Streptolysin-induced endoplasmic reticulum stress promotes group A Streptococcal host-associated biofilm formation and necrotising fasciitis

2.1 GAS forms biofilms on mammalian cells

To examine the potential host–pathogen interactions involved in GAS biofilm establishment during acute infection, we infected MEFs with GAS at a multiplicity of infection (MOI) of 10, using the clinical strain JS95 (M14 serotype; Hidalgo-Grass et al., 2004). We monitored GAS biofilm biomass accumulation over time by quantifying the biomass using a crystal violet (CV) staining protocol that distinguishes bacteria from mammalian cells (Negri et al., 2010; Figure 1a) and by enumerating colony forming units (CFUs; Figure 1b). Both assays revealed a significant increase in GAS biomass accumulation from 24 to 72 hr postinfection (hpi) and a steady decrease thereafter (Figures 1a,b). To ascertain whether GAS JS95 biomass accumulation represents a common trait during GAS–host interaction, we tested MEFs, human foreskin fibroblasts (HFFs) and human keratinocytes (HaCaTs) as well as GAS strains of other M-types for their ability to form cell-associated biofilms. Both HFFs and HaCaTs supported GAS biomass formation albeit with different kinetics than observed for MEFs (Figure 1c). Furthermore, both GAS MGAS5005 and JRS4 strains, of M1T1 and M6 serotypes, respectively, were able to form biofilms on MEFs (Figure 1d). Thus, the ability of GAS strain JS95 to accumulate biomass on top of mammalian cells may represent a general phenomenon in GAS–host interaction, although the biofilm formation kinetics is different among strains.

To examine the biofilm properties of mammalian cell-associated GAS, we first immunostained GAS infected MEFs with antibodies specific to GAS and imaged them by confocal laser scanning microscopy (CLSM) to study their spatial and temporal organisational dynamics (Figure 2a). GAS biofilm biomass on MEFs increased progressively between 24 and 72 hpi and decreased thereafter (Figure 2a), consistent with the biomass levels detected by CV staining and CFU enumeration (Figures 1a,b). Moreover, GAS biomass growing on MEFs at 72 hpi displayed a 3D architecture indicative of mature biofilms (Figures 2a and S1C,E; Monds, 2009), similar in appearance to GAS microcolony structures observed in pustule skin lesions from impetigo patients and in a mouse model of superficial human GAS skin infections (Akiyama et al., 2003). In addition, SEM revealed that GAS biofilms on MEFs appear in chains within thick and multilayered aggregates (Figure 2b), similar to previous reports of GAS biofilms in vivo (Roberts et al., 2010; Siemens et al., 2016). Although JS95 is capable of forming biofilm on polystyrene tissue culture plates (referred hereafter as plastic), biofilms on this abiotic substratum accumulate less biomass compared with biofilms on MEFs, display different developmental kinetics, and have a 3D architecture distinct from MEF-associated biofilms (Figs. S1A, B, C, D, E). Additionally, SEM images of abiotic-grown biofilms show reduced aggregation (Figs. S1F) compared with their MEF-grown counterparts (Figure 2b).

We next tested whether GAS growing on MEFs exhibited increased tolerance to antibiotics, a characteristic of biofilms (Baldassarri et al., 2006; Stewart & Costerton, 2001). We exposed preformed GAS biofilms growing on MEFs, or planktonic bacteria (both seeded with an initial inoculum of 1 × 10⁵ CFU), to increasing concentrations of ampicillin and enumerated CFU after 24 hr of antibiotic exposure. GAS biofilms were significantly more tolerant to ampicillin compared with planktonic cells (Figure 2c). Taken together, these observations show that GAS forms dense host-associated biofilms during acute infection that may contribute to disease progression by enhanced tolerance towards antibiotic treatment.

2.2 GAS streptolysin expression promotes biofilm formation on host cells

To identify GAS determinants important for biofilm formation on host cells, we examined the ability of isogenic mutants of JS95 WT to form biofilms on MEFs. The mutants chosen are defective in the expression of GAS virulence factors (Hynes & Sloan, 2016; Olsen & Musser, 2010; Walker et al., 2014), some of which have also been implicated in GAS biofilm formation (Fiedler et al., 2015; Young, Holder, Dubois, & Reid, 2016). We observed that deletion of emm14 encoding the M14 antiphagocytic surface protein and inactivation of the cysteine protease speB enhanced biofilm formation on MEFs (Figure S2A). This is consistent with previous studies where speB was shown to be associated with increased biofilm dispersal (Connolly, Braden, Holder, & Reid, 2011; Roberts et al., 2010). By contrast, inactivation of hasA, essential for hyaluronic acid capsule production, decreased biofilm formation significantly (Figure S2A). In addition, a strain lacking the expression of both SLO and SLS (Δslo,sagI⁻) was the most highly attenuated mutant for biofilm formation on MEFs (Figure S2A). To explore whether attenuated biofilm formation in the streptolysin and hasA mutants was related, we assessed streptolysin expression in a hasA⁻ mutant background by reverse transcription polymerase chain reaction. We observed that sagA, sagB, and slo were all expressed approximately 75% less in the hasA⁻ strain compared with WT background (Figure S2B). Thus, it is possible that a decrease in streptolysin activity in the hasA mutant is indirectly responsible for its attenuation in biofilm biomass. Because of this marked defect of the streptolysin-null strain, we focused the remainder of our study on the role of SLO and SLS in host-associated biofilm formation.

We next compared biofilm formation by JS95 (WT) and the streptolysin null mutant (Δslo,sagI⁻) over time on MEFs versus plastic. The amount of biomass produced on MEFs by WT biofilms was significantly higher than that produced by Δslo,sagI⁻ biofilms at all the time points examined (Figure 3a). By contrast, on plastic, Δslo,sagI⁻ produced significantly more biomass at 48 and 72 hpi compared with WT (Figure 3a). To ensure that both strains exhibit similar rates of growth in their planktonic state, we also enumerated CFU of WT and Δslo,sagI⁻ in the culture supernatants (nonbiofilm fraction) while growing on MEFs and observed no defect in planktonic growth (Figure S2C). To test whether GAS internalisation is responsible for reduced biofilm biomass of the double mutant, we quantified the internalisation of WT and Δslo,sagI⁻ bacteria every 2 hpi. In contrast to studies of an slo deletion mutant in GAS strain 188 (a derivative of the M type 3 WT strain 95077; Hakansson, Bentley, Shakhnovic, & Wessels, 2005) that displayed enhanced internalisation into keratinocyte lysosomes, we observed that internalisation of the double mutant into MEFs is abrogated as compared with WT JS95, and we conclude that reduced extracellular biofilm formation by the streptolysin null strain is not correlated with increased internalisation (Figure S3).

To gain more spatial insight into GAS biofilms on MEFs, we next examined the architecture of biofilms formed by WT, Δslo,sagI⁻, or Δslo,sagI⁻ harbouring plasmids complementing either slo or sagI. CLSM revealed that while WT GAS aggregation on MEFs was dense, Δslo,sagI⁻ displayed sparse aggregates on MEFs (Figure 3b). However, the corresponding complemented mutants expressing either SLO or SLS developed biofilm biomass similar to that of the WT, demonstrating that either SLO or SLS is sufficient to complement Δslo,sagI⁻ for biofilm formation (Figures 3b,c), indicating their redundant role in host-associated biofilms.

To determine whether the role of streptolysins in biofilm formation was strain or serotype-specific, we also examined streptolysin mutants in two additional GAS strains, M1 854 (Pinho-Ribeiro et al., 2018) and MGAS5005 (Baruch et al., 2014) of the M1 and M1T1 serotypes, respectively, that commonly cause acute infections. For both genetic backgrounds, we observed a significant decrease in biofilm biomass in the absence of the streptolysins, consistent with our previous observations in the M14 JS95 strain, and suggesting that our findings are not limited to the relatively rare M14 serotype (Figure S4).

2.3 Streptolysin-induced ER stress contributes to GAS biofilm formation on host cells

SLO and SLS are well-characterised cytotoxins secreted by GAS (Goldmann, Sastalla, Wos-Oxley, Rohde, & Medina, 2009). Lactate dehydrogenase (LDH) release into the culture media as a measure of streptolysin-mediated cytotoxicity revealed progressive cell death of WT-infected MEFs starting at 4 hpi. By 10 hpi, greater than 90% cell cytotoxicity was observed (Figure S5). By contrast, the Δslo,sagI⁻ mutant caused only 30% cytotoxicity in MEFs at the same time point (Figure S5). Consistent with this, we observed that WT-infected MEFs displayed markers of apoptosis as early as 6 hpi and necrosis by 8 hpi (Figure S6). By contrast, neither apoptosis nor necrosis occurred in MEFs infected with Δslo,sagI⁻ during the initial 12 hr of infection (Figure S6), indicating that these early cellular events influence later outcomes of biofilm formation. MEFs infected with the double mutant eventually undergo apoptosis at 24 hpi; however, this arises as a result of media acidification as a consequence of GAS proliferation and metabolism (data not shown).

GAS streptolysins SLO and SLS were earlier shown to cause ER stress in host cells leading to the release of asparagine (ASN) into the culture media, which GAS JS95 can sense and use to regulate its gene expression and rate of proliferation (Baruch et al., 2014). Because SLO or SLS contribute to biofilm formation, we hypothesised that streptolysin-mediated ER stress, which eventually leads to apoptosis (Li, Lee, & Lee, 2006), could similarly be involved in host cell-associated biofilm formation.

To test if ER stress is induced in GAS host-associated biofilms, we performed gene expression analysis of the ER stress markers Chop and Grp78 (Oslowski & Urano, 2011; Senft & Ronai, 2015; Zhang & Kaufman, 2008) in MEFs infected either with GAS or treated with the pharmacological ER stressor TG as a positive control (Shiraishi, Okamoto, Yoshimura, & Yoshida, 2006) or etoposide (ET), a DNA topoisomerase inhibitor that causes ER stress-independent apoptosis as a negative control (Jamil, Lam, Majd, Tsai, & Duronio, 2015). We observed more than a 2.5-fold increase in both Chop and Grp78 expression upon infection with WT GAS or single mutants Δslo or sagI⁻ (Figure 4a). However, Chop and Grp78 expression in MEFs infected with the streptolysin null mutant Δslo,sagI⁻ were not statistically different from untreated control cells, suggesting a redundant role of the two streptolysins in ER stress induction (Figure 4a). Furthermore, treatment of MEFs with purified recombinant SLO (r SLO) for 12 hr was sufficient to induce Chop and Grp78 (Figure 4b).

2.4 An ER stress-induced host-associated factor(s) contributes to GAS biofilm formation on host cells

GAS streptolysins SLO and SLS cause ER stress in host cells leading to the release of ASN into the culture media, which GAS JS95 can sense and use to regulate its gene expression and rate of proliferation (Baruch, Hertzog, et al., 2014). To determine if ER stress induction resulted in the release of a biofilm-promoting factor that could restore biofilm formation to the attenuated Δslo,sagI⁻ strain, we supplemented the double mutant with cell-free conditioned media (CM) from MEFs infected with GAS WT and streptolysin single and double mutants and assayed for biofilm formation. We observed a significant recovery in MEF-associated biofilm in the presence of CM obtained from MEFs infected with the WT and Δslo and sagI⁻strains but not the Δslo,sagI⁻ double mutant whereas CM obtained from untreated MEFs did not affect biofilm formation (Figure 5a). CM from r SLO treated MEFs was sufficient to restore biofilm formation to the double mutant strain (Figure 5b).

To examine if apoptosis of MEFs per se was responsible for promoting MEF-associated biofilm formation, we treated MEFs with TG (which induces ER stress-dependent apoptosis) and ET (which induces ER stress-independent apoptosis; R. Kim, Emi, & Tanabe, 2005) and tested if drug-treated CM were sufficient to restore biofilm formation to Δslo,sagI⁻ biofilms. CM of TG-treated MEFs, but not of ET-treated MEFs, promoted biofilm formation (Figure 5b), suggesting that ER-stress in particular, and not apoptosis in general, leads to the release of a host-associated biofilm promoting factor or factors that aids biofilm formation.

TG is a specific inhibitor of the ER lumen-associated SERCA pump that causes ER stress by Ca²⁺ dysregulation (Szegezdi, Logue, Gorman, & Samali, 2006). To confirm a role for ER stress in biofilm formation, we also tested tunicamycin, which induces ER stress by inhibiting protein glycosylation (Szegezdi et al., 2006). We observed that tunicamycin treatment, similar to TG treatment, generated a CM that could restore biofilm formation to the Δslo,sagI⁻ strain (Figure 5b). To further demonstrate the role of ER stress in biofilm formation, we pretreated MEFs 2 hr prior to GAS infection with the chemical chaperone 4-phenyl butyric acid (4-PBA) that protects cells from ER stress (H. J. Kim et al., 2013; Wiley et al., 2010). Under our assay conditions, 4-PBA pretreatment resulted in a delay in MEF cell death (Figure S7A). At 6 hpi, we observed fewer dead MEFs after pretreatment with 4-PBA compared with untreated cells. However, by 24 hpi, all MEFs were dead, indicating that early ER stress inhibition can delay MEF cell death but not abolish it. Nonetheless, we observed that GAS biofilms were modestly yet significantly reduced in biomass with 4-PBA pretreatment, suggesting that early ER stress inhibition may translate to lower biomass accumulation at 24 hpi (Figure S7B).

Because ASN is released by host cells upon induction of ER stress by SLO and SLS resulting in enhanced GAS proliferation (Baruch, Belotserkovsky, et al., 2014), we tested whether ASN is involved in MEF-associated biofilm formation. However, addition of increasing concentrations of ASN did not restore Δslo,sagI⁻ biofilm formation, indicating that under these assay conditions ASN is insufficient for complementation and that other host factors may be involved (Figure S8).

2.5 Streptolysin-mediated ER stress promotes GAS microcolony formation and spread in vivo

To assess the contribution of streptolysin-mediated ER stress in GAS microcolony and/or biofilm formation in vivo, we used a mouse model mimicking human GAS NF (Ashbaugh, Warren, Carey, & Wessels, 1998; Hidalgo-Grass, Dan-Goor, et al., 2004). We injected GAS subcutaneously and assessed bacterial burden and spread across the infected soft tissue at 12 hpi. Infection with WT resulted in high titre infection of more than 1 × 10¹⁰ CFU/g in the soft tissue (Figure 6a). To obtain high-resolution imaging of the bacteria and surrounding host tissue, we serially sectioned infected tissue biopsies, performed immunostaining, and visualised the stained sections by CLSM across the transverse anatomical plane. In WT-infected mice, GAS microcolonies were concentrated predominantly in the fascia (Figures 6b and S10A). By contrast, in tissues from mice infected with Δslo,sagI⁻, we recovered significantly fewer CFU and most commonly observed only sporadic bacteria in the fascia (Figures 6a,b). However, more extensive examination of multiple sections across the Δslo,sagI⁻ infected tissue revealed the presence of fascial microcolonies in only a few discrete regions of the tissue biopsy, whereas most other sections were entirely devoid of microcolonies (Figures 6c and S9A). This suggested that the streptolysin null mutant was unable to spread throughout the infected soft tissue and instead displayed localised aggregation. To quantify the lateral spread of microcolonies across the tissue, we measured the mean fluorescence intensity (MFI) from immunostained microcolonies in evenly spaced sections across the biopsied tissue from five mice. In contrast to WT-infected tissue in which GAS was evenly spread across the examined tissue, Δslo,sagI⁻ remained localised and did not spread (Figures 6d,e, S9A, and S10A).

Next, to address whether SLO- and SLS-induced ER stress contributes to microcolony formation in the host, we tested whether preinduction of ER stress in the infection site using TG promoted microcolony formation in soft tissue in vivo. Subcutaneous injection with TG administered 12 hr prior to inoculation with Δslo,sagI⁻ resulted in a ~100-fold increase in the number of viable bacteria recovered 12 hpi (Figure 6a) and restored the ability of Δslo,sagI⁻ to form microcolonies that could now disseminate laterally across the fascia (Figures 6c,f and S9B). Pretreatment with TG did not significantly alter the lateral or transverse distribution of WT GAS microcolonies (Figures 6h and S10B). However, pretreatment with TG increased the depth of microcolony spread across the transverse plane, from the fascia into the deep muscle region for Δslo,sagI⁻ (Figures 7a and S9B). Pretreatment of mice with ET did not affect CFU recovered from the soft tissue of WT or Δslo,sagI⁻ infected mice, compared with the recovery of those strains from phosphate-buffered saline (PBS)-pretreated mice (Figure 6a). In ET-pretreated mice, WT microcolonies and only few Δslo,sagI⁻ microcolonies were visible in the fascia and upper muscle region, respectively, and ET did not promote lateral spread of Δslo,sagI⁻ microcolonies (Figure 6g), although compared with the PBS-pretreated sections, there are more Δslo,sagI⁻ microcolonies present in the fascia (Figure 7a). ET (an apoptosis inducer) also causes host-tissue damage prior to GAS infection that may facilitate better GAS spread compared with PBS treatment alone. However, ET-associated spread is not as significant as when TG is preadministered (Figures 6f,g and 7a). Moreover, the total MFI / biopsy values as a combined measure of bacterial load across the entire lateral fascial span of each tissue biopsy showed that the WT-infected tissue had significantly greater bacterial load compared with streptolysin null-infected tissue. Upon TG, but not ET, pretreatment, we observed restoration of mutant-infected tissue MFI values to WT levels (Figure 6i).

Finally, to confirm that both GAS- and TG-induced ER stress followed by apoptosis were associated with infection in vivo, we stained adjacent tissue sections for bacteria and for markers of ER stress and apoptosis. Using antibodies against the ER stress marker protein CHOP (Li et al., 2006) or by performing TUNEL staining (Mockel et al., 2012), we observed that GAS, ER stress, and apoptosis predominated in the fascia in WT-infected mice (Figures 7a,b). In mice infected with Δslo,sagI⁻, we only detected apoptotic cells in the fascial tissue, with no sign of ER stress (Figure 7b). Pretreatment of mice with TG, followed by injection with WT GAS intensified the ER stress staining, which partially masked the few apoptotic cells present underneath (Figure 7b). Pretreatment with TG followed by challenge with Δslo,sagI⁻ increased the presence of ER stress markers and apoptotic cells in the fascia, together with the restoration of microcolony formation (Figures 7a,b). By contrast, pretreatment with ET followed by challenge with the WT or Δslo,sagI⁻ resulted in apoptotic cells in the upper layer of the muscle and the fascia, respectively, and little to no indication of ER stress, as expected (Figure 7b). The fascia of control mice injected with TG or ET, in the absence of GAS infection, stained positively for CHOP and TUNEL or for TUNEL only, respectively (Figure 7c). These findings show that markers of ER stress and apoptosis co-localize with sites of GAS microcolony formation. Together, these results establish that streptolysin-induced ER stress is important for GAS microcolony aggregation, distribution, and spread during the course of soft tissue infection in vivo.

4.1 Bacterial strains and growth conditions

All bacterial strains used in the study are described in Table S1. See Data S1 for GAS growth and culture conditions.

4.2 Cell culture and media

The cell lines used in this study were MEFs (Baruch, Hertzog, et al., 2014), HFFs (House, Vyas, & Sapolsky, 2011), and HaCaT human keratinocytes (Chan et al., 2017). Cells were cultivated in Dulbecco's modified eagle medium (DMEM + high glucose; Gibco, New York, USA) with 10% fetal bovine serum (FBS; PAA, Pasching, Austria) at 37 °C in 5% CO₂ in 24-well polystyrene plastic cell culture plates (Corning, New York, USA).

4.3 Cultivation of biofilms

Overnight cultures (16–18 hr) of GAS strains grown in Todd Hewitt yeast extract medium (Todd-Hewitt medium [Sigma-Aldrich, Missouri, USA] supplemented with 0.2% yeast extract (Becton, Dickinson, San Jose, USA) were diluted 20-fold in a 4:1 mixture of Todd Hewitt yeast extract media and DMEM supplemented with 10% FBS and grown until early log phase (OD₆₀₀ = 0.2). The bacterial cells were centrifuged at 5,000 g for 10 min, and the resulting pellet was washed twice in 1× PBS and then resuspended in 1× PBS to an OD₆₀₀ of 0.8, equivalent to 1 × 10⁸ CFU/ml. Finally, 5 × 10⁵ CFU of bacteria were seeded into 24-well polystyrene plastic cell culture plates (Corning, New York, USA) containing DMEM with 10% FBS for growth on the abiotic surface. To observe GAS growth on biotic surface, MEFs seeded at a density of 5 × 10⁴ MEFs/well in 24-well plates were infected with the same bacterial inoculum, at a multiplicity of infection of 10. The plates were incubated for static GAS growth at 37 °C with 5% CO₂ for the indicated times, with media change after every 24 hr to prevent accumulation of toxic metabolic byproducts released during cellular growth.

4.4 Biofilm quantitation

Biofilm biomass was quantified using CV staining of adherent biofilms on plastic, as previously described (Thenmozhi, Balaji, Kumar, Rao, & Pandian, 2011) with minor modifications for biofilms growing on mammalian cells (Negri et al., 2010). The detailed protocol is described in Data S1.

4.5 Immunofluorescence and microscopy

GAS biofilms were grown on 35-mm ibiTreat surface μ-Dishes (ibidi, Munich, Germany), either directly on the plastic surface or on MEFs seeded at 5 × 10⁴ cells/μ-Dish and immunostained with commercial antibodies, described in Data S1. All confocal microscopy was performed with an LSM 780 inverted laser scanning confocal microscope (Carl Zeiss, New York, USA), using a ×20 Plan-Apo 0.8 NA or ×63 Plan-Apo 1.4 NA oil immersion objective. Epifluorescence images were acquired with the Axio Observer Z1 Inverted Widefield microscope (Carl Zeiss, New York, USA) using a 20×/0.50 DICII and 63×/1.25 DICIII oil immersion objective. 3D reconstruction of biofilm images and quantification were performed using the Zen Black imaging software (Carl Zeiss, New York, USA).

4.6 Image analysis

Imaris (Bitplane AG, Belfast, United Kingdom) Version 8.4 was used to calculate biovolume of the biofilm. ImageJ (NIH, Maryland, USA) was used to quantify mean florescence intensity (MFI) of biopsy sections after immunostaining.

4.7 Cell death assays

Cell cytotoxicity was assessed using the LDH cytotoxicity determination kit (Clonetech, California, USA) according the manufacturer's directions. See Data S1 for additional details. Cell death was also assessed using a NucView 488 and RedDot 2 Apoptosis and Necrosis Kit (Biotium, California, USA). Culture media was removed from infected MEFs at the indicated time points, cells were washed with 1× PBS, and stained according to the manufacturer's protocol.

4.8 Animal experiments

GAS WT or Δslo,sagI⁻ mutant strains were injected in a mouse model of soft-tissue infection as described previously (Hidalgo-Grass et al., 2002; Hidalgo-Grass et al., 2006; Hidalgo-Grass, Dan-Goor, et al., 2004). TG or ET were injected subcutaneously into mouse soft tissue 12 hr prior to the bacterial infection. See Data S1 for additional details. All procedures were approved and performed in accordance with the Institutional Animal Care and Use Committee (IACUC) of National University of Singapore (NUS; IACUC protocol no: R14–0575) and Nanyang Technological University (NTU), School of Biological Sciences (ARF SBS/NIE [A0327]).

4.9 Statistical analysis

All experiments were repeated independently at least three times, with three technical replicates each (unless otherwise stated). Statistical significance for data from all experiments was determined using the GraphPad Prism software (Version 6.05 for Windows, California, United States). Data are expressed as mean ± SEM (unless stated otherwise), and p < 0.05 was considered significant (p < 0.05 (*); p < 0.01 (**); p < 0.001 (***); p < 0.0001 (****); “ns” denotes no statistical significance.