MGAS315 and MGAS10870 are fully sequenced invasive M3-type strains (Beres et al., 2002, 2010). MGAS315 was isolated from a patient with streptococcal toxic syndrome in the 1980's (Musser et al., 1991) and MGAS10870 was isolated from a patient with soft tissue infection in Ontario in 2002 (Beres et al., 2010). Additional strains from epidemiologically diverse M-types were used for comparison: MGAS6183 (M41), MGAS5005 (M1), and MGAS6143 (M28). The MGAS6143Δscl1-, MGAS6183Δscl1-, and MGAS10870Δscl1-inactivated mutant strains have been previously described (Han et al., 2006a; Caswell et al., 2007; Flores et al., 2015). Group A Streptococcus cultures were grown in Todd-Hewitt broth (Becton Dickinson and Co.) supplemented with 0.2% yeast extract (THY medium) and on Brain Heart Infusion agar (Becton Dickinson and Co.) at 37°C in an atmosphere with 5% CO₂. For antibiotic selection, erythromycin (5 μg mL⁻¹), chloramphenicol (10 μg mL⁻¹), kanamycin (200 μg mL⁻¹), and spectinomycin (100 μg mL⁻¹) were added to the medium.

Cloning experiments were performed in XL-1 Blue and TB1 E. coli cells, while protein expression experiments were performed in BL21 and TB1 E. coli cells grown in Luria-Bertani media (Difco Laboratories) at 37°C. For antibiotic selection, chloramphenicol (10 μg mL⁻¹), kanamycin (50 μg mL⁻¹), spectinomycin (100 μg mL⁻¹), and ampicillin (100 μg mL⁻¹) were added to the medium.

For trans-complementation experiments, plasmid pSL502, encoding full-length cell-associated Scl1.3FL protein, was electroporated into MGAS315 WT and MGAS6183Δscl1 electrocompetent cells. The pJRS525 vector was electroporated as a control. Transformants were selected on BHI agar containing 100 μg mL⁻¹ spectinomycin, and plasmids were re-sequenced. For MGAS10870Δscl1, which contains a spectinomycin resistance cassette in place of the scl1.3 allele, the spectinomycin resistance marker in pSL502 was replaced with a kanamycin resistance, generating the plasmid pSL518 and colonies were selected on BHI agar containing 150 μg mL⁻¹ kanamycin.

Since the resurgence of invasive GAS infections in the 1980s and an advent of molecular epidemiology fostered by large-scale sequencing, significant efforts have been made to define the molecular basis for the invasive phenotype of M3-type strains. In parallel to these advances, we identified two unique genomic traits in the scl1.3 locus, encoding streptococcal collagen-like protein 1: (i) the presence of an insertional sequence element, IS1548, in the promoter region and (ii) the presence of a nonsense mutation within the coding sequence of scl1.3 allele that was absent in other M-types.

Studies employing mga-inactivated mutants have indicated that Scl1 expression was positively regulated by the GAS-global transcriptional regulator, Mga (Rasmussen et al., 2000; Lukomski et al., 2001). Two putative Mga binding sites were identified upstream of the scl1 coding sequence in the M1-type strain SF370 and experiments demonstrated that the Mga binding site distal to scl1 was responsible for transcription activation (Almengor and McIver, 2004). Identical Mga binding sites I and II were also identified upstream of scl1.3 in the sequenced M3-type strain MGAS315 with IS1548 element inserted 38 bp upstream of the distal Mga I-binding site (Figure 1A). To determine if the IS1548 insertion was specific to M3 strains, we BLAST-searched this element in 45 completed GAS genomes representing 21 different M-types. We observed the presence of IS1548 in all strains searched with varying locations and occurrences from 1 to 12 per genome (Figure 1B). However, the IS1548 insertion upstream of the scl1.3 allele was only found in the sequenced genomes of M3-type strains MGAS315, SSI-1, and M3-b. Interestingly, a complete IS1548 element was not present upstream of scl1.3 in the recently reported genome of the M3 strain STAB902, which represents a non-invasive isolate (Soriano et al., 2014); instead, a 34-bp remnant of IS1548, including the inverted repeat and additional 14 bp, was found. Based on this bioinformatics data, we examined the presence of the IS1548 element upstream of scl1 by PCR in a panel of 40 M3-type strains, using primers located in the IS1548 and scl1.3 sequences (IS1548F and Scl1R, Table S1). All M3-type strains examined were positive for the IS1548-scl1.3 amplicon, while M1, M41, and M28-type controls were negative (Figure 1C), demonstrating a broad and conserved presence of the IS1548 insertion in this location among M3-type GAS.

The nonsense mutation in the 11th Gly-X-Y repeat in the Scl1.3-CL collagenous region (Figure 1A, red box), presumably results in a truncated secreted variant, which consists of the Scl1-V variable region and 10 Gly-X-Y repeats but lacks the cell wall anchor. This polymorphism was originally identified in five M3-type invasive GAS strains but was absent in 45 other strains analyzed, representing 20 different M-types (Lukomski et al., 2000). It was later shown in 98.7% of 479 sequenced invasive M3-type isolates (Flores et al., 2015). We performed targeted PCR amplification of scl1.3 from 46 additional M3-type strains and observed identical amplicon sizes in all of them (Figure S1A). Sequencing showed that all tested strains harbored an identical scl1.3 allele, containing 25 Gly-X-Y repeats in the collagenous domain with the null mutation in the 11th repeat (Data Sheet 1). A complete lack of genetic variation within scl1.3 is surprising and differs from the length variation that is commonly observed among scl alleles from other M-types (Lukomski et al., 2000; Rasmussen et al., 2000; Whatmore, 2001). These results demonstrate that the IS1548 insertion and null mutation we identified in the scl1.3 locus are unique to and conserved among M3-type GAS. Based on these results, we hypothesized M3 strains produce a truncated, but potentially biologically active Scl1.3 variant, which is secreted instead of being cell-attached.

Original reports showed scl1 transcripts in northern blots, as well as full-length Scl1 proteins (both cell-associated and cell-free fractions) in western blots for strains of emm types 1, 28, 52, and 41 (Lukomski et al., 2000, 2001; Rasmussen et al., 2000; Caswell et al., 2007). To analyze the expression of the truncated Scl1.3 protein in M3-type GAS, western blot analysis was performed on cell-wall (CW samples) and culture-supernatant (Sup samples) protein fractions of several M3-type strains grown to exponential phase (Figure 2A). The expected truncated Scl1.3 protein was not detected by anti-Scl1 antibodies, whereas the rScl1.3V positive control, corresponding to the V region of Scl1.3 variant, produced the expected immunoreactive band of ~8.3 kDa. In an additional control experiment, the same panel of M3 strains was tested for Scl1.3 on the cell surface using flow cytometry (Figure 2B). No shift in median fluorescence intensity was observed in M3 strains incubated with anti-Scl1 antibody compared to a secondary-only antibody control, indicating a lack of Scl1 on the cell surface among M3 strains.

Given the unique IS1548-scl1.3 location and the lack of truncated Scl1.3 products in culture supernatants, we investigated scl1.3 expression by qRT-PCR. Total RNA was isolated from exponential phase cultures of 5 M3-type strains, as well as from previously characterized control strains of emm types 1, 28, and 41 (Lukomski et al., 2000, 2001; Rasmussen et al., 2000; Caswell et al., 2007). Expression of scl1 from each strain was compared to scl1.3 transcription in M3 strain MGAS315 (Figure 2C). Each non-M3-type strain tested had significantly higher transcription level of respective scl1 allele compared to MGAS315. The M1 strain had the most increased expression by 21-fold, while M28 and M41 strains exhibited 6- and 8-fold higher scl1 expression, respectively. Interestingly, the scl1.3 transcripts examined in four additional M3-type MGAS strains 158, 335, 1313, and 10870 were significantly reduced as compared to MGAS315, with a range of 25- to 45-fold decrease. Overall, we observed that M3-type GAS harboring the IS element upstream of scl1.3 have drastically decreased scl1.3 transcript levels and lack the Scl1 protein product, as assessed by western blot and flow cytometry.

Since scl1 is regulated by the transcriptional activator Mga, we investigated whether the decreased scl1.3 transcripts in M3 strains were due to lack of mga gene expression in these strains. In parallel, we assessed transcript levels of the emm gene encoding the M surface protein, a key virulence factor regulated by Mga. For comparison, we included the M1 strain, which had significantly increased scl1 expression compared to the M3 strains. With the exception of MGAS335, which had significantly downregulated mga and emm expression, we found no significant differences in either mga or emm gene expression between MGAS315 and the other M3 strains or M1 strain (Figure 2D). Furthermore, M3 protein was highly expressed, as it was found in both cell wall and supernatant fractions, except for the MGAS335 strain, consistent with transcription data (Figure 2E). These results demonstrate that the striking downregulation of scl1.3 in M3-type GAS is not due to decreased mga expression or non-functional Mga protein, as emm is normally expressed in these strains.

Scl2 shares a similar structure with Scl1 but its biological function is poorly understood. One study demonstrated in a different M-type background that isogenic mutant devoid of Scl2.55 variant had lower adhesion to human skin fibroblasts (Rasmussen and Björck, 2001); however, the Scl2.3 variant present in M3-type GAS has not been investigated for expression and ECM binding. Therefore, we next assessed scl2.3/Scl2.3 expression in M3 strains. PCR amplification and sequencing showed that the majority of M3 strains contained in-frame scl2.3 allele (Figure S1B, Table S2).

Western blot analysis of cell wall (CW samples) and supernatant (Sup) protein fractions found Scl2.3 protein was expressed by M3 strains MGAS315, 10870, 158, 9517, and 1313, whereas samples obtained from strain MGAS335, which contains an out-of-frame scl2.3 allele, generated no immunoreactive band; rScl2.3V control produced the expected 10.1-kDa band (Figure 3A). Mass spectrometry confirmed the identity of the presumed immunoreactive Scl2.3-protein band from MGAS315 (Table S3). Consistently, Scl2.3 was detected on the surface of all five M3 strains containing in-frame scl2.3 alleles by flow cytometry, with a positive shift in median fluorescence intensity ranging in 67- to 131-fold change, as compared to the secondary-only antibody control (Figure 3B). We next compared the scl2-transcription level in M3 strain MGAS315 with scl2-transcription levels in M1-, M28-, and M41-type strains. In striking contrast to the pattern of scl1.3 transcription, scl2 transcripts were significantly decreased in the M1-type strain by 13-fold, as well as in the M28 (6-fold) and M41 (3-fold) strains (Figure 3C). Additionally, there was no significant difference between scl2 expression in MGAS315 and the M3 strains MGAS158, 335, 1313, and 10870 (Table S4). These results confirm that the M3 scl2.3 allele is transcribed at high levels, resulting in considerable expression of the Scl2.3 protein. These results suggested that Scl2 has an important biological function in M3 strains and warranted subsequent experiments assessing Scl2.3 function.

A wounded human skin equivalent, devoid of an inflammatory component, was utilized as a “mechanistic model” of GAS tissue colonization. The epidermal wound of each skin equivalent was infected with GFP-expressing M3-type invasive strain MGAS315 or M41-type non-invasive strain MGAS6183, and analyzed after 1–5 days by standard histopathology (H&E and Gram's stain) and using TPF microscopy. H&E of uninfected tissue controls harvested at day 0 showed the absence of a keratinocyte layer where the punch biopsy was performed (Figure 4A). Complete healing of the wound was observed after 5 days, with a newly-generated intact keratinocyte layer covering the punch biopsy site (Figure 4B). In contrast, tissue infected with either GAS strain exhibited delayed wound closure as late as day 5 post-infection (Figures 4E,H). H&E staining of skin equivalents infected with either M3 or M41 after 24 h revealed bacterial colonization of the exposed dermal surface, as well as bacterial invasion into puncture-associated defects extending deep into the dermal layer, largely located at the wound edges (Figures 4D,G). Notably, extensive bacterial growth and spread was largely confined to tissue crevices in the dermis, whereas the presence of large bacterial colonies directly below the wound bed was rarely seen on microscopic examination. By day 5 of infection, bacterial invasion via these dermal defects extended to the bottom of the dermis for both strains, presenting both vertical and lateral spread of bacteria (Figures 4E,H). Gram stain of infected tissues showed the formation of superficial colonies near dermal surfaces, as well as biofilm formation on the surface of exposed dermis (Figures 4J–M). Additionally, epidermal tissue neighboring the wound bed exhibited ~60% decreased thickness of the viable keratinocyte layer in tissues infected with MGAS315 and MGAS6183 by day 5, relative to uninfected tissues (Figures 4C,F,I). This suggests bacterial infection of the wound affects epidermal cells distant from the site of infection in this model.

TPF analysis was the performed on whole infected skin equivalents on day 5 post-infection in order to assess bacterial spread directly below the wound bed. This method allowed us to visualize bacterial structures within tissue that were not apparent in H&E or Gram-stained sections. Tissue microcolonies were observed in samples infected with M41 GAS (Figure 4N), whereas M3 cells had a scattered appearance (Figure 4P). TRITC-concanavalin A (TRITC-conA) was utilized to visualize glycocalyx associated with bacteria. TRITC-conA stain colocalized with bacterial microcolonies formed by the M41 GAS (Figure 4O, Figure S3), indicating microcolonies were encased in a glycocalyx, much like a classic biofilm. In contrast, TRITC-conA stain was associated with scattered M3-GAS chains located at the bottom cell layer in MGAS315-infected skin equivalents (Figure 4Q). These results indicate that M41 GAS, but not the M3 GAS, forms microcolonies in the tissue during a human skin infection that are encased in a glycocalyx, consistent with the observation that M41, but not M3, GAS forms robust biofilm in vitro on ECM coatings. In addition, M3 GAS disseminates in a form of scattered chains or single cells through the tissue.

It was previously reported that M3-type strains have no substantial biofilm formation in vitro on an inanimate surface (Oliver-Kozup et al., 2011). Here, we tested biofilm formation on cellular fibronectin (cFn) and laminin (Lm) coatings by a panel of representative M3 strains isolated from invasive cases of GAS disease, as compared to the non-invasive biofilm-capable M41-type model strain MGAS6183. As expected, wells coated with either cFn or Lm supported robust biofilm formation by the M41 strain, whereas significantly less bacterial biomass was measured for all M3-type GAS on both ECM coatings (Figure 5A). There was also no correlation between Scl2 expression and biofilm formation. We hypothesized that M3-type GAS, devoid of Scl1 adhesin, have decreased binding to cFn and Lm ECM components, thus, preventing the formation of tissue microcolonies, and that restoration of full-length Scl1.3 (Scl1.3FL) on the GAS cell surface will confer binding to host ECM, as well as biofilm capacity in vivo (Figure 5B).

To test this hypothesis, we: (i) constructed Scl1.3- and Scl2.3-derived recombinant proteins, (ii) characterized their structural organization, and (iii) assessed their ECM-binding capacities.

First, rScl1.3FL and rScl2.3 proteins were assessed for purity and integrity by SDS-PAGE (Figure 6A). The expected 20.6-kDa rScl1.3FL migrated at ~34 kDa, which is consistent with previous reports of aberrant migration of rScl proteins (Lukomski et al., 2000, 2001), whereas the rScl2.3 protein migrated according to the expected molecular mass of 16.2 kDa; both proteins were verified by mass spectrometry (Table S3). Rotary shadowed rScl1.3FL and rScl2.3 constructs, exhibited the characteristic lollipop-like structural organization (Figure S2), as observed for previously characterized rScl proteins (Xu et al., 2002; Han et al., 2006b). Interestingly, rScl1.3FL formed aggregates that were mediated by the intermolecular interactions between the globular domains; such interactions, however, were not observed between rScl2.3 molecules. The appearance of Scl1-Scl1 aggregates implies an attractive hypothesis that V-to-V-region interactions between the Scl1 molecules, but not between the Scl2 molecules, on the surface of neighboring GAS chains may support biofilm structure, as proposed in our model (Figure 5B).

It has been established that Scl1 proteins selectively bind cellular, but not plasma, fibronectin and laminin (Caswell et al., 2010), and that selective cFn binding is achieved by a unique binding mechanism involving the C-C′ loop of the extra domain A in cFn (EDA-cFn) (Oliver-Kozup et al., 2013). Here, we pre-incubated cFn with increasing concentrations of EDA-blocking IST-9 mAb (0.1, 1.0 μg), then added to wells containing immobilized rScl proteins and allowed for binding. We observed significant dose-dependent inhibition of cFn binding to rScl1.3FL, with a 36% inhibition at 0.1 μg and a 60% inhibition at 1.0 μg of IST-9 (Figure 6B). In addition, the rScl2.3 protein did not bind cFn, which is consistent with our previous data employing several different rScl2 proteins (Caswell et al., 2010). Surface plasmon resonance (SPR) measurements of binding affinity between rScl1.3FL and EDA-derived C-C′ cyclic peptide provided a dissociation constant of K_D = 63.3 μM (Figure 6C). To corroborate the rScl1.3FL-EDA binding affinity, an in-solution fluorescence binding assay was performed, in which the variation in tryptophan fluorescence of rScl1.3FL was recorded as a measure of positive binding to the C-C′ cyclic peptide. Tryptophan fluorescence emission at 333 nm showed a dose-response quenching upon addition of the C-C′ cyclic peptide and −Δ fluorescence intensity was plotted against peptide concentration (Figure 6D). Data were fitted with a 1:1 model of interaction, providing a K_D = 44.54 ± 9 μM, in agreement with SPR data (Russo et al., 2015).

We next assessed binding of rScl1.3FL and rScl2.3 proteins to laminin (Lm) by ELISA. We determined rScl1.3FL had significant Lm binding, whereas rScl2.3 had not (Figure 6E), consistent with previous findings that Scl1-derived recombinant proteins, but not the Scl2-derived, bind ECM proteins (Caswell et al., 2010). Collectively, these results demonstrate specific binding of Scl1.3FL to the EDA domain of cFn and to Lm and its capacity of being surface adhesin.

To assess the effect of cell-surface Scl1.3FL expression on the capacity to form biofilm, an in-trans complementation of two representative invasive M3 strains was performed, MGAS315 wild-type (WT) strain, naturally lacking Scl1.3 expression, and a previously generated scl1-inactivated mutant of MGAS10870 (10870Δscl1), with plasmids pSL502 (Sp^R) and pSL518 (Km^R), respectively, both encoding the full-length Scl1.3FL protein. As a control, MGAS315 was complemented with a shuttle vector pJRS525. The cell wall-associated expression of Scl1.3FL in complemented M3-type GAS was first tested by western blot analysis of the cell wall protein fractions and on the GAS-cell surface by flow cytometry. An expected ~35-kDa immunoreactive band in complemented strains was observed, which was absent in the parent strains (Figure 7A). Mass spectrometry analysis of the corresponding bands extracted from the gel confirmed they represented the Scl1.3FL protein, with five unique peptides identified with 17% sequence coverage for complemented MGAS315 and four unique peptides identified with 11% sequence coverage for complemented MGAS10870 (Table S3). A 6.8-fold increase in median fluorescence intensity of Scl1.3FL-complemented MGAS315 cells was measured by flow cytometry, as compared to the vector-complemented and WT control strains (Figure 7B). The Scl1.3FL-complemented 10870Δscl1 exhibited a 2.2-fold increase in median fluorescence intensity, as compared to the mutant control (Figure 7B). These results indicate the Scl1.3FL is indeed expressed and surface-exposed in the complemented M3-type strains.

Biofilm formation by the complemented and parent strains was then assessed after 24 h following crystal violet staining and using confocal laser scanning microscopy (CLSM). Scl1.3FL-expressing MGAS315 showed significantly increased biomass on both cFn and Lm coatings compared to the WT parent organism, as well as vector-complemented control; 3.6 or 2.5- and 2.5 or 2.1-fold OD₆₀₀ increases on cFn and Lm, respectively, were measured compared to MGAS315 WT or to vector-complemented MGAS315 (Figure 7C). Significantly thicker biofilm formed by Scl1.3FL-complemented MGAS315 was imaged by CLSM. We observed on average a 2.8-fold (P = 0.0002) and 2.6-fold (P = 0.0002) increased biofilm thickness on cFn and Lm, respectively, compared to MGAS315 WT, and a 2.0-fold (P = 0.0036) and 4.8-fold (P = 2.1 × 10⁻⁵) increased biofilm thickness on cFn and Lm, respectively, compared to vector-complemented MGAS315 (Figures 7D,E, representative images). Similar results were obtained for the complemented 10870Δscl1 mutant, which had 1.7-fold increase in biomass staining on both cFn and Lm compared to the MGAS10870 WT strain, and 2.2- and 2.6-fold increase in biomass staining on cFn and Lm, respectively, compared to the 10870Δscl1 mutant strain (Figure 7C). CLSM data, however, could not be rendered for MGAS10870 strains due to poor GFP expression for unknown reasons. Altogether, it was demonstrated that null mutation in the scl1 gene, which ablates surface Scl1.3 protein and is unique to M3-type GAS, is responsible for the decreased biofilm capacity since restoration of the full-length surface-exposed Scl1.3 adhesin significantly fosters stable biofilm formation.

In a previous study we showed that scl1.41-inactivation in a non-invasive biofilm-rich M41 strain MGAS6183 resulted in significantly decreased biofilm capacity, which was restored to wild-type level by complementation with surface Scl1.41 (Oliver-Kozup et al., 2013). Notably, this M41 strain expresses at least one additional major Fn-binding protein, protein F2 (Caswell et al., 2007), which binds both plasma and cellular fibronectin by a mechanism different from Scl1 (Sela et al., 1993). Here, we hypothesized that expression of rScl1.3FL in the heterologous M41 GAS will confer biofilm formation. Western immunoblotting of the cell wall protein fractions detected the ~35-kDa immunoreactive band, corresponding to full-length Scl1.3, associated with complemented cells, while the M41Δscl1 mutant and WT cells were signal-negative (Figure 8A). Mass spectrometry of the corresponding band extracted from the gel confirmed Scl1.3FL expression, with 3 unique peptides identified, covering 13% of the amino acid sequence (Table S3). In addition, we showed the expression of the larger Scl1.41 variant in the cell wall of M41 WT, but not in the Δscl1 mutant, by re-probing a portion of the blot with anti-rScl1.41 antibody, using rScl1.41 protein as a positive control.

Similarly to complementation with homologous Scl1.41 protein, the heterologous complementation of the M41Δscl1 mutant with Scl1.3FL also restored biofilm to M41 WT levels on cFn and Lm (Figures 8B–D). Crystal violet staining showed increased bacterial biomass (Figure 8B) and confocal microscopy revealed significantly thicker biofilms, on average 3-fold increased, as compared to the parental M41Δscl1 mutant strain on both cFn (Figure 8C) and Lm (Figure 8D) coatings (cFn, P = 0.0105; Lm, P = 0.0011).

We have previously shown that an M3 strain harboring a carrier scl1.3 allele, producing a shorter cell-attached Scl1.3 variant, had an attenuated phenotype in a murine model of necrotizing fasciitis (Flores et al., 2015). Here, we tested our hypothesis that Scl1 adhesin in biofilm-rich M28 and M41 background promotes stable colonization and localized infection, using a murine skin infection model. Hairless, immunocompetent SKH1 mice were subcutaneously infected with ~10⁹ GAS CFU of the M28 and M41 wild-type (WT) or their isogenic scl1-inactivated mutant (scl1) strains (Han et al., 2006a; Caswell et al., 2007), and mice were assessed for changes in gross pathology of the skin.

Skin lesions caused by both the WT and scl1 strains were observed as early as 48 h post-infection, and the lesions began to regress after day 7 with complete resolution of the lesion by the completion of the experiment. The areas of the skin lesions calculated for mice infected with the scl1 mutant strains were significantly larger than those of WT-infected mice (Figure 9A; 96 h time-point is shown); images of lesions of representative mice demonstrate differences in lesion severity between WT- and scl1-GAS infected mice for both the M28- and M41-infected groups (Figure 9B). Thus, the in vivo data, using GAS strains expressing surface Scl1 proteins and producing rich biofilms, support our hypothesis that decreased adhesion and biofilm formation, due to the absence of Scl1.3 on the surface of M3 strains, bears an inverse correlation to the invasive potential of the infecting GAS strain.

First, the insertion of IS1548 element in the promoter region of scl1 exclusively found in the M3-type strains is an intriguing observation, given the abundance of IS1548 insertions across GAS genomes representing a variety of M-types. IS elements are known to cause genomic rearrangements and affect the expression of important genes that contribute to bacterial virulence (Mahillon et al., 1999). For example, the insertion of IS1548 element (Granlund et al., 1998) into the scpB-lmb intergenic region in group B Streptococcus has been shown to upregulate lmb-gene transcription and Lmb-adhesin surface expression, resulting in enhanced laminin binding (Al Safadi et al., 2010). The IS1548 insertion in the scl1.3 promoter was conserved in the genomes of sequenced M3 strains MGAS315, SSI-1, and M3-b, as well as in 40 additional M3 strains tested by PCR. However, the recently sequenced STAB902 M3 strain (Soriano et al., 2014), which represents a non-invasive isolate, contained a 34-bp IS1548 remnant, suggesting that M3 circulating strains exist with this polymorphism. We hypothesize that the proximity of the IS1548 insertion to the Mga I binding site in M3 strains affects scl1.3 transcription. Comparative qRT-PCR analysis showed significantly higher levels of scl1 transcription, particularly in M1, as well as in M28 and M41 strains, relative to M3-type strains analyzed.

Variation in the Mga coding sequence and autoregulated mga promoter has been reported to significantly affect Mga-dependent gene expression and virulence properties in GAS (Flores et al., 2013; Cao et al., 2014; Sanson et al., 2015). For example, a 12-bp deletion of a single repeat in the VNTR region of the mga promoter is present in M3 carrier strains and absent in the invasive M3 strains, causing a downregulation of Mga expression and Mga-regulated genes in the carrier strains (Flores et al., 2013). We identified an analogous 12-bp deletion in the VNTR of the mga promoter of the M1-type MGAS5005 and in M3 strains, MGAS335 and MGAS1313, but not in MGAS315, MGAS158, and MGAS10870. It is likely that the variation observed in the VNTR region impacts Mga expression but could not, alone, explain differences in scl1 transcription patterns observed in this work. Consequently, qRT-PCR analyses showed mga and emm transcripts (emm is second downstream target of Mga within the Mga regulon) were expressed at similar levels in MGAS5005 and M3 strains, except for the MGAS335, indicating a mechanism of scl1 downregulation in M3 strains, which is independent from the level of Mga transcription. The presence of emm transcripts and M3-protein products indicate that Mga is present and functional in the majority of M3 strains tested, and therefore the polymorphism we observed in the mga promoter did not affect Mga expression or activity of Mga-controlled scl1.3 promoter. Altogether, our data hint at the insertion of the IS1548 in the downregulation of scl1.3 expression, specifically in M3-type GAS, although additional studies are required to firmly establish whether the IS element insertion is responsible.

It is intriguing that scl1 and scl2 expression patterns were drastically different between M1 and M3-type GAS, both associated with invasive infections. The scl1.3 transcript was decreased by 21-fold in MGAS315, and even further in the remaining M3 strains studied (additional 20- to 45-fold), as compared to the M1-type strain MGAS5005. The upregulation of scl1 in M1-type GAS has previously been shown in invasive M1 isolates, including MGAS5005, as compared to pharyngeal M1 isolates (Sumby et al., 2006). Studies comparing wild-type and isogenic scl1.1 mutants in M1-type GAS reported that Scl1.1 contributed to immune evasion by inhibiting neutrophil extracellular trap production and by protecting bacteria from the cathelicidin LL-37 (Döhrmann et al., 2014). We conclude the differences we observed in scl1 expression are related to different pathogenicity requirements for Scl1 protein in M1- vs. M3-type strains. In contrast, Scl2 expression is highly upregulated in M3-type strains, as compared to invasive M1-type GAS. While the majority of M3-type strains contain in-frame scl2.3 alleles, all 21 M1 strains we analyzed contained out-of-frame scl2.1 alleles (unpublished data). Although, the Scl2.3 human ligands are not known, it may have an unidentified biological function, which is important for pathogenesis of M3, but not M1, strains. Hence, we are reporting striking differences in the expression and features of the Scl1 and Scl2 proteins that evolved in the invasive M1 and M3 strains.

Scl1 mediates binding to human extracellular matrix components, cellular fibronectin and laminin, as well as biofilm formation (Caswell et al., 2010; Oliver-Kozup et al., 2011, 2013). Previous work identified that Scl1 binds specifically to the C-C′ loop of the type III-repeat EDA domain of cellular fibronectin (Oliver-Kozup et al., 2013). This represents a novel mechanism of fibronectin binding, which is distinct from the mechanism employed by other GAS fibronectin-binding proteins that bind to the N-terminal type I repeats of fibronectin (Yamaguchi et al., 2013). Here, we determined that rScl1.3FL binds to cellular fibronectin via the same C-C′ loop-dependent mechanism. It is important for our overall model to acknowledge that the EDA-cFn isoform is specifically produced during embryogenesis and during wound healing in adult tissue (Ffrench-Constant et al., 1989; Jarnagin et al., 1994; Singh et al., 2004), which indicates Scl1 evolved with a unique function for targeting wounded tissue, a presumed pathogen portal of entry. We also demonstrate significant binding of rScl1.3FL to laminin, a major basement membrane protein at the epidermal-dermal junction; none of those ECM ligands were bound by rScl2.3, consistent with previous knowledge that Scl1, but not Scl2, variants exhibit binding to cFn and Lm (Caswell et al., 2010). These studies show that full-length Scl1.3, if expressed on the M3-GAS cell surface, would bind ECM, contributing to tissue colonization.

The diminished Scl1.3 expression and poor biofilm formation, led us to hypothesize that lack of surface-expressed Scl1 in M3-type GAS reduces host ECM binding and stable microcolony formation in tissue, thus, shifting the balance toward invasive spread, augmented by other virulence factors expressed by these strains. Recombinant rScl1.3FL formed aggregates that were mediated by the V-to-V region interactions, which could represent a mechanism of biofilm and microcolony stabilization by Scl1 molecules on adjacent GAS cells. Two representative invasive M3 isolates, MGAS315 and MGAS10870, acquired biofilm formation on cFn and Lm when homologous complementation was performed with surface-attached Scl1.3FL. MGAS315 has been shown to contain a missense mutation in the covS gene, causing upregulation of CovRS-regulated virulence genes and enhanced virulence during subcutaneous infection of mice, compared to an isogenic strain containing the WT covS allele (Stetzner et al., 2015). Additionally, MGAS315 contains a mutation in the regulator of protein B allele, ropB, which produces a nonfunctional RopB variant, while MGAS10870 contains a wild-type covR/S and ropB alleles (Carroll et al., 2011), allowing us to demonstrate the effect of Scl1.3FL function in the presence of differing regulatory networks. Similarly, heterologous complementation with surface Scl1.3FL in a scl1.41-mutant of the non-invasive biofilm-capable strain M41-type MGAS6183, restored its biofilm capacity to the wild-type level. These results indicate that M3-derived Scl1.3FL variant has the full capacity to support biofilm formation to a similar degree as Scl1 from a divergent M-type. The robust biofilm observed in vitro on ECM coatings validates the concept that biofilm capacity combined with adherence to the surrounding host ECM would reinforce the formation of stable tissue microcolonies in vivo.

We observed inhibition of wound re-epithelization by GAS infection of wounded skin equivalents, as well as the thinning of the viable epidermal layer at sites distant from the infected wound. In addition to our study, others have reported changes in skin histopathology and wound healing, resulting from bacterial infections. An in vivo study has reported epidermal defects as a result of GAS infection in a humanized mouse model with human skin graft (Scaramuzzino et al., 2000). Previous study of M3-type GAS infection using a skin equivalent model showed that hyaluronic acid capsule interactions with CD44 receptor on keratinocytes induced intracellular signaling, resulting in cytoskeletal rearrangement and monolayer disruption (Cywes and Wessels, 2001). Infection of an in vitro skin model containing a burn wound with Pseudomonas aeruginosa caused a loss of the keratinocyte layer and basement membrane, while intact epidermis was observed in burned but uninfected tissue (Shepherd et al., 2009). Impairment of wound healing has also been demonstrated by staphylococcal infections. Infection of dermal wounds in rabbit ears with Staphylococcus aureus showed the formation of biofilm, production of a persistent, low grade inflammatory response, and significantly delayed wound healing (Gurjala et al., 2011). Similarly, delayed wound healing by both S. aureus and S. epidermidis biofilms was observed in a mouse model of cutaneous wounds (Schierle et al., 2009). The inhibition of wound healing we describe here is by and large consistent with reports by other laboratories generated using in vivo animal and in vitro human skin infection models.

Microcolonies have been identified within human streptococcal impetigo lesions (Akiyama et al., 2003) and in tonsils from patients with recurrent GAS pharyngeal tonsillitis (Roberts et al., 2012), and likely represent a superficial or persistent state of GAS colonization. However, streptococcal infections can result in invasive disease due to biofilm disruption and bacterial dissemination (Connolly et al., 2011a,b). We observed large rounded microcolonies formed in tissue during infection with M41 strain, while M3 GAS remained scattered throughout the tissue as single cells and chains. Microcolony formation has been previously observed with S. aureus infection in organotypic skin model (Popov et al., 2014). Moreover, wound infection in rabbit ears with S. aureus produced mature biofilms encased in exopolysaccharide, as revealed by concanavalin A staining (Gurjala et al., 2011). Similarly, we also demonstrated that microcolonies in M41-infected tissue were encapsulated in bacterial-associated glycocalyx. However, a lack of glycocalyx-encapsulated microcolonies was associated with infection by M3-type GAS.

These results support our hypothesis that biofilm-poor M3 strains are abolished in stable microcolony formation in vivo, in part due to a lack of the ECM-binding Scl1 protein and an overall lack of surface adhesins, although they likely express the FbaB protein, identified in M3 GAS to be involved in the adherence and invasion into epithelial and endothelial cells (Terao et al., 2001, 2002; Amelung et al., 2011).

Recent study reported that a small proportion of non-invasive M3-type strains (~1.3%) were found to harbor the scl1.3 “carrier allele,” which resulted from an in-frame deletion in the collagenous region, encompassing the null mutation, producing a shorter cell-attached Scl1.3 variant. This MGAS10870 strain containing the scl1.3 carrier allele was attenuated following intramuscular infection (Flores et al., 2015). In this study, we utilized the M28- and M41-type strains, representing biofilm-rich producers, for subcutaneous inoculation. We observed that scl1.28- and scl1.41-inactivated isogenic mutants produced significantly larger skin lesions as compared to the wild-type parent strains. These results, again, support the hypothesis that lack of Scl1 surface adhesin destabilizes focused nidus of infection, resulting in a shift toward increased tissue spread. However, previous studies performed in a M1 GAS background, utilizing scl1.1-mutants for subcutaneous infection, reported smaller skin lesions in the mutant groups, which likely reflects a differing predominant function of Scl1.1 in M1-type GAS (Lukomski et al., 2000; Döhrmann et al., 2014). Investigations using intranasal and intraperitoneal mouse infection models of Streptococcus pneumoniae have shown that culture-grown bacteria disseminated to the ear and lungs, while biofilm-grown bacteria stably colonized the nasopharynx (Blanchette-Cain et al., 2013; Marks et al., 2013). A similar study on Streptococcus pyogenes showed that bacteria grown in biofilms have downregulated virulence genes and tend to colonize the nasal associated lymphoid tissue of mice, while culture-grown bacteria had significantly increased dissemination and were more virulent in a septicemia model (Marks et al., 2014). Previous studies reported that inactivation of some GAS genes resulted in increased skin pathology produced by the mutants compared to their wild-type organisms, and these genes often encoded surface proteins, including SpyCEP (Sumby et al., 2008), Mrp (Boyle et al., 1998), protein F1 (Nyberg et al., 2004), and Spy0128, encoding a major pilus subunit (Crotty Alexander et al., 2010). Similarly, the covS mutant of group A streptococcal M1T1 strain with upregulated SpeB-protease activity was hypervirulent and had reduced capacity to bind human epithelial cells and fibronectin, and also to form biofilm due to increased cleavage of surface proteins (Hollands et al., 2010). Altogether, the concept that expression of a surface adhesin, such as Scl1, involved in biofilm formation and host tissue attachment, is inversely related to strain invasiveness has gained support from several studies, including this study.

We show the invasive M1- and M3-type GAS evolved scl1 and scl2 alleles with opposite expression patterns, with scl1 downregulated and scl2 upregulated in M3 compared to M1 GAS. We show M3-type GAS, devoid of surface-expressed Scl1.3, lacked biofilm formation on ECM coatings and microcolony formation during infection of in vitro wounded skin equivalent. Complementation with surface Scl1.3FL restored biofilm capacity of M3-type GAS on ECM coatings. Mouse infection with the isogenic scl1 mutants of biofilm-rich M28- and M41-type GAS produced larger lesions, supporting the role of Scl1 in a localized tissue infection. Lastly we developed a model for Scl1-mediated microcolony formation (Figure 10), whereby Scl1 expressed on the GAS surface strengthens host colonization by attachment to cFn and Lm expressed within wounded tissue, as well as biofilm formation via Scl1-Scl1 interactions, resulting in a local, stabilized microcolony. Conversely, when Scl1 is absent on the GAS cell surface, as is the case for M3-type GAS, bacteria lack stable anchoring in the surrounding host ECM, as well as structural strength within microcolony, promoting cells to disperse more freely.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.