larger number of variable amino acids, and so in effect the 3D pattern is diluted within the variability of the HVR (30).
However, once the 3D pattern was identified, it was recognizable in the primary sequence of M proteins of about 40 of the ~90 Strep A strains (16) that were shown to bind C4BP (33).
Based on this observation, we hypothesized that the typical antibody response was M typespecific simply because variable amino acids outnumber those in the conserved 3D pattern.
However, if an antibody were to bind amino acids of the conserved 3D pattern in one M type, it should then also recognize other M types that have this 3D pattern. Therefore, such an antibody would be M type cross-reactive. Notably, 15 of the 30 M protein HVRs in StreptAnova TM have the C4BP-binding 3D pattern, and correspondingly 20 of the ~50 cross-reactive M protein types elicited by StreptAnova TM have the 3D pattern (26,27), suggesting that at least some of the cross-reactivity of StreptAnova TM is due to recognition of the 3D pattern. Likewise, M type crossreactivity observed for three other multi-HVR immunogens may be explained by recognition of the 3D pattern (39)(40)(41). The composition of these three immunogens, which are pentavalent or hexavalent and mostly contain HVRs that are also in StreptAnova TM , is based on physicochemical properties rather than M type prevalence in North America and Europe as it is for StreptAnova TM (26). Together these results suggest that the C4BP-binding 3D pattern is capable of eliciting M type cross-reactive antibodies.
To test this hypothesis directly, we pursued a detailed proof-of-principle study. We used a short (34-amino acid) sequence from a single C4BP-binding M protein for immunization. M2 protein was chosen since its binding to C4BP has been studied in detail through mutagenesis (16). A 34-amino acid (aa) portion of M2 centered on the 3D pattern maintained full C4BPbinding affinity when fused to the canonical coiled-coil forming protein GCN4 (42). The antiserum evoked by the resulting immunogen, called M2G, was reactive against M2 and crossreactive against a number of C4BP-binding M types but not against M types that do not bind C4BP. The M2G antiserum was not cross-reactive against C4b or self-antigens. Reactivity and cross-reactivity of the M2G antiserum extended to M proteins displayed natively on the Strep A surface and resulted in the opsonophagocytic killing of Strep A strains.

Minimized C4BP-binding regions of M2 protein
Our previous structural studies used M2 N100 (Fig. 1a), a protein fragment consisting of the N-terminal 100 aa of the mature form of the protein (i.e., with its signal sequence cleaved), for co-crystallization with C4BPα1-2 (16). The structure revealed that the C4BP-binding region of M2 protein localized to a span of only 23 amino acids (aa 61-83) within the HVR (Fig. 1a). To limit immunoreactivity to the C4BP-binding amino acids of the M2 HVR, we first asked whether a short fragment of M2 protein constituting just the C4BP-binding amino acids would maintain C4BP binding. However, expression of M2 aa 61-83 by recombinant means in E. coli was poor and yielded insufficient quantities of protein for further experiments. We tried longer M2 fragments, either aa 42-86 (M242) or 53-86 (M253) (Fig. 1a). Amino acid 42 is the very Nterminus of mature M2 protein, and 53 and 86 are the first and last amino acids, respectively, that are ordered in the crystal structure of M2 bound to C4BPα1-2 (16). Both M242 and M253 were expressed recombinantly in sufficient quantities for further studies. However, neither M242 nor M253 bound His-tagged C4BPα1-2 above background levels (Figs. 1b, and S1a, b). A fragment of M22 protein (M22248, aa 42-248) was used as a positive control for C4BP-binding in this and other experiments. It seemed possible that M242 and M253 were too short to form a dimeric, α-helical coiled coil efficiently, a necessity for M protein to bind C4BP (16, 43). To overcome this problem, we fused short sequences from the ideal coiled-coil forming protein GCN4 (42) to M2 protein fragments, maintaining a continuous heptad register between the two (Table S1). We first tried sandwiching M2 aa 61-83 between single GCN4 heptads (Fig. 1a,   GM261G), but observed no binding to C4BP (Fig. S1c). Next, we tried longer GCN4 coiled-coil sequences of about three or four heptads (23 or 27 aa) fused to the C-terminus of M2 aa 53-86 or 61-83; these fusion constructs were called M253G and M261G, respectively (Table S1). While M261G bound C4BPα1-2 slightly above background level, M253G bound C4BPα1-2 well, with an affinity apparently higher than that of M22248 (Figs. 1b, c and S2a, b). For simplicity, we refer to M253G as M2G hereafter.
We then asked whether M2G recapitulated the C4BP-binding affinity of intact M2 protein. Isothermal titration calorimetry (ITC) was carried out and showed that the KD of C4BPα1-2 bound to intact M2 protein was identical to that of C4BPα1-2 bound to M2G, 4 µM ( Fig. 2, Table 1). Thus, M2G, which contains only 34 aa of M2, possessed the full C4BP-binding affinity of intact M2 protein. Furthermore, these results suggested that GCN4 aided the coiledcoil dimerization of this M2 region to restore its C4BP binding ability.

M2G as an immunogen
Having identified that M2G recapitulated C4BP-binding, we asked whether it was sufficient to evoke an immune response that was cross-reactive against M protein types that carry the 3D pattern. Rabbit polyclonal antibodies were raised against M2G and assayed for reactivity against various recombinant M proteins. While M2, M22, M28 and M49 HVRs all present similar 3D patterns of amino acids that are complementary to C4BP, these spatial patterns are exhibited differently in the heptad repeats of their primary sequences (16). The heptad patterns of M2 and M49 HVRs are similar to one another and belong to one subset, the M2/M49 sequence pattern; and the M22 and M28 HVR patterns are similar to one another and belong to a second subset, the M22/M28 sequence pattern (16). We chose M protein types from each pattern that are prevalent in human infectious disease epidemiology (44,45). For the M2/M49 group, these were M2, M49, M73, M77, and M89 proteins, and for the M22/M28 group, these were M4, M11, M22, M28, M44, and M81 proteins. As negative controls, we used M1, M5, and M6 proteins, which do not bind C4BP and lack the 3D pattern.
We expressed and purified constructs constituting the N-terminal 100 amino acids of the mature forms of these M proteins. Binding to C4BP had not been directly evaluated for some of the M proteins. For these an enzyme-linked immunosorbent assay (ELISA) for C4BP-binding was carried out (Fig. S3). All of the M N100 constructs (i.e., N-terminal 100 aa of the mature form) belonging to the M2/M49 or M22/M28 pattern bound C4BP, except for M77 N100 , which bound C4BP at the level of the negative control M5 N100 construct.
We next tested the reactivity and cross-reactivity of the M2G antiserum. As expected, the M2G antiserum recognized M2 N100 well, with an antibody titer that was significantly greater than that of pre-immune serum (>10 5 vs. <10 2 ) (Figs. 3 and S4). The M2G antiserum was cross-reactive with titers > 10 3 against all the M N100 constructs belonging to the M2/M49 pattern (i.e., M49, M73, and M89), except for M77 N100 , which as noted above did not bind C4BP. For the M22/M28 pattern, the M2G antiserum was cross-reactive with a titer > 10 3 against only M28 N100 . Statistically significant cross-reactivity was also seen for M11 N100 , but the titer was low (< 10 3 ).
For the remaining members of M22/M28 group (M22 N100 , M4 N100 , M44 N100 , and M81 N100 ), the titer of the M2G antiserum was low (< 10 3 ) and not significantly different from that of the pre-immune serum ( Fig. 3 and S4). The titer of the M2G antiserum was uniformly low against M N100 constructs of M proteins that are known not to bind C4BP (i.e., < 10 3 for M1, and < While autoreactivity is not attributed to M protein HVRs, this remains a general concern for vaccines based on M proteins (46). To evaluate the reactivity of the M2G antiserum against human tissues affected in Strep A autoimmune sequelae, western blot analysis was performed with normal adult human brain tissue lysate (HB) and heart tissue lysate (HH). Because autoreactivity can be due to portions of M proteins outside of the HVR, we also raised rabbit antibodies against intact M2 protein and compared the cross-reactivity of the M2G antiserum against that of the intact M2 protein antiserum. The M2G antiserum reacted against intact M2 protein but not HB or HH (Fig. S5a). In contrast, the antiserum raised against intact M2 protein reacted against intact M2 and both HB and HH (Fig. S5b). These results suggest that the M2G immunogen does not elicit cross-reactivity against human tissues, whereas intact M2 protein has the potential to do so. These results are also consistent with M protein HVRs not eliciting autoreactive antibodies (20).
We asked whether the pattern of M protein cross-reactivity described above was limited to a single rabbit or reproducible in a second rabbit. To this end, a second rabbit was immunized with M2G and the cross-reactivity of this second rabbit's M2G antiserum was examined (Figs. S6 and S7). The reactivity and cross-reactivity patterns were almost identical with a Pearson correlation coefficient of 0.998. Some small differences were evident. For example, for the M2/M49 pattern, the relative cross-reactivity to M89 N100 was lower in the second rabbit than in the first, and in the M22/M28 pattern, low-titer cross-reactivity to M44 N100 but not M11 N100 was statistically significant. Experiments were continued using the antiserum from the first rabbit.
As the M2G antigen contains sequences from both M2 protein and GCN4, we asked whether cross-reactivity was due to antibodies specific for the M2 as opposed to the GCN4 portion. We focused on M protein constructs that yielded the highest titers (> 10 3 ), namely M2 N100 , M49 N100 , M73 N100 , M89 N100 , and M28 N100 (Fig. 3). Reactivity of the M2G antiserum to these M protein constructs was competed with increasing concentrations of either M2G or M6G (Table S1). The latter consists of a portion of M6 protein (aa 56-89) fused to nearly the identical portion of GNC4 in M2G (Table S1). For all M N100 constructs except for M49 N100 , competition occurred with M2G but not M6G, even when M6G was used at >2-fold higher concentration (Fig. S8a). The integrity of M6G as a competitor was demonstrated by adhering His6-tagged M6G to ELISA plates and using soluble His6-tagged M6G as a competitor (Fig. S8b). These results indicated that for M2 N100 , M73 N100 , M28 N100 , and M89 N100 , cross-reactive antibodies were specific to the M2 portion of the M2G immunogen.
For M49 N100 , both M2G and M6G competed against the M2G antiserum for binding (Fig.   S8a), suggesting that some or all of the cross-reactive antibodies against M49 N100 were specific to the GCN4 portion of the M2G. M49 protein was eliminated from further studies.

Strep A surface binding and opsonophagocytic activity of M2G antiserum
To assess whether the M2G antiserum recognized M proteins on the bacterial surface, we carried out flow cytometry on whole, living Strep A strains of differing M types. In line with results using purified proteins, the M2G antiserum bound the surface of an M2 strain to a significantly higher extent than did the pre-immune serum (Fig. 4, Table S2). Likewise, the M2G antiserum displayed significant cross-reactivity compared to the pre-immune serum against Strep A M73, M89, and M28 strains. No cross-reactivity was seen against an M5 strain, which does not bind C4BP. Overall, these results confirm that the M2G antiserum recognizes M proteins in their native conformation on the Strep A surface.
Antibodies against M protein HVRs elicit opsonophagocytic antibodies (7)(8)(9)(10)(11). To verify that this is the case for the M2G antiserum, we evaluated whether the M2G antiserum promoted opsonophagocytic killing (OPK) of reactive and cross-reactive Strep A strains. For the OPK assay, we used cultured HL-60 cells differentiated to have a neutrophil-like phenotype, along with baby rabbit serum as the source of complement (47). While neutrophil-like HL-60 cells are not nearly as potent killers as primary neutrophils (48), they offer several advantages over the classical Lancefield assay (24). First and foremost, individual variation in complement and neutrophil activity is eliminated in the HL-60 assay, as is the existence of immunity against various M types (49). The HL-60 assay has been proposed as a standard means for evaluating Strep A vaccine candidates (47,49). We focused on the M types against which the greatest reactivity and cross-reactivity had been demonstrated -M2, M73, M89, and M28 -and used an M5 strain as a negative control. As a positive control for the assay, we used IVIG, a concentrated pool of human antibodies. We found that IVIG was most potent against M28 and M73 strains (100% and 94.6 ± 0.2% killing) followed by M2 (72.1 ± 4.5%), M89 (52.7 ± 4.2%), and M5 (30.6 ± 3.2%) strains (Fig. 5).

Discussion
We set out to test the hypothesis that directing the antibody response to the C4BPbinding 3D pattern in one M protein type would lead to cross-reactivity against other M protein types that share the 3D pattern. The design was to have the 3D pattern constitute as great a proportion of the immunogen as possible, and correspondingly, variable amino acids as minimal as possible. The well-studied M2 protein was chosen for these studies (16). Short segments of M2 protein containing the 3D pattern (e.g., aa 53-86) bound C4BPα1-2 poorly, likely because they formed unstable coiled coils. This finding may reflect the presence of charged amino acids at some of the core a positions of the coiled-coil heptad (16). Stabilization of coiled-coil structure through fusion of these M2 segments to portions of GCN4 (i.e., in M2G) restored C4BP binding.
While the 3D pattern in M2 protein appeared to start at aa 61 (16), inclusion of amino acids upstream of this position had a significant effect on C4BP binding, leading to a KD that matched that of intact M2 protein. In the crystal structure, these upstream amino acids contact a crystallographically related C4BPα1-2 molecule (16). Based on these results, it is likely that the contacts made by these upstream amino acids to C4BP are not a crystallization artifact but instead a bona fide interaction.
The M2G immunogen elicited a cross-reactive response against M73 and M89 proteins from the M2/M49 pattern. Cross-reactivity was also observed against M49 protein, but this was in part or entirely attributable to antibodies that were specific to GCN4 rather than M2 portions in the M2G immunogen. No notable sequence similarity exists between M49 and GCN4, and thus the basis for this surprising result requires further investigation. No cross-reactivity was seen to M77 protein, which despite having an unambiguous M2/M49 sequence pattern (16), did not bind C4BP. In the case of some M proteins (50) that M protein HVRs evoke opsonic antibodies (7)(8)(9)(10)(11). The range of killing of Strep A by HL-60 cells due to cross-reactivity seen here (22-44%) was similar or slightly better than that observed in another study (16-41%), in which peptides derived from multiple M protein HVRs served as the immunogen (51). Notably, the cross-reactivity we observed correlated better with C4BP-binding than sequence identity (Table S3). For example, M77 protein, which did not bind C4BP and was not recognized by the M2G antiserum, has 59% identity with the M2G immunogen, but M28 protein, which did bind C4BP and was recognized, has only 35% identity.
Opsonophagocytic killing of the M28 Strep A strain at 44% was the highest of all the crossreactive interactions observed.
These results provide evidence that the conserved C4BP-binding 3D pattern can elicit antibodies that cross-react against M protein types that have the 3D pattern, and promote the opsonophagocytic killing of such Strep A strains. Significantly, the recruitment of C4BP to the Strep A surface is an essential virulence trait for numerous Strep A strains (35,36,38), and thus, escape from a broadly protective antibody that targets the C4BP-binding 3D pattern through further sequence variation may be limited by pressure to maintain C4BP interaction during infection (30). In effect, the C4BP-binding 3D pattern is an Achilles' heel of many M protein types. These results provide impetus to pursue further experiments aimed at optimizing an immunogen based on the C4BP-binding 3D pattern.

Protein expression and purification
M proteins were expressed in Escherichia coli BL21 (Gold) and purified as previously described (14,16), except that imidazole was not included in the lysis buffer. C4BPα1-2 was expressed in E. coli Rosetta 2 (Novagen). The protein was purified and refolded as previously described (52) with minor modifications. Specifically, bacteria were lysed with a C-5 Emulsiflex (Avestin). After refolding and dialysis, C4BPα1-2 was applied to a HiTrap Q HP column (GE Healthcare) and eluted using a 0-1 M NaCl gradient in 50 mM Tris, pH 8.5.

Co-precipitation assays
His6-tagged C4BPα1-2 (150 μg) was mixed with M protein constructs (molar ratio 1:1. h. For detection, 100 μl TMB substrate (BD Biosciences) was added and incubated for 10 min (protected from light), followed by addition of 50 μl of 2 N sulfuric acid to stop the reaction. For determination of titers, the absorbance at 450 nm (A450) was measured and fit to a sigmoidal curve using GraphPad Prism. Antibody titers were defined as the reciprocal of the interpolated serum dilution level that yielded 50% of the maximum A450. Statistical analysis was performed using the Student's t-test to compare immune and pre-immune sera.
His6-M6G vs His6-M6G competition. ELISAs were carried out as above, except for the following.

Human tissue cross-reactivity
Twenty μg of normal adult human brain tissue lysate (Novus Biologicals) or heart tissue lysate (Novus Biologicals) was resolved on 4-20% gradient SDS-PAGE (Bio-Rad) and transferred to a PVDF membrane (Millipore) for immunoblotting. Membranes were blocked with 5% BSA in TBST at RT for 1 h, and then incubated with rabbit antisera (1:1000 dilution in 5% BSA/ TBST) at RT for 1 h. Membranes were washed three times by TBST for 5 min each. Membranes were then incubated with HRP-conjugated goat anti-rabbit IgG (H+L) (1:4000 dilution; Southern Biotech) at RT for 1 h, and SuperSignal west pico chemiluminescent substrate (Thermo Fisher Scientific) was then added. The resulting chemiluminescence was recorded on an ChemiDoc XRS+ imaging system (Bio-Rad).

Antiserum binding to S. pyogenes
S. pyogenes was grown to mid-logarithmic phase, washed in PBS, and blocked with 10% heatinactivated donkey serum in PBS (Sigma-Aldrich) at RT for 1 h. Heat-inactivated M2G antiserum or pre-immune serum was added to S. pyogenes to 1% final volume and incubated at
Prior to carrying out the OPK assay, S. pyogenes strains were passaged through HL-60 cells, as follows. S. pyogenes was grown to mid-logarithmic phase and then diluted in THB to 3,000-10,000 CFU/ml. Ten μl of S. pyogenes were incubated with 50 μL heat-inactivated normal rabbit serum (NRS, PRF&L) per well in a round-bottom 96-well plate (Corning) at RT for 30 min, followed by the addition of 40 μl of active baby rabbit complement (BRC, PelFreez) and 100 μl of differentiated HL-60 cells to each well. The plate was then sealed with aluminum film (AlumaSeal II AF100; Excel Scientific) and incubated at 37 ºC for 2 h with end-over-end rotation. The final concentration of BRC in the reaction mixture was 5-20% in OPS buffer, with the specific value dependent on the S. pyogenes strain (such that non-specific killing, as described below, was <35%). After 2 h incubation, the plate was placed on ice for 30 min to stop the activity of HL-60 cells. After mixing thoroughly, 10 μl from each well was spotted on THB agar plates, which were tilted immediately to spread the bacteria in drips across the plates.
The plates were incubated overnight at 37 ºC, and S. pyogenes colonies were recovered. This procedure was carried out a second time to yield a total of two passages for each strain.
The OPK assay was carried out as above with twice-passaged S. pyogenes. Bacteria      Table S2.         Data are presented as in Figure 3. Data are presented as in Figure S4. a. Binding as determined by ELISA of M2G antiserum to M2 N100 , M49 N100 , M73 N100 , M28 N100 , and M89 N100 , which were adhered to the wells of ELISA plates, was competed with increasing concentrations of M2G or M6G. Experiments were carried out in triplicate, and means and standard deviations are shown.
b. Binding as determined by ELISA of anti-His antibody to His6-M6G, which was adhered to wells of ELISA plates, was competed with increasing concentrations of soluble His6 -M6G.
Experiments were carried out in triplicate and absorbance was measured at 450 nm. The experiment was carried out in triplicate, and means and standard deviations are shown.   His6-M6G