Identification of distinct impacts of CovS inactivation on the transcriptome of acapsular group A streptococci

ABSTRACT Group A streptococcal (GAS) strains causing severe, invasive infections often have mutations in the control of virulence two-component regulatory system (CovRS) which represses capsule production, and high-level capsule production is considered critical to the GAS hypervirulent phenotype. Additionally, based on studies in emm1 GAS, hyperencapsulation is thought to limit transmission of CovRS-mutated strains by reducing GAS adherence to mucosal surfaces. It has recently been identified that about 30% of invasive GAS strains lacks capsule, but there are limited data regarding the impact of CovS inactivation in such acapsular strains. Using publicly available complete genomes (n = 2,455) of invasive GAS strains, we identified similar rates of CovRS inactivation and limited evidence for transmission of CovRS-mutated isolates for both encapsulated and acapsular emm types. Relative to encapsulated GAS, CovS transcriptomes of the prevalent acapsular emm types emm28, emm87, and emm89 revealed unique impacts such as increased transcript levels of genes in the emm/mga region along with decreased transcript levels of pilus operon-encoding genes and the streptokinase-encoding gene ska. CovS inactivation in emm87 and emm89 strains, but not emm28, increased GAS survival in human blood. Moreover, CovS inactivation in acapsular GAS reduced adherence to host epithelial cells. These data suggest that the hypervirulence induced by CovS inactivation in acapsular GAS follows distinct pathways from the better studied encapsulated strains and that factors other than hyperencapsulation may account for the lack of transmission of CovRS-mutated strains. IMPORTANCE Devastating infections due to group A streptococci (GAS) tend to occur sporadically and are often caused by strains that contain mutations in the control of virulence regulatory system (CovRS). In well-studied emm1 GAS, the increased production of capsule induced by CovRS mutation is considered key to both hypervirulence and limited transmissibility by interfering with proteins that mediate attachment to eukaryotic cells. Herein, we show that the rates of covRS mutations and genetic clustering of CovRS-mutated isolates are independent of capsule status. Moreover, we found that CovS inactivation in multiple acapsular GAS emm types results in dramatically altered transcript levels of a diverse array of cell-surface protein-encoding genes and a unique transcriptome relative to encapsulated GAS. These data provide new insights into how a major human pathogen achieves hypervirulence and indicate that factors other than hyperencapsulation likely account for the sporadic nature of the severe GAS disease.

IMPORTANCE Devastating infections due to group A streptococci (GAS) tend to occur sporadically and are often caused by strains that contain mutations in the control of virulence regulatory system (CovRS). In well-studied emm1 GAS, the increased produc tion of capsule induced by CovRS mutation is considered key to both hypervirulence and limited transmissibility by interfering with proteins that mediate attachment to eukaryotic cells. Herein, we show that the rates of covRS mutations and genetic clustering of CovRS-mutated isolates are independent of capsule status. Moreover, we found that CovS inactivation in multiple acapsular GAS emm types results in dramatically altered transcript levels of a diverse array of cell-surface protein-encoding genes and a unique transcriptome relative to encapsulated GAS. These data provide new insights into how a major human pathogen achieves hypervirulence and indicate that factors other than hyperencapsulation likely account for the sporadic nature of the severe GAS disease.
KEYWORDS Streptococcus, CovRS, acapsular G roup A Streptococcus (GAS) is among the leading causes of invasive bacterial disease in humans and has long served as a model organism for investigating serious bacterial infections (1). GAS causes a wide array of infections, ranging from uncomplicated pharyngitis and cellulitis to life-threatening diseases such as necrotizing fasciitis and streptococcal toxic shock syndrome (2). In part, this variation in disease is due to inactivating mutations in the control of virulence sensor kinase (CovS) which impacts the phosphorylation state and, hence activity, of its cognate control of virulence response regulator (CovR) (3)(4)(5). Phosphorylated CovR (CovR~P) primarily represses virulence factor production, and inactivation of CovS decreases CovR~P level leading to increased expression of virulence factor-encoding genes and thereby hypervirulent GAS strains (3,(6)(7)(8).
The major typing scheme for GAS is based on the hypervariable 5′ sequence of the emm gene, which encodes the key GAS cell surface virulence factor, M protein (9). The vast majority of work on the control of virulence two-component regulatory system (CovRS) system has been in the emm1 pandemic clone M1T1 in which the antiphago cytic M protein and hyaluronic acid capsule have been shown to be necessary for the emergence of hypervirulent GAS (3,(10)(11)(12)(13)(14)(15)(16). CovR~P directly represses the hyaluronic acid capsule operon (3,17), and the hyperproduction of capsule in the M1T1 background induced by CovS inactivation inhibits GAS adherence to epithelial cells, which, in turn, is thought to limit transmissibility of CovS-inactivated strains (13,18). Indeed, multiple studies have found that CovS-inactivated strains generally tend to cause only one or very limited numbers of infections (19)(20)(21).
The GAS hyaluronic acid capsule has long been considered a key virulence factor because of its critical role in inhibiting phagocytosis, and CovS inactiva tion improves M1T1 GAS survival during interaction with neutrophils, a process considered critical to the emergence of CovS-mutated strains (10,(22)(23)(24)(25)(26). However, whole-genome sequencing has recently identified GAS emm types that either lack the hasABC operon that encodes capsule synthesis proteins (emm4, emm22) or have conserved mutations in hasA that abolish capsule production (emm28, emm87) (27)(28)(29)(30)(31). Similarly, unencapsulated emm89 strains have been replacing encapsulated emm89 isolates over the past decade in numerous countries (32)(33)(34). GAS strains are divided into emm patterns (A-C, D, and E) based on emm region content, and all currently known capsule-negative GAS emm types are pattern E (9,35). Together, these pattern E, unencapsulated strains accounted for ~33% of invasive GAS disease in the most recent Centers for Disease Control and Preven tion (CDC) invasive GAS surveillance report (36).
We previously identified the emergence of CovS inactivation during infection in an emm4 GAS strain and found that the acapsular, CovS-inactivated strain was hypervirulent (37). Given the critical role of capsule in the emergence of CovS-inactivated strains in the M1T1 background, it was somewhat surprising to observe that CovS-inactivated strains could be identified in acapsular emm4 strains (29,37). Inasmuch as emm4 strains are relatively rare causes of invasive GAS infections, we sought to characterize the impact of CovS inactivation in other acapsular emm types. We found that CovS inactivation in emm28, emm87, and emm89 strains resulted in a unique transcriptional impact leading to upregulation of emm and the multigene activator-encoding gene (mga) along with downregulation of the pilus-encoding operon and the gene-encoding streptokinase (SKA). In addition, CovS inactivation reduced the adherence of acapsular GAS strains to tonsillar epithelial cells, an observation that could potentially explain the rarity of genetic clusters of CovRS-mutated, acapsular isolates. These data provide a new mechanism by which a major human pathogen achieves hypervirulence and suggest that factors other than hyperencapsulation may limit transmission of hypervirulent GAS.

Rate of covRS polymorphisms predicted to alter CovRS function is similar in encapsulated and acapsular invasive GAS strains
To compare the rates of CovRS mutations in acapsular vs. encapsulated GAS, we downloaded 2,455 publicly available genomes from National Center for Biotechnology Information (NCBI) of invasive GAS strains collected by the CDC using active surveillance (accessed 4 July 2022) (36). We then analyzed the covRS region and identified genetic polymorphisms in terms of predicted impact (e.g., synonymous, nonsynonymous, nonsense, frameshift, etc.). The most commonly identified covR polymorphisms that impacted protein composition were missense (97%) and nonsense (3%) mutations, whereas for covS, 58% were nonsynonymous polymorphisms, 24% were frameshifts, and 18% were nonsense mutations. There was no significant difference in the types of polymorphisms observed between encapsulated and acapsular strains (P-value = 0.9 by χ 2 analysis). Among all emm types with at least 20 isolates, the highest rates of predicted functionally significant covRS polymorphisms were observed for emm28 (35.7%). The rates of covRS polymorphisms predicted to alter function in the five most common encapsulated emm types (emm1, emm49, emm12, emm3, and emm82) vs. the five most common acapsular emm types (emm89, emm28, emm4, emm77, and emm87) were not significantly different (P = 0.8 by Student's t-test). Similarly, no significant difference was observed when comparing the rates of covRS polymorphisms in encapsulated vs. acapsular strains (18% vs. 23%, P-value = 0.43 by χ 2 analysis). These data indicate that the presence/absence of capsule does not significantly impact the rates of covRS polymor phisms among invasive GAS.

Genetic clustering of CovRS-mutated, invasive strains is rare in both encapsu lated and acapsular emm types
It has been theorized that the hyperproduction of capsule due to CovS inactivation hinders transmission of hypervirulent, encapsulated GAS (13). Thus, one might hypothe size that CovS-inactivated, acapsular GAS would not have such a transmission defect and could cause outbreaks. Indeed, we have previously described a family cluster of invasive, CovS-inactivated, acapsular emm87 GAS (38). We overlaid our covRS polymor phism data onto the core genome phylogenetic trees of common encapsulated (emm1, emm3, and emm12) as well as acapsular (emm28, emm87, and emm89) emm types to search for instances where strains with the same covRS polymorphism clustered together (Fig. 1). For a given emm type, the vast majority of strains with a specific covRS polymor phism occurred only a single time (259/319, 81%) (Table S1). There was no significant difference between encapsulated and acapsular strains in the rates at which a specific polymorphism occurred more than once in the same emm type (Fisher's exact test P-value = 0.41). For both encapsulated and acapsular GAS, the only instances of genetic clustering of three or more strains occurred for isolates with missense mutations (i.e., not FIG 1 Occurrence and genetic clustering of covRS mutations in encapsulated emm1 and acapsular emm89 invasive strains. A recombination-free, core genome alignment inferred maximum-likelihood phylogenetic tree was created from alignment of (A) 857 emm1 strains and (B) 562 emm89 strains. The occurrence of mutations in covR and covS is indicated on the inner and outer circles, respectively, and color coordinated for each specific mutation with corresponding amino acid change as detailed in the key. For frameshift mutations, the first altered amino acid is indicated followed by the distance to the premature stop (e.g., V25SfsX8 implies a frameshift that alters residues from amino acid 25 and results in a premature stop eight residues downstream).

Research
Article mSystems frameshifts or nonsense) ( Fig. 1; Fig. S1). Conversely, strains with frameshifts or nonsense mutations, which almost always occurred in CovS rather than CovR, were genetically heterogenous with the largest cluster being two strains ( Fig. 1; Fig. S1). For each specific covRS polymorphism within a given emm type, genetic clustering rarely occurred (10% of cases) and was not significantly different between encapsulated (10%) and acapsular strains (10%), P-value = 1.0 by Fisher's exact test (Table S1). Taken together, we conclude that these data suggest no significant impact of capsule presence on the capacity of covRS-mutated strains to cause outbreaks.

Characterization of CovS-inactivated, acapsular emm28, emm87, and emm89 strains
To gain insights into how CovS functions in acapsular GAS, we chose to study emm28, emm87, and emm89 strains because each is a common GAS emm type and the emm types are genetically diverse relative to each other (39). Each of the chosen wild-type strains, TSPY902 (herein called emm28-WT), TSPY1057 (herein called emm87-WT), and MSPY1 (emm89-WT), has been fully sequenced and is wild type for all major transcriptional regulators (40). CovS mediates the response of emm1 GAS to the immune antimicrobial peptide LL37 by lowering CovR~P levels and, thus, relieving CovR repression of critical virulence factor-encoding genes (41,42). We confirmed the activity of CovS in our wild-type acapsular strains through exposure to LL-37 which resulted in decreased levels of CovR~P ( Fig. 2A), increased transcript levels of the CovR~P-repressed gene prtS, and decreased transcript levels of the CovR~P activated gene grab (Fig. 2B).
PrtS encodes an IL-8 degrading protease, whereas grab encodes an alpha2 macroglobu lin-like binding protein. The direction of these transcript level changes (i.e., increase for prtS and decrease for grab following LL-37 exposure) was the same as observed for the well-characterized M1T1 (emm1) strain MGAS2221 (herein called emm1) ( Fig. 2A and B). We used non-polar insertional inactivation to remove the entire covS open reading frame from the parental acapsular strains (43) ( Table 1). For all three emm types, CovS inactivation did not significantly impact growth in nutrient-rich media nor phenotype on blood agar. Consistent with observations in emm1 strains, CovS inactivation in the acapsular strains significantly reduced the amount of CovR~P (Fig. 2C). CovR~P levels were similar for all three CovS-inactivated strains. However, the amount of CovR~P in the wild-type emm28 strain was lower compared to the wild-type emm87 and emm89 strains, perhaps due to a conserved CovS E226G polymorphism present in all sequenced emm28 strains (44). As observed in emm1 strains, inactivation of CovS in the acapsular GAS strains resulted in increased transcript levels of prtS and slo genes ( Fig. 2D and E), which encode the pore-forming toxin streptolysin O (Slo).
The impact of CovS inactivation on GAS virulence is heterogenous with hypervir ulence consistently observed for emm1 strains (6,11,46,47), whereas studies in non-emm1 GAS have shown hypervirulence, hypovirulence, or no virulence impact depending on the strain and virulence assay used (18,22,48,49). We analyzed the virulence of the emm28, emm87, and emm89 strains and their respective CovS-inacti vated mutants in an ex vivo human blood model (6,(50)(51)(52). Similar to emm1 strains (6,41), CovS inactivation in the emm87 and emm89 background survived better in blood compared to their wild-type counterparts (Fig. 3). However, we did not observe a statistically significant difference following CovS inactivation in the emm28 strain because the multiplication value for the emm28-wt strain was higher than emm87-wt or emm-89 wt, perhaps due to the lower baseline CovR~P observed in the emm28-wt strain.

Characterization of the CovS transcriptome in acapsular GAS
To gain further understanding into the function of CovS in acapsular GAS, we performed RNA-seq on wild-type and CovS-inactivated strains grown to mid-expo nential phase for the three chosen acapsular GAS emm types. Principal component analysis for each set of wild-type and CovS-inactivated strains showed clustering of samples by specific strains (i.e., CovS inactivated vs. wild type) (Fig. S2). Significant differential gene expression (DGE) was defined as a difference of 1.5-fold between the wild-type and CovS-inactivated strains with an adjusted Wald Test P-value ≤0.05. We identified 306, 276, and 199 significantly differentially expressed genes in CovS-inactivated mutants compared to wild-type emm28, emm87, and emm89 strains, respectively ( Fig. 4A; Table S2). There were 69 genes with significant DGE in all three emm types and an additional 131 having significant DGE in two of the three emm types (Fig. 4A). Of these 69 core genes, transcript levels were significantly increased for 45 genes (65%) and decreased for 24 genes Research Article mSystems (35%) following CovS inactivation, consistent with CovR primarily functioning as a repressor (3,17). Genes whose transcript levels increased following CovS inactiva tion in all three emm types included prtS (53), mac-1, which encodes an immuno globulin degrading enzyme (54), sagA, which is the first gene in the streptolysin S-encoding operon (55), emm, and the transcriptional regulator spxA2 (56,57). Genes whose transcript levels consistently decreased following CovS inactivation included speC, which encodes a superantigen (58), the cyclic AMP factor encoded by cfa (59), and ska, which encodes the plasminogen-activating enzyme streptoki nase (60). Genes whose transcript levels were increased in two of three emm types included mga (61), rivR, which encodes a RofA-like transcriptional regulator (61), and genes in the streptolysin O-encoding operon (nga/slo) (62). Conversely, genes whose transcript levels were decreased in two of three emm types following CovS inactivation included those in the pilus (63) and the remainder of the streptoly sin S-encoding operon. Finally, we observed large tracts of downregulated genes uniquely present in the emm28 strain which included genes in the prophage region as well as the so-called region of difference 2 (RD2) which shares significant homology with GBS ( Mann-Whitney U test is indicated by * (P < 0.05) and ** (P < 0.0005). findings for three genes that are upregulated upon CovS inactivation and found consistent results (Fig. 4C).

Comparison of the acapsular GAS transcriptome with genes bound by CovR in emm1 GAS
To gain insight into potential mechanisms by which CovS impacts acapsular GAS gene expression, we compared our acapsular transcriptome analyses to our previously published CovR chromatin immunoprecipitation and DNA sequencing (ChIP-Seq) results and CovS transcriptome data from the emm1 strain MGAS2221 (6,17). The largest category of genes directly regulated by CovR in MGAS2221 is CovR~P repressed (i.e., increased transcript levels following CovS inactivation), and we observed very similar transcript level patterns for such genes in each of the acapsular strains (Fig. 5A). These included well-described CovR directly regulated virulence factor-encoding genes as prtS, sclA which encodes a cell-surface collagen-binding protein (65), and mac-1. Transcript levels of the hasABC operon, which is present in emm87 and emm28 strains but does not produce capsule due to point mutations (66), were elevated following CovS inactivation similar to what has been observed for emm1 (6, 12, 13, 67). The main exception to that is, CovR~P-repressed are indicated in green, and genes downregulated in ∆covS mutants, that is, CovR~P-activated are indicated in red. White indicates differential gene expression that was not statistically significant, and gray lines indicate the absence of a specific gene in a strain. (C) Taqman qRT-PCR analysis of select transcripts. Data shown represent means ± standard deviations of two biological replicates, with two technical replicates, done on two separate days.
Research Article mSystems this concordance was ska, whose transcript level is increased by CovS inactivation in emm1 GAS (6, 11, 67) but was decreased by CovS inactivation in each of the acapsu lar strains (Fig. 5A). Similarly, we observed similar transcript level changes for directly CovR~P activated genes from MGAS2221 (i.e., decreased transcript levels following CovS inactivation) for the acapsular strains (Fig. 5B). Of these genes, only braB, which encodes a putative branched-chain amino acid permease protein (68), showed increased transcript levels in CovS-inactivated emm28 but was unchanged in emm87 and emm89 ∆covS strains. Fig. 5C   Research Article mSystems significant transcript level variation following CovS inactivation in the emm1 strain but did have transcript level variation following CovS inactivation in one or more of the acapsular strains. These included such important GAS virulence factor-encoding genes as the streptolysin S operon, emm, and mga. Finally, Fig. 5D shows genes whose transcript levels were impacted by CovS inactivation in one or more acapsular strains but are either absent or not bound by CovR in strain MGAS2221. These included the pilus operon, enn and sof, which encode cell-surface proteins, as well as mf2, which encodes an actively secreted DNase cotranscribed with speC, as well as SpxA2, which encodes a transcriptional regulator previously shown to impact expression of CovR-regulated genes through an unclear mechanism (69). Of the 69 core genes in the acapsular CovS transcriptome, 62 are present in MGAS2221 and 32 of these (52%) were CovR bound in MGAS2221. Taken together, we conclude that direct CovR regulation likely accounts for a majority of the acapsular CovS transcriptome and that CovS inactivation results in key distinction between encapsulated emm1 strain MGAS2221 and acapsular GAS which will be further discussed below.

CovR directly represses ska in acapsular and encapsulated strains yet CovS inactivation leads to differential impact on ska transcript levels
The first area of distinction we sought to investigate was the differential impact of CovS inactivation on ska transcript levels between the emm1 and acapsular strains that were also observed upon LL37 treatment (Fig. 6A). There are two major variants of the ska gene with the acapsular strains all harboring pattern 1 and emm1 harboring pattern 2 (70,71). The ska promoters are approximately 90% conserved between the emm1 and acapsular strains with the major difference being a 21-bp insertion approximately 170 bps upstream of the ATG start codon and 120 bps from the major CovR binding site identified in vitro (72) (Fig. S3). To determine whether CovR directly repressed ska in the acapsular strains, we inactivated CovR in all three emm types as well as assayed for CovR-mediated enrichment at the ska promoter using chromatin immunoprecipita tion followed by qPCR (ChIP-qPCR). Similar to observations in emm1, CovR inactivation increased ska transcript levels in each of the acapsular strains, and we detected DNA enrichment of the ska promoter by ChIP-qPCR for both emm28 and emm89 strains (Fig.  6B). The ska transcript level pattern in acapsular strains (i.e., CovR repression but CovS activation) resembles those described for both speB and grab and has been postulated to be due to increased binding of nonphosphorylated relative to phosphorylated CovR and upon inactivation of CovS and CovR in emm1, emm28, emm87, and emm89 GAS strains. Data shown are mean ± standard deviation from two biological replicates, with two technical replicates, done on two separate days. Unpaired Student t-test was used to determine statistical significance relative to wild type (**, P < 0.005; ****, P < 0.0001). Binding of CovR~P to promoter of (B) ska and (C) 2221-0159 was analyzed by ChIP-qPCR in emm28 and emm89 wild-type and CovS-inactivated strains. Data shown are mean ± standard deviation of at least two biological replicates in duplicate.
Research Article mSystems (7,73). Thus, we next tested the hypothesis that lowering CovR~P levels via CovS inactivation would increase CovR-mediated enrichment at the ska promoter. Contrary to our hypothesis, CovS inactivation either decreased (emm28) or did not significantly change (emm89) CovR-mediated enrichment as measured using ChIP-qPCR. To ensure the validity of our ChIP experiment, we confirmed that CovS inactivation resulted in the expected decrease in CovR enrichment at the promoter of 2221_0159 for both emm28 and emm89 GAS (Fig. 6C) as has previously been published for emm1 (17). These data suggest that CovR directly represses the ska promoter in both encapsulated emm1 and acapsular strains but that factors beyond changes in CovR-DNA binding may account for the differential impact on ska transcript levels of altering CovR~P levels in encapsulated vs. acapsular strains.

The mga/emm locus is upregulated in acapsular CovS-inactivated strains
RNA-seq revealed that CovS inactivation significantly increased the transcript levels of emm in all three acapsular strains and of mga in emm28 and emm89 (Fig. 7A). These findings are in contrast to emm1 GAS in which CovR binds to the emm and mga promoters, but there is no impact of CovS inactivation on mga or emm transcript levels (17). The percent homology for the mga and emm promoter in the acapsular relative to emm1 GAS is ~60% and 75%, respectively, indicating potential variation in cis-regulatory elements between the acapsular and emm1 strains. We first confirmed our RNA-seq CovS inactivation data using qRT-PCR ( Fig. 7B and D) and also found that treatment of all three acapsular strains with LL-37 resulted in similar alteration of emm and mga transcript levels as observed for CovS inactivation (Fig. 7B and D). The lack of an increase in mga transcripts upon CovS inactivation or LL-37 treatment in the emm87 strain might result from a 10nt deletion in the mga promoter of emm87 that overlaps a putative CovR binding site as well as the -35 promoter element (Fig. S4). Inactivation of CovR significantly increased emm transcript levels for all three acapsular types and mga transcript levels for emm28 and emm89 ( Fig. 7B and D). Given that emm is directly regulated by Mga, the increase in emm, but not mga in the emm87 strain, suggested that CovR may be directly regulating emm in addition to mga. Consistent with this hypothesis, we found that CovR binds the promoters of both mga and emm as measured by ChIP-qPCR in the emm28 and emm89 acapsular strains ( Fig. 7C and E). Although we did observe a slight decrease in CovR-mediated DNA enrichment for the emm89 and emm28 strains at the emm and mga promoters following CovS inactivation, this difference was not statistically significant for either strain ( Fig. 7C and E). Thus, these data indicate that CovR directly regulates both emm and mga in the acapsular strains and that CovS inactivation increases emm (all three strains) and mga (emm28 and emm89) transcript levels but is not clear whether a decrease in CovR at the emm and mga promoter drives this variation.

CovS inactivation broadly increases transcript levels of cell-surface proteinencoding genes
Our transcriptome data indicated that CovS inactivation led to consistent upregulation of genes encoding the cell-surface proteins serum opacity factor (Sof ), which is a fibronectin-binding protein, and the fibrinogen-binding SfbX. These genes are encoded by a single operon and are absent in emm1 GAS (Fig. 8A). Alignment of the sof/sfbX region showed significant homology for the upstream and downstream genes between the acapsular and emm1 strains, suggesting either en bloc gene gain or gene loss leading to the observed arrangements. Divergently transcribed from sof is a gene previously called grm for gene regulated by Mga, which we have identified as directly regulated by CovR in emm1 GAS (17,74). The CovR binding site in the grm promoter is conserved between emm1 GAS and the acapsular strains, which could potentially place the sof/sfbX promoter under CovR control. The transcript level patterns induced by CovS inactivation of grm and the sof/sfbX operon are highly similar in all three acapsular strains (Fig.  8B), suggesting that CovR may be directly regulating both. We confirmed that CovS inactivation consistently upregulated sof and grm by qRT-PCR (Fig. 8C). Overall, there Research Article mSystems were 10 putative cell-surface protein (i.e., those that contain LPXTG motifs) encoding genes consistently upregulated following CovS inactivation in the acapsular strains (prtS, mac-1, sclA, sclB, fbpA, scpA, enn, emm, sfbX, and sof). We had previously reported an increase in the abundance of several of these cell-surface proteins in acapsular emm4∆covS strains accompanied by a rough cell surface covered by protruding material (37). Thus, we used electron microscopy to evaluate how CovS inactivation impacted the Student t-test was used to determine statistical significance relative to wild type (ns, not significant; *, P < 0.05; **, P < 0.005; ***, P < 0.0005; ****, P < 0.0001). Fold enrichment of the (C) mga and (E) emm promoter relative to ldh in emm28 and emm89 wild-type and CovS-inactivated strains analyzed by ChIP-qPCR. Data shown are mean ± standard deviation of at least two biological replicates in duplicate.

Research Article mSystems
cell surface of acapsular GAS and found a marked increase in cell-surface projections suggestive of increased amount of cell-surface proteins (Fig. 8D).

CovS inactivation decreases transcript levels of pilus operon-encoding genes and adherence to human tonsillar epithelial cells in acapsular GAS
The finding that genes of the pilus operon are downregulated by CovS inactivation in acapsular strains was intriguing given that pilus-related genes have not been observed as being impacted by CovS in encapsulated GAS, such as emm1 or emm3. The arrange ment of the fibronectin and collagen-binding proteins and trypsin-resistant antigens (FCT) region from emm28, emm87, and emm89 GAS (FCT4 variant) along with RNA-seq transcript level data is shown in Fig. 9A and B. We identified significant genetic hetero geneity among the FCT region for the acapsular strains except for the last two genes of the pilus operon, srtB and fctB, which encode a sortase and minor pilus subunit, respectively (Fig. 9A). Thus, we used fctB to measure pilus gene transcript levels, and consistent with the RNA-seq data, qRT-PCR showed significantly lower fctB transcript levels in all three emm types following CovS inactivation (Fig. 9C). Conversely, CovR inactivation either increased fctB transcript levels (emm28 and emm87) or caused no change (emm89) (Fig. 9C). The divergent impact of CovR vs. CovS inactivation on pilus operon transcript levels suggested that the impact of CovS inactivation on FCT region genes might be indirect. Consistent with this hypothesis, we did not observe significant CovR-mediated DNA enrichment of any of the promoters within the FCT locus. The GAS relative to wild type. Data shown are mean ± standard deviation from two biological replicates, with two technical replicates, done on two separate days. Unpaired Student t-test was used to determine statistical significance relative to wild type (*, P < 0.05; **, P < 0.005; ***, P < 0.0005; ****, P < 0.0001). (D) Adherence of wild-type (black symbols) and CovS-inactivated (red symbols) emm28, emm87, and emm89 strains to tonsillar epithelial cells. Statistical significance was determined by Mann-Whitney test (ns, not significant; ***, P < 0.0005; ****, P < 0.0001).
Research Article mSystems pilus is important for GAS adherence to host cells, including tonsillar epithelium (75,76). Thus, we analyzed the impact of CovS inactivation on the adherence of emm28, emm87, and emm89 strains to primary tonsillar epithelial cells (HTEpiC). For emm28 and emm87, CovS inactivation significantly reduced adherence (Fig. 9D). For emm89, the observed reduction in adherence between wild-type and the ∆covS strain was not statistically significant due to large intrastrain variation (Fig. 9D). These results are in accord with the relative magnitude of transcriptional changes that were observed for the FCT locus genes upon CovS inactivation (Fig. 9B).

DISCUSSION
The recent recognition that about 30% of GAS strains lack hyaluronic acid capsule was surprising given that capsule has long been identified as central to many critical aspects of GAS pathogenesis such as inhibition of phagocytosis (31,36,77). Moreover, immune avoidance resultant from increased capsule production arising from abrogated CovR repression is considered a major contributor to CovS-dependent GAS hypervirulence (10,78). In turn, this upregulation of capsule has been suggested to mask critical cell-surface adherence molecules necessary for the initial stages of GAS host-pathogen interaction, which provides a potential mechanism for the general lack of transmission of CovS-inactivated strains (13). Herein, we extended understanding of CovS inactivation to common, but relatively understudied, acapsular GAS emm types through a combination of genomic, transcriptomic, and functional studies that challenge current paradigms regarding the dynamics between CovS inactivation and capsule production and the impact of CovS inactivation on GAS pathophysiology.
A key finding of our work was that acapsular emm types causing invasive infec tions have similar rates of CovRS variation, including complete inactivation of CovS, relative to encapsulated strains. Relative to encapsulated GAS, one might have expec ted to observe less CovRS variation for acapsular strains given the lack of advantage induced by hyperproduction of capsule. These findings suggest that CovRS variation in invasive, acapsular GAS still confers a fitness advantage during host-pathogen interac tion, probably as a result of the increased production of a broad array of cell surface proteins and secreted toxins and are consistent with previous studies of CovS inactiva tion in emm4 and emm87 backgrounds (37,52). It was recently identified that the M protein from emm87 is able to bind and confer resistance to the important human antimicrobial peptide LL-37 (79) and improves survival in human blood (80) such that the upregulation of emm induced in acapsular strains following CovS inactivation may have similar immune avoidance properties as that conveyed by augmented capsule. Additionally, CovS inactivation significantly increased mga transcript levels which likely further increased the expression of Mga-activated genes encoding such key cell-surface proteins as M protein, C5a peptidase, and SclA (74). Interestingly, we identified direct CovR binding to the mga promoter in the acapsular strains as well as in emm1 GAS, but only in the acapsular strains did CovS inactivation increase mga transcript levels (17). Of note, hyperphosphorylation of CovR by abrogating CovS phosphatase activity diminished mga transcript levels in both emm1 and emm3 GAS consistent with CovR having some activity at the mga promoter even in encapsulated strains (6). Given the marked difference in mga promoter composition between emm1/3 and the acapsular strains studied herein, we hypothesize that variance in cis-regulatory elements may underlie differential impacts of varying CovR~P levels on mga transcript levels.
A second important aspect of our study was our identification of similar low rates of genetic clustering of invasive, covRS-mutated GAS strains irrespective of capsule status. These data are consistent with recent CDC publications that studied genetic clustering of either all invasive GAS (i.e., not just strains with CovRS mutations) (81) or CovS-inac tivated strains in Colorado (21) which did not identify any particular predilection for clustering based on capsule status. Previous genomic studies have found that most CovRS-altered strains cause a single invasive infection suggesting limited transmissibility (19)(20)(21)52). For emm1 GAS, it was shown that hyperproduction of capsule reduces adherence to epithelial cells, thus rendering a fitness cost for CovRS inactivation (13). If such a hypothesis is correct, then one might have expected to observe clusters of invasive infections due to genetically related acapsular strains with the same CovRS variation. Our finding of increased multiplication and decreased adherence following CovS inactivation is consistent with previous data from members of our group studying CovS inactivation in a different emm87 strain (52). The identification that CovS inacti vation resulted in a decrease in pilus operon transcript levels provides a mechanism for the reduced adherence observed in both studies as well as a parallel pathway for limiting transmission of acapsular GAS analogous to augmented capsule production in encapsulated GAS. The precise role of the pilus in GAS pathogenesis has been difficult to clearly delineate due to variation in both pilus content and pilus expression (75,82). An overall theme, however, has emerged where pilus production positively impacts the establishment of GAS colonization at mucosal surfaces through augmented attachment to epithelial cells but negatively impacts invasive GAS disease by facilitating neutro phil-mediated killing (76,83). Thus, the decreased pilus transcript levels resulting from CovS inactivation may positively impact acapsular GAS interaction with human immune components but decrease the subsequent transmission of CovS-inactivated isolates. Alternative or additional explanations for the lack of transmission of CovS-inactivated strains have also been proposed (18) including poor growth under nutrient-limited conditions such as human saliva (12).
The consistent decrease in ska transcript levels following CovS inactivation in the acapsular strains stands in stark contrast to the increase in ska expression observed in CovS-mutated, encapsulated GAS (6,11,46). Given that streptokinase converts human plasminogen to plasmin which subsequently degrades extracellular matrix components, an increase in SKA activity is considered a key mechanism by which CovS inactivation augments GAS invasiveness (16). Bernard et al. previously noted that inactivation of the regulator of CovR (RocA), which diminishes CovR~P levels, also reduced ska transcript and SKA activity in emm28 strains, although they were unable to determine a mechanism (84). Herein, we show that ska transcript levels are reduced by CovS inactivation in all three studied acapsular emm types but that CovR remains a direct repressor of ska expression in emm1 and acapsular GAS. Thus, for acapsular GAS, the ska gene is regulated in a fashion similar to grab, which is directly repressed by CovR yet shows reduced transcript levels following CovS inactivation. Given that we did not observe increased CovR-mediated enrichment of ska promoter DNA in CovS-inactivated strains, the mechanism of the varied impact of CovS inactivation on ska transcript levels in encapsulated vs. acapsular strains remains enigmatic. Additionally, it is not clear what evolutionary advantage of decreased SKA activity would be engendered specifically in acapsular CovS-inactivated strains, but further study of this difference may shed new light on GAS pathogenesis.
Our finding that the emm28 strain had lower baseline CovR~P levels relative to emm87, emm89, and emm1 strains adds to growing information regarding the appa rently atypical nature of the CovRS system in this particular emm type (44). Bernard et al. noted that emm28 strains have an unusually high number of RocA polymorphisms relative to other emm types and demonstrated a dramatic impact of RocA on the emm28 transcriptome (85). Moreover, an analysis of over 2,000 emm28 strains identi fied numerous CovRS polymorphisms that were associated with a markedly altered transcriptome and increased virulence in a mouse necrotizing fasciitis model (44). In concert with the Bernard et al. study, we found that emm28 strains had the highest rates of CovRS polymorphisms (~35%) for invasive strains collected by the CDC and that the emm28 CovS transcriptome was larger than either the emm87 or the emm89 strains. This finding is somewhat counterintuitive given that the impact of CovS inactivation on GAS gene expression is thought to occur by decreasing CovR~P levels (4) and that the decrease in CovR~P following CovS inactivation was least for the emm28 strain. Indeed, for CovR~P-repressed genes such as prtS and sclA, the transcript level increases were lowest following CovS inactivation in emm28 compared with emm87 and emm89.
However, we identified large tracts of downregulated genes in the emm28ΔcovS strain which were present in gene regions unique to emm28, including RD2. RD2 was likely acquired from GBS, contains known and putative transcriptional regulators, and has previously been shown to alter the GAS transcriptome (86). In GBS, CovR serves to silence expression of recently acquired DNA (87), and we hypothesize that CovRS influence on RD2 expression may contribute, in part, to the magnitude of the CovS transcriptome in GAS, much of which is likely indirect in nature given the large number of contiguous genes with similar transcript patterns despite being in distinct operons.
In conclusion, we present herein the most comprehensive analysis to date of CovS inactivation in acapsular GAS. We found that CovRS variation is common among acapsular GAS but that similar to encapsulated GAS, transmission of CovRS-mutated, acapsular GAS appears to be relatively rare, perhaps due to unique downregulation of pilus components induced by CovS inactivation. Our data suggest that CovS inactivation likely impacts virulence of acapsular strains through different mechanisms given the unique differences in gene regulation observed in acapsular ∆covS GAS strains. Further study of how acapsular GAS achieves hypervirulence in the absence of capsule may provide novel insights into GAS pathogenesis.

Genome analysis
Paired-end short-read data were retrieved from NCBI using the sra-toolkit-v3.0.0 fasterq-dump script. Short-reads were then aligned to reference capsular (RefSeq #: emm1: NZ_CP043530.1, emm3: NZ_CP033815.1, and emm12: NZ_CP009612.1) and acapsular (emm28: NC_007296.2, emm87: NZ_CP007560.1, and emm89: NZ_CP013839.1) GAS complete genomes using the snippy-v4.6.0 (Seemann T, snippy, GitHub: https:// github.com/tseemann/snippy) core SNP phylogeny pipeline. A core SNP alignment was then used as input to Gubbins-v3.2.1 (88), which can estimate regions of high recom bination through measuring variance in SNP densities. A core SNP inferred maximumlikelihood phylogeny was built downstream with IQTree-v2.0.3 (89) as the Gubbins tree-model option using a generalized-time reversible nucleotide substitution model and gamma distribution to model rate heterogeneity. Branch values were tested using UFBoot (n = 1,000) and SH-Test (n = 1,000) Variants of covR and covS from the Snippy-core pipeline were extracted and merged using bcftools-1.15. The R library VariantAnnotation-1.42.1 was used to annotate and predict coding changes from reference genomes. A custom R script was adapted to annotate and extract variants using the above R library (S Selvalakshmi, VariantAnno tator, GitHub: https://github.com/Selvalakshmi27/VariantAnnotator). The pipeline uses merged vcf file output from bcftools and a reference genbank file as inputs. In case of a gff3 file as an input reference file, agat_convert_sp_gxf2gxf.pl script from agat-1.0.0 was used to index and sort gff3 files before using the VariantAnnotator pipeline. The output is a csv file containing the predicted coding changes of the variants along with the consequences of the change (synonymous/nonsynonymous). Nonsynonymous covR and covS mutations that occurred in two or greater isolates of each respective emm type were mapped onto respective phylogenies and visualized using iTOL-v6 (90).

Bacterial strains, growth conditions, and mutant generation
Strains were grown in a nutrient-rich standard laboratory medium [Todd-Hewitt broth with 0.2% yeast extract (THY)] at 37°C with 5% CO 2 . Phenotypic characterization was performed following growth on sheep blood agar plates. Nonpolar insertional mutagen esis with a spectinomycin resistance cassette was employed to obtain isogenic covS and covR mutants in representative strain of emm28, emm87, and emm89 GAS as described previously (Table 1) (39,43). The emm1 MGAS2221 and its isogenic ∆covS and ∆covR mutants has been previously described (12).

Detection of CovR phosphorylation status
Recombinant CovR was produced and phosphorylated as described previously (6). GAS lysates were prepared from mid-exponential cultures, separated on 10% Phos-Tag SDS-polyacrylamide gels, and unphosphorylated/phosphorylated CovR proteins were detected by using a polyclonal anti-CovR antibody as described previously (6) using an Odyssey imaging system.

Transcript level analysis and RNA-seq
RNA was purified from various GAS strains grown to mid-exponential phase (OD 600~0 .5), and TaqMan real-time qRT-PCR was performed as described previously (6,91,92) (primers and probes listed in Table S3). All strains were collected in duplicate on at least two separate occasions and analyzed in duplicate.
For RNA-seq analysis, strains were grown on two different days to mid-exponen tial phase in THY, and RNA was isolated as for TaqMan qRT-PCR. RNA library prepara tion followed by RNA-seq was performed by the MD Anderson core genome center to generate raw paired-end RNA reads (75 × 2). RNA-seq quality control was done on untrimmed reads using fastqc-v0.11.8. Untrimmed QC'd reads were aligned to each respective complete assembly [NC_007296 (Streptococcus pyogenes MGAS6180) for emm28 strain TSPY902, NZ_CP007560 (S. pyogenes NGAS743) for emm87 strains TSPY1057, and CP013840.1 (S. pyogenes MGAS27061) for emm89 strain MSPY1] using the STAR aligner (93). The HTSeq python package (94) was used for enumerating read counts of uniquely mapped reads using the *Aligned.sortedByCoord.out.bam STAR output file along with a reference gtf file. RNA-seq analysis was performed as described previously (37,91,95). Briefly, differential expression between wild-type and CovS mutant strains was performed using the DESeq2 package-v1.30.1 (96). Read counts from HTSeq were normalized by estimating size factors using the median ratio method employed in estimateSizeFactors function. Genes with less than one per million normalized reads mapped for all samples were excluded from downstream analyses. The DESeq2::nbinom WaldTest function is used to test for differential expression, employing a Wald test that tests the significance of coefficients from a negative binomial generalized linear model fit to each gene and is corrected for multiple comparisons using a false discovery rate. A fold change of 1.5× with an alpha parameter = 0.05 was used for these calculations. A principal component analysis was used to visualize the clustering of expression profiles by sample.

Growth in human blood
Lancefield assays were performed as previously described (97) under a protocol approved by the Committee for the Protection of Human Subjects at the McGovern Medical School at UTHealth Houston. Blood samples from four healthy, nonimmune, adult donors were used for each strain, and assay was performed in triplicate. Strains were grown to mid-exponential phase (OD 600 ~ 0.5) in THY, and cells were pelleted and resuspended in phosphate-buffered saline (PBS). To assay growth, ~100 CFUs of each strain were added to 300 µL of blood, incubated for 3 h, and dilutions plated on blood agar plates for enumeration. Multiplication factors were calculated by dividing the number of CFU per milliliter after 3 h of incubation by the initial inoculum.

Chromatin immunoprecipitation and SYBR qRT-PCR
Chromatin immunoprecipitation was performed for emm28 and emm89 strains as described previously (17,92). Briefly, GAS strains were grown in THY medium to mid-exponential phase, proteins were cross-linked to DNA, and cells were subsequently harvested. Cell pellets were resuspended in lysis buffer, sonicated in a Diagenode Bioruptor Plus machine, and CovR-bound DNA fragments were immunoprecipitated using a polyclonal antibody directed against the N-terminal domain of CovR (CovR ND ) (17). Both ChIP and input DNA were purified. Enrichment of selected promoters in ChIP Research Article mSystems samples derived from wild-type and CovS-inactivated strains relative to input DNA was assessed by SYBR qRT-PCR as described previously (17) using primers listed in Table S3. The fold enrichment of promoters of interest was normalized to that of the ldh promoter region (not CovR regulated). Measurements were done in duplicate on at least three biological (n = 3) samples.

Microscopy
The cell surface morphology of wild-type and CovS-inactivated strains was analyzed using transmission electron microscopy. Cells were prepared, fixed, and sectioned as described previously (37). Images were obtained using a Jeol 1200 transmission electron microscope with a Gatan digital camera.

Adherence assay
Adherence of wild-type and CovS-inactivated strains to human tonsillar epithelial cells was assayed as described previously (52). Briefly, mid-exponential GAS cells (OD 600 ~ 0.5) were washed, resuspended in PBS, and used to infect wells seeded with 0.5-1 × 10 6 epithelial cells at an MOI of 10 in 4 technical replicate wells. Plates were incubated at 37°C for 90 min, washed, and lysed with Trypsin. Dilutions were plated on blood agar plates to enumerate, and percent adherence was calculated based on CFUs recovered on blood agar plates. The experiment was repeated on four different days.

DATA AVAILABILITY STATEMENT
The RNAseq data has been submitted to the GEO depository (GSE230158) and will be publicly available.

ADDITIONAL FILES
The following material is available online.