Single Amino Acid Replacements in RocA Disrupt Protein-Protein Interactions To Alter the Molecular Pathogenesis of Group A Streptococcus

Group A Streptococcus (GAS) is a human-specific pathogen and major cause of disease worldwide. The molecular pathogenesis of GAS, like many pathogens, is dependent on the coordinated expression of genes encoding different virulence factors. The control of virulence regulator/sensor (CovRS) two-component system is a major virulence regulator of GAS that has been extensively studied. More recent investigations have also involved regulator of Cov (RocA), a regulatory accessory protein to CovRS.

ABSTRACT

Group A Streptococcus (GAS) is a human-specific pathogen and major cause of disease worldwide. The molecular pathogenesis of GAS, like many pathogens, is dependent on the coordinated expression of genes encoding different virulence factors. The control of virulence regulator/sensor (CovRS) two-component system is a major virulence regulator of GAS that has been extensively studied. More recent investigations have also involved regulator of Cov (RocA), a regulatory accessory protein to CovRS. RocA interacts, in some manner, with CovRS; however, the precise molecular mechanism is unknown. Here, we demonstrate that RocA is a membrane protein containing seven transmembrane helices with an extracytoplasmically located N terminus and cytoplasmically located C terminus. For the first time, we demonstrate that RocA directly interacts with itself (RocA) and CovS, but not CovR, in intact cells. Single amino acid replacements along the entire length of RocA disrupt RocA-RocA and RocA-CovS interactions to significantly alter the GAS virulence phenotype as defined by secreted virulence factor activity in vitro and tissue destruction and mortality in vivo. In summary, we show that single amino acid replacements in a regulatory accessory protein can affect protein-protein interactions to significantly alter the virulence of a major human pathogen.

INTRODUCTION

Group A Streptococcus (GAS) is a human-specific pathogen that causes a number of diseases ranging in severity from relatively innocuous bacterial pharyngitis (“strep throat”) to life-threatening necrotizing fasciitis (“flesh eating” disease) (1, 2). Despite over a century of efforts by many investigators, no licensed GAS vaccine is available (2, 3), which may, in part, be due to the complex regulation of its many putative and proven virulence factors and surface-exposed proteins (3–27).

One major virulence regulator of GAS is the control of virulence regulator/sensor (CovRS) two-component system. CovRS is a negative regulator of virulence, regulating approximately 10 to 15% of the GAS transcriptome (27–31). That is, when activated, CovRS decreases expression of many GAS genes and decreases virulence. In turn, CovRS inactivation significantly increases virulence (27, 29, 31–36). Since its identification over 20 years ago (32, 37, 38), the molecular mechanism of CovRS gene regulation has been extensively studied. Although the in vivo stimulus for CovRS activation remains unknown, CovR activation in vitro via phosphorylation by CovS can be modulated by cationic magnesium (Mg²⁺) and the human antimicrobial peptide LL-37 (39–41). Many studies have investigated the effect of single nucleotide polymorphisms and amino acid changes in CovRS, identifying key residues and protein domains important for virulence factor gene regulation (27, 33, 35, 41–45).

More recent study of virulence regulation in GAS has involved regulator of Cov (RocA), a regulatory accessory protein to the CovRS two-component system (25, 46–61). RocA is a positive regulator of CovRS (46). That is, RocA modulates the phosphatase activity of CovS (35, 62) to increase CovR phosphorylation (51, 53, 59, 61, 62), leading to decreased virulence factor gene expression and decreased strain virulence. Inactivation of RocA by gene deletion or truncation increases virulence factor expression and strain virulence (25, 47, 51–54, 56–59, 61). In contrast to truncation and deletion mutations that can have substantial effects on protein structure, we recently investigated the effect of single amino acid replacements in RocA on GAS molecular pathogenesis (25). We discovered that many different naturally occurring amino acid replacements result in a significantly altered global transcriptome, secreted virulence factor activity in vitro, and virulence in vivo (25). However, the molecular basis of RocA-CovRS interactions remains unknown.

While some aspects of RocA molecular pathogenesis have been discovered (25, 53, 56, 59, 61), many key knowledge gaps remain. First, the physical interaction between RocA and CovS is hypothesized to occur through their N-terminal transmembrane domains (56, 59, 61); however, mutations, including truncation, to the C terminus of RocA unexpectedly result in a RocA deletion-like virulence phenotype (25, 51). That is, the C-terminal cytoplasmic domain has a yet undefined function. Second, demonstration of a direct physical interaction between RocA and CovS has been dependent on the experimental methodology employed (59, 61). Third, the number, location, and boundaries of the N-terminal transmembrane helices are unknown. Whereas some published studies predict six transmembrane helices (55, 56, 58, 61), others predict seven (25, 46, 57, 59). A detailed understanding of RocA topology is necessary for downstream mechanistic and translational studies bearing on RocA-CovRS protein interactions. Fourth, although many different amino acid replacements in RocA are proven to alter GAS virulence (25, 57), none result in identical effects on genome-wide transcriptomes, virulence factor activities in vitro, and virulence phenotypes in mice and nonhuman primates (25). The molecular basis for different amino acid replacements conferring different virulence phenotypes is crucial to understanding the RocA-CovRS interaction specifically and accessory protein-regulatory protein interactions in general.

To begin determining the potential role of single amino acid replacements in altering RocA-CovRS molecular interactions, we addressed the aforementioned knowledge gaps. Our topology studies demonstrate that RocA is a membrane protein with seven N-terminal transmembrane helices. Additionally, using a bacterial adenylate cyclase-based two-hybrid (BACTH) experimental methodology, and in contrast to previous reports (61), we demonstrate that RocA directly interacts with itself and CovS, but not CovR. Finally, we developed a model to explain how single amino acid replacements in RocA alter its interaction with itself and CovS to significantly affect gene expression, secreted virulence factor activity in vitro, and strain virulence in vivo.

RESULTS

Multiple in silico membrane topology algorithms predict a consensus RocA membrane topology.

RocA is composed of two major predicted domains, a functionally important transmembrane domain in the N terminus and a putative nonfunctional histidine kinase ATPase domain in the C terminus (Fig. 1) (25, 46, 52, 56, 57, 59). Previous investigators speculated that either six (55, 56, 58, 61) or seven (25, 46, 57, 59) transmembrane helices exist, but the exact number has not been experimentally determined. Furthermore, the overall topology of the N-terminal transmembrane domain of RocA has not been experimentally determined. To understand the mechanism by which single amino acid replacements alter the functionality of RocA, knowledge of the precise protein topology is needed.

FIG 1.

The predicted membrane topology model of RocA has a seven-transmembrane helical architecture and a putative C-terminal histidine kinase ATPase domain (residues 346 to 447) (57). A composite of eight different in silico algorithms was used (see Fig. S1 for details). The N and C termini are identified. Residues between the purple lines are predicted to be in transmembrane helices by all algorithms, whereas residues between a red and purple line are predicted to be in transmembrane helices in at least one, but not all, algorithms. Residues in the red shaded box were predicted to be an eighth transmembrane helix by one in silico algorithm (Fig. S1), but the prediction was not supported by experimental evidence. Native cysteine residues are colored blue. Residues used in membrane topology experiments are identified. Relevant results of SCAM^TM and protein fusion assays are indicated.

We used multiple in silico protein topology algorithms to predict the membrane topology of RocA. The algorithms included Phobius (63), Philius (64), OCTOPUS (65), PolyPhobius (66), SPOCTOPUS (67), SCAMPI (68), MEMSAT-SVM (69), and Phyre2 (70). These algorithms use a combination of hidden Markov models, artificial neural networks, and physical data parameters and support vector machine-based approaches on well-defined training sets to predict transmembrane segments in proteins based on amino acid sequence and context and protein homology. The consensus topology predicts that RocA has seven transmembrane helices in the N-terminal half of the protein, the N terminus is located extracytoplasmically, and the C terminus is located in the cytoplasm (Fig. 1). Although none of the protein topology algorithms predicted identical transmembrane helical boundaries (see Fig. S1 in the supplemental material), each was generally consistent with respect to transmembrane helical boundaries. The one outlier was MEMSAT-SVM, which predicted an eighth transmembrane helix in the C-terminal half of RocA (Fig. 1 and Fig. S1). Next, we used the modeled data, predicting a seven-transmembrane helical architecture, to guide experiments seeking to precisely determine the membrane topology of RocA.

SCAM^TM recapitulates the predicted in silico topology of RocA.

We applied the substituted-cysteine accessibility method as applied to transmembrane orientation (SCAM^TM) methodology as a first approach to determine the membrane topology of the N-terminal transmembrane domain of RocA (71). In this methodology, naturally occurring or strategically engineered cysteine residues in membrane proteins are differentially labeled based on their cellular localization and accessibility to the water-soluble, thiol-specific labeling reagent 3-(N-maleimido-propionyl)biocytin (MPB). Accessible extracytoplasmic cysteine residues can be labeled by treating intact cells with MPB. Accessible cytoplasmic cysteine residues can be labeled by first treating intact cells with nondetectable, water-soluble 4-acetamido-4′-maleimidylstilbene-2,2′-disulfonic acid (AMS) to block extracytoplasmic cysteine residues, followed by treatment of AMS-blocked sonicated cells with MPB. Importantly, SCAM^TM uses the full-length protein expressed in GAS, allowing for topological analysis in the native host environment without disrupting the normal protein conformation.

First, we determined the location of the six naturally occurring cysteine residues in RocA (C135, C144, C181, C304, C338, and C371; Fig. 1). Since the in silico algorithms predicted that each is located in a transmembrane helix (C135, C144, and C181) or the cytoplasm (C304, C338, and C371), we hypothesized that wild-type RocA would only be labeled under conditions that allowed MPB to access the cytoplasm. We used a wild-type rocA allele containing a FLAG-tag epitope at the C terminus (59, 61). Whole-genome sequencing confirmed the absence of spurious mutations in this isogenic clone. No difference in GAS growth in nutrient-rich broth or secreted NAD⁺-glycohydrolase (SPN) activity was observed due to the addition of the FLAG-tag epitope (Fig. S2) (53, 56). As seen in Table 1 and Fig. S3A, both of the labeling agent treatment samples contained RocA-FLAG protein, and RocA-FLAG was only labeled with MPB under conditions that allowed the labeling reagent to access cytoplasmic cysteine residues. Based on these data, we conclude that all naturally occurring accessible cysteine residues in the wild-type RocA-FLAG protein (C304, C338, and C371) are located in the cytoplasm. Thus, SCAM^TM can be used to differentiate between strategically engineered cytoplasmic and extracytoplasmic cysteine residues in RocA to subsequently map its membrane topology.

TABLE 1.

Experimentally determined cellular location for RocA residues, as determined by SCAM^TM

RocA varianta	MPB labeling		RocA-FLAGd		Interpretatione
	Extracytoplasmicb	Cytoplasmicc	Extracytoplasmic	Cytoplasmic
WT	No	Yes	Yes	Yes	Cyt
E3C	Yes	Yes	Yes	Yes	Ext
L34C	No	Yes	Yes	Yes	Cyt
H60C	Yes	Yes	Yes	Yes	Ext
K88C	No	Yes	Yes	Yes	Cyt
S121C	Yes	Yes	Yes	Yes	Ext
Q160C	No	Yes	Yes	Yes	Cyt
L192C	Yes	Yes	Yes	Yes	Ext
Q225C	No	Yes	Yes	Yes	Int

^aRocA allele assayed. All variants had a C-terminal FLAG-tag.

^bDefined as band present for avidin labeling by MPB without sonication.

^cDefined as band present for avidin labeling by MPB when pretreated with AMS and sonication.

^dPresence of RocA-FLAG protein in indicated sample, as determined by an anti-FLAG-tag antibody.

^eLocation of engineered cysteine residue, as defined in “RocA variant” column. For WT, location of native cysteine residues that were labeled by MPB. Cyt, cytoplasmic; Ext, extracytoplasmic.

Based on the in silico predictions, we engineered eight cysteine residue mutants in RocA-FLAG to determine the number and localization of the N-terminal transmembrane helical boundaries—E3C, L34C, H60C, K88C, S121C, Q160C, L192C, and Q225C. The first mutant (E3C) was created to determine the N terminus locale, the next 6 mutants (L34C, H60C, K88C, S121C, Q160C, and L192C) were designed to define the intervening loops, and the final mutant (Q225C) was created at the boundary between the N-terminal transmembrane and C-terminal cytoplasmic domains (56, 59). Whole-genome sequencing confirmed the absence of spurious mutations in each cysteine engineered mutant strain. No difference in GAS growth in nutrient-rich broth or secreted SPN activity was observed due to the introduction of the engineered cysteine residues (Fig. S4). As seen in Table 1 and Fig. S3B, RocA-FLAG mutants E3C, H60C, S121C, and L192C had labeling profiles suggestive of an extracytoplasmic localization. The RocA-FLAG mutants L34C, K88C, Q160C, and Q225C had labeling profiles suggestive of a cytoplasmic localization (Table 1 and Fig. S3B).

RocA-PhoA-LacZα protein fusions are consistent with the in silico algorithm predictions and SCAM^TM results.

As a second complementary method to experimentally determine RocA topology, we used a PhoA-LacZα protein fusion system in Escherichia coli (72–79). We designed an expression construct to fuse RocA to PhoA-LacZα at selected amino acids based on the in silico algorithm consensus topology predicting a cytoplasmic or extracytoplasmic location. Using this experimental strategy, extracytoplasmic fusions result in a protein fusion with high alkaline phosphatase activity and low β-galactosidase activity, whereas cytoplasmic fusions result in a protein fusion with low alkaline phosphatase activity and high β-galactosidase activity. Additionally, β-galactosidase and alkaline phosphatase activities can be measured simultaneously using this protein fusion approach, allowing for normalization of the resulting activities and accurate localization determination (72). RocA protein fusions were designed to occur within the six predicted intervening loops (L34, H60, K88, S121, Q160, and L192; Fig. 1). To determine if an eighth transmembrane helix exists and to identify the location of the C terminus, we also designed RocA protein fusions to occur at A360 and D451, respectively (Fig. 1).

We constructed RocA protein fusions in plasmid pKTop, transformed them into E. coli strain DH5α, and performed β-galactosidase and alkaline phosphatase assays (78, 79). Sanger sequencing confirmed that the correct protein fusions were generated with no spurious mutations. The results are shown in Table 2 and Fig. S5. In general, the RocA-PhoA-LacZα data supported the in silico prediction model (Fig. 1). Our results indicated that residues H60 and S121 are located extracytoplasmically, and residues K88, Q160, A360, and D451 are located in the cytoplasm, as predicted. However, assays using residues L34 and L192 indicated an indeterminate localization for these residues (Table 2 and Fig. S5A). That is, we could not determine whether L34 and L192 were cytoplasmic or extracytoplasmic with high certainty with this methodology. To better define the loops containing amino acids L34 and L192, four additional neighboring residues were chosen to create new protein fusions (K24, L44, C181, and R201; Fig. 1). A protein fusion at Q225 was also created to determine the end boundary of the seventh transmembrane helix. Our results demonstrated that K24, L44, C181, R201, and Q225 are located in the cytoplasm using this methodology (Table 2 and Fig. S5B).

TABLE 2.

Normalized activity ratios (NARs) and experimentally determined cellular location for RocA residues

RocA residuea	Initial exptl design		Additional exptl design
	NARb	Locationc	NAR	Location
K24			0.03	Cyt
L34	0.77	Ind	0.98	Ind
L44			0.03	Cyt
H60	7.74	Ext	5.96	Ext
K88	0.01	Cyt	0.02	Cyt
S121	4.03	Ext	2.12	Ext
Q160	0.02	Cyt	0.02	Cyt
C181			0.09	Cyt
L192	0.55	Ind	0.85	Ind
R201			0.28	Cyt
Q225			0.22	Cyt
A360	0.09	Cyt	0.13	Cyt
D451	0.04	Cyt	0.06	Cyt

^aTerminal residue of RocA followed by PhoA-LacZα reporter.

^bNAR calculated as relative alkaline phosphatase activity/relative β-galactosidase activity (79).

^cLocation determined by NAR. Cyt, cytoplasmic; Ext, extracytoplasmic; Ind, indeterminate.

Summary of the RocA protein topology experiments.

Taken together, the in silico algorithm predictions, SCAM^TM data, and RocA-PhoA-LacZα protein fusion data demonstrate that RocA has a seven-transmembrane helical architecture with an extracytoplasmically located N terminus and a cytoplasmically located C terminus (Fig. 1). The seven transmembrane helices span from (i) residues 3 to 34 extracytoplasm to cytoplasm, (ii) residues 34 to 60 cytoplasm to extracytoplasm, (iii) residues 60 to 88 extracytoplasm to cytoplasm, (iv) residues 88 to 121 cytoplasm to extracytoplasm, (v) residues 121 to 160 extracytoplasm to cytoplasm, (vi) residues 160 to 192 cytoplasm to extracytoplasm, and (vii) residues 192 to 225 extracytoplasm to cytoplasm (Fig. 1). The stated amino acid residues do not mark the exact transmembrane helical boundaries but serve as approximations based on the overall experimental data (Tables 1 and 2). In summary, for the first time, we have experimentally determined the membrane topology of RocA, which will facilitate investigation into the effect of specific amino acid changes on RocA protein-protein interactions.

RocA interacts with RocA and CovS, but not CovR.

After experimentally determining the membrane topological architecture of RocA, we next investigated whether specific amino acid changes could alter RocA protein-protein interactions. Previous attempts to demonstrate a direct interaction between RocA and CovRS have had variable success depending on the experimental methodology used (59, 61). While RocA-RocA and RocA-CovS interactions have been demonstrated by coimmunoprecipitation using cell membrane preparations, no direct interaction between RocA and RocA, CovS, or CovR using two-hybrid assays has been demonstrated (59, 61). We used bacterial adenylate cyclase-based two-hybrid (BACTH) assays to first assess wild-type RocA, CovS, and CovR protein-protein interactions (80–82). Briefly, RocA, CovS, and CovR were fused to either T18 or T25, two complementary fragments of the catalytic domain of the Bordetella pertussis adenylate cyclase. In this assay, interaction between two proteins in E. coli brings T18 and T25 into close proximity, leading to a functional adenylate cyclase that generates cAMP. Interactions are then quantitated by β-galactosidase activity. Sanger sequencing confirmed that the correct protein fusions were generated with no spurious mutations.

Since the BACTH constructs used in a previous study only included one set of T18/T25 protein fusions (61), we sought to reassess the usefulness of BACTH for testing RocA-CovRS interactions using all possible combinations of T18/T25-RocA/CovS/CovR fusions (Table S1 and Fig. S6). In agreement with previously published data, we observed direct interactions of CovS and CovS, CovR and CovR, and CovR and CovS (Fig. 2 and Fig. S7) (61). Additionally, our results demonstrated that RocA interacted with itself (RocA) and CovS but not CovR (Fig. 2). Interaction between RocA and RocA or RocA and CovS was only observed when T18/T25 was fused to the C terminus of RocA (Fig. 2 and Fig. S6). That is, consistent with the experimentally determined topology of RocA, only RocA fusions with cytoplasmic T18/T25 (fused at the C terminus) demonstrated direct protein-protein interactions. We reproduced these results using a second strain of E. coli (Fig. S7).

FIG 2.

RocA interacts with itself and CovS, but not CovR. (A and B) Representative BACTH assays performed in E. coli strain DHM1 for wild-type RocA, CovS, and CovR homodimers (A) and heterodimers (B). The white bar represents the positive control (zip, GCN4 leucine zipper motif), and striped bars represent the negative controls (E, empty plasmid). The red line indicates the positive interaction threshold. A red asterisk indicates a positive interaction between the two assayed proteins. pKT25, N-terminal T25 protein fusion; pKNT25, C-terminal T25 protein fusion; pUT18, C-terminal T18 protein fusion; pUT18C, N-terminal T18 protein fusion.

Amino acid changes in RocA differentially alter the interaction of RocA with itself and CovS.

After demonstrating a direct interaction between the wild-type RocA and CovS proteins, we next tested the hypothesis that single amino acid replacements in RocA can differentially alter its interaction with itself and CovS. We first investigated the RocA P97L, G184W, R258K, and T442P amino acid replacements. The amino acid replacements were selected because they have very well-known effects on the GAS global transcriptome and virulence in mice and nonhuman primates (25). Our results demonstrated that amino acid changes P97L, G184W, and T442P disrupted RocA-RocA interactions, whereas amino acid change R258K did not disrupt RocA-RocA interaction (Fig. 3A). RocA with amino acid change T442P disrupted interaction with CovS, whereas RocA with amino acid changes G184W and R258K did not (Fig. 3B). RocA with amino acid change P97L had an intermediate RocA-CovS interaction phenotype (Fig. 3B). Taken together, the BACTH data suggest that specific amino acid changes in RocA have different effects on RocA functionality by altering the RocA-RocA and/or RocA-CovS physical interaction.

FIG 3.

Single amino acid replacements in RocA alter the interaction between RocA and itself and CovS. (A and B) Representative BACTH assays performed in E. coli strain DHM1 for RocA homodimers (A) and RocA-CovS heterodimers (B). The RocA amino acid replacements assayed are indicated (WT, wild-type RocA). The white bar represents the positive control (zip, GCN4 leucine zipper motif), and striped bars represent the negative controls (E, empty plasmid). The red line indicates the positive interaction threshold. A red asterisk indicates a positive interaction between the two assayed proteins. pKT25, N-terminal T25 protein fusion; pKNT25, C-terminal T25 protein fusion; pUT18, C-terminal T18 protein fusion; pUT18C, N-terminal T18 protein fusion.

Polymorphisms in the C terminus of RocA are function-altering.

A previous study of isogenic rocA polymorphism mutants found that mutation of the RocA C terminus resulted in a rocA deletion-like phenotype (25). The results were unexpected, since the C terminus of RocA had not been previously implicated in RocA functionality (56, 59). We therefore sought to further characterize the role of the RocA C terminus in GAS molecular pathogenesis. In addition to the previously studied T442P mutation, we identified three other naturally occurring mutations of interest in the C terminus of RocA, including V420I, T442I, and Q443* (glutamine 443 replaced with a stop codon; Fig. 4A) (25, 57). We constructed isogenic mutant strains, and whole-genome sequencing confirmed the absence of spurious mutations. To determine if C-terminal RocA amino acid changes alter the growth phenotype, we grew the parental wild-type strain, isogenic ΔrocA deletion mutant strain, and four C-terminal rocA mutant strains in nutrient-rich Todd-Hewitt broth with yeast extract (THY). While the growth curves of the RocA V420I and T442I mutant strains were similar to the parental wild-type strain, the RocA T442P and Q443* mutant strain growth curves were more similar to the isogenic ΔrocA strain that has a shortened lag phase (Fig. 4B) (25).

FIG 4.

Polymorphisms in the C terminus of RocA are function-altering. (A) Protein map of RocA highlighting C-terminus polymorphisms (inset, amino acids 300 to 451). Predicted domains of the RocA protein are indicated (HATPase, histidine kinase ATPase domain). Predicted functional domains of the putative histidine kinase domain (N box, F box, G box) are identified (46). (B) Growth curve in nutrient-rich THY broth. (C) SPN activity assay. (D) SLO activity assay. (E) SKA activity assay. Data are shown as the mean ± standard deviation (n = 3). ***, P < 0.001; ****, P < 0.0001; one-way analysis of variance (ANOVA) with Tukey’s multiple-comparison test compared to the parental wild-type (WT) strain; #, P < 0.05; ####, P < 0.0001; one-way ANOVA with Tukey’s multiple-comparison test compared to the isogenic ΔrocA deletion mutant strain. (F) Representative BACTH assays performed in E. coli strain DHM1 for RocA homodimers and RocA-CovS heterodimers. The RocA amino acid replacements assayed are indicated (WT, wild-type RocA). The white bar represents the positive control (zip, GCN4 leucine zipper motif), and striped bars represent the negative controls (E, empty plasmid). The red line indicates the positive interaction threshold. A red asterisk indicates a positive interaction between the two assayed proteins. pKT25, N-terminal T25 protein fusion; pKNT25, C-terminal T25 protein fusion; pUT18, C-terminal T18 protein fusion; pUT18C, N-terminal T18 protein fusion.

Next, to test the hypothesis that amino acid changes in the RocA C terminus result in altered virulence, we performed in vitro virulence factor activity assays. The RocA V420I and T442I isogenic mutant strains had wild-type-like secreted SPN, streptolysin O (SLO), and streptokinase (SKA) activities (Fig. 4C to E). In contrast, the ΔrocA and RocA T442P and Q443* isogenic mutant strains had significantly increased SPN and SLO activity, and significantly decreased SKA activity, compared to the parental wild-type strain (Fig. 4C to E). As previously observed (25), the RocA T442P isogenic mutant strain had significantly decreased SKA activity compared to the isogenic ΔrocA deletion mutant strain (Fig. 4E). The SKA activity of the RocA Q443* isogenic mutant strain did not significantly differ from the secreted SKA activity of the isogenic ΔrocA deletion mutant strain (Fig. 4E). These results are consistent with the BACTH analysis that demonstrated that RocA T442P and Q443* variant proteins do not directly interact with themselves or CovS (that is, they behave like rocA deletion mutants), whereas RocA V420I interacts with itself and CovS, similar to the wild-type RocA protein (Fig. 4F). Interestingly, RocA T442I did not directly interact with itself or CovS (Fig. 4F).

Based on the in vitro virulence factor assay data, we hypothesized that RocA amino acid replacements T442P and Q443* increase strain virulence. To test this hypothesis, we compared the virulence of the parental wild-type strain, the isogenic ΔrocA deletion mutant strain, and four C-terminal rocA mutant strains in a mouse model of necrotizing myositis (13, 25, 57, 83, 84). Consistent with our hypothesis, the isogenic ΔrocA deletion and RocA T442P and Q443* mutant strains caused significantly increased mortality and larger lesions with more tissue destruction than the parental wild-type strain and wild-type-like RocA V420I and T442I mutant strains (Fig. 5).

FIG 5.

Polymorphisms in the C terminus of RocA result in altered virulence in a mouse model of necrotizing myositis. (A) Kaplan-Meier survival curve for mice infected in the right hindlimb with the indicated strain (n = 40 mice/strain). *, P < 0.05; **, P < 0.01; log rank test compared to the parental wild-type (WT) strain. (B) Representative gross lesions of the infected limb on day 2 postinoculation (n = 4 mice/strain). Abscesses are encompassed by red circles. (C) Representative microscopic lesions of the infected limb on day 1 postinoculation (n = 4 mice/strain). The necrotic lesions are encompassed by black boxes. Hematoxylin and eosin staining was used. Original magnification, ×4.

Altogether, the data demonstrate that the C terminus of RocA, in addition to the N-terminal transmembrane domain, is important for RocA protein-protein interactions and molecular pathogenesis.

DISCUSSION

In recent years, virulence regulation in GAS has focused on transcription regulators such as the CovRS two-component system and its regulatory accessory protein RocA (25, 46–61). Two key knowledge gaps bearing on the molecular pathogenesis of RocA, and accessory proteins in general, are (i) how accessory proteins interact with two-component system proteins and (ii) how naturally occurring amino acid replacements alter this interaction. Here, we demonstrate that RocA is a seven-transmembrane helix regulatory accessory protein (Fig. 1) that interacts with itself (RocA) and CovS, but not CovR (Fig. 2). Different amino acid replacements in RocA had different effects on RocA-RocA and RocA-CovS interactions (Fig. 3 and 4). The disrupted protein-protein interactions lead to significant changes in gene expression, secreted virulence factor activity in vitro, and virulence in vivo (Fig. 5 and 6) (25).

FIG 6.

Single amino acid replacements in RocA alter RocA-RocA and RocA-CovS interactions, leading to altered gene expression and virulence. Under normal conditions (wild-type [WT] RocA, upper left), RocA (red) interacts with itself and CovS (black), resulting in WT virulence factor expression and secreted activity in vitro and virulence in vivo. The RocA G184W amino acid change (brown, upper right) causes a loss of RocA dimerization but retention of the RocA-CovS interaction, resulting in decreased virulence factor expression and activity in vitro but WT-like virulence in vivo. The RocA P97L amino acid change (orange, lower right) causes a loss of the RocA-RocA interaction and a transient RocA-CovS interaction, resulting in increased virulence factor expression and activity in vitro but WT-like virulence in vivo. The RocA T442P (purple, lower left) causes a loss of RocA-RocA and RocA-CovS interactions, resulting in increased virulence factor expression and activity in vitro and increased virulence in vivo.

We confirmed in silico algorithm predictions of RocA membrane topology using two independent experimental methodologies. Some previously published studies predicted six transmembrane domains (55, 56, 58, 61). Here, results from our topology analyses using two independent methodologies demonstrate that RocA has seven transmembrane helices (Fig. 1 and Tables 1 and 2). SCAM^TM assays demonstrated that the N-terminal half of RocA has seven transmembrane helices and the N terminus is located extracytoplasmically (Fig. 1 and Table 1). The SCAM^TM results were confirmed with RocA-PhoA-LacZα protein fusions. The RocA-PhoA-LacZα protein fusions expressed in E. coli recapitulated the SCAM^TM data, except for indeterminate results with transmembrane helices 1 and 6. Two major limitations to the RocA-PhoA-LacZα protein fusion methodology, which could lead to misfolding and ambiguous results, are that (i) it does not use the full-length protein to determine residue locations and (ii) it is not performed in the native host bacterium. That is, using the RocA-PhoA-LacZα protein fusion methodology, transmembrane helices 1 and 6 are located in the cytoplasm, and association with the cell membrane and membrane insertion of transmembrane helices 1 and 6 are likely dependent on the presence of the adjacent transmembrane helices 2 and 7, respectively. Numerous other factors, beyond amino acid sequence, may also contribute to the membrane topology of transmembrane helices (85–93).

Two recent studies have demonstrated an interaction between RocA and CovS proteins using coimmunoprecipitation methodologies with cell membrane preparations (59, 61). However, two-hybrid assays failed to demonstrate direct RocA-RocA or RocA-CovS interactions (61). Here, using a BACTH assay, we unambiguously demonstrated that RocA interacts with itself and CovS in intact cells (Fig. 2). In the previous study, Jain et al. were unable to demonstrate a direct RocA-RocA or RocA-CovS interaction in a BACTH assay. One possible reason is due to the use of N-terminal protein fusions (61). Given that the N terminus of RocA is extracytoplasmically located (Fig. 1), the N-terminal RocA-T18/T25 protein fusion would also be located extracytoplasmically, rendering a negative result regardless of any protein-protein interaction. Consistent with the membrane topology of RocA (Fig. 1), the C-terminal RocA-T18/T25 protein fusions used in our study demonstrated direct RocA-RocA and RocA-CovS interactions (Fig. 2). Understanding the topological architecture was crucial to making these discoveries and guiding downstream experiments using RocA variants with single amino acid replacements.

A majority of the CovR/CovS-T18/T25 protein fusion construct pairs demonstrated a positive interaction between CovS and CovS, CovR and CovR, and CovR and CovS (Fig. 2 and Fig. S7). However, only a C-terminal RocA-T18 protein fusion construct was able to directly interact with CovS (Fig. 2 and Fig. S7 and S8). The two plasmid backbones used for the assay differ in their replication origins and copy numbers (82). pUT18 and pUT18C have a ColE1 high-copy-number replication origin, whereas pKT25 and pKNT25 have a p15A low-copy-number replication origin (82). Direct interaction between RocA and CovS was observed only when RocA was expressed from a high-copy-number plasmid (pUT18) and CovS was expressed from a low-copy-number plasmid (pKT25/pKNT25; Fig. 2). Thus, we speculate that the RocA-CovS interaction only occurs when RocA is present in an excess of CovS, suggesting that RocA and CovS do not merely heterodimerize. That is, under normal conditions, RocA-CovS complexes form only when the stoichiometry of RocA to CovS favors RocA. This discovery can explain the apparent dosage effect of RocA observed in our study and others (Fig. S2 and S4) (56). A well-defined dosage effect is observed in other regulatory systems, such as the LiaFSR two-component system from Bacillus subtilis, where the accessory protein LiaF is present in excess of the histidine kinase LiaS (94), and in GAS, where the MtsR-PrsA-SpeB virulence axis depends on relative amounts of each protein present (13).

Previous studies have demonstrated that RocA is predicted to fold in a manner similar to histidine kinases; however, RocA does not possess histidine kinase activity (46, 52, 61). RocA modulates CovR phosphorylation by inhibiting the phosphatase activity of CovS (35, 62). Since RocA retains the ability to homodimerize like other histidine kinases (Fig. 2), but it does not possess enzymatic activity for two-component system signaling, there is likely a physiological role for homodimerization. Results from our two-hybrid protein interaction assay suggest that a specific stoichiometry of RocA to CovS influences protein-protein interaction. RocA could possibly interact with CovS in both monomeric and dimeric forms. Therefore, we speculate that RocA homodimerization adds an additional level of regulatory control. One or more, yet to be determined, factors could also directly or indirectly contribute to the RocA-CovRS virulence axis.

Different amino acid changes in RocA resulted in differentially altered RocA-RocA and RocA-CovS physical interactions. The RocA P97L and G184W single amino acid replacements are putatively located in transmembrane helices and generated different protein interaction profiles that corresponded to different virulence phenotypes as defined by secreted virulence factor activity and mortality in mice (Fig. 3 and 6) (25). The RocA G184W amino acid replacement resulted in a loss of interaction between RocA and RocA, but not between RocA and CovS (Fig. 3). A loss of RocA interaction with itself could increase the relative number of RocA G184W protein units available in the membrane to interact with CovS, resulting in increased CovR phosphorylation and decreased virulence factor gene expression and secreted activity in vitro (Fig. 6) (25). In comparison, the RocA P97L amino acid replacement resulted in a loss of interaction between RocA and RocA, and interaction between RocA P97L and CovS was only observed for one of the two plasmid construct pairs (Fig. 3). A possible explanation for an intermediate phenotype is that the RocA P97L-CovS interaction is more transient than that for the wild-type RocA protein. A transient interaction with CovS could explain the observed discrepant virulence phenotype of an isogenic GAS strain with the RocA P97L variant (25). RocA P97L results in transient decreases in CovR phosphorylation secondary to transient loss of CovS phosphatase activity inhibition, resulting in increased virulence factor gene expression and secreted activity in vitro but wild-type-like virulence in vivo. In support of this idea, a recent dual transcriptome sequencing study of GAS-infected nonhuman primate muscle found rocA transcripts to be significantly increased in vivo compared to in vitro conditions (3). Together, these findings suggest that a RocA-RocA interaction is necessary for normal RocA-mediated CovRS regulatory functions but not needed for an interaction with CovS. Further investigation is warranted to define the host factor(s) that possibly lead to increased rocA expression and/or stabilize the RocA-CovS physical interaction in vivo.

Previous research has suggested that the transmembrane domain of RocA, by itself, is sufficient, and the cytoplasmic domain of RocA is dispensable, for RocA-mediated CovRS regulatory functions (56, 59). Our previous study of the RocA T442P variant demonstrated that mutation to the supposed nonfunctional cytoplasmic domain results in a rocA deletion-like phenotype (25). Here, we studied naturally occurring amino acid changes in the C terminus of RocA (Fig. 4). We demonstrated that loss of the last nine amino acids (RocA Q443*) is sufficient to elicit a RocA deletion-like phenotype (Fig. 4 and 5), further suggesting a regulatory role for the RocA cytoplasmic domain. BACTH assays demonstrated that the RocA T442P and Q443* variant proteins could not interact with themselves (RocA) or CovS (Fig. 4), which would lead to decreased CovR phosphorylation and the resultant virulence phenotype for these variants (Fig. 4 and 5). Interestingly, the RocA T442I amino acid replacement also resulted in a loss of RocA-RocA and RocA-CovS interactions despite having a wild-type-like virulence phenotype (Fig. 4). Taken together, the data suggest that threonine 442, both the residue itself (T442P and T442I) and the relative positioning of this amino acid (T442I and Q443*), is crucial for a wild-type RocA-CovRS regulatory pathway. One possibility is that another, yet to be determined, factor in the RocA-CovRS virulence axis interacts in this region. Recent investigation of SpdC, an accessory protein to the WalKR two-component system in Staphylococcus aureus, demonstrated that SpdC interacts with several other two-component system proteins besides WalKR (95). In a similar manner, RocA may interact with other proteins besides CovS, and single amino acid replacements in RocA may alter these interactions. For example, the RocA T442P variant has significantly decreased expression of fasX, which may possibly be due to altered interaction with FasBCA or other proteins (25, 99). Further investigation is warranted to better define the global RocA protein interactome.

In summary, we demonstrate that RocA, a regulatory accessory protein to the virulence regulatory CovRS two-component system, is a seven-transmembrane helix protein that directly interacts with itself (RocA) and CovS. Single amino acid replacements in both the transmembrane and cytoplasmic domains result in altered RocA-RocA and/or RocA-CovS interactions that affect downstream gene expression, secreted virulence factor activity in vitro, and virulence in vivo (Fig. 6) (25). Additionally, based on studies of RocA C-terminal polymorphisms, other unidentified factors may directly or indirectly contribute to the RocA-CovRS virulence axis. In total, our study provides evidence that, under normal conditions, a full-length, wild-type RocA protein is needed to interact with itself (RocA) and CovS at a tightly regulated stoichiometry for wild-type gene regulation, and single amino acid replacements in a regulatory accessory protein can drastically alter the virulence of a major human pathogen.

MATERIALS AND METHODS

Strains and culture conditions.

We used serotype M28 GAS strain MGAS28426 because it is genetically representative of serotype M28 GAS strains circulating globally (27), it has a wild-type allele for all major virulence factors and transcription regulators, including rocA and covRS, and it has been used in multiple animal virulence models (25, 27, 57). Information about this strain and its derivatives is provided in Table S1 (25, 57, 78, 96). GAS strains were grown in Todd-Hewitt broth supplemented with 0.2% yeast extract (THY), with chloramphenicol (10 μg/ml) as needed. For standard cloning and PhoA-LacZα reporter fusion studies, we used E. coli strain DH5α. For bacterial adenylate cyclase-based two-hybrid (BACTH) assays, we used E. coli strains DHM1 and BTH101. E. coli strains were grown in LB broth at 37°C with agitation, with ampicillin (100 μg/ml) and/or kanamycin (50 μg/ml) as needed.

In silico modeling and prediction of RocA membrane topology.

The wild-type serotype M28 RocA protein sequence (GenBank accession number AAX72469.1) was used to model and predict the membrane topology of RocA. Multiple algorithms were used to determine a consensus topology (Fig. 1 and Fig. S1) that was used for codon selection in membrane topology experiments. The algorithms used included Phobius (http://phobius.sbc.su.se/) (63), Philius (http://www.yeastrc.org/philius/pages/philius/runPhilius.jsp) (64), OCTOPUS (http://octopus.cbr.su.se/) (65), PolyPhobius (http://phobius.sbc.su.se/poly.html) (66), SPOCTOPUS (http://octopus.cbr.su.se/index.php) (67), SCAMPI (http://topcons.cbr.su.se/) (68), MEMSAT-SVM (http://bioinf.cs.ucl.ac.uk/psipred/) (69), and Phyre2 (http://www.sbg.bio.ic.ac.uk/phyre2/html/page.cgi?id=index) (70). All membrane topology predictions were accessed on 17 January 2020.

Construction of cysteine-engineered RocA-FLAG-tag plasmids and strains.

Based on the predicted membrane topology of RocA (Fig. 1), cysteine residue codons were engineered into a wild-type rocA gene allele with a C-terminal FLAG-tag epitope on plasmid pDC123 expressed in the serotype M28 GAS isogenic ΔrocA deletion mutant strain (25, 57). Briefly, we introduced the FLAG-tag epitope and engineered cysteine residues into the wild-type rocA allele by PCR using primers FLAGRocA-F and the corresponding cysteine primer-R and FLAGRocA-R and the corresponding cysteine primer-F (Table S2) (25, 53). The two resulting fragments were fused together by combinatorial PCR using primers RocA-FLAG_fwd and RocA-FLAG_rev1. Plasmid pDC123 was amplified using primers pDC123_fwd1 and pDC123_rev, and the two resulting products were circularized using the NEBuilder HiFi DNA assembly kit (New England BioLabs, Ipswich, MA). The resulting plasmid was transformed into the serotype M28 GAS isogenic ΔrocA deletion mutant strain (25, 57). The genomes of all transformants were sequenced to verify correct insertion of the engineered cysteine residues and FLAG-tag epitope and absence of spurious mutations.

To ensure that the engineered cysteine residues and FLAG-tag epitope did not result in an altered growth or secreted NAD⁺-glycohydrolase (SPN) activity phenotype, we generated two additional plasmids as described above, pDC123, expressing a wild-type rocA allele, and pDC123, expressing a wild-type rocA allele with a C-terminal FLAG-tag epitope. These plasmids were transformed into the isogenic ΔrocA deletion mutant, and the resulting transformants were sequenced to verify correct plasmid and genome sequences.

Membrane topology studies: substituted-cysteine accessibility method as applied to transmembrane orientation (SCAM^TM).

The membrane topology of the N-terminal transmembrane helices of RocA was determined using SCAM^TM as previously described (71), with minor modifications. Strains carrying plasmids with the engineered cysteine residues and FLAG-tag epitope in rocA were grown to the mid-exponential growth phase, pelleted by centrifugation, washed with phosphate-buffered saline (PBS), and resuspended in buffer A (100 mM HEPES-KOH, pH 8.1, 250 mM sucrose, 25 mM MgCl₂, and 0.1 mM KCl). For labeling of accessible extracellular cysteine residues, cells were treated with 3-(N-maleimido-propionyl)biocytin (MPB; 100 μM) for 5 min, followed by quenching with 20 mM β-mercaptoethanol. Samples were then sonicated at an amplitude of 15% for 1 min. For labeling of accessible intracellular cysteine residues, cells were pretreated with nondetectable, water-soluble 4-acetamido-4′-maleimidylstilbene-2,2′-disulfonic acid (AMS; 5 mM) for 30 min in the dark with end-over-end rotation to block accessible extracellular cysteine residues. Cells were pelleted and washed twice with buffer A to remove excess AMS and then resuspended in buffer A. AMS-pretreated samples were then treated with MPB (100 μM) for 5 min in total as follows: 1 min with sonication at an amplitude of 15% and 4 min without sonication. Samples were quenched with 20 mM β-mercaptoethanol.

All samples were centrifuged (65,000 × g, 10 min, 4°C), and the resulting membranes were resuspended in buffer A with 20 mM β-mercaptoethanol. Solubilization buffer (50 mM Tris-HCl, pH 8.1, 2% SDS, and 0.1 mM EDTA) was added, and the samples were vortexed for 15 min at room temperature, incubated at 37°C for 15 min, and vortexed again for 15 min at room temperature. Samples were diluted with IP1 (50 mM Tris-HCl, pH 8.1, 0.15 M NaCl, 0.1 mM EDTA, 2% nonaethylene glycol monododecyl ether, and 0.4% SDS) and pelleted. The resulting supernatant was transferred to a spin column and incubated overnight with anti-FLAG-tag antibody (Abcam, Cambridge, MA; catalog number ab2493; 1:100) with rocking at 4°C.

Protein A/G-agarose affinity resin was added to the samples for a 90-min incubation at 4°C. The resin was washed with IP1, IP2 (50 mM Tris-HCl, pH 8.1, 1 M NaCl, 0.1 mM EDTA, 2% nonaethylene glycol monododecyl ether, and 0.4% SDS) and 10 mM Tris-HCl (pH 8.1). Protein from the samples was extracted in 2 × SDS sample buffer (vortexed for 15 min at room temperature, incubated at 37°C for 15 min, and vortexed again for 15 min at room temperature, followed by elution via centrifugation). Samples were subjected to SDS-PAGE, transferred to nitrocellulose membranes, and visualized with avidin-horseradish peroxidase (HRP) as described below. Regardless of stain intensity or protein amount, the qualitative presence of MPB labeling was interpreted as the cysteine being located extracellularly, and the qualitative presence of AMS labeling was interpreted as the cysteine being located intracellularly. Since interpretations are only made by comparing sample pairs, differences in expression and/or labeling efficiency between different engineered cysteine mutant proteins were not determined.

Western immunoblot analysis.

For determination of RocA-FLAG protein expression, whole-cell lysates prepared from cells at the mid-exponential growth phase were subjected to SDS-PAGE and transferred to nitrocellulose membranes. Membranes were blocked with 5% milk in PBS with 0.1% Tween 20 (PBS-T) for 1 h before incubation with anti-FLAG-tag antibody (Abcam, Cambridge, MA; catalog number ab2493; 1:5,000) for 2 h. Membranes were washed with PBS-T three times before detection using SuperSignal West Pico Plus chemiluminescent substrate (Thermo Fisher Scientific, Waltham, MA). For determination of MPB labeling, samples were subjected to SDS-PAGE and transferred to nitrocellulose membranes. Membranes were blocked with 5% bovine serum albumin (BSA) in Tris-buffered saline (TBS) overnight before incubation with avidin-HRP (1:5,000 in 0.3% BSA in TBS) for 2 h. Membranes were washed with 0.3% BSA in TBS twice, 0.5% IgePal CA-630 in TBS twice, and TBS once before detection using SuperSignal West Femto maximum sensitivity substrate (Thermo Fisher Scientific, Waltham, MA).

Membrane protein topology studies: PhoA-LacZα protein fusions.

The membrane topology of RocA was determined using a PhoA-LacZα reporter system (79). Reporter plasmid pKTop was used (78) to create RocA fragment PhoA-LacZα protein fusions after selected RocA amino acids (Fig. 1). The rocA fragments were PCR amplified from genomic DNA of strain MGAS28426 using the appropriate primers (Table S2), and pKTop was cleaved using BamHI. Both the linearized vector and PCR-amplified fragment were transformed into E. coli DH5α cells as previously described (97). Plasmids (Table S1) were verified for the correct insert sequence by Sanger sequencing (Table S2).

β-galactosidase and phosphatase assays were conducted using E. coli DH5α cells containing pKTop-RocA protein fusion plasmids as previously described (79), with minor modifications. Briefly, a single colony from an LB agar plate with kanamycin and 0.1% glucose was inoculated into LB broth with kanamycin and 0.1% glucose overnight. The overnight culture was diluted 1:100 into fresh LB broth with kanamycin and 0.1% glucose and grown to the mid-exponential growth phase. Isopropyl-β-d-1-thiogalactopyranoside (IPTG; 1 mM) was added, and cultures were incubated for 1 h.

For the β-galactosidase assay, cells were resuspended in M63 medium (100 mM potassium phosphate monobasic, 15 mM ammonium sulfate, 1.7 μM ferrous sulfate, and 1 mM magnesium sulfate, pH 7.0) and permeabilized with chloroform and 0.05% sodium dodecyl sulfate (SDS). Permeabilized cells were incubated with 0.15% O-nitrophenyl-β-galactoside (ONPG) in PM2 buffer (70 mM sodium phosphate dibasic, 30 mM sodium phosphate monobasic, 1 mM magnesium sulfate, 0.2 mM manganese sulfate, and 100 mM β-mercaptoethanol, pH 7.0) at 37°C for 1 h. Absorbance was measured, and β-galactosidase enzymatic activity was determined as previously described (79).

For the phosphatase assay, cells were washed with wash buffer (10 mM Tris-HCl, pH 8.0, and 10 mM magnesium sulfate), resuspended in PM1 buffer (1 M Tris-HCl, pH 8.0, 0.1 mM zinc chloride, and 1 mM iodoacetamide) and permeabilized with chloroform and 0.05% SDS. Permeabilized cells were incubated with 0.15% p-nitrophenyl phosphate (pNPP) in 1 M Tris-HCl, pH 8.0, at 37°C for 1 h. Absorbance was measured, and phosphatase enzymatic activity was determined as previously described (79).

To determine the localization of each protein fusion, the normalized activity ratio (NAR) was calculated as previously described (79). An NAR of >2 was considered highly likely to be extracellular, an NAR of <0.5 was considered highly likely to be intracellular, and an NAR between 0.5 and 2 was considered to be indeterminate, as is typically done (79).

Bacterial adenylate cyclase-based two-hybrid (BACTH) assays.

Interactions between RocA, CovR, and CovS were assessed using BACTH assays (80–82). Plasmids pKNT25 and pKT25, and pUT18 and pUT18C (Euromedex, Souffelweyersheim, France) were used for constructing T25 and T18 fusions, respectively. The rocA, covR, and covS alleles were PCR amplified from genomic DNA of serotype M28 GAS strain MGAS28426 or the respective isogenic rocA mutant strain (Table S1) using the appropriate primers (Table S2), and plasmids were cleaved using BamHI. Both the linearized vector and PCR-amplified fragment were transformed into E. coli DH5α cells as previously described (97). Plasmids (Table S1) were verified for the correct insert sequence by Sanger sequencing (Table S2). Once verified, plasmids were transformed into either DHM1 or BTH101 cells for the assay. PCR analysis was used to verify the correct plasmid backbones and inserts after transformation.

For quantification of the interaction between protein fusions, a β-galactosidase assay was performed as previously described (82). Briefly, a single colony from an LB agar plate with kanamycin, ampicillin, and 0.1% glucose was inoculated into LB broth with ampicillin, kanamycin, and 0.5 mM IPTG and grown overnight. Cultures were diluted 5-fold with M63 medium and permeabilized with chloroform and 0.05% SDS. Permeabilized cells were added to PM2 buffer containing 0.1% ONPG. The reaction mixture was incubated at 37°C for 30 min, absorbance was measured, and β-galactosidase enzymatic activity was determined as previously described (82). Plasmids containing GCN4 leucine zipper motifs (Euromedex, Souffelweyersheim, France) were used as a positive control, and empty plasmids were used as a negative control. Proteins were considered to have a positive interaction if the resulting β-galactosidase activity was at least five times higher than the negative controls (82).

Generation of isogenic C-terminal rocA polymorphism strains.

Isogenic strains containing C-terminal rocA polymorphisms (25, 57) were generated in the serotype M28 parental wild-type strain MGAS28426. The isogenic RocA T442P mutant strain has been previously described (25). The additional isogenic C-terminal rocA polymorphism strains (RocA V420I, T442I, and Q443*) were constructed by allelic exchange (the allele of interest was cloned from a clinical isolate with the naturally occurring rocA polymorphism) as previously described (22, 25). Primer sequences are listed in Table S2. Whole-genome sequence analysis of the isogenic C-terminal rocA mutant strains confirmed the expected rocA polymorphism and absence of spurious mutations.

In vitro virulence factor activity assays.

SPN and streptolysin O (SLO) activity were measured as previously described using mid-exponential growth phase culture supernatants (84). Streptokinase (SKA) activity was measured as previously described using mid-exponential growth phase cell-free culture supernatants (15, 25, 57).

Mouse model of necrotizing myositis.

Mouse necrotizing fasciitis/myositis studies were performed as previously described (13, 25, 57, 83, 84). Immunocompetent 4-week-old female CD1 mice (Envigo Laboratories, Houston, TX) were randomly assigned to treatment groups and inoculated in the right lower hind limb with 5 × 10⁸ CFU of the indicated bacterial strain suspended in PBS (n = 40 mice per strain). Mice were monitored at least once daily, and mortality was determined using the NIH Guide for the Care and Use of Laboratory Animals (98). Survival comparisons were made using the log-rank test. For gross and histologic evaluation, mice (n = 4 mice per strain per time point) were sacrificed on day 1 or day 2 postinoculation, and the limbs were processed using standard methods (13). All animal experiments were approved by the Institutional Animal Care and Use Committee of the Houston Methodist Research Institute (Houston, TX; AUP-0318-0016). Sample sizes were determined with a power calculation.

Statistical analyses.

All statistical analyses were performed using Prism 8 (GraphPad Software, Inc., La Jolla, CA) with three biological replicates, and P < 0.05 was considered statistically significant.