Analysis of Polymorphic Residues Reveals Distinct Enzymatic and Cytotoxic Activities of the Streptococcus pyogenes NAD+ Glycohydrolase*

Background: Two phenotypic variants of the Streptococcus pyogenes NAD+ glycohydrolase SPN exist among clinical isolates. One lacks NADase activity. Results: There are 9 polymorphic residues. Three residues are responsible for differences in NADase activity; however, both variants have equivalent cytotoxicity. Conclusion: SPN is a multifunctional toxin. Significance: Learning how evolution has adapted multiple SPN activities is crucial for understanding its contribution to tissue tropism. The Streptococcus pyogenes NAD+ glycohydrolase (SPN) is secreted from the bacterial cell and translocated into the host cell cytosol where it contributes to cell death. Recent studies suggest that SPN is evolving and has diverged into NAD+ glycohydrolase-inactive variants that correlate with tissue tropism. However, the role of SPN in both cytotoxicity and niche selection are unknown. To gain insight into the forces driving the adaptation of SPN, a detailed comparison of representative glycohydrolase activity-proficient and -deficient variants was conducted. Of a total 454 amino acids, the activity-deficient variants differed at only nine highly conserved positions. Exchanging residues between variants revealed that no one single residue could account for the inability of the deficient variants to cleave the glycosidic bond of β-NAD+ into nicotinamide and ADP-ribose; rather, reciprocal changes at 3 specific residues were required to both abolish activity of the proficient version and restore full activity to the deficient variant. Changing any combination of 1 or 2 residues resulted in intermediate activity. However, a change to any 1 residue resulted in a significant decrease in enzyme efficiency. A similar pattern involving multiple residues was observed for comparison with a second highly conserved activity-deficient variant class. Remarkably, despite differences in glycohydrolase activity, all versions of SPN were equally cytotoxic to cultured epithelial cells. These data indicate that the glycohydrolase activity of SPN may not be the only contribution the toxin has to the pathogenesis of S. pyogenes and that both versions of SPN play an important role during infection.

The Streptococcus pyogenes NAD ؉ glycohydrolase (SPN) is secreted from the bacterial cell and translocated into the host cell cytosol where it contributes to cell death. Recent studies suggest that SPN is evolving and has diverged into NAD ؉ glycohydrolase-inactive variants that correlate with tissue tropism. However, the role of SPN in both cytotoxicity and niche selection are unknown. To gain insight into the forces driving the adaptation of SPN, a detailed comparison of representative glycohydrolase activity-proficient and -deficient variants was conducted. Of a total 454 amino acids, the activity-deficient variants differed at only nine highly conserved positions. Exchanging residues between variants revealed that no one single residue could account for the inability of the deficient variants to cleave the glycosidic bond of ␤-NAD ؉ into nicotinamide and ADPribose; rather, reciprocal changes at 3 specific residues were required to both abolish activity of the proficient version and restore full activity to the deficient variant. Changing any combination of 1 or 2 residues resulted in intermediate activity. However, a change to any 1 residue resulted in a significant decrease in enzyme efficiency. A similar pattern involving multiple residues was observed for comparison with a second highly conserved activity-deficient variant class. Remarkably, despite differences in glycohydrolase activity, all versions of SPN were equally cytotoxic to cultured epithelial cells. These data indicate that the glycohydrolase activity of SPN may not be the only contribution the toxin has to the pathogenesis of S. pyogenes and that both versions of SPN play an important role during infection.
Within a single bacterial species, allelic variation of a specific gene can arise through the process of niche specialization. For bacterial pathogens, this can occur when a generalist population diverges into distinct subpopulations with strong tropism for different hosts or for different tissues within the same host. Diversity emerges as continued selection pressure results in variants of virulence genes that function to increase fitness for infection of a particular niche. Understanding the functional consequences of these changes can provide important insights into how a specific virulence factor contributes to exploitation of a niche and its role in pathogenesis.
Variation in the Streptococcus pyogenes (group A streptococcus) NAD ϩ glycohydrolase (SPN 3 ; also known as NGA) toxin has been associated with niche specialization. This Gram-positive pathogen is one of the most versatile pathogens of humans capable of causing both superficial and invasive diseases, including pharyngitis, impetigo, and necrotizing fasciitis, as well as postinfection sequelae such as rheumatic fever and acute glomerulonephritis. Part of this versatility can be attributed to its ability to secrete a multitude of proteins that affect host cell function in numerous ways (1). One of these is SPN, which was originally identified by its ability to cleave the nicotinamide-ribosyl bond of ␤-NAD ϩ to produce nicotinamide and adenosine diphosphoribose. All S. pyogenes strains examined to date possess the gene encoding SPN (2). However, more recent studies have shown that these spn alleles exhibit diversity. Furthermore, SPN is evolving under positive selection and is diverging into multiple subtypes, including subtypes that lack its signature NAD ϩ glycohydrolase (NADase) activity. Neither the function of the NADase-inactive subtypes nor the molecular basis of their loss of activity is well understood.
However, it is understood that although there is not a clear consensus as to whether SPN subtypes have any association with isolates that can cause invasive disease there is a strong association between SPN subtype and tissue tropism. Most cases of S. pyogenes infection are superficial and occur at one of two tissue sites: the throat (pharyngitis) or the skin (impetigo). Substantial epidemiological evidence indicates that there are distinct subpopulations of strains that are specialized for infection of just one of these two tissues (specialists). There is also a distinct subpopulation (generalists) that readily infects either tissue. Analysis of a collection of 113 strains that was assembled to maximize diversity revealed that intact alleles of spn were found in all strains and that NADase-active and -inactive haplotypes were equally prevalent (2). Of interest, tissue and throat specialist strains correlated with NADase-inactive SPN, whereas generalist strains correlated with NADase-active SPN. The reason underlying these associations is unknown. Furthermore, because there is evidence that the NADase activity of SPN can contribute to pathogenesis, the prevalence of NADase-inactive SPN is unclear.
The observation that NADase-inactive SPN remains under positive selection suggests that it does contribute to pathogenesis. Support for this idea comes through analysis of an endogenous competitive inhibitor of the NADase activity of SPN known as immunity factor for SPN (IFS). The gene encoding IFS is located immediately adjacent to spn, and the ability of S. pyogenes to produce NADase-active SPN is absolutely dependent on IFS, which acts to inhibit self-toxicity that may arise from any SPN molecules that inadvertently fold prior to their export from the streptococcal cell (3). As expected for an essential activity, ifs has very little sequence divergence in these strains. However, NADase-inactive SPN haplotypes are associated with nonfunctional truncated alleles of ifs that are undergoing a pattern of random nucleotide change characteristic of a pseudogene. Thus, although ifs loses selective constraint and becomes non-functional, the same degradation does not occur for NADase-inactive spn, which remains under positive selection.
During infection, SPN undergoes complicated interactions with host cells that provide some clues to its role. Although S. pyogenes is an extracellular pathogen, SPN is delivered into the host cell cytoplasmic compartment by a process termed cytolysin-mediated translocation (CMT). In this process, S. pyogenes first attaches to a host cell and exports SPN via the general secretory pathway. SPN is then translocated across the host cell membrane by an unknown mechanism that requires a second streptococcal protein, the cholesterol-dependent cytolysin streptolysin O (SLO) (4). Analyses of CMT have revealed that SPN has multiple domains, including an N-terminal domain required for translocation and a C-terminal enzymatic domain (5). Both active and inactive SPN haplotypes undergo CMT, and S. pyogenes mutants defective for expression of either SPN or SLO have a reduced cytotoxic effect for host epithelial cells. How these activities may contribute to niche specialization is not known.
Multiple enzymatic activities have been attributed to NADase-active SPN that could contribute to cytotoxicity, including ADP-ribosyltransferase, ADP-ribosyl cyclase, and NADase activities (6,7). The structure of the enzymatic domain of SPN has recently been solved and shows that it is related to the ADP-ribosyltransferase family of bacterial toxins (8). However, a re-evaluation of the enzymatic activity of SPN using highly purified recombinant protein has shown that SPN functions strictly as an NADase (9). Interestingly, comparison of NADase-active and -inactive subtypes reveals that the inactive haplotypes segregate into only two Bayesian clusters, and these differ only minutely from the active haplotypes, varying at only 9 -10 amino acid residues of 454 total. Of these, only 5-6 polymorphic residues are located in the enzymatic domain, and all were identified as sites that are evolving under positive selection (2).
Although minor, these polymorphisms contribute to a significant difference in enzymatic activity and may affect pathogenesis and niche specialization. Thus, to gain greater insight into the function of SPN, its two phenotypic subtypes, and the forces that are driving their evolution, we conducted a detailed analysis of the effect of polymorphism on enzymatic activity and the contribution of enzymatic activity to cytotoxicity. These studies revealed that differences in enzymatic activity were the result of polymorphism at not one but rather multiple residues. Furthermore, this analysis revealed an unexpected NADaseindependent cytotoxic activity that was retained by the enzymatically inactive haplotypes. Together, these data reveal that SPN has multiple activities and provide insight into its molecular evolution.

EXPERIMENTAL PROCEDURES
Bacterial Strains and Media-The studies with S. pyogenes utilized the M serotype 6 strain JRS4 (10) and derivatives, including SPN1 (⌬spn; Ref. 4) and Joy1, which expresses SPN J4 with a C-terminal influenza hemagglutinin (HA) epitope tag. Chromosomal DNA from strain HSC5 (11) was used a template in PCRs for molecular cloning of SPN H5 . Other molecular cloning and protein expression studies utilized Escherichia coli TOP10 (Invitrogen). Routine culture of S. pyogenes and E. coli was conducted using Todd Hewitt yeast extract and Luria-Bertani media, respectively, as described (9). Where appropriate, antibiotics were used at the following concentrations: chloramphenicol, 7.5 g/ml for E. coli and 3 g/ml for S. pyogenes; erythromycin, 500 g/ml for E. coli and 1 g/ml for S. pyogenes; and carbenicillin, 50 g/ml for E. coli.
Manipulation of DNA-E. coli was transformed using the method of Kushner (12), and S. pyogenes was transformed by electroporation (13). Plasmid DNA was isolated by standard techniques, and all enzymes, including restriction endonucleases, ligases (New England Biolabs), and polymerases (Pfx, Invitrogen), were used according to the manufacturers' recommendations. All site-specific mutations described in the text were generated by PCR with the mutagenic oligonucleotide primers listed in supplemental Tables S1 and S2 and the DpnI digestion method to degrade template DNA using a commercial kit (QuikChange XL kit, Stratagene). Fidelity of all DNA sequences generated by PCR was verified by DNA sequence analyses performed by commercial vendors (SeqWright, Galveston, TX; GeneWiz, South Plainfield, NJ).
Expression of SPN derivatives in S. pyogenes-Plasmids for expression of SPN in S. pyogenes were based on pJOY3 (5) and include pJOY7, which expresses SPN J4 with a C-terminal HA epitope tag. A fragment containing SPN H5 was amplified from HSC5 chromosomal DNA using primers oJOY5 and oJOY6 (supplemental Table S1) and introduced between the EcoRI and BstEII sites of pJOY7. The resulting plasmid (pJOY39) expresses SPN H5 with a C-terminal HA tag (supplemental Table S1). The various site-specific mutations were then introduced into pJOY7 and pJOY39 by PCR using the primers listed in supplemental Table S1. For expression and analysis, each plasmid was used to transform the SPN deletion strain SPN1, and the resulting strains were analyzed by the assays described below. Plasmid constructions are summarized in supplemental Table S1.
Expression and Purification of Recombinant SPNs-PCR was used to introduce the various site-specific mutations of interest into a derivative of the E. coli expression plasmid pBAD (Invitrogen) that encodes a His 6 Table S2. Expression and purification of the His 6 -tagged recombinant proteins was performed as described previously (9). As a final step, the purified proteins were dialyzed at 4°C against a buffer consisting of 50 mM potassium phosphate and 100 mM sodium chloride. Purity of the 48.6-kDa proteins was routinely assessed by SDS-PAGE and staining with Coomassie Brilliant Blue, and protein concentrations were determined using a BCA assay (Pierce) with a BSA standard.
Single Copy S. pyogenes SPN Expression Strains-SPN J4 and SPN H5 were amplified by PCR from templates pJOY7 and pJOY39, respectively, using primers oJOY180 (5Ј-CGGTG-GTTTACTCGAGAAACAAAAAAGTAACATTAGC-3Ј) and oJOY6 (see supplemental Table S1). The resulting fragments were inserted between the XhoI and PstI sites of pJRS233 (14) to produce pJOY109 and pJOY110. Site-directed mutagenesis of these two plasmids was performed to generate spn mutants described in supplemental Table S3. These plasmids were then used to replace the native spn allele in JRS4 by a standard method (15) to generate the designated strains in supplemental Table S3. Chromosomal sequences were verified by DNA sequence analysis of PCR products generated using appropriate primers.
Tests for ifs Essentiality-The requirement for ifs to support viability of strains expressing SPN of differing NADase activities was assessed by two complementary mutational strategies. In the first, an attempt was made to replace ifs with the aad9 spectinomycin resistance gene. In the second, an attempt was made to insertionally inactivate ifs using an integrational plasmid. Both strategies were based on established methods using plasmids with conditional temperature-sensitive replication (14). The construction of the two mutagenic plasmids and steps involved in mutagenesis are described in detail in the supplemental Experimental Procedures. Supplemental Fig. S3, and Tables S4 and S5. It was concluded that ifs was essential in a given strain if both strategies failed to generate an ifs-deficient mutant. In successful ifs mutants, DNA sequence analyses of spn and its promoter were conducted to confirm that each mutant retained the expected spn allele.
␤-NAD ϩ Glycohydrolase Activity of SPN from S. pyogenes-A fluorometric assay was used to analyze the ␤-NAD ϩ glycohydrolase activity in culture supernatants from S. pyogenes strains expressing the various SPN constructs (9). Concentrations of SPN in culture supernatants were normalized by immunoblotting to detect an HA epitope tag as described previously (9). Specific activities of each SPN mutant are reported relative to wild type JRS4 as described (5). Data presented are derived from three independent experiments, each performed in triplicate.
␤-NAD ϩ Glycohydrolase Activities of Recombinant Enzymes-Rates of ␤-NAD ϩ cleavage were determined by analytical HPLC as described (supplemental Fig. S2) (9). In initial trials, it was found that inconsistencies in results obtained from several of the low activity enzymes could be eliminated by the inclusion of NaCl in the reaction mixture (100 mM). Because this had no effect on the K m of SPN J4 , all analyses were conducted using this modification. Briefly, the molar concentrations of SPN indicated in the text were incubated with 2.5 mM ␤-NAD ϩ in 100 l of reaction buffer (50 mM potassium phosphate, 100 mM sodium chloride, pH 7.4, 0.5 mg/ml BSA) at 37°C. Reactions were quenched by the addition of 100 l of 20% ice-cold perchloric acid and incubated on ice for 30 min to precipitate protein and BSA. Precipitates were removed by centrifugation, supernatants were diluted into 980 l of deionized water, and the mixture was subjected to reverse-phase HPLC (SunFire, C 18 column, Waters; 5 m, 4.6 ϫ 250 mm) developed isocratically with 1% (v/v) acetonitrile in buffer A (10 mM diammonium phosphate buffer, pH 6.4) with a flow rate of 1.5 ml/min over 22 min. Quantitation of reaction products and determination of initial reaction rates and other kinetic parameters were performed as described in detail elsewhere (9). Data presented are derived from three independent experiments, each performed in triplicate.
End Point ␤-NAD ϩ Glycohydrolase Activity Assay-A single end point assay to determine the amount of uncleaved ␤-NAD ϩ remaining following reaction with various SPN derivatives was conducted a follows. Enzyme (20 pmol) was incubated in reaction buffer (50 mM potassium phosphate, 100 mM sodium chloride, pH 7.4, 0.5 mg/ml BSA) with 1 mM ␤-NAD ϩ for 1, 5, 7, and 20 h at 37°C. Reactions were quenched by chromatography using a centrifugal column (Millipore; 3-kDa exclusion). The flow-through fraction was diluted into deionized water and analyzed by reverse-phase HPLC as described above. Data presented are derived from three independent experiments, each performed in triplicate.
Infection of Epithelial Cells-Analysis of CMT and cytotoxicity were assessed following infection of A549 epithelial cells conducted as described in detail elsewhere (4). Briefly, various S. pyogenes strains were cultured overnight, back-diluted, grown to midlogarithmic phase, and used to infect confluent monolayers of A549 cells. Following incubation for 285 min, cells were lysed by the addition of saponin to a final concentration of 0.1%, the lysate was fractionated by ultracentrifugation, and the concentration of SPN in the cytosolic fraction was determined by immunoblotting to detect the HA epitope tag. Efficiencies of CMT were quantitated relative to a strain expressing SPN J4 as described (5). Cytotoxicity of various SPNexpressing S. pyogenes strains was assessed by the ability of infected cells to exclude the membrane-impermeable fluorescent probe ethidium homodimer-1 (LIVE/DEAD, catalog number L3224, Invitrogen) as described previously (4). Data presented represent the mean and S.E. derived from at least three independent experiments.
Computational Methods and Statistical Analyses-For molecular modeling, we used the PyMOL Molecular Graphics System (Version 1.5.0.1, Schrödinger, LLC) and the structure of the enzymatic domain of SPN (Protein Data Bank code 3PNT) determined by Smith et al. (8). Where indicated, differences between observed experimental mean values were tested for significance using the Tukey-Kramer multiple comparisons test. The null hypothesis was rejected for p Ͻ 0.05.

RESULTS
SPN Diversity-As discussed above, all S. pyogenes genomes analyzed to date possess the gene encoding SPN. However, it has been recognized that this population has distinct functional heterogeneity, including haplotypes that lack the signature NADase activity of SPN. Our prior analysis of 113 genomes from a strain collection assembled to reflect diversity indicated that these SPN haplotypes grouped into four Bayesian clusters, two of which lacked detectable NADase activity (2). For the present study, we chose two strains (JRS4 and HSC5) whose SPN haplotypes were representative of the most common NADase-active and -inactive Bayesian clusters (clusters 2 and 3, respectively). A direct comparison between NADase-active SPN J4 (from strain JRS4) and NADase activity-deficient SPN H5 (from strain HSC5) revealed the presence of 9 polymorphic residues, each of which has previously been shown to be under positive selection in the activity-deficient haplotypes (2). Of these, three were located in the translocation domain with the remaining residues distributed throughout the enzymatic domain (Fig. 1). As is typical, the gene encoding the NADaseactive enzyme (SPN J4 ) is associated with an intact gene for IFS, the endogenous inhibitor of SPN, whereas that of the activitydeficient protein (SPN H5 ) is associated with a truncated IFS gene ( Fig. 1) (2).
Multiple Polymorphic Residues at the C-terminal End of SPN Contribute to Enzymatic Activity-To determine which polymorphic residues may affect NADase activity, mutations were made to switch residues in NADase-active SPN J4 to their counterparts from activity-deficient SPN H5 . To provide a uniform background for comparison, the mutant genes were expressed from a plasmid vector that was introduced into a derivative of strain JRS4 (SPN1) engineered to contain an internal deletion of its native spn chromosomal locus (4). All proteins were found to be expressed at equivalent levels following analysis of cell-free culture supernatants by Western blotting (data not shown). These supernatants were then subjected to a standard end point titer assay to determine the relative levels of NADase activity produced by each protein. Individual swaps of the residues located in the translocation domain had no significant effect on the ability of the proteins to cleave a ␤-NAD ϩ substrate (R103H, G136R, and M195I; Fig. 2) as compared with the native SPN J4 (Fig. 2, Unmod). Similarly, swaps of the 3 N-terminal residues of the enzymatic domain (L199I, Q253H, and L280V; Fig. 2) also did not affect activity. In contrast, swaps of the 7th and 9th polymorphic residues (R289K and I374V) resulted in significant reductions in NADase activities (p Ͻ 0.01; Fig. 2). Consistent with prior reports (2, 16), a swap of the 8th polymorphic residue (G330D) resulted in an undetectable level of NADase activity in these supernatants (Fig. 2). These data indicate that the activity of SPN J4 is influenced to varying degrees by 3 different C-terminal polymorphic residues.
Conversion to an NADase-active Enzyme Requires 3 Residues-Reciprocal exchanges of the 3 C-terminal resides affecting activity of SPN J4 were then made in NADase-inactive SPN H5 . Starting with the most N-terminal amino acids, a swap at residue 289 of SPN H5 (K289R) did not increase the NADase activity of the resulting protein above background levels (Fig. 2). Also, despite its dramatic effect on reduction of SPN J4 activity, the reciprocal exchange at residue 330 of SPN H5 (D330G) resulted in only a very modest increase in NADase activity to levels less than 3.0% of SPN J4 (Fig. 2). An exchange of both of these C-terminal residues resulted in a protein with a significant increase in activity over the single swap proteins (p Ͻ 0.01); however, the NADase activity of this protein was only at levels equivalent to ϳ30% of SPN J4 (Fig. 2, K289R ϩ D330G). Activity equivalent to that of SPN J4 was obtained by the reciprocal swap of all 3 C-terminal residues (Fig. 2, K289R ϩ D330G ϩ V374I). Thus, all 3 of these residues make a critical contribution to the functional difference between these two naturally occurring variants of SPN.
SPN J4 Single and Double Residue Swap Proteins Have Detectable Activity-In our prior analysis of 113 clinical isolates, all alleles from NADase-active Bayesian clusters 1 and 2 contained the identical C-terminal residues at the positions analyzed above for SPN J4 (2). Similarly, all activity-deficient cluster 3 proteins had the identical polymorphisms at these positions found in SPN H5 (2). Thus, no naturally occurring intermediates containing just one or any combination of two of these polymorphisms were observed. To gain greater insight into this distribution, the properties of the engineered intermediate proteins were analyzed further. Recombinant versions of the intermediates were purified, and their reaction products following incubation with ␤-NAD ϩ were analyzed by a more sensitive HPLC-based assay. After a 1-h reaction under these conditions, essentially all ␤-NAD ϩ was converted to nicotinamide and ADP-ribose by SPN J4 , whereas all ␤-NAD ϩ remained uncleaved following reaction with SPN H5 (Fig. 3, A and B). In contrast to the assay described above, the R289K protein did not appear different from unmodified SPN J4 , and the G330D protein readily consumed substrate although to a much lower extent than SPN J4 (Fig. 3, A and B). The ability of a double swap protein with these substitutions (R289K/G330D) to consume substrate was further reduced than by either single swap, although it was still more active than SPN H5 (Fig. 3, A and B). Similarly, the introduction of the I374V along with G330D resulted in a more inactive enzyme than the G330D protein alone (Fig. 3, A and B). Only when all three polymorphisms were combined in the same protein did activity become indistinguishable from SPN H5 (Fig. 3, A and B). When combined with the results described above, these data suggest that the single and double swap proteins retain at least partial activity.
The NADase-deficient Bayesian Cluster 4 Protein Has Partial Activity-As discussed above, Bayesian cluster 4 proteins represent a second distinct cluster of NADase-deficient SPN in clinical isolates with a truncated immunity factor (supplemental Fig. S1). These cluster 4 proteins differ from the activityproficient enzymes at 2 important residues. First, they share the G330D polymorphism with the cluster 3 SPN H5 (Fig. 4A). Second, instead of the lysine for arginine substitution found at residue 289 in SPN H5 , the cluster 4 proteins substitute an asparagine (Fig. 4A). Finally, unlike the cluster 3 proteins, the cluster 4 proteins do not differ from the activity-proficient (SPN J4 ) proteins at residue 374 (Fig. 4A). Single and double swaps of these polymorphic residues were introduced into SPN J4 followed by purification and analysis utilizing an HPLC-based assay as describe above. Similar to the cluster 3 R289K polymorphism, the cluster 4 R289N single swap protein retained an appreciable ability to consume substrate (Fig. 4B). However, when combined with G330D, the resulting double swap protein failed to consume substrate and was indistinguishable from SPN H5 (Fig. 4B). Nevertheless, upon an extended period of incubation (up to 20 h), the cluster 4 double swap protein did demonstrate some activity, whereas both SPN H5 and its corresponding triple swap protein (R289K/G330D/I374V) failed to consume any ␤-NAD ϩ (Fig. 4C). Taken together, these data indicate that polymorphism at two positions contribute to the NADase-deficient property of the cluster 4 protein, although it is more active than the cluster 3 protein.
Each C-terminal Polymorphism Reduces the Rate of Enzymatic Activity-To gain greater insight into the effect of each polymorphic residue on enzymatic activity, the kinetic properties of the various swap proteins were analyzed. Analysis of the SPN J4 derivative modified at residue 289 revealed a modest (Ͻ2-fold) decrease in k cat (Table 1, R289K and R289N). However, despite being able to cleave substrate at a fast rate, alterations at this position did have a marked effect on catalytic efficiency for the R289K protein, resulting in over a 10-fold increase in K m and an ϳ20-fold decrease in efficiency (k cat /K m ; Table 2). The modification at residue 330 was more inhibitory, resulting in over a 5,000-fold decrease in k cat (Table 1, G330D), and in combination with other cluster 3 polymorphisms resulted in up to an 8-fold further decrease in k cat (Table 1). In the enzyme modified with both of the polymorphic residues of the cluster 4 protein, k cat was reduced nearly 60,000-fold, whereas for the enzyme with all 3 of the cluster 3 residues, k cat was below the limit of the assay (reduced over 2 ϫ 10 8 -fold versus SPN J4 ; Table 1). For all proteins containing the G330D polymorphism, it was not possible to determine K m at the concentrations of ␤-NAD ϩ obtainable in the assay. Furthermore, attempts to measure NAD ϩ binding to the NADase-inactive FIGURE 2. Three residues are required to restore NADase activity to SPN H5 . Polymorphic residues in SPN J4 were changed to the corresponding residue in SPN H5 , and those in SPN H5 were changed to their counterparts in SPN J4 as indicated by the gray and black bars, respectively, for comparison with the unmodified proteins (Unmod). The genes encoding the mutations were expressed from a plasmid (supplemental Table S2) introduced into a derivative of JRS4 (SPN1) with a deletion of its resident gene. NADase activities were determined from cell-free culture supernatants as described under "Experimental Procedures" and are presented relative to SPN J4 . "BL" indicates that activity was below the limit of detection (Ͻ0.5%). A double asterisk indicates significantly less NADase activity when compared with unmodified SPN J4 (p Ͻ 0.01). Data shown are the mean and S.E. (error bars) derived from at least three independent experiments. SPN were unsuccessful in part due to a high K m . Taken together, these data implicate a role for these 3 polymorphic residues in influencing substrate affinity.
IFS Is Not Essential in Strains Encoding Minimally Active SPN-To protect its own pool of intracellular ␤-NAD ϩ , S. pyogenes encodes IFS, an inhibitor of the NADase activity of SPN (3). However, in strains with an NADase-negative allele, the gene encoding IFS has accumulated mutations and has become a pseudogene (2,3). Thus, to gain insight into the co-evolution of IFS and SPN, it was of interest to determine the requirement for IFS in strains of intermediate NADase activity. A test for essentiality was conducted using two different mutational strategies to determine whether the IFS gene could be replaced or insertionally inactivated in strains expressing various spn alleles. Consistent with prior results (3), it was not possible to replace or inactivate ifs in a strain expressing SPN J4 (JRS4; Table  3 and supplemental Table S5). However, when spn was deleted from this strain, mutations in ifs were readily obtained (⌬SPN; Table 3 and supplemental Table S5). Similarly, ifs was not essential for expression of the activity-deficient triple swap protein or any of the intermediate strains expressing a low activity SPN (Table 3 and supplemental Table S5). The exception was the intermediate strain of highest NADase activity (R289K) for which no ifs insertion or replacement mutants could be recovered (Table 3 and supplemental Table S5). These results show that IFS is not absolutely required for expression of NADaseactive SPN but is required when a certain threshold of activity is exceeded.
NADase-deficient SPN Is Cytotoxic to Eukaryotic Cells-The observation that both clusters of NADase activity-deficient SPN are maintained under positive selection whereas the gene encoding IFS degrades into a pseudogene (2) suggests that the cluster 3 and cluster 4 proteins make an important contribution to pathogenesis that is independent of NADase activity. One documented function for NADase-proficient SPN is that it is cytotoxic for mammalian cells when translocated into their cytosolic compartment via CMT (17,18). To gain insight into the role of NADase activity in this context, the native spn allele of JRS4 (Fig. 5A) was modified so that various SPN proteins containing an influenza HA epitope tag (Fig. 5B) would be expressed from an identical host background for comparison with a derivative with an in-frame deletion (SPN1; Fig. 5A and Ref. 4). These included a triple mutant of SPN J4 (R289K/ G330D/I374V) and SPN H5 (Fig. 5B), which were as efficiently translocated into A549 cell cytosols following infection as the SPN J4 derivative (Fig. 5C, inset). Surprisingly, although the deletion strain was minimally cytotoxic as expected (SPN1; Fig.   5C), the strains expressing both NADase activity-deficient proteins were as cytotoxic as the strain expressing the SPN J4 protein with nearly all A549 cells showing evidence of membrane compromise following 7 h of infection (Fig. 5C). These data suggested that NADase activity could be uncoupled from cyto-  Fig. 1A). B, polymorphic residues from NADase activity-deficient cluster 4 were introduced into SPN J4 as indicated and analyzed as described in Fig. 3. Shown are representative HPLC chromatograms indicating the products of enzymatic cleavage of ␤-NAD ϩ following a 1-h reaction. The ␤-NAD ϩ control is a reaction that does not include enzyme. C, relative activity of selected proteins following the extended incubation periods shown in the figure. Data presented are the mean and S.E. (error bars) derived from at least three independent experiments. toxic activity, revealing that NADase activity is not the sole function of SPN that contributes to cytotoxicity. To confirm this, the biglutamic acid motif previously shown to be essential for catalysis (9) was exchanged for glycines in all three of these proteins, and cytotoxicity was assessed. Again, all derivatives demonstrated an ability to damage A549 cells as compared with the deletion strain (Fig. 6). The strains expressing the proteins derived from either the triple mutant or SPN H5 were not as cytotoxic as the SPN J4 derivate, but this likely reflects a lower level of translocation into the host cell cytosols (Fig. 6, inset).

DISCUSSION
Understanding the functional consequences of polymorphism can provide insight into how a specific virulence factor has been adapted in disparately evolving lineages, including those that are diverging with regard to niche selection. The present analysis of SPN has revealed that differences in enzy-matic activity between haplotypes cannot be attributed to a single polymorphic residue; rather, the transition from an enzymatically active to an inactive protein required changes at multiple positions among polymorphic residues universally conserved in both activity-deficient haplotype clusters 3 and 4. Furthermore, despite a precipitous loss of NADase activity, these variants retained their cytotoxic activities for cultured epithelial cells. These data provide insight into the evolution of SPN by demonstrating that this toxin has at least two distinct activities: an NADase activity whose loss has been selected for in certain lineages and an NADase-independent cytotoxic activity. An important and conserved role for this latter activity may explain why SPN remains under positive selection even in those lineages that lack NADase activity.
Most analyses of selection with regard to niche specialization in S. pyogenes have focused on events that have been facilitated by this organism's high rate of recombination. Prominent among these are horizontal gene transfer events, including interspecies transfer of pathogenicity islands (19) and toxin alleles (20), recombination between orthologous regulatory genes (21), and the gain or loss of adhesins at defined loci (22,23). However, the facts that the SPN haplotypes lacking NADase activity are found only in S. pyogenes, that intermediate variants containing only a subset of the activity-associated polymorphic alleles have never been observed, and that these alleles are always associated with a degraded variant of IFS (2) suggest that recombination may not have played a major role in the evolution of these variants. Thus, how these variant haplotypes may have descended from the activity-proficient versions is not clear. It can be speculated that because multiple residues are involved and because the two haplotypes share features in common adaptation may have followed a stepwise course. Both NADase-inactive haplotypes share the D330G polymorphism, suggesting that a change at this position may be the progenitor of both alleles. However, as shown in this study, intermediate combinations of these activity-associated residues, including the G330D change, retain partial activities. Given the extremely high NADase activity of SPN, the partial activities observed in the engineered intermediate forms remain relatively high when compared with other classes of enzymes that share this activity, including NAD ribosyltransferases and NAD cyclases (9). This suggests that additional alterations are required for complete loss of NADase activity. Because both variant haplotypes share polymorphism at residue 289, this may have been the next residue to come under selection. In the case of the Bayesian cluster 4 variant, the R289N variation further reduced NADase activity to levels that are not toxic for the streptococcal cell because this haplotype no longer requires a functional immunity protein for its expression.
The evolution of IFS is likely intimately linked with that of SPN. Consistent with the scheme described above, IFS is no longer essential when paired with an SPN that has the G330D polymorphism. The development of this variant would then release ifs from selective pressure to begin its transition to a pseudogene. If correct, this model implies several things about SPN-IFS co-evolution. First, the fact that ifs has become a pseudogene in the absence of robust NADase activity indicates that inhibition of this activity is its principal function rather than  (2). The number indicates the cluster(s) from which the polymorphism is derived. c Determined using the HPLC-based assay modified as described in the text. k cat values of variants were derived from data in supplemental Fig. S1. Data represent mean Ϯ S.E. from at least three independent experiments. d Relative to SPN J4 .  Ϫ No a Recipient of IFS-inactivating plasmids. Strains were generated as described in supplemental Table S3. b Defined as described in Table 1 as -fold change in k cat relative to SPN J4 as follows: ϩϩϩ, equal to SPN J4 ; ϩϩ, 1-2-fold lower; ϩ, 10 3 -fold lower; Ϫ, 10 8 -fold lower. c Defined by the ability to recover a viable mutant with a deletion or insertion mutation in ifs as described under "Experimental Procedures." Yes, IFS is essential; No, IFS is dispensable.
having any additional contributions to virulence, including SPN secretion or the CMT injection process. Second, the fact that S. pyogenes can tolerate a low level of NADase activity in the absence of IFS indicates that alterations to ifs itself were not a major driver of the continuing evolution of SPN to an activitydeficient phenotype. Instead, the lack of intermediates for the SPN NADase-deficient haplotypes in the S. pyogenes population structure suggests that the near to complete loss of NADase activity in these lineages was driven exclusively by selective pressure imparted by some virulence-related function associated with how SPN interacts with host cells.
Although their origins may not be clear, insight into the mechanistic role of polymorphism in function can be gained from analysis of the structure of SPN. The three-dimensional structure of the enzymatic domain of SPN has recently been determined, revealing that it is structurally related to the broad family of NAD ϩ ribosyltransferases (8). These proteins share a conserved core structure consisting of seven ␤-strands arranged in two perpendicular ␤-sheets that bracket the ␤-NAD ϩ binding pocket. In addition, SPN and the NAD ϩ ribo-syltransferases share an active site ADP-ribosyl turn-turn motif that contains a catalytically essential glutamate residue (Glu-391 for SPN) that lies at the base of a bowl-like substrate binding pocket (Fig. 7A). The three positions of interest are all situated in the wall of the substrate binding pocket, although two of these are distal to the active site and are located near the lip of the bowl. One of these is Gly-330, which hydrogen bonds with Gln-216 from the opposing surface to form one of the outside walls of the bowl (Fig. 7B). Molecular modeling of the glutamate polymorphism reveals significant steric clashes with Gln-216 that likely lead to a significant distortion of this wall, possibly leading to a reorientation of the two halves of the enzyme relative to each other. Consistent with this, the G330D change had the highest impact on NADase activity as compared with changes at the other two sites and likely radically altered substrate binding as the K m was raised to a level that was difficult to determine with precision. On a different face of the wall, Arg-289 interacts with Asp-286 and Lys-288 to form a surface that molecular modeling suggests may be involved in positioning the ribose moiety of the ␤-NAD ϩ substrate in the active site cleft (Fig. 7C). Although the lysine polymorphism in the cluster 3 variant SPN also represents a highly conserved substitution, the reduced ability of lysine to form hydrogen bonds likely results in some disruption of the extensive network of interactions contributed by arginine at this position. Consistent with this, the lysine substitution had the least effect on NADase activity but was still substantial, resulting in a near 20-fold decrease in enzymatic efficiency. The much greater decrease in NADase activity found in the cluster 4 variants can be explained by the considerable loss of interactions that would result from the substitution of glutamine. The third position of interest, Ile-374, is located at the bottom of the bowl near catalytic Glu-391. However, whereas the side chain of Glu-391 extends into the lumen of the bowl, the side chain of Ile-374 is buried in the hydrophobic core (Fig. 7D). Molecular modeling of the I374V substitution reveals extensive steric clashes with Phe-199 and Tyr-243 from two adjacent helices that likely would distort the position of the side chain of Glu-391 to an orientation less favorable for interaction with the glycosidic bond of the ␤-NAD ϩ substrate.
The fact that adaptation produced two distinct variant haplotypes suggests that this event was subject to considerable selective pressure. Why this process involved selection at multiple residues is not clear because a single mutation at residue Glu-391 would lead to a drastic reduction in NADase activity. It is possible that significant structural constraint was imposed by a dual requirement for loss of NADase activity and selection for the cytotoxic activity. However, this scenario would imply a larger conserved role for Glu-391 beyond cytotoxicity because mutation at this position did not influence cytotoxicity for cultured epithelial cells. A pathway of multiple changes may have been necessary because any single mutation resulted in a less stable protein. Some evidence for this comes from the observation that the enzymatic activity of some of the engineered intermediate forms was more erratic using low salt conditions that had no effect on activity of the SPN J4 enzyme. However, all engineered intermediate forms were readily expressed by either S. pyogenes or E. coli, and none exhibited any obvious differences with regard to solubility or degradation relative to the unmodified proteins. Thus, a more detailed understanding of adaptation will require uncovering the mechanism of the NADase-independent cytotoxic activity.
Multifunctionality is not an uncommon property of many bacterial toxins that are recognized to be subject to ongoing adaptation (24). For example, the VacA vacuolating cytotoxin of Helicobacter pylori is an important determinant of colonization, persistence, and pathogenesis of its human host and exhibits considerable diversity. Like SPN, VacA diversity is restricted to specific regions of the toxin and generates a wide range of toxin efficacy from variants that produce extensive vacuolation, to those that only produce vacuolation in a limited range of cultured cells, to variants that fail to cause any detectable cytotoxicity in in vitro assays (24,25). Also like SPN, variation has a functional consequence, although for VacA this appears to correlate with the risk of developing certain complications of long term colonization like peptic ulcer disease or gastric cancer (26,27). However, VacA is multifunctional and in addition to cytotoxicity has been reported to alter membrane permeability, to target and damage mitochondria, and to activate numerous host cell signaling pathways (24). Similar to SPN, it is thought that this multifunctional nature likely explains why VacA is found in essentially all H. pylori strains that infect humans despite extensive variation in vacuolating ability (27).
A common property of multifunctional toxins exhibited by VacA is the ability to cause distinct cellular effects by acting at different cellular locations (24). Whether SPN acts at different cellular locations is not clear. Translocated SPN is found exclusively in the cytoplasmic compartment (4, 28), although the presence or absence of SPN in defined mutants has been reported to affect the invasiveness of S. pyogenes during infection of cultured epithelial cells (18), a property that could contribute to niche selection. SPN itself is a multidomain protein and in addition to its NADase domain has an N-terminal domain that is essential for its translocation (5). Although the structure of this domain has not been solved, homology modeling indicates that it adopts a "jelly roll" fold common to many carbohydrate-binding proteins, which suggests that it may promote interaction at the cell surface (5). The role of this domain in cytotoxicity remains to be determined. Prior reports suggested that SPN may have multiple enzymatic activities that could produce differential cellular effects, including ADP-ribosyltransferase and cyclase activities (6,7). However, more refined analyses of highly purified recombinantly produced SPN have failed to confirm these additional activities (9). Removing the NADase activity does not affect cytotoxicity, calling into question the contribution of the NADase activity in cell death. However, support for a role for NADase activity in cytotoxicity comes from studies in E. coli and yeast that show FIGURE 6. Catalytic biglutamic acid residues are not required for cytotoxicity. The biglutamic acid residues involved in catalysis (see Fig. 1) were changed to glycines in SPN proteins SPN J4 , SPN TM (R289K/G330D/I374V), and SPN H5 at the chromosomal spn locus in strain JRS4 to generate strains SPN J4-GG , SPN TM-GG , and SPN H5-GG , respectively. Refer to supplemental Table  S3 for detailed construction of these strains. The strains expressing the proteins listed in parentheses shown in the figure were used to infected A549 cells, and cytotoxicity was monitored at the indicated times as described for that expression of SPN in the absence of IFS, the endogenous inhibitor of its NADase activity, is lethal (9,17). In contrast, expression in yeast of an activity-deficient haplotype is not lethal (9). Yeast may lack a target required for the NADaseindependent activity, or this may be due to the observation that full penetrance of the cytotoxic effect on epithelial cells requires a synergistic interaction with the pore-forming activity of the streptococcal streptolysin O protein (29). Taken together, these data suggest two possible modes of SPN toxicity.
The studies described here have expanded our understanding of SPN structure and function and the molecular basis for its evolution. However, several important questions remain, including the structural and molecular basis of its NADaseindependent cytotoxic activity and how its various activities contribute to tissue tropism. Several studies have suggested that the number of S. pyogenes isolates that demonstrate NADase activity has been increasing since the late 1980s and that the activity-proficient SPN haplotypes have entered into some lineages that are now more frequently isolated from invasive disease (7,16,30). Thus, a detailed understanding of SPN function and evolution will further our understanding of how S. pyogenes populations continue to evolve.