Sequenced M1 strains SF370 (Ferretti et al., 2001), MGAS2221 (Sumby et al., 2006), MGAS5005 (Sumby et al., 2006), and 5448 (Walker et al., 2007), and M3 strain MGAS315 (Beres et al., 2002) were used in this study. These GAS strains and their derivatives were grown in Todd-Hewitt broth supplemented with 0.2% yeast extract (THY) at 37°C in 5% CO₂. Tryptose agar with 5% sheep blood and THY agar were used as solid media.

MGAS2221 was cultured in THY starting with an optical density at 600 nm (OD₆₀₀) of 0.05 at 37°C in 5% CO₂. The continued culture was done by adding 1 mL of the prior culture to 11 mL of THY and grown for 12 h. This serial passage continuous culturing process was repeated for 15 cycels. Following the 15th passage bacteria were harvested at an OD₆₀₀ of 0.3, diluted and plated on THY agar. Randomly picked colonies were streaked on plates, grown to mid-exponential growth, and stored frozen until analysis for detection of the M protein by Western blotting analysis. Gene deletion mutants of SF370, MGAS2221, MGAS5005, and MGAS315 were obtained in a two-step in-frame gene deletion procedure as described (Zhu et al., 2009). The step in which mutants might lose M protein production was the growth of isolates from the first crossover in THY for >8 passages (one passage = 0.05 to 0.7 in OD₆₀₀) that allowed the second crossover event to occur for the generation of gene deletion mutants.

The cell surface M protein of M1 GAS isolates was detected by Western blotting, as previously reported (Zhou et al., 2013). Briefly, bacteria in 10 mL culture with an OD₆₀₀ of 0.2 were harvested by centrifugation, suspended in 0.2 mL phosphate-buffered saline (PBS), and digested with 10 μg PlyC (Nelson et al., 2006) at room temperature for 1 h. The supernatants containing the released cell wall proteins were diluted with 1:60 1x SDS-PAGE loading buffer, boiled, and 10 μL of each sample was resolved by SDS-PAGE. Proteins were transferred from SDS-PAGE gel to nitrocellulose membrane (Immobilon-NC, Millipore Corporation) with Towbin transfer buffer using a Trans-Blot Semi-Dry Transfer Cell (Bio-Rad Laboratories) at 15 V for 40 min. The membrane was blocked with 3% bovine serum albumin–0.1% Tween 20 in 20 mM Tris–HCl, pH 8.0, for 1 h and incubated for 1 h with 1:2000 anti-M1 antisera that was kindly provided by Dr. James Dale at University of Tennessee Health Science Center. The membrane was then rinsed twice and washed three times for 15 min with 0.1% Tween 20 in PBS. The membrane was incubated with goat anti-rabbit IgG (heavy + light chain) horseradish peroxidase (HRP) conjugate secondary antibodies for 1 h and rinsed and washed as described above. Antigen–antibody reactivity was visualized by enhanced chemiluminescence. Western blots usually shows two bands for M protein and the higher MW band is M protein with attached peptidoglycan fragments.

Data for emm mRNA levels in Supplementary Table S1 were associated with our previous publications (Zhu et al., 2009; Liu et al., 2012; Li et al., 2013; Zhou et al., 2013; Liu et al., 2015; Stetzner et al., 2015; Feng et al., 2017). Levels of emm and gyrA (control) mRNA were measured by using TaqMan quantitative RT-PCR assays with specific probes and primers as reported previously (Zhou et al., 2013) or using the All-in-One SYBR qPCR mix from GeneCopoeia (Stetzner et al., 2015). The transcription data for MGAS2221, GAS1806, M405, M406, and M497 were measured using the TaqMan quantitative RT-PCR as described by Zhou et al. (2013). The data were for early exponential growth phase at optical density 0.2 of GAS in THY. Control reactions that did not contain reverse transcriptase revealed no contamination of genomic DNA in any RNA sample. All RNA samples were assayed in triplicate, and the levels of emm mRNA were compared using the ΔΔCT method with normalization to the mRNA levels of the gyrA gene and presented with normalization to that of corresponding wild-type or control strain.

The mga gene of MGAS2221 was cloned into pDCBB-RFA (Liu et al., 2013) with the Gateway Cloning Technology according to the manufacturer’s manual. pDCBB (Treviño et al., 2009) was modified by inserting the blunt-ended reading frame cassette A (RFA) into pDCBB at the EcoRV site, resulting in pDCBB-RFA (Liu et al., 2013). The mga gene was PCR amplified using Phusion DNA polymerase (New England BioLabs, Ipswich, MA) and the primers 5′- GGGGACAAGTTTGTACAAAAAAGCAGGCTagaagggtatacaaggtaatg-3′ and 5’-GGGGACCACTTTGTACAAGAAAGCTGGGTgtttttgagttgctacagtta-3′. The sequences in the capital letters were attB sequences for the BP clonase reaction. The PCR product was cloned into the donor vector pDONR221 using the BP clonase, yielding pDONR221-mga. The mga gene in pDONR221 was transferred into pDCBB-RFA by the LR clonase, yielding pDCBB-mga. pDCBB-mga or vector control pDCBB was introduced into M^- variants via electroporation.

DNA sequencing of amplified PCR products was performed using the BigDye Terminator v3.1 Cycle Sequencing Kit and an Applied Biosystems 3130 genetic analyzer. Sequence data were analyzed using the software Sequencer 5.1 from the Gene Codes Corporation. To sequence the whole mga gene, a 2,454-bp fragment containing mga was amplified using Phusion DNA polymerase and the paired primers 5′-GTTGTACCATAACAGTCAAAC-3′/5′-TTTCAAGTTCTTCAGCTCTC-3′. The primers used for sequencing were the two PCR primers and additional primers, 5′-AACGAATCAAGTTAACTGAGC-3′, 5′-TCCTAAACTTAAAGAACTGTG-3′, 5′-TGTCACGATCACATCATACTG-3′, and 5′-TTTAACAGTGTTGGTAATTTC-3′. To sequence the c.1574C[8] region of M1 and M1T1 GAS isolates, a 469-bp fragment was amplified and sequenced using the primers 5′-TTTAAACATCAGCTTTGCAGA-3′ and 5′-TCTTCTATAACTTCCCTAGGA-3′. To sequence the c.1592C[8] region of M3 GAS isolates, a 457-bp fragment was amplified and sequenced using the primers 5′–TTTTTAAACATCAGCTCTGCAGA–3′ and 5′–CATTAACACTCCTAGCATCTG–3′.

M^- variants were grown in THY, harvested at the mid-exponential growth phase (OD₆₀₀ of 0.4) and washed three times with and resuspended in pyrogen-free Dulbecco’s phosphate-buffered saline (DPBS). Five 5-weeks-old female C57BL/6 J mice were injected subcutaneously with 0.2 mL of GAS suspension in DPBS with an OD₆₀₀ of 0.8. The mice were euthanized with CO₂ at day 4 after inoculation. Skin infection sites were collected and homogenized in DPBS using a Kontes pestle, and plated at appropriate dilutions. Randomly picked 70 colonies were grown and analyzed for detection of M protein production using the western blotting analysis as described above. The mouse experimental procedures were carried out in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Research Council (2011).

The DNA sequences of the mga gene and its protein sequences were analyzed for M1, M3, M12, M14, M23, and M49. Complete and incomplete GAS genomes as of October 10, 2019 were downloaded from the NCBI GenBank database.¹ M type of GAS strains was determined by aligning their sequences with the emm sequences in an emm genotype database.² Because there were no polymorphisms in the length of Mga-2 of M49 GAS, we just listed all of 158 M1, 106 M3, 106 M12, 6 M14, and 2 M23 strains that were available at the time for analysis (Supplementary Tables S2–S4 and Table 1). The nucleotides of mga and emm in each genome were extracted by mapping to the mga and emm locus using blast (Camacho et al., 2009). Multiple sequence alignments of mga DNA sequences and Mga protein sequences were performed using ClustalO (Sievers et al., 2011).

Reverse genetic analysis is a commonly used approach to define the function and phenotype of target genes. We use a two-step strategy to delete genes in-frame in GAS (Zhu et al., 2009). A suicide plasmid is constructed to contain the upstream and downstream flanking fragments of target gene that are joined to each other. The first step involves the first homologous crossover between one flanking fragment in the suicide plasmid and GAS chromosome, yielding chloramphenicol-resistant merodiploid transformants. The second step passes one transformant in liquid THY medium or on THY agar plate without chloramphenicol selection for more than 8 times to allow the second crossover between the other flanking fragment of the integrated suicide plasmid and its homologous sequence of GAS chromosome, resulting in gene deletion mutants. During genetic manipulation or any process that entails bacterial DNA replication, spontaneous random mutations can occur. To rule out the presence of a potential spurious mutation grossly altering the expression of known major virulence factors, western blotting analysis and/or qRT-PCR analyses were performed to check M protein production and the expression of M protein gene, hasA, spyCEP, and sse genes of gene deletion mutants. We have previously shown that genetic manipulation of GAS to construct gene deletion mutants can independently of the gene being targeted lead to the spontaneous loss of emm gene expression and M protein production (Zhou et al., 2013). To evaluate if loss of M protein production during such processes is a general phenomenon common to many GAS strains of M1 and other M types, we analyzed gene deletion mutants of M1 strains SF370, MGAS2221, MGAS5005, and 5,448 by the western blotting analysis and gene deletion mutants of M3 strain MGAS315. As shown in Figure 1, independent of the gene targeted for deletion, gene deletion mutants of M1 GAS strains MGAS2221, MGAS5005, and SF370 could lose M protein production. In total, 43 of 92 gene deletion mutants of M1 GAS strains MGAS5005, MGAS2221, 5448, and SF370 lost M protein production (Supplementary Table S1). As listed in Supplementary Table S1, 13 M⁺ and 11 M^- gene deletion mutants of them had relative mRNA levels of the emm gene that were obtained for our previous publications (Zhu et al., 2009; Liu et al., 2012; Li et al., 2013; Zhou et al., 2013; Liu et al., 2015; Stetzner et al., 2015; Feng et al., 2017). The emm mRNA levels in 9 of the 10 M^- gene deletion mutants were 1 to 5.5% of those in their parent strains whereas 1 M⁻, GAS1778 (5448Δsda1), and all the 11 M⁺ deletion mutants has similar levels of emm mRNA with those in their parent strain. Two of 9 gene deletion mutants of M3 strains MGAS315, GAS1191 and GAS1028, down-regulated emm3 transcription by about 99%, and western blotting analysis was not performed for the M3 gene deletion mutants because anti-M3 protein antibody was not available (Supplementary Table S1).

The second crossover to achieve gene deletion occurred during growing first-crossover tranconjugants in THY medium or on THY agar plates. The 44 M^- gene deletion mutants were all among 70 mutants that were obtained by using THY medium for the second crossover whereas all 31 gene deletion mutants that were obtained by using THY agar were all positive in M protein production (Supplementary Table S1). These data suggest that M^- variants have an advantage over M⁺ GAS to survive in THY.

To determine whether M^- variants of M1 GAS arise during culture ex-vivo in THY, strain MGAS2221 was passed in THY twice a day for 7 days, randomly chosen colonies after the last passage were analyzed for M protein production using western blotting analysis. Fifty colonies from the MGAS2221 culture after 14 passes were tested for M protein production, and 30 of them had no detectable M protein by Western blotting analysis. Figure 1B shows a representative Western blot in the analysis. Thus, M⁻ M1 GAS variants arise during in vitro growth, a phenomenon similar to that described for M12 GAS strains (Simpson and Cleary, 1987).

Most of 10 M^- variants with emm transcript data had dramatic downregulation in transcription of the emm gene (Supplementary Table S1). Since Mga is the primary regulator of emm expression any polymorphism in the mga gene or the promoter region of mga and/or emm that reduces the capacity of Mga to activate emm transcription could contribute to loss of the M protein expression. To look for such polymorphisms, a PCR amplicon containing the mga gene and its 379-bp upstream and 470-bp downstream sequences from one M⁻ variant and its parent strain MGAS2221 were sequenced by the Sanger DNA sequencing. A single C deletion was found in the polycytidine tract that starts at base 1,571 and has 8 repeats of C near the end of the mga gene of M1 GAS (Figure 2A). This polycytidine tract of wild-type mga is designated as c.1574C[8] in which 1,574 and 8 represent the starting base of the polycytidine tract and the number of the C repeats, respectively. The 1C deletion in the c.1571C[8] tract leads to the c.1571C[7] mga variant. The c.1571C[7] mga variant gene in M1 GAS has an altered reading frame of the mga gene, and the mutated mga gene is read through the intergenic region between the mga and emm1 genes and the emm1 gene, leading to a protein variant that contains the amino acid sequence of Mga, 62 amino acid residues encoded by the intergenic sequence, and M1 Protein. Transformation of the c.1571C[7] mga variant with plasmid pDCBB-mga, but not the plasmid vector control, restored the M protein production (Figure 2B).

To determine whether the c.1571C[7] mga variant was prevalent among our M⁻ strains, a 820-bp DNA fragment covering 349-bp of 3′ part of mga, 184-bp mga/emm intergenic region, and 287-bp of the 5′ part of emm was sequenced for 37 gene deletion mutants with diminished M protein production. The results are presented in Supplementary Table S1. 37 of the 43 M⁻ mutants of M1 strains MGAS2221, MGAS5005, 5,448, or SF370 had the c.1571C[7] mga variant whereas 6 mutants had the c.1574C[8] tract of mga. One of the 5 M^- mutants with the c.1571C[8] mga, GAS1099, had A-to-G mutation at base 74 upstream of emm that is referred as -74A > G of emm. This mutation changed the-10 box TACAAT of the emm promoter to TGCAAT and had reduction of emm mRNA by 97.9%. The whole mga gene of the 5 remaining M^- variants with c.1571C[8] of mga was sequenced. GAS874 had missense mutation of A to C at base 292, c.292A > C, leading to the threonine-to-proline mutation at residue 98 of the Mga protein, Mga T98P, and had reduction of emm mRNA levels by 94.5%. The other 4 M^- variants, GAS1499, GAS1081, GAS1185, and GAS1778, had the wild-type Mga and were not further characterized. The mga gene of 10 M^- variants from in vitro MGAS2221 culture were also sequenced. Eight of them had c.1571C[7]; isolate 499 had mga c.866A > C missense mutation that led to Mga T289P; and isolate 521 had wt c.1571C[8] mga (Table 2). Thus, the conversion of wt c.1571C[8] mga into the c.1571C[7] variant was a common cause for M^- variants arisen during genetic manipulation and simple in vitro culturing of M1 GAS strains.

The mga gene of M3 GAS strains also has the polycytidine tract starting at base 1,592 that is referred as c.1592C[8]. Two gene deletion mutants of emm3 GAS strain MGAS315, GAS1191 and GAS1028, had reduction of emm mRNA levels by 99% and 1C deletion in the c.1592C[8] tract of mga (Supplementary Table S1). This 1-C deletion in the c.1592C[8] tract did not lead to the Mga-M3 fusion protein but led to an Mga variant with 52 extra amino residues in comparison with the wild-type Mga protein of M3 GAS. Unlike the M1 c1571C[7] mga variant, the c.1592C[7] mga variant of M3 GAS does not encode Mga-M protein fusion. This difference is due to the deletion of G at base 148 of the intergenic sequence between the mga and emm genes in M3 GAS in comparison with M1 GAS (Supplementary Figure S1).

A question of interest is whether the conversion of c.1571C[8] mga to c.1571C[7] mga is reversible. Since the M protein contributes to protection of GAS against the host innate immune response, in vivo growth would provide an environment to allow the conversion of c.1571C[7] mga to c.1571C[8] mga. To test this idea, two mice were subcutaneously infected with GAS1285, an M⁻ MGAS2221 ΔsagA mutant with the c.1571C[7] mga variant. GAS isolates were recovered from skin infection sites 4 days after inoculation, and 13 GAS colonies randomly chosen from each of mouse 1 and 2 for detecting M protein production by western blotting analysis. Three and four isolates among the analyzed isolates from mouse 1 (Figure 3A) and mouse 2 (western blot not shown) restored M protein production. DNA sequencing found that all 7 M⁺ isolates had c.1571C[8] mga (Table 3) whereas one M⁻ isolate from the infection was sequenced and still had the c.1571C[7] mga variant.

We also tested GAS1806, an M⁻ c.1571C[7] variant of MGAS2221, in subcutaneous infection of 5 mice that are numbered from 3 to 7 in Table 3. Four of 20 isolates from mouse 3, 1 of 8 isolates from mouse 4, and 2 of 8 isolates from mouse 5 restored M protein production whereas 10 isolates from mouse 6 or mouse 7 did not have detectable M protein production. Three of the 4 M⁺ isolates from mouse 3 restored the c.1571C[8] tract of mga, and the other M⁺ revertant had an additional C deletion in the c.1571C[7] tract, resulting in 6 C at the polycytidine tract of mga that is designated as c.1571C[6]. The M⁺ isolate in mouse 4 restored the c.1571C[8] tract of mga whereas both M⁺ isolates in mouse 5 had the c.1571C[6] mga variant (Table 3). The additional 1C deletion of the c.1571C[7] tract of mga in the M⁻ variant shifted the reading frame, converting the Mga-62 AA-M1 protein fusion protein into an Mga variant of 542 aa residues, which was just 13 amino acid residues longer than the wild-type Mga protein of M1 GAS (Figure 3B). This c.1571C[6] mga variant apparently restored Mga function and thus the M protein production.

To test whether the c.1571C[7] and c.1571C[6] mga variants lost and regained function, emm mRNA levels were measured by real-time RT PCR for MGAS2221 (wt control), GAS1806 (c.1571C[7] mga variant of MGAS2221), and isolates M497, M405, and M406 from mouse infection with GAS1806. M497, M405, and M406 had c.1571C[6] (M497), c.1571C[7], and c.1571C[8] mga variants. As shown in Figure 3C, GAS1806 and M405 lost emm transcription by 98% whereas M497 and M406 restored emm transcription to 90 and 100% of that in MGAS2221. Thus, the c.1571C[7] mga variant is nonfunctional and can be reversed into functional c.1571C[8] mga by gaining 1\u00B0C or converted into functional c.1571C[6] mga variant by losing another C during mouse infection. The results indicate that the polycytidine tract at the 3′ end of the mga gene in M1 GAS can switch between c.1571C[8] or c.1574C[6] and c.1574C[7] to turn on and off M protein production, respectively.

If the polycytidine tract-based switch of mga for phase variation in M protein production functions during infection, there should be polymorphism at the polycytidine tract of the mga gene in clinical isolates. M12 GAS frequently shows M protein phase variation in vitro (Simpson and Cleary, 1987). So we first examined the mga sequences of M12 GAS genomes in the NCBI genome database. The mga gene and protein sequences of 106 available M12 GAS genomes were aligned. As shown in Figure 4A, there are four different types of polymorphisms related to the number of the C repeats starting at base 1574 and polymorphism at base 1657: (1) functional c.1574C[8] mga and base G at position 1657 (1657G) (wt), (2) nonfunctional c.1574C[7] mga and 1657G (Variant 1), (3) functional c.1574C[6] mga and 1657G (Variant 2), and (4) functional mga variant with c.1574C[7] and 1657A (Variant 3). Wild-type Mga of M12 GAS has 530 amino acid residues. The nonfunctional c.1574C[7]/1657G Variant 1 encodes a nonfunctional Mga-M protein fusion protein with a 62-aa bridge encoded by the intergenic sequence between the mga and emm genes, the functional c.1574C[6]/1657G variant has 543 amino acid residues and is similar with the c.1571C[6] mga variants of MGAS2221 that were from the c.1574C[7] mga variant of MGAS2221 in mouse infection; and the G-to-A nonsense mutation at base 1657 of the c.1574C[7] mga variant converts Variant 1 into c.1574C[7]/1657A Mga variant that is 21 aa-residue longer than the wt Mga protein (Figure 4B) (Table 4). Among 106 M12 genomes in the NCBI genome database, there are 25 wt mga; 3 Variant 1, 38 Variant 2; and 40 Variant 3. These data are summarized in Table 4, and the mga polymorphisms at the polycytidine tract and base 1657 of 106 M12 genomes are listed in Supplementary Table S2. It is interesting that two of the three strains with nonfunctional c.1574C[7]/1657G mga variant 1, ATCC 11434 (Tanaka et al., 2020) and MGAS2096 (Sarkar et al., 2018), were isolated from patients with acute poststreptococcal glomerulonephritis, and the infection nature of the third strain, NCTC8300, is not known. The data indicate that the length polymorphism of Mga is caused through arising the nonfunctional c.1574C[7]/1657G mga variant and restoring the function of the nonfunctional c.1574C[7]/1657G mga variant by the additional C deletion at the c.1574C[7] tract or the G-to-A mutation at base 1657. These mga polymorphisms among clinical M12 isolates also indicate that M12 GAS undergoes phase variation in M protein production through the slipped-strand mispairing of the mga polycytidine tract.

Among 158 M1 GAS strains (Supplementary Table S3), 5 M1 GAS strains, FDAARGOS, CCUG-4207, CCUG-4207-W1, NCTC8198, and AP1, had the functional c.1571C[6] mga variant, and three M1 strains, CCUG-47803, SPY8157, and GA41345, had the nonfunctional c.1571C[7] mga variant (Table 4 and Supplementary Table S3). Other M1 strains in the NCBI genome database have the wt c.1571C[8] mga. The frequency of the c.1571C[6] mga variant among the sequenced M1 GAS isolates, 5/158, is less than that among the sequenced M12 GAS isolates, 38/106 (Table 4). Four of 6 M14 GAS strains in the GenBank database have the wt c.1574C[8] mga encoding Mga with 530 amino acid residues, and two other strains have the c.1574C[6] mga variant for Mga with 543 amino acid residues (Table 1). One of two M23 strain has the c.1574C[6] mga variant for Mga with 543 amino acid residue, and another has 1C addition at the c.1574C[8] tract and 1C deletion at the c.1555C[6] tract, leading to c.1555C[5]/c.1573C[9] mga variant that encode Mga with 530 amino acid residues (Table 1). The c.1574C[6] mga variant in clinical M1, M14, and M23 GAS isolates indicates that the nonfunctional c1574C[7] mga variants of M1, M14, and M23 GAS arise during infection.

M3 GAS strains in the GenBank database have four polymorphisms of M3 mga that are related to the polycytidine tract near the 3′ end of mga (Supplementary Table S4). 106 strains have the wt c.1592C[8] mga; 5 strains have a missense mutation at base 1599 of the c.1592C[8] tract from C to T or c.1598C > T; 1 strain has an addition of 1C to the tract to result in c.1592C[9] mga variant; and two strains, A842 and A843, had 1C deletion at the c.1592C[8] tract and 1C addition at the c.1573C[6] tract to result in c.1592C[7]/c.1573[7] mga variant. It is unlikely that the two events in strains A842 and A843 occurred at the same time. These polymorphisms of M3 mga indicates the slipped-strand mispairing within the c.1592C[8] tract of M3 mga also occurs in hosts.

The M49 mga gene belongs to the mga-2 cluster that does not have the polycytidine tract. Except for a three amino acid residue insertion near the C terminus of Mga in M49 strain K36294, all the Mga sequences of the other 29 M49 genomes do not show length polymorphism. The Mga sequences of the other mga-2, such as M89 mga, do not have the polycytidine tract, which is shown in the sequence alignment of mga of M1 strain MGAS5005 and M89 strain KUN-0012590 (Murase et al., 2020) (Supplementary Figure S2). M89 strains do not have polymorphism in the length of the Mga protein. These results indicate that the polycytidine tract of mga-1, but not mga-2, causes polymorphism in the length of Mga proteins.